Cognitive Biases and Nonverbal Cue Availability in Detecting Deception

HUMAN COMMUNICATION research
Human Communication Research ISSN 0360-3989
ORIGINAL ARTICLE
Cognitive Biases and Nonverbal Cue Availability in Detecting Deception

Judee K. Burgoon1, J. Pete Blair2, & Renee E. Strom3
1 Center for the Management of Information, University of Arizona, Tucson, AZ 85719 2 Department of Criminal Justice, Texas State University, San Marcos, TX 78666 3 Department of Communication Studies, St. Cloud State University, St. Cloud, MN 56301
In potentially deceptive situations, people rely on mental shortcuts to help process information. These heuristic judgments are often biased and result in inaccurate assessments of sender veracity. Four such biasestruth bias, visual bias, demeanor bias, and expectancy violation biaswere examined in a judgment experiment that varied nonverbal cue availability and deception. Observers saw a complete videotaped interview (full access to visual, vocal, and verbal cues), heard the complete interview (vocal and verbal access), or read a transcript (verbal access) of a truthful or deceptive suspect being questioned about a mock theft and then rated the interviewee on information, behavior, and image management and truthfulness. Results supported the presence of all four biases, which were most evident when interviewees were deceptive and observers had access to all visual, vocal, and verbal modalities. Deceivers messages were judged as increasingly complete, honest, clear, and relevant; their behavior as more involved and dominant; and their overall demeanor as more credible, with the addition of nonverbal cues. Deceivers were actually judged as more credible than truthtellers in the audiovisual modality, whereas best discrimination and detection accuracy occurred in the audio condition. Results have implications for what factors influence judgments of a senders credibility and accuracy in distinguishing truth from deception, especially under conditions where senders are producing messages interactively. doi:10.1111/j.1468-2958.2008.00333.x
Cognitive biases, nonverbal cue availability, and deception detection
One of the most well-documented claims in the deception literature is that humans are poor detectors of deception. A recent meta-analysis reveals that although people show a statistically reliable ability to discriminate truths from lies, overall accuracy rates average 54% or only a little above chance (Bond & DePaulo, 2006). A primary causal mechanism cited for biased judgments of deception and credibility is reliance on heuristic social information processinga nonanalytic orientation to
Corresponding author: Judee K. Burgoon; e-mail: jburgoon@cmi.arizona.edu This article was accepted under the editorship of Jim Dillard.
572
Human Communication Research 34 (2008) 572599 2008 International Communication Association
J. K. Burgoon et al.
information processing in which only some informational cues are carefully considered (Chaiken, 1980; Todorov, Chaiken, & Henderson, 2002). As mental shortcuts, people invoke cognitive heuristics, simple decision rules that arise from conventional beliefs and expectations and that are used repeatedly in daily interactions (Tversky & Kahneman, 1974). These mental shortcuts may yield biased information processing and faulty judgments of others veracity (Fiedler, 1993). Four especially salient and potentially interrelated biases are truth bias (the tendency to overestimate others truthfulness), visual bias (the tendency to place more reliance on visual than vocal, linguistic, and other forms of social information), demeanor bias (the tendency to judge some senders communication styles as credible irrespective of their actual truthfulness), and expectancy violations bias (the tendency to judge unusual behavior as deceptive). Together, these biases may account not only for poor detection of deception but also more generally for judgments of communicator credibility. The interrelationships among these biases have not been investigated previously. It may be that some are subordinate to, or artifacts of, others. The visual bias, for example, may be the product of demeanor and expectancy violations biases or it may be a product of other factors such as the information richness of the medium. Thus, a central objective of the investigation to be reported was to examine the interrelationships among these biases and their ultimate impact on veracity judgments. A second objective was to test these biases when judgments are applied to the kinds of message exchange that typify normal, ongoing interaction. The Bond and DePaulo (2006) meta-analysis, though quite comprehensive, included very few studies in which the stimuli that were judged when produced under fully interactive conditions, that is, ones in which senders engaged in ongoing and interdependent social interaction with the intended targets of their deceit.1 Given that deception typically is embedded in ongoing interaction rather than judged in isolation, and given that judgments made of naturalistic interaction differ from those made of brief, experimentally controlled stimuli (Motley & Camden, 1988), knowledge of how people make veracity judgments should be founded on the kinds of stimuli they normally encounter rather than on brief, decontextualized snippets. That the bulk of experimental stimuli have been less than 60 seconds in length (see Bond & DePaulo, 2006; DePaulo et al., 2003) renders most of the extant literature mute as to what happens beyond the rst minute of interaction. It may be that as a deceptive episode unfolds, deception becomes more difcult to detect because deceivers capitalize on the features of interpersonal interaction to regulate their performances more effectively and thus evade detection (Burgoon & Buller, 2004). Conversely, messages intended to deceive interlocutors might be more transparent in their intent and therefore more readily detected as observers gain extended exposure to the subtleties of deceptions enmeshed in the ongoing conversational context and they consider simultaneous or serial incongruities in different information streams, such as when a pleasant face accompanies a strained voice. The Bond and DePaulo (2006) meta-analysis results suggest such an explanation.
573
The issue of what biases influence judgments of a persons veracity under conditions of interactive message production was examined in a factorial experiment in which the stimuli to be judged were interviewees who had been questioned about a mock theft. Observers judged a truthful or deceptive interview under one of the three modalities: text, audio, or audiovisual (AV). The modality manipulation tested how the addition or deletion of visual and vocal nonverbal demeanor cues affected judgments. Observers in the text condition had access only to a transcript of an interview and so had no access to nonverbal demeanor cues. Observers in the audio condition heard a recorded interview and so had access to both words and voice, thus exposing them to vocalic demeanor cues and to possible channel discrepancies. Those in the AV condition watched a videotaped interview and so, in addition to words and voice, had access to visual nonverbal cues as well as to any discrepancies among the three channels. Observers judged interviewee communication and decided if the interviewee was innocent or guilty. Other design features were also introduced to maximize the ecological validity of the results. The mock theft task, coupled with monetary incentives for success, was expected to heighten interviewees motivation and arousal and hence produce samples of behavior more akin to what transpires in higher stakes, real-world deception than is commonly achieved in laboratory deception experiments. Moreover, deceptive interviewees were not constrained to produce outright lies; they could employ whatever strategies they chose to enact, including ambiguity, concealment, equivocation, and other forms of obfuscation. Though not the primary thrust, this investigation also has relevance to new media in that it speaks to how judgments of communicator veracity vary according to the medium in which receivers access anothers messages. To the extent that some media foster or inhibit biased information processing more than others, users may select media according to how well they suit their impression management aims. This holds as much for senders who may use media for ulterior motives as for receivers who are seeking to form the most accurate judgments of others.
Literature review and hypotheses
Everyday truth judgments must often rely on stereotypical knowledge that is detached from the assessment of authentic cues (Fiedler, 1993). Though cognitive heuristics often lead to efcient and correct decisions, they can just as easily lead to biased judgments. The latter case is of interest here. Pared-down processing is especially common when receivers are unmotivated or have limited cognitive resources to appraise carefully a senders communicative behavior and so become cognitive misers, expending the least possible amount of cognitive effort necessary to arrive at a judgment (Fiske, 1993; Fiske & Taylor, 1991). Processing deceptive messages should be less taxing for observers than for participants, inasmuch as observers are freed from the complex multitasking that occupies conversational participants (Buller & Burgoon, 1996). Nonetheless, the tendency to eschew full analytical energy
574
should still be present among observers, for whom the consequences of making erroneous judgments are small. The current experiment centered on the kinds of biases that might operate in routine, day-to-day judgments of anothers veracity, in other words, when cognitive miserliness might be most probable.
Truth bias Of the four biases investigated here, the truth bias is the most cited and documented one in the deception literature (e.g., Kraut & Higgins, 1984; Levine, Park, & McCornack, 1999; McCornack & Parks, 1986; OSullivan, Ekman, & Friesen, 1988; Zuckerman, DeFrank, Hall, Larrance, & Rosenthal, 1979; Zuckerman, DePaulo, & Rosenthal, 1981). There are at least two different conceptualizations of truth bias. One is as an a priori belief, expectation, or presumption that reects the oft-observed tendency to assume communicators are truthful most of the time (Clark & Clark, 1977; OSullivan, 2003). This presumption of truthfulness, which might be labeled a truthfulness heuristic, nds roots in Grices (1989) principle of cooperative discourse. It also comports with what Gilbert and colleagues (Gilbert, Krull, & Malone, 1990; Gilbert, Pelham, & Krull, 1988) described as a Spinozan view of human information processing in which all incoming information is initially tagged as truthful and only subsequently revised if something occasions the need for appraisal and revision. The other conceptualization follows common usage for the term bias in psychometric literature and statistics, where a bias represents a departure from the true state of affairs (e.g., a biased sample statistic over- or underestimates the true mean value of a population) and therefore is inaccurate by definition. Put in deception terms, a truth bias reflects a tendency to judge more messages as truths than lies, independent of their actual veracity (McCornack & Parks, 1986; Zuckerman, DePaulo et al., 1981). When judging anothers veracity, it results in an overestimate of actual number of truths relative to the base rate of actual truthfulness; a lie bias reects an underestimate of the same. Conceptualized in this manner, truth biases may be a byproduct of, or closely aligned with, leniency and positivity biases. Presence of a truthfulness heuristic and/or truth bias has been amply documented in a variety of contexts (e.g., Anolli, Balconi, & Ciceri, 2003; Buller, Burgoon, White, & Ebesu, 1994; Buller, Strzyzewski, & Hunsaker, 1991; McCornack & Parks, 1986; Stiff, Kim, & Ramesh, 1992; Vrij & Mann, 2001). People rating message veracity consistently exhibit a tendency to judge most messages as truthful, even when the base rate of deception is varied (Levine, Kim, Park, & Hughes, 2006; Levine et al., 1999). The rst hypothesis sought to replicate this tendency to err in the direction of truthfulness when judging message veracity but to extend it to interactive message production with the aforementioned modications to methods (unconstrained, naturalistic, and motivated discourse production by senders; longer stimuli to judge). These methodological features pose a more stringent test of truth bias in that motivated, extended discourse could make deception more detectable or introduce statistical error variance that would mitigate judgmental bias. The hypothesis
575
posited that observers err in the direction of judging more messages as true than the base rate of truthful and deceptive stimuli being judged (Hypothesis 1).
Visual and demeanor biases The visual bias is a tendency to assign primacy to visual information over other forms of social information (DePaulo & Rosenthal, 1979; Noller, 1985; Stiff et al., 1989). Extensive research on channel reliance has shown systematic differences in judgments of messages with text only, audio only, video only, and AV delivery (Burgoon, 1985, 1994; DePaulo, Rosenthal, Green, & Rosenkrantz, 1982; DePaulo, Zuckerman, & Rosenthal, 1980). Observers attend more closely to facial than to body or voice cues (Bauchner, Kaplan, & Miller, 1980; Buller et al., 1991; Ekman & Friesen, 1974) despite the fact that facial cues typically are the least diagnostic in identifying deception (Feldman, 1976; Hocking, Bauchner, Kaminski, & Miller, 1979; Zuckerman, Larrance, Spiegel, & Klorman, 1981). Stiff et al. (1989) advanced two explanations for the visual cue primacy effect: A distraction hypothesisthat nonverbal visual cues distract from processing diagnostic (reliable) verbal informationand a situational familiarity hypothesisthat reliance shifts primarily to verbal content (as compared to using both verbal and nonverbal information) when the situation is familiar. Experimental results attested to a visual primacy effect: Visual cues had a substantial impact on judgments of truthfulness, vocal cues had a significant though weaker effect, and verbal variations did not alter judgments. A second experiment also showed that reliance on nonverbal cues was greater in the unfamiliar than in the familiar circumstance. However, two design features of the Stiff et al. (1989) study introduce some equivocality to the conclusions. Actors followed a tight script rather than producing the kinds of natural discourse present in normal deceptive interviews. Also, of the six cues that were manipulated, gaze aversion and audible pauses are stereotypical cues, whereas adaptors, postural shifts, speech errors, and silent pauses can be reliable (though by no means ever-present) indicators of deceit. Unclear, then, is if ratings of deceptiveness reected accurate detection or stereotypic judgments. The current experiment was designed to untangle and clarify the effects of availability of nonverbal cues, including vocal ones, about which Stiff et al. (1989) had no hypotheses. To replicate the ordering found by Stiff et al., we predicted that judgments of a persons truthfulness increase ordinally with nonverbal cue availability from text (verbal-only) to audio (verbal 1 vocal) to AV (verbal 1 vocal 1 visual) presentations (Hypothesis 2).2 These predictions beg the question of exactly why the presence of visual information is biasing. After all, visual primacy in itself does not guarantee biased judgments; bias should result only if observers attend to incorrect rather than correct cues. We believe there are multiple, and not mutually exclusive, causal mechanisms at work, among them qualitative features of senders communication style. Interpersonal deception theory (IDT; Buller & Burgoon, 1994, 1996; Burgoon & Buller, 2004) holds that deceivers engage in three classes of strategic communication that make detection of deceit difcult. Information management concerns the ways in
576
which deceivers manipulate the verbal contents of their messages by altering the quantity, quality, clarity, and directness of message elements (Burgoon, Buller, Guerrero, A, & Feldman, 1996; McCornack, Levine, Morrison, & Lapinski, 1996). Relative to truthful messages, deceptive messages may be briefer with sparser details; less clear and straightforward; and more indirect, depersonalized, and irrelevant. Behavior management concerns the control of specic nonverbal behaviors that accompany verbal messages. Deceivers strive to create the appearance of conversational normalcy and involvement (Burgoon, Buller, Floyd, & Grandpre, 1996; Burgoon, Buller, White, A, & Buslig, 1999; White & Burgoon, 2001). Image management concerns more global efforts to present the self in a favorable light and to reduce ones apparent culpability should deceit be discovered (DePaulo, 1992). The components of strategic communication together can be viewed as synonymous with demeanor bias, which is the tendency for some communicators to evoke general impressions of honesty and truthfulness irrespective of their actual veracity (Zuckerman, Larrance, et al., 1981).3 Access to visual nonverbal cues should magnify demeanor bias by exposing observers to those controllable cues such as smiles, facial expressions, eye contact, gestures, spacing, and dress that are deceivers stock in trade when attempting to put their best face forward (Bond, Kahler, & Paolicelli, 1985; Riggio, Tucker, & Throckmorton, 1987). Notwithstanding variability in social skills, motivated deceivers should employ a variety of strategies to mount a credible performance, and their strategic advantage should be realized most strongly when a richer array of social cues is available to observers. Although visual modalities conceivably could benet receivers over senders by affording them rich veins of social information to mine, the task of interpreting multiple streams of information increases cognitive demands that ironically may make receivers more prone to rely on heuristics precisely because there are so many channels and cues to decode. Comparatively, biases may be less pronounced when fewer nonverbal channels are available because differences in sender demeanor bias are liable to be less marked and to have less chance of fostering familiarity and trust in message recipients or observers. We therefore hypothesized that demeanor bias varies with nonverbal cue availability such that (a) information management, (b) behavior management, and (c) image management increase ordinally from text to audio to AV (Hypothesis 3). Verifying these predictions presents a methodological conundrum because a deceivers goal is to approximate the normal communication of truthtellers (Buller & Burgoon, 1996; Burgoon, Buller, Floyd, et al., 1996). If successful, deceivers may match the levels of, say, involvement and perceived trustworthiness of truthtellers, making any demeanor differences due to deception imperceptible. If, however, modality alters attention to visual cues, then modality may interact with deception to affect what observers perceive. We therefore posed as a research question: Do truth and deception interact to inuence information management, behavior management, and image management (RQ1)?
577
Expectancy violations bias Although the demeanor bias focuses on credibility-inducing behavior, the fourth bias focuses on suspicion-provoking behavior. Expectancy violations bias is the tendency to infer deception from abnormal, fishy-looking behavior (Bond et al., 1992). Expectancy violations theory (Burgoon, 1983; Burgoon & Burgoon, 2001) postulates that deviations from normative behaviors are arousing and divert attention to the unexpected act. IDT and many other theories of deception (A & Weiner, 2004; Johnson, Grazioli, Jamal, & Berryman, 2001; Swets, 2000) assert that deceptive behavior is often unexpected, anomalous, or deviant. Tests of IDT have shown that deceptive performances often suffer some initial impairment but improve over time as deceivers strategically repair their communication (Buller & Burgoon, 1994; Burgoon, Buller, White, et al., 1999; Burgoon, Buller, & Floyd, 2001), which should mitigate expectancy violations. Thus, evidence of an expectancy violations bias would imply that despite senders efforts to manage their performance, they still inadvertently give off signs of deceit that are detected by receivers. Any such signs should appear differentially according to which nonverbal and verbal channels are available to observers. Access to visual, vocal, and verbal cues could create more expectancy violations because three different channels of informationvisual, auditory, and verbalare more difficult for senders to coordinate and may expose observers to more suspicion-arousing channel discrepancies. Vocal cues can be very reliable indicators of deceit (DePaulo et al., 2003) possibly because they deviate from customary vocal patterns and escape deceivers self-monitoring. Text de facto lacks channel discrepancies, but odd verbal behavior might become more glaring without the distractions of nonverbal cues. These alternatives led us to pose as a research question: Does modality interact with deception to produce judgments of negative expectancy violations (R2)? Detection accuracy under different modalities Although detection accuracy reports often combine truth and deception detection within the same estimates, it is important to distinguish deception detection accuracy from truth detection accuracy, which may differ markedly (Burgoon, Buller, Ebesu, & Rockwell, 1994; Levine et al., 1999; Vrij & Mann, 2001). False alarms (judging truths as deception) and false negatives (judging deception as truths) can also be calculated (Green & Swets, 1966). As regards deception detection accuracy, the picture that emerges so far is of individuals entering communicative situations with strong proclivities to view others as truthful, to be drawn to visual information more so than other nonverbal social cues, and, when accessing visual cues, to fall victim to senders strategic efforts to manage their messages and overall demeanor. The only bias working to benet receivers is expectancy violations due to channel discrepancies or to the sheer number of cues that could be at odds with normative social patterns. The net result of these various biases should be to yield very poor detection accuracy under the visual modality.
578
Text, too, should produce poor detection accuracy, but two countervailing forcesthe diagnosticity of verbal cues and detachment between sender and receivershould net it higher accuracy than modalities with nonverbal cues. First, verbal cues are not inherently inscrutable. Deception ought to be, and is, detectable from textual features (Vrij, 2000; Zhou, Burgoon, Twitchell, & Nunamaker, 2004). That said, accuracy may be attenuated somewhat by the fact that untrained detectors lack familiarity with linguistic clues to deception and tend to favor stereotypical cues over valid ones (Buller et al., 1994; Zuckerman & Driver, 1985). Second, text fails to elicit the same sense of connection and involvement with message senders that happens when nonverbal cues are present (e.g., Burgoon et al., 19992000; Burgoon, Stoner, Bonito, & Dunbar, 2003; Ramirez & Burgoon, 2004). This detachment may introduce greater objectivity but also may dampen overall attentiveness to social information, again causing text-based judgments to suffer some inaccuracy but to a lesser extent than AV-based judgments. In the middle are judgments based on the combination of vocal and verbal cues. The voice is a rich source of social information. Its ability to promote involvement and intimacy often evokes positive responses that could be truth biasing. For example, Atoum and Al-Simadi (2000) found that speakers were judged as more honest and attractive when the speaker could be heard (i.e., in an AV or audio modality) than when just seen (in a video-only modality). Yet, the voice also lacks many of the known stereotypical (and incorrect) cues that people rely upon to make veracity judgments. The absence of stereotypical cues may encourage judges to attend to more reliable indicators of veracity such as pitch, hesitancies, and response latencies. Hence, audio-based judgments may attain greater detection accuracy. The Bond and DePaulo (2006) meta-analysis supports these conclusions, reporting lowest deception detection accuracy in a visual-only mode, better accuracy with verbal transcriptions, and best with audio or AV modalities. (Among visual cues, detectability is worse from the face only or body only than the combination of the two.) Thus, access to visual cues, especially facial ones, impairs detection. The authors concluded that detection is better when deception can be heard and worse when it can be seen.4 Recent experiments in computer-mediated deception point to similar results under conditions where targets of deception rendered judgments following extended interaction (e.g., Boyle & Ruppel, 2003; Burgoon et al., 2003). In the latter study, for example, participants discriminated best between truths and lies in the audio modality and fared worse when visual cues were present (the face-toface modality). Accuracy was lowest in the text condition, where deceivers were actually rated as more trustworthy than truthtellers. Accordingly, we hypothesized that deception detection is more accurate with audio (verbal 1 vocal) than text (verbal-only) or AV (verbal 1 vocal 1 visual) presentations (Hypothesis 4). As regards truth detection accuracy, the paucity of empirical evidence led us to pose as a last research question: Does truth detection accuracy vary by modality (RQ3)? One possibility is that the greater detachment and tempered judgments with text might result in less accuracy when nonverbal cues are absent than present. This
579
speculation coincides with a previous nding that video modalities are better than text-based modalities at truth detection (Porter, Campbell, Stapleton, & Birt, 2002).
Method Participants The sample consisted of 51 undergraduate students at a large university in the Midwestern United States who received extra credit for participation in a study of interviews conducted via new media. Each participant was randomly assigned to a deceptive or truthful interview to judge and to one of the three cue availability conditions, resulting in 17 observers per condition. Stimulus materials The AV, audio, and text files for this study were derived from a mock theft experiment conducted by Burgoon and Blair (Burgoon, Blair, & Hamel, 2006; Burgoon, Marett, & Blair, 2004). In the mock theft study, participants were randomly assigned to the role of thieves or innocent bystanders. Thieves were asked to take a wallet from a classroom on an assigned day and then to deceive during an interview about the theft. Innocents were simply told that a theft would take place in their classroom and were asked to respond truthfully during the interview. Motivation was induced by offering participants $10 if they could convince the interviewer of their innocence. They also could win another $50 if they were the most successful at appearing credible. (Interviews from a low-motivation condition were excluded from the stimulus pool so that only motivated deception was judged.) Trained interviewers followed a structured interview protocol that began with some preliminary questions (personal background, education, and work experiences) then turned to the theft. Nine questions were modeled after the Behavioral Analysis Interview, a procedure that is used routinely in criminal investigations (Inbau, Reid, Buckley, & Jayne, 2001). Questions included items such as, Did you take the wallet? Do you know where the wallet is now? Walk me through what happened from the time that you arrived at class until now and What do you think should happen to [the person who took the wallet]? The theft-related responses averaged 158 words, clearly enough length to qualify as interactive. Interviews were videotaped at 30 frames per second with a Prosumer quality Canon digital camera. It was essential that only high-quality recordings be included so as to prevent recording artifacts influencing judgments. A total of 17 recordings (nine innocent and eight deceptive subjects) met the criteria of acceptable video and audio quality. These videos were then converted into Windows media files for the audio and AV conditions. The interviews were transcribed for the text condition. One approach to conducting judgment studies is to present each observer a series of brief excerpts from multiple interviews. To obtain the advantages of observing a lengthier and interactive sample of behavior, we opted instead to have each observer judge a single interview. Like other judgment experiments, comparability
580
across conditions was achieved by using the same set of interviews in each modality. Important to note is that interviewee behavior was always produced under a face-toface condition so that what observers heard in the audio condition or read in the text condition was the voice and/or words that were originally also accompanied by all the other nonverbal cues available in face-to-face circumstances. This procedure assured a valid assessment of demeanor bias because senders could utilize all their normal self-presentational strategies when encoding their interview replies.
Procedure Participants reported to a computer laboratory where they were seated at a computer station, completed a consent form, and received experimental instructions. The instructions explained that they would be observing an interviewee being questioned about the theft of a wallet and that interviewees would be making a plea of innocence, regardless of whether they were innocent or guilty. The participants task was to determine whether the person that you will be observing is telling the truth about his/her innocence or whether the statement they are giving is deceptive. They then saw, heard, or read the interview on the computer, completed a brief Web-based questionnaire about the interview, were debriefed about the purpose of the study, given credit, and thanked for their time. Dependent measures Participants first rated interviewee communication on information, behavior, and image management. Information management was assessed with a multidimensional scale developed by Burgoon, Buller, Guerrero, et al. (1996) to measure quantity (completeness), quality, clarity, and directness/relevance of interviewee verbal responses. Ratings were made on 24 7-point Likert-format scales (e.g., The interviewee gave very brief answers). Coefcient alpha reliabilities were .81 for quantity, .80 for quality, .85 for clarity, and .76 for directness. Because of high intercorrelations among the four dimensions (average r = .71), the four dimensions were also analyzed unidimensionally. Behavior management consisted of 14 semantic differential items measuring involvement and dominance that had been used in previous investigations (e.g., Burgoon, Buller, Floyd, et al., 1996). Reliabilities were .88 for dominance and .68 for involvement. Image management was assessed with a multidimensional measure of credibility that has been well validated in previous investigations (e.g., Burgoon, 1976; McCroskey & Young, 1981). Coefcient alpha reliabilities for the respective dimensions were .90 for character, .75 for competence, .78 for sociability, and .86 for composure. The four measures were highly correlated with an average truth estimate: character, r(51) = .70, p , .001; competence, r(51) = .48, p , .001; composure, r(51) = .52, p , .001; and sociability, r(51) = .46, p , .001. The very large effect sizes indicate a strong association between general impressions of credibility and attributions of truthfulness on specic questions. As a further estimate of
581
credibility, a projected trust measure was created consisting of four 7-point Likert-format scales that asked observers if they would choose the interviewee as a roommate, job candidate, house sitter for pets, or date for a friend. These four items were combined into a single trust measure with a reliability of .86. To assess whether interviewee behaviors violated expectations negatively, participants completed seven expectedness and valence measures, taken from Burgoon and Walther (1990), on the 7-point Likert format. Coefcient alpha reliabilities were .76 and .78. Due to high intercorrelation (r = .84), these measures were also combined into a unidimensional version. To assess bias and detection accuracy, the last part of the questionnaire asked participants to rate, on a 010 scale, how truthful they thought the interviewee was in answering seven of the questions in the interview and to check off whether they thought the interviewee was guilty or innocent of taking the wallet. The dichotomous measure of guilt assessed truth bias, calculated as the aggregate deviation of the dichotomous judgments from the base rate of truthful and deceptive stimuli to be judged. Judgments were compared to actual guilt or innocence to calculate one measure of accuracy.5 The truthfulness ratings were averaged together for a mean truth estimate. The absolute value was a second gauge of bias; the relative differences across conditions served as a second measure of accuracy.
Results
All hypotheses were tested with alpha set at .05, one-tailed. Power for full-sample binomial tests was .78; for tests within modalities, it was .45. Power of factorial F tests and simple effect t tests to detect medium effect sizes (Glasss d = .50) was approximately .53 for deception effects and .45 for modality effects (Kraemer & Thiemann, 1987; Lenth, 2006).
Hypothesis 1: Truth bias Hypothesis 1 predicted that observers err in the direction of judging too many messages as truthful. On the dichotomous judgments, 67% of the participants indicated that they thought that the interviewee was truthful and 33% judged the interviewee as deceptive. A binomial test confirmed that these estimates were significantly different from the expected percentages of 53% and 47%, respectively (p = .004, onetailed). On the 10-point truthfulness scale, the mean judgment was 7.58 (SD = 1.58), which was signicantly higher (more truthful) than the expected median scale value of 5.30, t(50) = 5.78, p , .001. These results support Hypothesis 1. Observers judgments were biased in favor of truth. Hypothesis 2: Visual bias Hypothesis 2 predicted that the truth bias observed in Hypothesis 1 would increase ordinally with the addition of vocal and then visual cues. A planned contrast revealed an ordinal increase in the proportion of truthful judgments across
582
modalities, t(48) = 2.23, p , .05. Judgments of truth (innocence) increased ordinally from 47% with the text presentation to 71% with the addition of vocal cues to 82% with the addition of visual cues. (Binomial tests conducted within each modality conrmed that judgments of truthfulness in the visual and vocal conditions, respectively, were signicantly different than the expected value, p = .003 and p = .043, one-tailed; the text condition did not differ from the expected value.) A repeated measures analysis of variance on the truthfulness ratings for the three theft-specific questions produced a near-significant main effect for modality, F(2, 45) = 3.07, p = .056, partial h2 = .12. A planned contrast with lambda coefcients of 21, 0, and 11 was signicant, t(48) = 2.46, p = .009, one-tailed. The mean truth estimates on the 10-point scale were 8.01 (SD = 2.28) for the AV condition, 7.03 (SD = 1.98) for the audio condition, and 6.25 (SD = 2.17) for the text condition. Taken together, these analyses support Hypothesis 2. Truth bias was greatest when visual cues were present.
Hypothesis 3 and RQ1: Demeanor biases Hypothesis 3 predicted that demeanor bias, measured as (a) information management, (b) behavior management, and (c) image management, would increase ordinally with the addition of vocal and then visual nonverbal cues. RQ1 asked if these relationships are moderated by deception. Information management was initially tested with the composite measure. A 2 3 3 analysis of variance produced a main effect for modality, F(2, 51) = 5.05, p = .011, partial h2 = .18, which was qualied by a modality by deception interaction, F(2, 51) = 4.34, p = .019, partial h2 = .16. Follow-up univariate analyses on the four separate dimensions produced signicant main effects on all dimensions except directness and modality by deception interactions on quality and directness. Although the overall pattern showed the hypothesized ordinal increase (text, M = 4.14; audio, M = 4.59, AV, M = 5.33), the patterns differed within truth and deception. Under truth, the ordering from highest to lowest was audio then AV then text. Under deception, AV was higher than text and audio, as conrmed by a simple effect test using contrast codes of 21, 0, and 11, t(21) = 3.33, p = .003. Thus, the general trends conformed to Hypothesis 3ainterviewees were perceived as increasingly complete, truthful, clear, direct, and relevant with the addition of nonverbal cuesbut deception moderated results. The patterns for each of the four dimensions can be seen in Figures 1a through 1d. See Table 1 for all means. Multivariate analysis of the behavioral management dimensions of involvement and dominance produced a signicant interaction between modality and deception, Wilks l = .74, F(4, 88) = 3.55, p = .010, partial h2 = .14, and a nonsignicant main effect, Wilks l = .86, F(4, 88) = 1.74, p = .148, partial h2 = .07. Univariate analyses also produced signicant interactions for both measures and a main effect for dominance. As seen in Figures 1e and 1f, the predicted ordinal increase held true in the deception condition but not the truth condition. Simple effect tests within deception were signicant for both involvement, t(21) = 3.60, p , .001, one-tailed, and dominance, t(21) = 4.29, p , .001, one-tailed.
583
(a)
Quality
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text Audio
(b)
7.0 6.0
Quantity
Mean Rating
Mean Rating
5.0 4.0 3.0 2.0 1.0 Text Audio

Deception Deception Truth
AV
AV
Modality
Modality
(c)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Clarity
(d)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Directness
Mean Rating
Mean Rating
Audio
AV
Audio
AV
Modality
Modality
(e)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Involvement
(f)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Dominance
Mean Rating
Mean Rating
Audio
AV
Audio
AV
Modality
Modality
Figure 1 Continued on next page.
584
(g)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Projected Trust
(h)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Character
Mean Rating
Mean Rating
Audio
AV
Audio
AV
Modality
Modality
(i)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Sociability
(j)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Competence
Mean Rating
Mean Rating
Audio
AV
Audio
AV
Modality
Modality
(k)
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Text
Composure
Mean Rating
Audio
AV
Modality
Figure 1 Effects of deception and modality on information management dimensions of (a) quality, (b) quantity, (c) clarity, and (d) directness; behavior management dimensions of (e) involvement and (f) dominance; and image management dimensions of (g) projected trust, (h) character, (i) competence, (j) sociability, and (k) composure.
585
Table 1 Means and Standard Deviations for All Dependent Measures, by Modality and Deception Deception Text M Truth estimate Information management Dominance Involvement Expectedness & valence Character Sociability Composure Competence Projected trust 6.75 3.93 3.51 3.50 3.92 3.72 3.75 3.43 3.56 2.78 SD 2.38 1.19 0.87 0.85 1.03 1.06 1.18 1.36 1.50 1.37 Audio M 6.04 3.97 4.31 4.75 4.26 4.53 4.56 4.60 4.63 3.59 SD 1.93 1.03 0.86 0.77 1.22 0.53 0.82 1.25 0.58 0.67 FtF M 8.63 5.82 5.09 5.00 5.71 5.38 5.81 5.03 4.88 5.03 SD 2.34 1.19 0.36 0.87 0.69 0.94 0.73 1.01 0.69 1.08 Truth Text M 5.81 4.35 4.16 4.59 4.13 3.86 4.89 3.71 4.00 2.47 SD 2.01 0.93 1.05 1.04 1.31 0.79 0.87 0.96 1.50 1.13 Audio M 7.93 5.21 4.44 4.48 5.17 5.31 5.31 4.80 4.61 4.39 SD 1.65 0.91 0.77 0.78 1.10 1.16 1.04 1.44 1.27 1.03 FtF M 7.48 4.85 3.96 4.19 4.81 4.42 4.58 4.20 4.11 4.50 SD 2.22 1.32 0.89 1.00 1.29 1.59 0.79 1.11 0.86 0.89
The analyses for image management produced a multivariate modality by deception interaction on the ve credibility measures, F(8, 84) = 2.24, p = .032, partial h2 = .18, and a near-signicant main effect, F(8, 84) = 1.78, p = .093, partial h2 = .15. As with information and behavior management, the deception condition, but not the truth condition, conformed to the predicted ordinal increase from text to audio to AV, as conrmed by simple effect tests: character, t(21) = 3.78, p , .001; competence, t(21) = 2.60, p = .009; sociability, t(21) = 4.43, p , .001; composure, t(21) = 2.64, p = .007 (all one-tailed). Deceivers actually earned higher credibility ratings than did truthtellers under AV. Comparatively, truthtellers earned highest ratings under audio. A univariate analysis of projected trust produced only a signicant main effect for modality, F(2, 51) = 18.05, p , .01, partial h2 = .45, and conformed to predictions. In sum, perceptions of strategic communication increased with the addition of nonverbal channels of information when interviewees were deceptive but not when interviewees were truthful. Truthtellers regularly were judged most favorably in the audio presentation, that is, when judges had access to verbal and vocal cues. By contrast, deceiver communication was judged as the most complete, truthful, clear, direct, relevant, and dominant, and deceivers themselves were judged as the most trustworthy, sociable, competent, and composed in the AV presentation, that is, when judges had access to the additional cues. When judges only had verbal information, the same deceivers received the lowest ratings. (An exception was that deceivers earned a higher projective trust rating than truthtellers in both the AV and the text modalities, indicating that in both of these modalities, receivers are at risk of being deluded.) These combined results are strongly supportive of deceivers benefiting from the addition of visual nonverbal cues, in line with the demeanor bias
586
hypothesis. That the pattern was restricted to deceivers implies that deceivers were more proactive than truthtellers in managing their demeanor.
Research question 2: Expectancy violations bias RQ2 asked if deception and modality interact to affect expectedness and valence judgments. Univariate analysis of variance produced a significant modality main effect on the combined expectedness and valence measure, F(2, 51) = 5.03, p = .01, partial h2 = .18, and a near-signicant interaction effect between modality and deception, F(2, 51) = 2.78, p = .07, partial h2 = .11. Again, the deception condition showed the ordinal increase from text to audio to AV, but the truth condition did not. To truly analyze whether negative violations were perceived, expectedness needs to be crossed with valence, as shown in Figure 2, where the six
6.00 Condition Audio/Deception Audio/Truth 5.50 Text/Deception Text/Truth Video/Deception Video/Truth 5.00
Positive Confirmation
Valence
4.50
Negative Violation
4.00
Negative Confirmation
3.50
4.00
4.50
5.00
5.50
Expectedness
Figure 2 Expectedness and valence of deception by modality conditions.

587
experimental conditions are arrayed. The left-hand quadrants represent unexpected behaviors; the right-hand quadrants represent expected behavior. The upper quadrants represent positively valenced behaviors; the bottom quadrant represents negatively valenced behaviors. The graph indicates that observers were most favorable toward deceptive interviewees when they had full visual, vocal, and verbal access, rating them higher than all other interviewees, including truthtellers, on valence and expectedness. This result is consistent with the results for demeanor bias. Comparatively, ratings were sufciently low for deceivers in text-based and audio presentations to qualify as negative violations; truthtellers under text also received ratings that qualied as negative violations.
Hypothesis 4: Detection accuracy Hypothesis 4 predicted that detection of deception would be the most accurate in the audio condition and lower in the text and AV conditions. RQ3 asked if deception interacts with modality to affect accuracy. The results are best understood against the backdrop of the overall accuracy, which was 47%. By deception condition, only 29% of actual deceptive interviews were judged as deceptive (71% false negatives) and 63% of truthful interviews were judged as truthful (37% false positives). The overall accuracy and deception detection rates are markedly different from the 54 and 47% rates reported in the Bond and DePaulo (2006) meta-analysis, though only the latter approaches statistical significance (binomial test p = .056, one-tailed). The dichotomous measure, when analyzed by modality, revealed that observers in the audio and text conditions were correct in judging 38% of the deceptive interviewees as guilty, whereas in the AV condition, only 13% of the guilty parties were correctly judged as deceptive. These differences, however, failed to achieve statistical significance, x2(2) = 1.61, p = .45. The pattern of means for the overall accuracy rates (i.e., including truth detection accuracy) conformed to predictions 35% accuracy in text, 59% in audio, and 47% in AVbut also failed to achieve statistical significance, x2(2) = 1.89, p = .39. Analysis of the truth estimate data produced a near-significant deception by modality interaction, F(2, 45) = 2.76, p = .07, partial h2 = .11. Simple effect tests within each modality produced a signicant difference in truth and deception ratings within the audio condition, t(15) = 1.75, p = .05, one-tailed, but not in the text and AV conditions (see Figure 3). In fact, the deception condition means were actually higher than the truth condition means in the latter two conditions. Hypothesis 4 thus received limited support and RQ3 was answered with a partial yes. Discussion
This investigation is important in several respects. First, unlike most previous judgment studies, biases and detection accuracy were examined under fully interactive conditions. Use of lengthier interviews as stimuli availed observers (as well as senders) of the dynamic adjustments that characterize extended discourse and that might
588
10
Estimated Marginal Means
0 Text Audio AV
Modality
Figure 3 Effects of deception and modality on truth estimates.
reveal deceptive intentions. As well, the stimuli were drawn from dialogic rather than monologic responses. As postulated in IDT, the presence of natural discursive turntaking permits senders opportunities to adapt strategically to their interlocutor, improve their self-presentations, and evade detection (Burgoon et al., 2001). Such adaptations would have been absent in less interactive or more abbreviated responses. Second, this is the rst investigation to examine multiple cognitive biases in the same experiment so that their relationships to one another and to multiple outcome measures could be assessed. Because each judge viewed a single interview rather than a series of excerpts, it was possible to collect from them more extensive dependent measures, including both deception-related and credibility-related outcomes and to illuminate the close interconnection between deceit and the more superordinate construct of credibility. Finally, this work is germane to new media, specically to social information processing in computer-mediated communication. Insights into how people assess text-based as compared to multimodal messages can advance understanding of differential uses and responses to computer-mediated modes of communication such as e-mail, text chat, or audioconferencing.
Cognitive biases The human ability to accurately judge anothers veracity is often sabotaged by cognitive biases. The current investigation confirmed the presence of truth, visual, demeanor, and expectancy violations biases that enabled deceivers to evade detection. These same biases that are responsible for poor deception detection also contribute to message senders being judged as trustworthy, competent, sociable, and composed, thus also enlightening what information processing patterns foster credibility.
589
Observers showed a marked tendency to bias judgments in favor of truth. Compared to the 53% of all stimuli that were actually truthful, observers judged 67% to be truthful, and the average truth estimate was far above the midpoint of the scale. These results reinforce what has been a consistent finding in the literature, namely, that people are highly inclined to trust the communication of others and unlikely to question those judgments unless faced with some major deviation that triggers a reevaluation. The current findings extend this conclusion to messages generated under fully interactive conditions. According to media richness theory and social information processing theory (Daft & Lengel, 1984, 1986; Walther & Parks, 2002), differences in availability of social information in different channels should affect deception detection. Textbased messages and transcripts only avail the receiver of verbal information (save for efforts to add in nonverbal information through such features as capitalization and emoticons). Auditory channels add vocalic cues. AV modalities add kinesic, proxemic, physical appearance, and (sometimes) environmental information. We hypothesized that observers judging an AV presentation would exhibit the most visual and demeanor biases, that is, the truth bias would be most aggravated in the AV condition, and the AV condition would be most associated with strategic manipulation of message content, style, and overall demeanor. Results bore out our predictions, especially for deceivers. The truth bias was intensied by modalities that gave observers access to nonverbal cues. Despite the fact that the same verbal content was present in all three modality conditions, the addition of nonverbal vocal and visual cues increasingly led observers to judge senders interview answers as truthful. Following IDT postulates of strategic communication by deceivers, we also hypothesized that observers would succumb to a demeanor bias with increasing availability of nonverbal social cues. Results confirmed that deceivers (but not truthtellers) overall communication was judged more favorably on measures of information, behavior, and image management with increasing availability of nonverbal cues. The communication of deceptive interviewees was seen as the most complete, honest, clear, direct/relevant, involved, dominant, credible, trustworthy, expected, and positively valenced in the AV condition. The demeanor bias is only valid to the extent that an honest-appearing presentation leads observers to make faulty attributions about anothers veracity; that is, there must be differences between truthtellers and deceivers or else the bias devolves to a straight social skills variable in which some people are more skillful communicators than others. Had we found the same pattern of behavior for both truthtellers and deceivers, we would have been left with questionable support for the demeanor bias. However, the repeated interactions between deception and modality and associated differential patterns across modalities for deceivers versus truthtellers imply that judgments were not exclusively a function of structural modality features per se but also of the self-presentations that deceivers were able to craft using all the kinesic, physical appearance, proxemic, and vocalic features at their disposal. Deceivers elicited ordinal increases in favorability from the text to the audio to the AV condition, whereas truthtellers elicited a nonmonotonic pattern such that favorability was highest under audio.
590
These results further imply that deceivers visual presentation was a major focus of their strategic efforts, and those efforts succeeded in winning more favorable assessments from observers, inasmuch as in the AV modality, deceivers earned higher ratings than truthtellers on all measures. In the audio modality, the reverse was true: Truthtellers earned higher ratings than deceivers on virtually all measures (except involvement, dominance, and competence). In the text modality, most judgments likewise favored truthtellers over deceivers, although deceivers were rated as most truthful and worthy of future trust. The full complexion of the findings, then, points to the visual modality as the most likely to elicit judgments favorable to deceivers. Advocates of increasing bandwidth in new technologies to give participants full multimodal access may, ironically, confer advantages on deceivers. Instead of improving social presence, which such moves are intended to foster, they may simply reduce peoples ability to detect anothers insincerity, ulterior motives, or deceit. Comparatively, use of leaner modalities may offer the best prospect for inhibiting biased information processing. Future research on visual biasing could productively examine three additional factors. One is the degree of vividness of visual stimuli, which could affect the extent to which they galvanize attention. A second is whether the continuous presence of static and slow signals (such as ones physical appearance and posture) permits deeper and repeated scrutiny as compared to dynamic and transitory vocal signals that must be processed in real time and are then gone. A third is the proportion of reliable versus unreliable indicators within the total stimulus pool. If a disproportionate amount of incorrect cues are present due to senders efforts to manage their visual presentations, then the addition of visual cues should introduce error. This is essentially the distraction argument advanced by Stiff et al. (1989) that receivers are distracted by the visual cues and so ignore more reliable cues available in other channels. As regards the expectancy violations bias, we hypothesized that atypical behaviors would lead to deceivers communication being judged as a negative violation. The results were only partially supportive. Deception under both text and audio conditions was judged as a negative violation, which implies that deceptive performances can give themselves away by their departures from normative standards for content, language, and voice. Were these the only conditions to qualify as expectancy violations, we would regard the hypothesis as largely supported. However, the truthful responding via text was also among the least expected and desirable combinations. This finding bolsters claims elsewhere about the likely dampening of feelings of involvement, connection, and trust associated with text-based communication (Burgoon, Bonito, & Kam, 2006). At the same time, this nding conrms that the expectancy violations bias is not conned to communicative behavior but may also be applicable to communication channels over which such behavior is transmitted. The results for the deception/AV condition place a further qualification on the expectancy violations bias. Communication in this condition was judged to be the most normal and positively valenced of any of the combinations, that is, it was a positive confirmation. This makes sense when considered within the context of
591
the demeanor bias results. Such findings could only be obtained if deceivers were more successful than truthtellers in promulgating an attractive image in the AV condition and if adding visual nonverbal cues enhanced their demeanor relative to the exact same performances in the audio and text conditions. At the same time, the results indicate that abnormal behavior by itself is not the only basis for biased judgment; behavior that is judged as exceedingly normal and appropriate can also lead to biased judgment. The expectancy violations results demonstrate the utility of arraying communicative behavior and modalities according to expectations and evaluations. Yet, they also raise questions about whether negatively valenced, unexpected behavior should be regarded as a bias, inasmuch as only the text but not the audio condition produced detection inaccuracies. Put differently, negative violations can be quite diagnostic under the correct conditions (Bond et al., 1992). They can alert receivers to anomalies that are in fact sound indicators that something is amiss. Like positive conrmations, they are only biasing to the extent that observers attend to the wrong, stereotypic indicators rather than to diagnostic ones. The inaccuracies in the AV condition are a reminder as well that expected behaviors can also lead to erroneous judgments.6
Detection accuracy The generally poor ability of receivers to detect deception in this study is consistent with previous research. The poor detection accuracy rates overall (47%) and within the deception condition (29%) suggest that detectability may even worsen when judging messages generated interactively. Detection accuracy was also somewhat sensitive to modality. On the continuous measure of truthfulness (but not a dichotomous one), observers accurately discriminated truthful from deceptive interviewees when in the audio condition. Their counterparts in the text and AV conditions did not succeed in making such discriminations. In fact, observers showed a tendency to regard deceptive interviews as more truthful than truthful ones in the nonverbally leanest and richest conditions. This pattern of findings supports the hypothesized accuracy of deception detection when observers have access only to audio (and verbal) information. Our findings that truth bias and accuracy vary by modality have important ramifications for the detection of deception. It appears that false-positive and false-negative rates can vary by modality without having a large impact on accuracy. It may be that the biases inherent in different modalities would make certain modalities preferable for different detection tasks. For example, our criminal justice system values protection of the innocent; therefore, this system would want as few false positives as possible. The truth bias inherent in the AV condition might reinforce its desirability for courtroom use. A low false-negative rate might be desired in other circumstances. For example, a single error in intelligence analysis could have profound implications for national security. Thus, the reduced truth bias found in text or audio conditions might be preferable for intelligence assessment tasks. In light of
592
the high rate of detection accuracy in the audio condition, audio detection may represent the best of all options and has the further advantage of requiring less investment in bandwidth for the messages being transmitted. Be it first responders, police detectives, job recruiters, or friends unmasking lies by friends, use of a voiceonly modality such as the telephone for questioning might prove to be more advantageous than a face-to-face confrontation.
A theory of nonverbal cue availability deception detection A wealth of empirical evidence now documents that social information processing and ability to detect deception vary according to access to nonverbal channels of information. Can a single theory account for these effects? Probably not. Modalities have multiple influences on sender behavior and receiver perception. That said, we propose that a strategic communication perspective supplies a partial explanation in that visual media present receivers with a preponderance of well-practiced and managed sender behaviors intended to produce a credible front. The sheer amount of social information to be processed also can result in erroneous judgments. Media that only afford access to senders words reduce the processing task for receivers and include some useful linguistic indicators of deceit, but again the preponderance of cues is likely to be deliberate, especially if senders are motivated and have had opportunities to plan, rehearse, or edit their responses. In between are audio modalities that add to verbal cues a mix of highly diagnostic and less controlled vocal cues. The greater proportion of diagnostic indicators, coupled with some diminution in the truth bias, would account for the better discrimination between truth and deception in the audio condition. Observers recognition of expectancy-violating deceptive behaviors in this condition is consistent with this interpretation. To conclude, deception detection is a complex task that is fraught with cognitive biases. Nonverbal cues, especially visual ones, lead detectors astray. Detectors can improve their accuracy by attending more closely to vocal information and relying upon audio modalities to discriminate between truth and deception. Continued exploration of when biases are most pronounced and what can mitigate them will aid not only in better detection of deception but also better understanding of how humans come to trust the veracity of others. Notes
1 Interactivity was coded for 50 studies. It was dened as senders not interacting if lying while alone or to a passive observer; all other cases were deemed interactive. Less than 9% of the pairwise comparisons that were analyzed came from cases where senders interacted with the person who was to judge their veracity. The vast majority came from cases where senders told their lies to someone else (58%), such as giving a single reply to an interviewer, or where they did not interact with anyone (33%). Median length of sender messages was brief at 52 seconds.
593
4 5
In the Bond and DePaulo (2006) meta-analysis, judgments of truthfulness from withinstudy comparisons follow a different ordering, with video-only messages judged as less truthful than audio-only, AV, or text messages. We surmise that the absence of any verbal content upon which to base a veracity judgment in the video-only condition resulted in indecision or neutrality. It should be noted that unlike other biases, demeanor bias derives not from a cognitive proclivity among receivers but rather from features of the senders communication that systematically elicit biased judgments. Previous meta-analyses and studies (Burgoon, 2005; DePaulo et al., 1980; Zuckerman et al., 1981) have reported different orderings of conditions. Truth bias has been measured in a variety of ways. For example, Burgoon and colleagues (Burgoon et al., 1994, 2003; Dunbar, Ramirez, & Burgoon, 2003) have measured bias as the deviation of receiver estimates of truthfulness from sender reports of actual truthfulness such that a positively signed score reected truth bias and a negatively signed score, a lie bias. McBurney and Comadena (1992) measured truth bias as the extent to which the average truthfulness rating across multiple trials of truths and lies fell toward the high end of the rating scale. Here, we opted for objective comparison to the sample base rate. Per signal detection theory, bias is generally considered to be independent from accuracy. That is to say, one can achieve the same accuracy level while showing very different biases. For example, imagine a sample of materials in which 50% of the materials are truthful and 50% are deceptive. One could obtain 50% accuracy while exhibiting either a complete truth bias (e.g., all materials judged as truthful) or complete deception bias (e.g., all materials judged as deceptive). Thus, varying bias scores are compatible with a variety of accuracy scores in samples that are roughly balanced such that increased bias may accompany increased accuracy or decreased accuracy (Swets, 2000).
References
A, W. A., & Weiner, J. A. (2004). Toward a theory of motivated information management. Communication Theory, 14, 167190. Anolli, L., Balconi, M., & Ciceri, R. (2003). Linguistic styles in deceptive communication: Dubitative ambiguity and elliptic eluding in packaged lies. Social Behavior and Personality, 31, 687710. Atoum, A. O., & Al-Simadi, F. A. (2000). The effect of presentation modality on judgments of honesty and attractiveness. Social Behavior and Personality, 28, 269278. Bauchner, J. E., Kaplan, E. P., & Miller, G. R. (1980). Detecting deception: The relationship between available information to judgmental accuracy in initial encounters. Human Communication Research, 6, 251264. Bond, C. F., Jr., & DePaulo, B. M. (2006). Accuracy of deception judgments. Review of Personality and Social Psychology, 10, 214234. Bond, C. F., Jr., Kahler, K. N., & Paolicelli, L. M. (1985). The miscommunication of deception: An adaptive perspective. Journal of Experimental Social Psychology, 21, 331345.
594
Bond, C. F., Jr., Omar, A., Pitre, U., Lashley, B. R., Skaggs, L. M., & Kirk, C. T. (1992). Fishy-looking liars: Deception judgment from expectancy violation. Journal of Personality and Social Psychology, 63, 969977. Boyle, R. J., & Ruppel, C. P. (2003). The impact of media richness, suspicion, and perceived truth bias on deception detection. Proceedings of the 36th annual Hawaii International Conference on System Sciences, Big Island, HI. Los Alamitos, CA: IEEE. Buller, D. B., & Burgoon, J. K. (1994). Deception: Strategic and nonstrategic communication. In J. A. Daly & J. M. Wiemann (Eds.), Strategic interpersonal communication (pp. 191223). Hillsdale, NJ: Erlbaum. Buller, D. B., & Burgoon, J. K. (1996). Interpersonal deception theory. Communication Theory, 6, 203242. Buller, D. B., Burgoon, J. K., White, C. H., & Ebesu, A. S. (1994). Interpersonal deception: VII. Behavioral proles of falsication, concealment, and equivocation. Journal of Language and Social Psychology, 13, 366395. Buller, D. B., Strzyewski, K. D., & Hunsaker, F. G. (1991). Interpersonal deception: II. The inferiority of conversational participants as deception detectors. Communication Monographs, 58, 2540. Burgoon, J. K. (1976). The ideal source: A reexamination of source credibility measurement. Central States Speech Journal, 27, 200206. Burgoon, J. K. (1983). Nonverbal violations of expectations. In J. Wiemann & R. Harrison (Eds.), Nonverbal interaction: Vol. 11. Sage annual reviews of communication (pp. 1177). Beverly Hills, CA: Sage. Burgoon, J. K. (1985). The relationship of verbal and nonverbal codes. In B. Dervin & M. J. Voight (Eds.), Progress in communication sciences (pp. 263298). Norwood, NJ: Ablex. Burgoon, J. K. (1994). Nonverbal signals. In M. L. Knapp & G. R. Miller (Eds.), Handbook of interpersonal communication (2nd ed., pp. 344390). Beverly Hills, CA: Sage. Burgoon, J. K. (2005). The future of motivated deception detection. In P. Kalbeisch (Ed.), Communication yearbook 29 (pp. 4995). Mahwah, NJ: Erlbaum. Burgoon, J. K., Blair, J. P., & Hamel, L. (2006, June). Factors inuencing deception detection: Impairment or facilitation? Paper presented to the annual meeting of the International Communication Association, Dresden, Germany. Burgoon, J. K., Bonito, J. A., Bengtsson, B., Ramirez, A., Jr., Dunbar, N., & Miczo, N. (19992000). Testing the interactivity model: Communication processes, partner assessments, and the quality of collaborative work. Journal of Management Information Systems, 16(3), 3356. Burgoon, J. K., Bonito, J. B., & Kam, K. (2006). Communication and trust under face-to-face and mediated conditions: Implications for leading from a distance. Manuscript submitted for publication. Burgoon, J. K., & Buller, D. B. (2004). Interpersonal deception theory. In J. S. Seiter & R. H. Gass (Eds.), Readings in persuasion, social inuence and compliance-gaining (pp. 239264). Boston: Allyn & Bacon. Burgoon, J. K., Buller, D. B., Ebesu, A. S., & Rockwell, P. (1994). Interpersonal deception: V. Accuracy in deception detection. Communication Monographs, 61, 303325. Burgoon, J. K., Buller, D. B., & Floyd, K. (2001). Does participation affect deception success? A test of the inter-activity effect. Human Communication Research, 27, 503534.
595
Burgoon, J. K., Buller, D. B., Floyd, K., & Grandpre, J. (1996). Deceptive realities: Sender, receiver, and observer perspectives in deceptive conversations. Communication Research, 23, 724748. Burgoon, J. K., Buller, D. B., Guerrero, L., A, W., & Feldman, C. (1996). Interpersonal deception: XII. Information management dimensions underlying deceptive and truthful messages. Communication Monographs, 63, 5069. Burgoon, J. K., Buller, D. B., White, C. H., A, W., & Buslig, A. L. S. (1999). The role of conversational involvement in deceptive interpersonal interactions. Personality and Social Psychology Bulletin, 25, 669686. Burgoon, J. K., & Burgoon, M. (2001). Expectancy theories. In P. Robinson & H. Giles (Eds.), Handbook of language and social psychology (2nd ed., pp. 79101). Sussex, UK: John Wiley & Sons. Burgoon, J. K., Marett, K., & Blair, J. P. (2004). Detecting deception in computer-mediated communication. In J. F. George (Ed.), Computers in society: Privacy, ethics & the Internet (pp. 154166). Upper Saddle River, NJ: Prentice-Hall. Burgoon, J. K., Stoner, G. M., Bonito, J. A., & Dunbar, N. E. (2003, January). Trust and deception in mediated communication. Proceedings of the 36th Annual Hawaii International Conference on System Sciences. Big Island, HI. Los Alamitos, CA: IEEE. Burgoon, J. K., & Walther, J. B. (1990). Nonverbal expectancies and the consequences of violations. Human Communication Research, 17, 232265. Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality & Social Psychology, 39, 752766. Clark, H., & Clark, E. (1977). Psychology and language. New York: Harcourt Brace Jovanovich. Daft, R. L., & Lengel, R. H. (1984). Information richnessA new approach to managerial behavior and organization design. Research in Organizational Behavior, 6, 191233. Daft, R. L., & Lengel, R. H. (1986). Organizational information, media richness and structural design. Management Science, 32, 554571. DePaulo, B. M. (1992). Nonverbal behavior and Self-presentation. Psychological Bulletin, 111, 203243. DePaulo, B., Lindsay, J., Malone, B., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129, 74118. DePaulo, B. M., & Rosenthal, R. (1979). Ambivalence, discrepancy, and deception. In R. Rosenthal (Ed.), Skill in nonverbal communication: Individual differences (pp. 204248). Cambridge, MA: Oelgeeschlager, Gunn & Hain. DePaulo, B. M., Rosenthal, R., Green, C. R., & Rosenkrantz, J. (1982). Diagnosing deceptive and mixed messages from verbal and nonverbal cues. Journal of Experimental Social Psychology, 18, 433446. DePaulo, B. M., Zuckerman, M., & Rosenthal, R. (1980). Detecting deception: Modality effects. In L. Wheeler (Ed.), Review of personality and social psychology (pp. 125162). Beverly Hills, CA: Sage. Dunbar, N. E., Ramirez, A., Jr., & Burgoon, J. K. (2003). Interactive deception: Effects of participation on participant-receiver and observer judgments. Communication Reports, 16, 2333. Ekman, P., & Friesen, W. V. (1974). Detecting deception from the body or face. Journal of Personality and Social Psychology, 29, 288298.
596
Feldman, R. S. (1976). Nonverbal disclosure of teacher deception and interpersonal affect. Journal of Educational Psychology, 68, 807816. Fiedler, K. (1993). Training lie detectors to use nonverbal cues instead of global heuristics. Human Communication Research, 20, 199223. Fiske, S. T. (1993). Social cognition and social perception. Annual Review of Psychology, 44, 155194. Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York: McGraw-Hill. Gilbert, D. T., Krull, D. S., & Malone, P. S. (1990). Unbelieving the unbelievable: Some problems in the rejection of false information. Journal of Personality and Social Psychology, 59, 601613. Gilbert, D. T., Pelham, B. W., & Krull, D. S. (1988). On cognitive busyness: When person perceivers meet person perceived. Journal of Personality and Social Psychology, 54, 733740. Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics. New York: Wiley. Grice, P. (1989). Studies in the way of words. Cambridge, MA: Harvard University Press. Hocking, J. E., Bauchner, J., Kaminski, E. P., & Miller, G. R. (1979). Detecting deceptive communication from verbal, visual, and paralinguistic cues. Human Communication Research, 6, 3346. Inbau, F. E., Reid, J. E., Buckley, J. P., & Jayne, B. P. (2001). Criminal interrogations and Confessions. Gaithersburg, MD: Aspen. Johnson, P. E., Grazioli, S., Jamal, K., & Berryman, G. (2001). Detecting deception: Adversarial problem solving in a low base rate world. Cognitive Science, 25, 355392. Kraemer, H. C., & Thiemann, S. (1987). How many subjects? Statistical power analysis in research. Thousand Oaks, CA: Sage. Kraut, R. E., & Higgins, E. T. (1984). Communication and social cognition. In R. S. Wyer, Jr. & T. K. Srull (Eds.), Handbook of social cognition (Vol. 3, pp. 88127). Hillsdale, NJ: Erlbaum. Lenth, R. V. (2006). Java applets for power and sample size [computer software]. Retrieved February 12, 2007, from http://www.stat.uiowa.edu/~rlenth/Power Levine, T. R., Kim, R., Park, H. S., & Hughes, M. (2006). Deception detection accuracy is a predictable linear function of message veracity base-rate: A formal test of Park and Levines probability model. Communication Monographs, 73, 243260. Levine, T. R., Park, H. S., & McCornack, S. A. (1999). Accuracy in detecting truths and lies: Documenting the veracity effect. Communication Monographs, 66, 125144. McBurney, D. L., & Comadena, M. E. (1992, November). An examination of the McComack and Parks model of deception detection. Paper presented at the annual meeting of the Speech Communication Association, Chicago, IL. McCornack, S. A., Levine, T. R., Morrison, K., & Lapinski, M. (1996). Speaking of information manipulation: A critical rejoinder. Communication Monographs, 63, 8392. McCornack, S. A., & Parks, M. (1986). Deception detection and relationship development: The other side of trust. In M. L. McLaughlin (Ed.), Communication yearbook 9. Beverly Hills, CA: Sage. McCroskey, J. C., & Young, T. J. (1981). Ethos and credibility: The construct and its measurement after three decades. Central States Speech Journal, 32, 2434.
597
Motley, M. T., & Camden, C. T. (1988). Facial expression of emotion: A comparison of posed expressions versus spontaneous expressions in an interpersonal setting. Western Journal of Speech Communication, 52, 122. Noller, P. (1985). Video primacyAnother look. Journal of Nonverbal Behavior, 9, 2847. OSullivan, M. (2003). The fundamental attribution error in detecting deception: The boy-who-cried-wolf effect. Personality and Social Psychology Bulletin, 29, 13161327. OSullivan, M., Ekman, P., & Friesen, W. V. (1988). The effect of comparisons on detecting deceit. Journal of Nonverbal Behavior, 12, 203215. Porter, S., Campbell, M. A., Stapleton, J., & Birt, A. R. (2002). The inuence of judge, target, and stimulus characteristics on the accuracy of detecting deceit. Canadian Journal of Behavioural ScienceRevue Canadienne Des Sciences Du Comportement, 34, 172185. Ramirez, A. Jr., & Burgoon, J. K. (2004). The effect of interactivity on initial interactions: The inuence of information valence and modality and information richness on computer-mediated interaction. Communication Monographs, 71, 422477. Riggio, R. E., Tucker, J., & Throckmorton, B. (1987). Social skills and deception ability. Personality and Social Psychology Bulletin, 13, 568577. Stiff, J., Kim, H., & Ramesh, C. (1992). Truth biases and aroused suspicion in relational deception. Communication Research, 19, 326345. Stiff, J., Miller, G., Sleight, C., Mongeau, P., Garlick, R., & Rogan, R. (1989). Explanations for visual cue primacy in judgments of honesty and deceit. Journal of Personality & Social Psychology, 56, 555564. Swets, J. A. (2000). Signal detection theory and ROC analysis in psychology and diagnostics: Collected papers. Mahwah, NJ: Erlbaum. Todorov, A., Chaiken, S., & Henderson, M. D. (2002). The heuristic-systematic model of social information processing. In J. P. Dillard & M. Pfau (Eds.), The persuasion handbook: Developments in theory and practice (pp. 195211). Thousand Oaks, CA: Sage Publications. Tversky, A., & Kahneman, D. (1974). Judgment under uncertaintyHeuristics and biases. Science, 185, 11241131. Vrij, A. (2000). Detecting lies and deceit: The psychology of lying and the implications for professional practice. New York: John Wiley. Vrij, A., & Mann, S. (2001). Who killed my relative? Police ofcers ability to detect real-life high stake lies. Psychology, Crime & Law, 7, 119132. Walther, J. B., & Parks, M. R. (2002). Cues ltered out, cues ltered in: Computer mediated communication and relationships. Thousand Oaks, CA: Sage. White, C. H., & Burgoon, J. K. (2001). Adaptation and communicative design: Patterns of interaction in truthful and deceptive conversations. Human Communication Research, 27, 937. Zhou, L., Burgoon, J. K., Twitchell, D., & Nunamaker, J. F., Jr (2004). Automating linguistics-based cues for detecting deception in text-based asynchronous computer-mediated communication. Group Decision and Negotiation, 13, 81106. Zuckerman, M., DeFrank, R. S., Hall, J. A., Larrance, D. T., & Rosenthal, R. (1979). Facial and vocal cues of deception and honesty. Journal of Experimental Social Psychology, 15, 378396.
598
Zuckerman, M., DePaulo, B. M., & Rosenthal, R. (1981). Verbal and nonverbal communication of deception. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 14, 159). New York: Academic Press. Zuckerman, M., & Driver, R. (1985). Telling lies: Verbal and nonverbal correlates of deception. In A. W. Siegman & S. Feldstein (Eds.), Nonverbal communication: An integrated perspective (pp. 129147). Hillsdale, NJ: Erlbaum. Zuckerman, M., Larrance, D. T., Spiegel, N. H., & Klorman, R. (1981). Controlling nonverbal displays: Facial expressions and tone of voice. Journal of Experimental Social Psychology, 17, 506524.
599
Les biais cognitifs et la disponibilit des indices non verbaux dans la dtection du mensonge Judee K. Burgoon, University of Arizona J. P. Blair, Texas State University Renee E. Strom, St. Cloud State University Rsum
Dans les situations potentiellement trompeuses, les gens se fient sur des raccourcis mentaux afin d'aider traiter l'information. Ces jugements heuristiques sont souvent biaiss et ont pour rsultat des valuations errones de l'honntet de l'metteur. Quatre de ces biais (le biais de vrit, le biais visuel, le biais comportemental et le biais de violation des attentes) furent examins dans une exprience de jugements qui variait en disponibilit des indices non verbaux et en mensonge. Les observateurs ont vu un entretien complet enregistr sur vido (accs complet aux indices visuels, vocaux et verbaux), entendu l'entretien complet (accs vocal et verbal) ou lu une transcription (accs verbal) d'un suspect honnte ou trompeur, interrog propos d'un faux vol. Ils ont ensuite class l'interview selon des critres d'information, de comportement, de gestion de l'image et d'honntet. Les rsultats appuient la prsence de chacun des quatre biais, qui taient le plus vidents lorsque les interviews mentaient et que les observateurs avaient accs toutes les modalits visuelles, vocales et verbales. Avec l'ajout des indices non verbaux, les messages des menteurs taient jugs comme tant de plus en plus complets, honntes, clairs et pertinents; leurs comportements comme tant plus complexes et dominants; leur comportement gnral comme plus crdible. Les menteurs taient en fait jugs plus crdibles que les personnes honntes dans la modalit la plus complte (indices visuels, vocaux et verbaux), tandis que la plus grande exactitude dans la discrimination et la dtection s'est produite chez les gens n'ayant eu accs qu' l'enregistrement audio. Les rsultats ont des implications pour les facteurs qui influencent les jugements de la crdibilit d'un metteur et l'exactitude dans la distinction entre la vrit et le mensonge, surtout dans des conditions o les metteurs produisent les messages de faon interactive. Mots cls : mensonge, comportement non verbal, communication interpersonnelle, crdibilit, confiance, modalit, CMO
Kognitive Befangenheit und nonverbale Hinweisverfgbarkeit beim Aufdecken von Tuschung Judee K. Burgoon, University of Arizona J. P. Blair, Texas State University Renee E. Strom, St. Cloud State University
In potentiellen Tuschungssituationen greifen Menschen auf mentale Abkrzungen zurck, die ihnen helfen, Informationen zu verarbeiten. Diese heuristischen Urteile sind oft befangen und resultieren in einer fehlerhaften Beurteilung der Aufrichtigkeit des Senders. Vier solcher Befangenheiten Wahrheitsbefangenheit, visuelle Befangenheit, Verhaltensbefangenheit und Erwartungsverletzungsbefangenheit untersuchten wir in einem Beurteilungsexperiment mit variierter nonverbaler Hinweisverfgbarkeit und Tuschung. Beobachter sahen ein
aufgezeichnetes Video (visueller, vokaler und verbaler Zugang), hrten ein Interview (vokaler und verbaler Zugang) oder lasen ein Manuskript (verbaler Zugang) eines wahrheitsgemen oder tuschenden Verdchtigen, der bezglich eines Entwendungsdiebstahls verhrt wurde. Danach beurteilten die Teilnehmer diesen hinsichtlich der Informationen und Verhaltensweisen, des Imagemanagement und der Wahrhaftigkeit. Die Ergebnisse sttzen die Existenz aller vier Befangenheiten, die sich am deutlichsten zeigten, wenn Interviewte tuschten und die Beobachter Zugang zu allen visuellen, vokalen und verbalen Modalitten hatten. Die Botschaft des Tuschenden wurde als zunehmend vollstndig, ehrlich, klar und relevant, sein Verhalten als strker involviert und dominant, und sein allgemeines Verhalten als glaubwrdiger beurteilt, wenn nonverbale Hinweise ergnzt wurden. Tuschende wurden in der AV-Variante sogar als glaubwrdiger beurteilt als jene, die die Wahrheit sagten. Die beste Unterscheidung und Entdeckungsgenauigkeit herrschte in der Audio-Kondition vor. Die Ergebnisse zeigen auf, welche Faktoren die Beurteilung der Glaubwrdigkeit eines Senders und die Genauigkeit bei der Unterscheidung von Wahrheit und Tuschung beeinflussen; insbesondere unter Bedingungen, in denen der Sender die Botschaft interaktiv produziert.
Los Prejuicios Cognitivos y La Disponibilidad de la Clave No Verbal en la Deteccin del Engao Judee K. Burgoon, University of Arizona J. P. Blair, Texas State University Renee E. Strom, St. Cloud State University Resumen
En situaciones potencialmente engaosas, la gente confa en los atajos mentales para ayudarse en el procesamiento de informacin. Estos juicios heursticos son a menudo tendenciosos y dan como resultado evaluaciones imprecisas acerca de la veracidad del emisor. Cuatro de esos prejuicios prejuicio sobre la veracidad, prejuicio visual, prejuicio sobre el comportamiento, y prejuicio sobre la violacin de expectacin fueron examinados en un experimento de juicio variando la disponibilidad de la clave no verbal y el engao. Los observadores vieron una entrevista completa grabada en video (con acceso pleno a las claves visuales, vocales y verbales), escucharon la entrevista en su totalidad (acceso a lo vocal y verbal), leyeron una transcripcin (acceso a lo verbal) de un sospechoso veraz mentiroso cuestionado sobre un presunto robo, luego clasificaron al entrevistado acerca de la informacin, el comportamiento, el manejo de la imagen y la veracidad. Los resultados respaldaron la presencia de los 4 prejuicios, que fueron ms evidentes cuando los entrevistados mintieron y los observadores tuvieron acceso a las modalidades visuales, vocales, y verbales. Los mensajes de los impostores fueron juzgados como ms completes, honestos, claros, y relevantes; sus comportamientos fueron ms involucrados y dominantes; y sus comportamientos en general fueron ms crebles, con el aditamento de las claves no verbales. Los impostores fueron juzgados actualmente como ms crebles que aquellos que decan la verdad en la modalidad audio visual, mientras que la mayor discriminacin y certeza de deteccin ocurri en la condicin auditiva. Los resultados tienen implicancias sobre qu factores influyen los juicios sobre la credibilidad del emisor de un mensaje y la certeza para distinguir la verdad de la mentira, especialmente bajo condiciones en la cuales los emisores producen mensajes en forma interactiva. Palabra claves: decepcin, comportamiento no verbal, comunicacin interpersonal, credibilidad, confianza, modalidad, CMC
Judee K. Burgoon J. P. Blair Renee E. Strom St. Cloud
Judee K. Burgoon, University of Arizona J. P. Blair, Texas State University Renee E. Strom, St. Cloud State University
, . . (, , , ) . (, , ) , ( ) , ( ) . , , . , , , 4 . ,
, . . AV . , . , .

Cognitive Biases and Nonverbal Cue Availability in Detecting Deception

Uploaded by

Copyright:

Available Formats

Cognitive Biases and Nonverbal Cue Availability in Detecting Deception

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cognitive Biases and Nonverbal Cue Availability in Detecting Deception

Uploaded by

Copyright:

Available Formats

What are the four cognitive biases examined in the study?

What are the four cognitive biases examined in the study?

What conditions did observers view the interviews under?

What conditions did observers view the interviews under?

HUMAN COMMUNICATION research

Human Communication Research ISSN 0360-3989