verbal and nonverbal cues in the perception of...

Journal of Personality and Social Psychology1978, Vol. 36, No. 4, 380-391

Verbal and Nonverbal Cues in the Perception of Lying

Robert E. KrautDepartments of Psychology and Sociology

Cornell University

Two experiments examined strategies observers use to see through self-presenta-tions. In the first, five male actor subjects lied or told the truth in simulatedjob interviews. Forty-one observers were moderately accurate in judging theactors' truthfulness. Actors were consistently good or poor liars, but judges werenot consistently good or poor. When actors lied, they gave less plausible, shorteranswers with longer latencies. Observers seemed to use the plausibility andlatency, as well as an answer's vagueness and consistency and an actor's smiling,postural shifting, and grooming, to determine whether he was lying. The secondstudy experimentally manipulated the content of an answer and a nonverbal cue.Observers were more likely to judge a female job applicant as lying when heranswers were self-serving. A long hesitation before an answer made observersmore suspicious of an already self-serving answer and more certain of the truthof an already forthright one.

According to sociological social psycholo-gists such as Mead (1934), Goffman (1959),and Turner (1968), we continually play rolesand manage the impressions of ourselves thatwe give off. We continually strive to give theimpression that we are acting consistentlywith the expectations associated with thepositions we occupy, and most of the time weget away with it. Most actors sincerely believein the roles they play, incorporating them intotheir identities, and most audiences believethat actors are what they portray (Jones &Nisbett, 1971). Indeed, Goffman (19S9) ar-gues that most social behavior occurs in whatcould be called a conspiracy mode. In thismode, we act as if we believed that a personand his or her role are identical, that the per-

This research was supported in part by NationalInstitute of Mental Health Grant MH 30041. Por-tions of this article were presented at the annualmeeting of the American Psychological Association,Washington, D. C., September 1976. William Hem-sath, Teresa Canal, Sharon Rich, Mark McNamara,Mary Chay, Linda Kupfer, and Ellen Schecter pro-vided valuable help in the research.

Requests for reprints should be sent to RobertKraut, Departments of Psychology and Sociology,Uris Hall, Cornell University, Ithaca, New York14853.

son has a legitimate right to the role, and thathe or she is adequately meeting the expecta-tions associated with it. Even if we are suspi-cious about the performance of a role partner,at least for the duration of a social interaction,we act as if we believe.

Occasionally, however, social behavior op-erates in an agonistic or sparring mode. Inthis case, the interaction resembles a contestin which the actor tries to present himself inone way and the audience tries to see throughthis presentation to the actor's real qualities.Presidential debates, job and psychiatric in-terviews, labor-management negotiating ses-sions, first dates, defense testimony, and policeinterrogations are often performed in the spar-ring mode.

The focus for this research is determininghow an observer spars. That is, how does anobserver discount, disregard, convert, or in-terpret an actor's manifest behavior to infera different underlying reality? When and howdoes the observer decide that an actor islying?

Social psychologists have suggested twogeneral rules that observers may use when theyare trying to see through self-presentationsand sort facade from reality. The first mightbe called the ulterior motive rule, a variant of

Copyright 1978 by the American Psychological Association, Inc. 0022-3S14/78/3604-0380?00.75

380

CUES IN THE PERCEPTION OF LYING 381

Kelley's discounting rule (Kelley, 1967). Ac-cording to this rule, one should discount anactor's behavior as a reflection of his or hertrue nature to the extent that the behaviorfurthers the actor's short-term self-interests.In their classic experiment, Jones, Davis, andGergen (1961) showed that observers use thisrule; they learned less about a job applicantfrom his behavior when the behavior was con-sistent with job requirements than when it wasinconsistent with them. Walster, Aronson, andAbrahams (1966), Jones and Wortman(1973), and Frankel and Morris (1976) haveall presented data and made observations con-sistent with the ulterior motive rule: Obvi-ously self-serving behavior is believed less.

The second major rule that observers mayuse to decide if an actor's behavior reflectshis or her true nature might be called thecontrollability rule. According to this rule, oneshould believe most in those aspects of a per-son's performance that the person is leastable to deliberately and consciously control(Goffman, 1959). If one cannot control it,one cannot fake it. Professionals whose taskis to distinguish truth from lie have systemati-cally looked at difficult-to-control behaviorsto determine if their clients were lying. Freudand the clinical psychologists and psychia-trists who have followed him have used thecontrollability rule by relying on dreams, as-sociations, forgettings, slips of the tongue, andother difficult-to-control behaviors to discoverproblems their patients might be hiding bothfrom the clinicians and from themselves. Po-lice use galvanic skin response (GSR) andother relatively uncontrollable physiologicalresponses in mechanized lie detectors, andtechnically less sophisticated interrogatorshave used similar, if more primitive, tech-niques for centuries (Lykken, 1974).

Research on lie detection, with few excep-tions, has assumed that uncontrollable behav-ior reveals the truth about a potential liar andhas focused on the actors and ways in whichtheir uncontrollable behavior might accuratelyreflect their internal states. A research tradi-tion in psychophysiology has tried to estab-lish the accuracy of polygraph and other liedetector techniques (see Lykken, 1974, for areview). Assuming that lying is more stressful

than truth telling because of the difficulty ofthe task, the guilt associated with deception,the fear of being found out (Davis, 1961), orthe knowledge of secret information (Lykken,1974), a polygraph should be able to measurephysiological concomitants of stress. In clini-cal psychology, the search has been for ob-servable nonverbal and other uncontrollablebehaviors that clinicians could use to assess apatient's emotional state (e.g., Kasl & Mahl,1965; Krout, 1935).

Very little systematic work has examinedwhether naive observers use the controllabilityrule to determine if someone is lying and, ifthey do, what cues they use in making theirjudgments. The exceptions to this are in workby Ekman and Friesen, by Krauss, and in thework reported below. In a series of articles,Ekman and Friesen (1965, 1967, 1969, 1974)have shown that naive observers can distin-guish between actors who are experiencingpleasant emotions or unpleasant ones whenthey view the actors' bodies, hands, and feet,but not when they view only faces and heads.Actors who were lying used fewer illustratorhand movements and had higher pitchedvoices (Ekman, Friesen, & Scherer, 1976).Research by Krauss and his colleagues(Streeter, Krauss, Geller, Olson, & Apple,1977; Krauss, Geller, & Olson, Note 1) showsthat naive observers can identify deception innonemotional interviews. They were more ac-curate when they judged from a video imageonly, especially when the interviewee did notknow that he or she could be observed; theywere next most accurate at judging from audiotape and least accurate at judging from bothaudio and visual images simultaneously. Whenthe interviewees lied, their speech had ahigher fundamental pitch and was judged tobe more nervous and less fluent. Observerswere more likely to judge less fluent, less seri-ous, less empathic, and more nervous answersas lies.

The present report describes two studies inwhich observers distinguished truths from lies.The first was exploratory and correlational.Spontaneous, simulated job interviews werevideotaped, and characteristics of the answersthat might have been associated with truthful-ness judgments were measured. The second

382 ROBERT E. KRAUT

study was hypothesis testing and experimen-tal. Both the content and nonverbal style ofone segment from a job interview were experi-mentally manipulated to test observers' useof the ulterior motive and controllability rules.

Experiment 1: Component Analysis of Lyingand Lie Detection

Method

Overview

This exploratory research first asked whether ob-servers agreed with each other and were accurate inevaluating the truthfulness of another person's state-ments and then tried to identify some of the cues in-fluencing their judgments. Five college undergradu-ates each participated in a 15-minute-long simulatedjob interview in which they lied and told the truthabout equally often. The interviews were videotaped.Each actor watched his own videotape and indicatedthe extent to which he was lying during each answer.Between 12 and 18 observers watched each tape andcompeted for a prize by judging the extent to whichthe actor was lying during each answer. Other judgesrated verbal, paralinguistic, and nonverbal character-istics of each answer.

Subjects and Procedures

Five male college undergraduates who were aboutto apply for a dormitory advisor position were re-cruited for an experiment on interviewing strategies.They participated in a simulated job interview for adormitory advisor position.

The interviewer was a graduate student in psy-chology with experience in clinical interviewing. Al-though topics had been selected in advance, the inter-view was unstructured. The interviewer asked ques-tions about the applicant's academic background,personal characteristics, friends, drug use, counselingand organizational experiences, and his potential re-sponses to such problems as depressed freshmen anddrug selling in the dormitory.

Interviewees were instructed to role play the partof a jo.b applicant, framing all their answers so thatthey gave the impression that they wanted and werequalified for the job and that they were sincere andhonest. Since they were about to apply for a similarjob within the month, there was an element of real-ism in the simulation. To heighten this realism andincrease subjects' motivation to lie and tell the truthconvincingly, the experimenter emphasized the use-fulness of this simulated interview for improving ap-plicants' job-seeking skills and promised them the op-portunity to watch their own performances on video-tape.

Behind the interviewer, in the candidate's line ofsight, was a pair of signal lights that instructed thecandidate to either lie or tell the truth on an answer.

After a S-minute warm-up in which the lights alwayssignaled "truth," the lights changed at the beginningof a question from "truth" to "lie" randomly, witha 50% probability.

Each interview lasted about 15 minutes. The inter-views were videotape-recorded, with the interviewees'knowledge. Four of the interviewees were photo-graphed from the waist up and the other from thechest up. A digital timer was also filmed, as an index-ing system, using split-screen techniques.

Deception Judgments

Following the interview, actors viewed their ownvideotape and indicated via a switch whether at thatmoment they had been lying or telling the truth orwhether the interviewer had been asking a question.As they made judgments about their interview inreal time, a PDF Lab 8/e computer sampled the posi-tion of the switch once per second and recorded upto 900 seconds of their judgments. These judgmentsbecame the standard against which all observers'judgments were compared for accuracy. This pro-cedure of determining when a lie occurred has twoadvantages over merely recording the experimentalcondition of an answer or asking for actors' global,retrospective judgments of the truth of an answer. Itcan deal conveniently with cases in which the sub-jects did not follow instructions or cases in whichthey combined truthful and deceptive elements in asingle extended answer. However, the latter was rare.Actors switched from truth to lie or the reverse dur-ing an answer in less than 10% of their answers.

Observers were recruited through campus postersoffering a $25 prize for the person best able to ac-curately distinguish lies from truth when viewing ajob interview. Recruiting observers to compete for asubstantial prize increased their interest and the im-portance of the task for them. Observers judged tworandomly selected videotapes in a dark, soundproofroom on a large video monitor. They were given anaccurate description of the manner in which thevideotapes were made. They saw 2 minutes ofwarm-up in which an actor was telling the truth andthen judged the extent to which he was lyingthroughout the rest of the interview. They madetheir judgments using the same apparatus and pro-cedures used by the actors previously. Speed and ac-curacy in making decisions were stressed. Both theapparatus and special instructions may have promptedobservers to make judgments on the basis of cuesearly in an answer and to underuse the context fol-lowing them. As a result, the present research mayunderestimate observer accuracy, compared to a morerelaxed, real-world setting, and may overestimate theimportance of some nonverbal behavior, like speechlatency, as a cue that observers use in judging decep-tion.

Between 12 and 18 observers judged each interview,and each interview contained between 14 and 38question-answer sequences. Because of unreliability inthe speed of the videotape playback equipment andthe time it took observers to begin making judgments


of an answer after listening to the question, shortanswers of less than 6 seconds have been excludedfrom these analyses. With these answers excluded,the mean answer length was 29 seconds over all fiveinterviews and ranged from a mean of 44 seconds forthe most wordy interviewee to 18 seconds for themost reticent. Over the five interviews, 106 question-and-answer sequences were analyzed.

Were observers either consensual or accurate atjudging deception? We can take a question-answersequence as the unit of analysis. Let us classify as alie any answer in which the actor indicated he liedat any time and look at the proportion of observerswho judged it a lie at the moment at which themaximum number of observers were making judg-ments. There was consensus among observers in 54%of the 106 answers, where consensus means that ob-servers significantly agreed among themselves at the5% level or less according to the sign test. There wasalso substantial accuracy. The correlation between theactors' judgments and the proportion of observersagreeing with them was .49 (p < .001).

We can also take as a unit of analysis one ob-server judging one actor's total interview and usethe phi correlation, based on the total number ofseconds that an actor and an observer agreed in judg-ing the actor's interview, as a measure of accuracy.The mean phi correlation was significantly greaterthan 0 for four of the five interviews, and for thefive interviews it ranged from .39 to .08. On theother hand, the correlation between one observer'stwo accuracy scores did not differ significantly from0 (r = — .06). These results mean that some actorswere consistently good or poor liars across differentobservers, but observers were not consistently good orpoor perceivers across different actors.

In summary, these data show moderate interob-server agreement and observer accuracy. For this re-sult to occur, observers must have been attending tosimilar or, at least, correlated cues and processingthem similarly. The search for these .cues and therules observers use to process them seems warranted.

Stimulus Characteristics

Both the prior literature on deception and on non-verbal indicators of anxiety and negative affect andobservers' introspections suggest that the followingcues may be associated with either the actual truth ofan answer or observers' judgments of its truthfulness:

1. Verbal, content cues, (a) The social desirabilityof an answer (e.g., Baskett & Freedle, 1974); (b) con-sistency between a focal answer and prior answers,that is, 'Consistency .between the answer and knowl-edge about a particular actor; (c) plausibility of ananswer, that is, its consistency with common senseand knowledge about the world generally, withoutspecific knowledge of a particular actor; (d) con-creteness, specificity, and the amount of detail in ananswer.

2. Auditory and paralinguistic cues, (a) Fluencyof speech (e.g., Boomer, 1963; Dibner, 19S6); (b) re-sponse latency (e.g., Baskett & Freedle, 1974; Krauss

et al., Note 1); (c) fundamental pitch (e.g., Streeteret al., 1977); (d) variations in pitch and loudness(e.g., Friedhoff, Alpert, & Kurtzberg, 1964; Scherer,1974); (e) length of answer (e.g., Mehrabian, 1971).

3. Visual cues, (a) Smiles (e.g., McClintock &Hunt, 1975); (b) gaze avoidance (e.g., Exline, 1970;McClintock & Hunt, 1975); (c) postural shifts (e.g.,McClintock & Hunt, 1975; Mehrabian, 1971); (d)self-manipulation (e.g., McClintock & Hunt, 197S);(e) hand and foot movements (e.g., Ekman & Friesen,1974; Krout, 1954; Luria, 1932; Mehrabian, 1971;Sainsbury, 1958); (f) tongue showing and lip tuck-ing (e.g., Smith, Chase, & Bieblich, 1974).

The exploratory research reported here examinedonly a few of these potential cues:

1. Concreteness. The extent to which the answerwas filled with detail rather than vague and lackingin detail, compared to what one would expect in re-sponse to a particular question; the extent to whichit elaborated or specified the details surrounding themajor facts.

2. A priori plausibility. The extent to which ananswer was plausible or likely to have been said by atypical college-student job candidate rather than im-plausible or unlikely. For example, a student statingthat he is friendly, studious, and unathletic is quiteplausible, but one stating that he does not know hisfather's occupation is quite implausible.

3. Consistency. The extent to which an answerwas consistent rather than inconsistent with thosethat came before in the interview; the extent towhich the context and tone of a particular answersupported or contradicted facts or general impressionsfrom earlier answers. For example, if the actor hadgiven the impression during some answers of beingshy and withdrawn and then said he partied a lot orgave the impression of being very friendly and so-cial, he would have contradicted himself. If he saidthat he hated parties, he would have been respondingconsistently. Seven raters were given descriptions ofthe concreteness, plausibility, and consistency dimen-sions and were trained on sample interviews. Usingonly transcripts, they then rated on 7-point scaleseach answer from each interview on each dimension.The rating of an answer for each dimension was theirmean judgment.

4. Length. Length of the answer in seconds.5. Percent hesitation. The time from the end of

the interviewer's question to the beginning of theactor's substantive answer divided by the length ofthe answer. The actor's substantive answer was thefirst words that he spoke, excluding filled pauses suchas "Well, let me see," "I think," "you know," "ah,""um," and the like.

6. Percent speech error. The number of noticeablebreaks in speaking indicated by stuttering, repetitionof phrases, broken phrases, or the insertion of non-substantive phrases divided by the length of theanswer. This set of speech errors combines Kasl andMahl's (1965) ah, repetition, and stuttering cate-gories. Both the hesitation and speech error measureswere coded by one trained rater working from audiorecordings and verbatim transcripts of the interviews.

384 ROBERT E. KRAUT

7. Percent smiling. The seconds of smiling dividedby the length of the answer. This measure was codedby two to five trained raters using event recordersand watching only the video portion of each inter-view. The mean Pearson correlation reliability esti-mate of the number of seconds of smiling during ananswer for the five interviews was .80.

8. Percent postural shift. The seconds in whichthe interviewee moved his whole body back andforth or rotated it in a swivel chair divided by thelength of the interview. This measure was coded by-three trained raters using event recorders and watch-ing only the video portion of each interview. Themean Pearson correlation reliability estimate of thenumber of seconds of postural shifting during an an-swer for the five interviews was .69.

9. Percent grooming. The seconds in which the in-terviewee touched and manipulated his body or facewith his hand divided by the length of the interview.Scratching, smoothing, and stroking are examples,but holding one's chin stationary in one's hand is not.Because one actor was photographed from the chestup and other actors had their hands offscreen forvarious lengths of time, the grooming measure repre-sents only grooming and self-manipulation seen byobservers but not all the grooming that actually oc-curred during an interview. This measure was codedby three trained raters using event recorders andwatching only the video portion of each interview.The mean Pearson correlation reliability estimate ofthe number of seconds of grooming during each an-swer for the five interviews was .77.

Results

Figure 1 summarizes the relationshipsamong the truth of an answer as judged by

the actor, the stimulus characteristics of theanswer as judged by independent raters, andobservers' judgments of the answer's truth.The actor's judgment measure is coded 0 for alie and 1 for a truthful answer. The observers'judgment measure is the mean percentage oftime that observers judged an answer as trueminus the mean percentage that they judgedit as a lie.

Figure 1 shows that when actors told thetruth, they gave longer, more plausible an-swers with shorter hesitations before theiranswers. Observers judged more plausible an-swers with shorter hesitations as true, and theuse of these cues increased their accuracy. Inaddition, observers were more likely to judgeas true consistent and concrete answers inwhich the actor smiled less, shifted his posi-tion less, and groomed himself more. The useof these cues did not increase observers' ac-curacy because they were unrelated to the ac-tual truth of an answer.

It is interesting to note that the plausibilityof an answer was the single best predictor ofboth its actual truth and observers' judgmentsof its truth. This verbal cue, a characteristicof the content of the answer, was measuredby judges who rated transcripts of the inter-views while denied paralinguistic and visualinformation. Although nonverbal cues addedto the prediction, they may not have had the

STIMULUS CHARACTERISTICS

CONCRETENE6S

PLAUSIBILITY

CONSISTENCY

HESITATION

TRUTH OF ANSWERCACTORS' JUDGMENT5

SPEECH ERROR

SMILE

POSTURAL SHIFT

GROOM

.IE .LENGTH

- .ZDX

TRUTH - LIECOBSERVERS' JUDGMENT3

Figure 1. Pearson correlations for truth of answer, stimulus characteristics, and observers' judgments.


overwhelming importance that some research-ers (e.g., Archer & Akert, 1977) have ascribedto them.

Forty-one percent of the variance in ob-servers' judgments can be accounted for bythe truth of the answers and rated stimuluscharacteristics: The multiple correlation coef-ficient is .64. Thirty-two percent of the vari-ance in observers' judgments can be accountedfor by the measured stimulus characteristic ofeach answer: The multiple correlation coeffi-cient is .56. Eighteen percent of the variancein observers' judgments can be accounted forby the actual truthfulness of an answer: Thepoint biserial correlation is .43. Of this 18%,which we might think of as accuracy variance,10% is mediated by measured stimulus char-acteristics, and another 8% is not. The partialcorrelation of the actual truth of an answerand observers' judgments of its truth, withthe measured stimulus characteristics held sta-tistically constant, is .31.

To summarize in less technical languagethis partitioning of the variance in observers'judgments, observers were moderately accu-rate in judging whether an actor was lying ortelling the truth. In addition, one can predictreasonably well whether observers thought anactor was lying or telling the truth on an an-swer by knowing the measured stimulus char-acteristics of each answer. However, observers'accuracy was based only in part on these stim-ulus characteristics, since holding them con-stant statistically through partial correlationsleft about 46% of the accuracy variance stillunaccounted for.

Why is it that the measured stimulus char-acteristics can account for more of the ob-servers' truthfulness judgments but less oftheir accuracy? One possibility is that thepresent experimental situation is a relativelyaffect-free one, in which the guilt and fear ofexposure that characterize lying outside thelaboratory were not sufficiently aroused. Sincenonverbal affect cues themselves did not dif-ferentiate the actors' lying from truth telling,observers could not use them to increasetheir accuracy. However, it follows from themoderate accuracy of observers' truthfulnessjudgments that they were basing their judg-ments on some public behaviors, perhaps asyet unmeasured. It may be that they were

using subtler, second-order cues, such as thebreakdown of the normal patterning of con-versation, as well as the first-order cues mea-sured, such as speech errors or postural shifts.If these subtler cues were less controllable bythe actors, they could be the basis of ob-servers' accuracy. That is, observers may besensitive to violations of some structural rulesof social interaction (e.g., Birdwhistle, 1970;Duncan & Fiske, 1977) as well as to the oc-currence of a few meaningful nonverbal be-haviors (e.g., Dibner, 1956; Mehrabian,1971). However, the view that observers aresensitive to subtle violations of normal pat-terns becomes less plausible when one con-siders that they were not even sensitive toindividual differences in the frequency of be-havior, as demonstrated by the analysis ofstandardized measures discussed below.

Standardizing the measures of stimuluscharacteristics for differences between actorsdid not increase these cues' ability to predictobservers' deception judgments. Actors showedlarge differences in how often they displayeda nonverbal behavior. For example, the meanpercentage of time that different actors smiledranged from less than 1% to over 12%, andthe mean percentage of time they spent shift-ing in their seat ranged from less than 7% toover 41%. Yet, using scores standardized forindividual differences in a regression analysisaccounted for 22% of the variance in ob-servers' deception judgments, versus 23% forthe nonstandardized scores. Observers seemedless sensitive to idiosyncratic patterns of be-havior and deviations from them than onemight expect (cf. Ekman & Friesen, 1974).

Experiment 2: Pauses and Contexts

The research just described, as well as pre-vious research (e.g., Baskett & Freedle, 1974;Ekman & Friesen, 1974), has shown that ob-servers can use both verbal and nonverbalcues to judge the truthfulness of a statement.It is unlikely that they are continually moni-toring all potential cues, if only because ofthe information overload this would cause.Verbal statements are simultaneously the mosteasily accessible sources of information abouta speaker and yet the least reliable, since theyare so easily controlled. One way that ob-

386 ROBERT E. KRAUT

servers could cope with the potential informa-tion overload and the potential unreliabilityof verbal statements is to apply the ulteriormotive rule and the controllability rule se-quentially. That is, they can accept at facevalue all that speakers say about themselvesuntil a self-serving, implausible, contradictory,or otherwise suspicious statement is made. Atthis point, observers can shift down a level ofcontrollability and seek evidence confirmingtheir suspicions from information sources lesssubject to the speaker's conscious control.

The experiment described below examinedthe ways in which observers might use thesetwo rules in judging a stimulus person'struthfulness. One hypothesis is a conceptualreplication of Jones et al. (1961): Observerswill believe a stimulus person's self-servingstatements less than those that are not self-serving. The second hypothesis is about ob-servers' attempts to deal with the tension be-tween the accessibility and reliability of ver-bal statements. Observers will use relativelyuncontrollable nonverbal cues to modify theirjudgments of the truthfulness of an answerbased on its verbal content. Specifically, anonfluency in a stimulus person's speech, suchas a long hesitation before answering, will bea cue to an observer that the stimulus personis lying when the hesitation precedes an al-ready suspicious, self-serving statement, butnot when it precedes a self-damaging one.

Method

Overview

Seventy-four subjects listened to a S-minute excerptfrom a simulated interview in which a female ap-plicant applied for a job as a dormitory counselor.When the male interviewer asked if the candidatesmoked marijuana, he gave the impression that heeither strongly opposed its use or supported its use.The job candidate answered either that she did notsmoke it and found its use distasteful or that shesmoked it recreationally several times a week. Heranswer was preceded by either a 7-second pause ora 1-second pause. Subjects estimated the job can-didate's use of marijuana and related attitudes, in-dicated their confidence in these estimates, and judgedthe candidate's honesty in the interview.

Stimulus Materials

The excerpt that subjects heard was taken fromtape recordings of a spontaneous but simulated job

interview in which a candidate applied for a positionas a dormitory counselor. The interviewer's and in-terviewee's instructions were to act as if the simula-tion were an actual job interview, with the addi-tional requirement that they give the impression thatthey either approved of marijuana use in one ver-sion of the interview or disapproved of its use inthe other. The interview was conducted with inter-viewer and interviewee separated, communicatingwith each other through microphones and earphones.

The S-minute excerpt that subjects heard wascreated by editing and splicing together three ques-tions and answers from the tape recording of theinterview and rerecording the edited version. Thefirst and third questions and answers were the samefor all experimental conditions. In the first, the in-terviewer asked about the candidate's experience inhandling discipline problems in a dormitory. Thecandidate discussed students' use of forbidden elec-trical appliances and claimed to have handled prob-lems successfully by gaining her students' respect andappealing to their empathy for fellow students. Inthe third question, the interviewer asked how thecandidate would deal with suspected drug selling inthe dormitory. She replied that she would remind thesuspected seller of the seriousness of this action andof the severe university and legal sanctions andthreaten to report the seller to a dean.

The second question and answer and the candi-date's paralinguistic behavior prior to this answerwere experimentally manipulated. Both versions ofthe interviewer's question asked if the candidatesmoked marijuana. One version of the question gavethe impression that he disapproved of marijuana useand the other that he approved of its use. The anti-marijuana question was phrased as follows:

We've just recently been having trouble with drugsin the dorm. A lot of the kids, especially freshmen,have been getting really offended when they haveroommates who smoke dope or when they go to adorm party and there's dope being smoked, andthey have a perfect right to be. The dorm is es-sentially their home, and they should be able toallow what they want to occur in their home . . .What's more, we've been getting some complaintsfrom parents who seem to be worried that this easyaccess to drugs and having role models around whogive the impression that smoking is OK and that. . . students don't have to spend much time inacademics are going to lead their kids into thesame sort of thing. I kind of agree with whatthose parents are saying. I am lucky that I don'thave kids, but if I did, I might be worried thesame way. In a second I'm going to ask you whatyou think you would be doing if you found outthat there was drug selling in your dorm. But be-fore I do, let me ask you this: Do you use drugs?

The promarijuana question was phrased in the fol-lowing way:

One of the problems that we're really having inthe dorms now is the use of drugs. I really don't


care what kids do in the dorms for .their own per-sonal use, but we've been getting a lot of com-plaints . . . both from the kids themselves and justrecently from some of their parents, about peoplewho are using drugs and especially grass in thedorms. There are a bunch of kids that must havecome from convents in the midwest or something—I don't know—but they get really upset when theyget assigned a roommate who smokes or they goto a party where there is smoking. And they alsohave been complaining to their parents and it getsback to the dean . . . We may have to figure outsomething to do about the indiscreet use of drugsin the dorm, and I'm going to ask you what youthink you might do. But before I do that, just letme ask you whether you use drugs?

The job candidate answered this question by say-ing either that she did not currently smoke marijuanaand found its use distasteful or that she smoked sev-eral times a week.1 The antimarijuana answer was:

No, I had at one point in time used it on a coupleof occasions. (Q: What sort of thing?) Pot, grass.I found it distasteful, frankly; I don't smoke other-wise. (Q: you mean cigarettes?) I don't smokecigarettes. I found the process distasteful, for start-ers. I didn't enjoy it. So frankly, I saw no sense tocontinuing it. As well as I have seen people . . .when I was in college, which was when I triedsome, use it a lot and become potheads and dis-sipate what I thought was some good potential, andso I guess I never got really interested in it. Drink,sometimes, but not very often.

The promarijuana answer was:

Yes, I do use drugs. Probably on the average oftwice, maybe three times, a week. (Q: What sort ofthing?) I smoke grass. And that's about it. But Ido, and feel, of course in so far as I do live withone other roommate and it's not bothersome toher, that it doesn't matter. I enjoy it. It helps meto relax some. So it's something I do twice or threetimes a week.

Silence Manipulation

The paralinguistic cue was manipulated by in-serting a 7-second partially filled pause between theinterviewer's marijuana question and the candidate'sanswer. Four seconds of blank tape, an "uh" spokenby the candidate and taken from another of her an-swers, and 3 seconds of blank tape were spliced intothe interview, starting at the last sounds of thequestion. This length of silence seems to be at thelimits of those that appear in normal conversation, atleast with a fluent speaker like the present job candi-date. Thus, the silence, according to pretest results,was noticed by virtually all subjects but did not ap-pear unnaturally long to them.2 In the other version,no silence was inserted between the question and an-swer, and the naturally occurring hesitation of ap-proximately 1 second was retained.

Subjects and Procedures

Seventy-four undergraduates were randomly as-signed to listen to one version of the interview andmake judgments about the interviewer and candidateand rate the confidence of their judgments. Subjectswere reminded that people often try to manage theimpression they give off and that their task was todiscover what a job candidate was really like afterlistening to a brief excerpt from a job interview. Inaddition, they were informed that the candidate hadpreviously been a dormitory counselor and thereforehad considerable experience both in performing thejob and in job interviews. Only the 11 subjects whoasked were told that the interview was simulated.They were spread approximately evenly across condi-tions. Subjects listened to the interview over ear-phones. The experimenter giving instructions wasblind to the subjects' experimental conditions through-out their participation.

Results

The dependent measures represent twogroups. The first comprised subjects' judg-ments about the amount the candidate smokedmarijuana and attitudes that are consistentwith this, such as her attitude toward mari-juana and her political ideology. The secondgroup consisted of judgments of the candi-date's honesty, how well subjects thought theyknew the candidate based on the interview,the amount she distorted her views in order toimpress the interviewer, the amount that shelied about her marijuana smoking, and hercandor and honesty in the interview.

Subjects' judgments of the amount that thecandidate smoked marijuana and related at-titudes were influenced almost exclusively bywhat the candidate said. Subjects believedthat she smoked more, F(l, 66) = 114.8; hada more positive attitude toward marijuana, F(1, 66) = 123.8; and had a more liberal po-litical ideology, F(l, 66) = S3.4, when shesaid that she smoked marijuana than whenshe said that she did not. These results seem

1 Neither answer was true. The .candidate, in fact,smoked marijuana socially, neither finding its use dis-tasteful nor smoking it regularly.

2 It is interesting to note that a 7-second pausewithout the interjected "uh" gave the impressionthat the attention of the candidate had wandered andthat she had suddenly returned it to the interview atthe start of her answer. With the interjected "uh,"the silence gave the impression of a person in thought,making up her mind what to say.

388 ROBERT E. KRAUT

Table 1Observers' Judgments of an Experimentally Manipulated Job Interview

M

Antimarijuana question Promarij uana question

No pause Pause No pauseF tests

PauseQues-

Anti Pro Anti Pro Anti Pro Anti Pro tion Pause AnswerDependent measure answer answer answer answer answer answer answer answer (Q) (P) (A) Q XA PXQ PXA

Candidate'smarijuana attitude

Confidence injudging candidate'smarijuana attitude

Candidate's honestyCandidate's truthful-

ness aboutmarijuana use

How well knowncandidate is onbasis of interview

How much candidatedistorted view toplease interviewer

Interviewer'smarijuana attitude

2.20

4.903.20

2.80

5.20

3.30

2.10

5.12

4.873.62

4.38

5.75

2.83

3.00

2.40

4.002.70

2.00

4.90

3.30

2.20

5.11

6.334.89

4.78

4.00

2.00

1.67

2.44

3.563.89

4.00

5.44

2.78

3.12

5.00

5.102.82

3.20

4.90

3.59

2.90

2.33

3.673.00

2.67

5.78

2.78

4.33

5.56

6.004.11

3.89

4.44

3.00

3.78 18.7

123.0

16.3 4.44.2 4.7 10.0

10.0 7.1 5.0

3.6 4.3 3.8 2.9

5.7

7.4

Note. All means were based on 7-point scales and have been recoded so that a higher number indicates more favorable marijuanaattitudes (strongly disapproves to strongly approves), greater confidence in making judgments (very low confidence to very highconfidence), greater honesty (very dishonest to very honest), greater truthfulness (completely lying to completely truthful) , moreknowledge of the candidate (not at all known to very well known), and less distortion (very much to not at all). There were 1and 66 degrees of freedom for each analysis of variance. Only the results of F tests significant at the .05 level or lower are included.

only to reflect the subjects' repetition of theinformation they received, without substantialcognitive work on their part (see Table 1).

Two sets of results show that observersused the ulterior motive rule to judge thecandidate's honesty. As shown in Table 1,the candidate was seen as more honest andtruthful, and observers believed they knew herbetter and were more confident in judging hermarijuana attitude, when she admitted smok-ing marijuana several times a week than whenshe denied smoking it. When subjects wrotebrief statements describing how they madetheir judgments, 53% of them discussed thesocial desirability of marijuana smoking, andall of these subjects indicated that marijuanasmoking decreased one's chances of obtainingthe counselor job and not smoking increasedthem. Thus, subjects believed that the candi-date was more honest when she said shesmoked marijuana, because in doing so sheworked against her own self-interest.

The second set of findings provides moreconvincing evidence that subjects were usingthe ulterior motive rule to judge the job can-didate's truthfulness. Within limits, subjectsshould believe that the candidates' agreementwith a superior is ingratiating (Jones, 1964).

If subjects were using the ulterior motiverule, they should have discounted the agree-ing statements as possibly motivated by self-interest and put trust in the disagreeing state-ments, even though the content and deliveryof these answers were identical, since theywere copies of the same piece of audiotape. Infact, subjects did this on all appropriate de-pendent measures. While, in general, the anti-marijuana answer was believed less than thepromarijuana answer, the antimarijuana an-swer following an antimarijuana question anda promarijuana answer following a promari-juana question were believed least. In thiscase, the candidate was believed less whenshe agreed with a powerful other.

The most interesting results involve thepause. Consistent with the controllability rule,subjects used nonverbal cues to judge thetruthfulness of a statement, but in ways morecomplex than might be first expected. The re-sults can be best summarized by the statementthat the nonverbal cue acted as an amplifierfor the verbal content.

The 7-second pause increased subjects' sus-picion of the candidate when they were al-ready suspicious. Compared to subjects whoheard only the candidate's denial of marijuana


use, subjects who heard a long pause and thenthe denial thought the candidate had beenless candid and lied more in the interview, be-lieved they knew her less well on the basis ofthe interview, and were less confident in mak-ing a judgment about her marijuana attitude.On the other hand, when they already thoughtthe candidate was being truthful, the pausefurther increased this belief. Compared tosubjects who only heard the candidate admitto smoking marijuana, subjects who heard along pause and the admission thought she hadbeen more candid in the interview and morehonest about her marijuana use, believed theyknew her better on the basis of the interview,and were more confident in making judgmentsabout her marijuana attitudes (for all com-parisons, .01 < p< .05).

This interaction of the silence and the can-didate's attitude manipulations suggests thatsubjects were sensitive to irregularities in asample of speech and then used available cluesto search for the meaning of the irregularity.Goldman-Eisler (1968) has shown that hesita-tions in speech are multiply caused. In thepresent experiment, the pause served as anindicator to subjects that the candidate'sthought and speech processes were disturbedfor some reason—that some private and in-ternal processing was delaying the start ofher answer; the verbal content following ithelped provide the reason. If the answer fol-lowing the pause was an obviously self-servingone, the answer was interpreted as a lie, andthe pause was interpreted as the time neededto create the lie. This consistency of motiveand opportunity convinced more subjects thata lie was being told than did either alone. Ifthe statement following the pause was a self-damaging one, the pause provided the timeboth to decide to say it and to phrase it inthe least damaging way. If the candidateneeded to think before her admission and stilladmitted marijuana use, she must have con-sidered her admission to be especially damag-ing. According to the ulterior motive rule,especially damaging statements are especiallytruthful. Thus, it seems likely that subjectsused the candidate's statements to make non-verbal behavior less ambiguous and then usedthe nonverbal behavior to further comment onthe statement.

The interaction of the silence and inter-viewer's attitude manipulations works simi-larly. The interviewer's questions seemedmore leading when they were followed by apause. Both interactions suggest that the in-terpretative process used by subjects is re-cursive rather than linear and additive.

Discussion

A danger in research on impression man-agement and on the detection of lying is totreat verbal and nonverbal cues associatedwith deception as if they were analogous tocues associated with emotion and, thereby, un-derestimate the importance of the context inproviding them with meanings (cf. Morris,1977). It is unlikely that any behaviors areinvariably linked to deception attempts. Itmay be more helpful to think of the informa-tion that signals deception to an audience ascomprising two fundamentally different types,performance cues and motivational cues,which differ in the extent to which they areseparable from the specifics of a deception at-tempt and the extent to which they depend oncontext to provide them with meaning. Withperformance cues, the audience perceives thatan actor has failed to adequately control someaspect of his deceptive performance. As suchthey are inseparable from the specifics of aperformance. Ekman and Friesen's (1969)leakage and deception cues are subsets of theperformance cue category. With leakage cues,some of the to-be-hidden information is trans-mitted despite the actor's deception attempt.The information may be transmitted eitherverbally or nonverbally and, depending on thecontent, can often be interpreted indepen-dently of context. With deception cues, theactor also loses control of his performance, butwithout revealing the content of his decep-tion. Deception cues can also be verbal or non-verbal, and because they can be caused bymany stressors other than lying, they oftenneed a context to be interpretable.

Motivational cues are situations or verbalcontexts in which deception is likely. Unlikeperformance cues, which are inseparable fromwhat a particular actor does, these motiva-tional cues often have their origins in socialnorms outside the actor's performance. These

390 ROBERT E. KRAUT

norms provide the standards against which aparticular performance is judged. For ex-ample, the need for a favorable self-presenta-tion, and therefore the likelihood of a decep-tion attempt, is built into job interviews, po-lice-citizen encounters, and leading questions.An audience compares the behavior demandedby a situation with an actor's performance todetermine if deception could plausibly haveoccurred.

In Experiment 1, the cues used by observerscan all be seen as instances of performancecues. The experimental instructions providedmotivation to the actors to lie. Thus, behav-ior that seemed to observers to indicate thatthe actor had failed to control his performanceon either a verbal or nonverbal level was evi-dence of deception, although it may not havebeen in other contexts. In Experiment 2, aperformance cue, the long pause, was seen asevidence of deception only in an appropriatemotivational context, when it preceded a self-serving answer.

In summary, the results from these experi-ments show that (a) observers were moder-ately accurate in judging whether an actor waslying or telling the truth in a simulated jobinterview; (b) actors were consistently goodor bad liars, but observers were not consist-ently good or bad at detecting lies; (c) actors,when they lied, were more likely to give lessplausible and shorter answers with longerhesitations prior to starting; and observersseemed to use the plausibility and the lengthof hesitation, as well as the vagueness andlack of detail of the answer and its consistencywith earlier statements, the amount the actorsmiled, shifted and groomed, and other, asyet unidentified, cues to determine whetherthe actor was lying. The first two of thesecues helped them make accurate statementswhereas the last five did not. (d) In addition,a nonverbal cue can amplify the suspicious-ness of a statement, with a long pause preced-ing a statement causing a prior self-servingstatement to be believed less and a prior self-damaging statement to be believed more.

The research described here, in which naivesubjects made judgments of the truthfulnessof either simulated or experimentally con-trived job interviews, has the advantage thatthe experimenter could specify the information

that the observer received, either by coding itor manipulating it. Both have the disadvan-tage that the simulated or contrived interviewformat probably played havoc with the cuesassociated with deception in nonexperimentalsettings. This may be especially true becauseboth experiments used simulations withinsimulations, experimentally induced lies withina mock job interview. In Experiment 1, espe-cially, actors may not have behaved as theywould if the deception attempt had beenmore consequential.

The problems of ecological validity andsmall numbers of actor subjects should makeone cautious about generalizing from thepresent results. On the other hand, the simul-taneous search for cues that actors are emit-ting and cues that observers are receiving isvaluable and can be made more so by extend-ing the search to more natural settings. In ad-dition, supplementing the present passive, in-formation-processing approach with an in-vestigation of observers' attempts to activelyprobe deception would make research on de-ception, particularly, and person perception,generally, more realistic.

Reference Note

1. Krauss, R. M., Geller, V., & Olson, C. Modalitiesand eves in the detection of deception. Paper pre-sented at the meeting of the American Psychologi-cal Association, Washington, D.C., September 1976.

References

Archer, D., & Akert, R. M. Words and everythingelse: Verbal and nonverbal cues in social interpreta-tion. Journal of Personality and Social Psychology,1977, 35, 443-449.

Baskett, G., & Freedle, R. O. Aspects of languagepragmatics and the social perception of lying. Jour-nal of Psycholinguistic Research, 1974, 3, 112-131.

Birdwhistle, R. Kinesics and context. Philadelphia:University of Pennsylvania, 1970.

Boomer, D. S. Speech disturbances and body move-ment in interviews. Journal of Nervous and Men-tal Disease, 1963,136, 263-266.

Davis, R. C. Physiological responses as a means ofevaluating information. In A. D. Biderman & H.Zimmer (Eds.), The manipulation o.f human be-havior. New York: Wiley, 1961.

Dibner, A. S. Cue-counting: A measure of anxiety ininterviews. Journal of Consulting Psychology, 1956,20, 475-478.

Duncan, S., & Fiske, D. W., Face-to-face interaction.Hillsdale, N.J.: Erlbaum, 1977.


Ekman, P. Differential communication of affect byhead and body cues. Journal of Personality andSocial Psychology, 1965, 2, 726-735.

Ekman, P., & Friesen, W. V. Head and body cues inthe judgment of emotion: A reformulation. Per-ceptual and Motor Skills, 1967, 24, 711-724.

Ekman, P., & Friesen, W. V. Nonverbal leakage andclues to deception. Psychiatry, 1969, 32, 88-106.

Ekman, P., & Friesen, W. V. Detecting deceptionfrom the body or face. Journal of Personality andSocial Psychology, 1974, 29, 288-298.

Ekman, P., Friesen, W. V., & Scherer, K. R. Bodymovement and voice pitch in deceptive interaction.Semiotica, 1976, 16, 23-27.

Exline, R. V. Visual interaction: The glances ofpower and preference. In J. Cole (Ed.), NebraskaSymposium on Motivation (Vol. 17). Lincoln:University of Nebraska Press, 1970.

Frankel, A., & Morris, W. Testifying in one's owndefense: The ingratiator's dilemma. Journal ofPersonality and Social Psychology, 1976, 34, 475-480.

Friedhoff, A. J., Alpert, M., & Kurteberg, R. L. Anelectro-acoustic analysis of the effects of stress onvoice. Journal of Neuropsychiatry, 1964, 5, 266-272.

Goffman, E. The presentation of self in everyday life.Garden City, N.Y.: Doubleday Anchor, 1959.

Goldman-Eisler, F. Psycholinguistics: Experimentsin spontaneous speech. New York: Academic Press,1968.

Jones, E. E. Ingratiation. New York: Appleton-Cen-tury-Crofts, 1964.

Jones, E. E., Davis, K. E., & Gergen, K. J. Role-playing variations and their information valuefor person perception. Journal of Abnormal andSocial Psychology, 1961, 63, 302-310.

Jones, E. E., & Nisbett, R. E. The actor and the ob-server; Divergent perceptions of the causes of be-havior. Morristown, N.J.: General Learning Press,1971.

Jones, E. E., & Wortman, C. Ingratiation: An at-tributional approach. Morristown, N.J.: GeneralLearning Press, 1973.

Kasl, S. V., & Mahl, G. F. The relationship of dis-turbances and hesitations in spontaneous speech toanxiety. Journal of Personality and Social Psy-chology, 1965, 1, 425-433.

Kelley, H. H. Attribution theory in social psychology.In D. Levine (Ed.), Nebraska Symposium on Mo-tivation (Vol. 15). Lincoln: University of Ne-braska Press, 1967.

Krout, M. H. An experimentation attempt to deter-mine the significance of unconscious manual sym-bolic movements. Journal of General Psychology,1954,57, 121-152.

Luria, A. R. The nature of human conflicts. London:Liveright, 1932.

Lykken, D. T. Psychology and the lie detector in-dustry. American Psychologist, 1974, 29, 725-739.

McClintock, C. C., & Hunt, R. G. Nonverbal indi-cators of affect and deception in an interview set-ting. Journal of Applied Social Psychology, 1975, 5,54-67.

Mead, G. H. Mind, self, and society. Chicago: Uni-versity of Chicago Press, 1934.

Mehrabian, A. Nonverbal betrayal of feeling. Journalof Experimental Research in Personality, 1971, S,64-73.

Morris, D. Nonverbal leakage: How you can tell ifsomeone's lying. New York, October 17, 1977, pp.43-46.

Reid, J. E., & Inbau, F. E. Truth and deception.Baltimore, Md.: Williams & Wilkins, 1966.

Sainsbury, P. Gestural movement during psychiatricinterviews. Psychosomatic Medicine, 1958, 17, 458-469.

Scherer, K. R. Acoustic concomitants of emotionaldimensions. In S. Weitz (Ed.), Nonverbal com-munication. New York: Oxford University Press,1974.

Smith, W. J., Chase, J., & Bieblich, A. K. Tongueshowing: A facial display of humans and otherprimate species. Semiotica, 1974, 11, 201-246.

Streeter, L. A., Krauss, R. M., Geller, V., Olson, C.,& Apple, W. Pitch changes during attempted decep-tion. Journal of Personality and Social Psychology,1977, 35, 345-350.

Turner, R. Role: Sociological aspects. In D. Sills(Ed.), International encyclopedia of the social sci-ences. New York: Macmillan, 1968.

Walster, E., Aronson, E., & Abrahams, P. On increas-ing the persuasiveness of a low prestige communi-cator. Journal of Experimental Social Psychology,1966, 2, 325-342.

Received March 28, 1977 •

verbal and nonverbal cues in the perception of...

Documents