Theorizing About Deception

Abstract

This essay provides a unifying commentary concluding this special issue on new theories of deception. Information Manipulation Theory 2 (IMT2) and Truth-Default Theory (TDT) offer perspectives of deception that contrast with many past and current approaches. Key points of difference between these new theories and prior works include how deception is defined, the centrality of deception cues, the role of stakes in deception, the importance of communication media or channel, and whether or not deception is intrinsically more cognitively effortful than truthtelling. IMT2 and TDT shift focus away from deception cues and toward situated, contextualized information and communication content. Theory-data consistency is argued to be paramount. The overarching goal of these new theories is to chart the course for future deception research.

Keywords

theory deception lying Truth-Default Theory Information Manipulation Theory

Let us begin this concluding essay with an expression of gratitude. We are honored to have our new theories appear in the Journal of Language and Social Psychology, and we are deeply grateful to Howie Giles for his support in making this special issue possible. We also appreciate the insightful commentaries by Cole (2014), Greene (2014), Harwood (2014), Van Swol (2014), Verschuere and Shalvi (2014), and Walczyk (2014). The diversity of perspectives offered by these astute and accomplished scholars is invaluable. We are extraordinarily pleased with how this special issue turned out, and we hope that readers find this collection of essays valuable.

To understand the motivation behind both Information Manipulation Theory 2 (IMT2) and Truth-Default Theory (TDT) that resulted in this special issue, we must first provide the historical back-story. When we began our careers as deception researchers in the 1980s, we lamented the lack of theory. Deception scholarship, at that point in time, was largely variable-analytic or “dust-bowl” in nature. Rather than being guided by undergirding theory, researchers focused their attention on specific and narrow (albeit important) empirical questions. How can we improve detection accuracy? What individual differences variables predict proclivity to deceive? Does a set of consistent, cross-individual, behavioral correlates to deception exist? The result was a domain pockmarked by particularization: each scholar settled into a comfortable micro-niche and gathered data set after data set, without questioning underlying presumptions or considering other concerns.

As the years have passed, we’ve witnessed the emergence of several theories and frameworks dedicated to addressing these old questions. But rather than making things better, in many ways things have gotten worse. Why? Because deception scholars remain beholden to what McCornack (1997) described as “hopeful myths”—beliefs about deception that are intuitively pleasing but empirically false. As Fiedler and Walka (1993) first articulated more than 20 years ago, most deception scholars use “intuitive but debateable assumptions about the manner in which dishonesty is manifested in communication and discovered by communication partners” to guide their theory construction and research designs (pp. 199-200). The result is that extant deception theory lacks verisimilitude and fruitful agenda setting.

Worse yet, the data resulting from “programmatic, theory-driven” deception research routinely depart from theoretical predictions. As several large-scale meta-analyses have made clear, a startling truth about deception research is that theory-inconsistent findings are the norm rather than the exception. But rather than sparking a wholesale questioning of theories and underlying assumptions, scholars instead have adapted their theories to be nonfalsifiable. No matter what the findings, they always either are embraced as “confirmation” or dismissed as methodological artifacts.

We believe that enough is enough. More precisely, the time is long overdue for deception scholars to abandon their old presumptions, embrace the observable characteristics of deception, and set about generating viable theories that explain these characteristics (McCornack, 1997). For examples, and as useful start-points, deception theories (and scholars advancing them) need to provide satisfying accounts of three key meta-analytic findings that define the literature:

Research has failed to identify reliable behavioral cues that differentiate between honest and deceptive communication (DePaulo et al., 2003). Not only are such differences typically small, but they are heterogeneous across studies, individual communicators, and messages (DePaulo et al., 2003; Levine et al., 2005) and findings have become systematically weaker over time as evidence has accumulated (Bond, Levine, & Hartwig, 2014).

Deception detection accuracy hovers around slightly-better-than-chance regardless of factors such as training, professional experience, age of judge, communication media involved, presence or absence of interaction, relationship between sender and judge, planned or spontaneous messages, and sender motivation (Bond & DePaulo, 2006).

Truth-bias is remarkably robust (Bond & DePaulo, 2006; Levine, Park, & McCornack, 1999).

Beyond explaining these past findings, deception theories need to pave future paths forward, by addressing two critical concerns: how exactly deceptive discourse is produced, and how deception detection accuracy can be improved.

Information Manipulation Theory 2 (IMT2; McCornack, Morrison, Paik, Wisner, & Zhu, 2014) and Truth-Default Theory (TDT; Levine, 2014) reflect our efforts to address these concerns and improve the state of deception theory in doing so. In this concluding essay, our aim is not to debate specific claims made in the commentaries. We accept the critical observations along with the praise. Instead, our aim is to comment on critical issues related to the past, present, and future of theory on deception. Thus, this essay is not a rejoinder but, instead, a unifying commentary explaining the common vision offered by IMT2 and TDT.

Theoretical Commitments

In many ways, IMT2 and TDT are wildly contrastive. In terms of the mechanics of theory construction, TDT was developed more abductively, whereas IMT2 was deductively derived. IMT2 is a production theory, whereas TDT focuses more on detection. TDT is farther along in terms of theory testing than IMT2. Nevertheless, despite these differences, we share a common set of theoretical commitments, a unified perspective on what should be expected from theory, and a united vision of the future of deception theory.

First off, we flatly reject the desirability of theory for theory’s sake. The aim is not more theory, but better theory. Theory is not inherently valuable. Instead, theory is valuable to the extent that it makes predictions that turn out to be correct, offers explanations that enhance valid understanding, and leads to new knowledge that would be overlooked absent the theory. Some theories, however, are at odds with the data, offer misguided explanations of phenomena, fail to produce useful applications, and blind its advocates to meaningful advances.

On what basis can we distinguish good theory from bad? In seeking to develop new theory, there is no shortage of priorities that might take precedence. One might seek elegance, precision, comprehensiveness, explanatory power, parsimony, heuristic value, applicability to real-world problems, or any number of other desirable qualities. We hope our theoretical contributions score well on criteria such as these. But we believe that three criteria trump all others in terms of importance; and we suggest a fourth criteria that also should be considered a high priority, albeit in a different way.

As a first requisite criterion, a theory must be logically coherent. A theory that contradicts itself can be dismissed without further consideration. Consumers of theory, therefore, should always consider the posited propositions in tandem, and ask: Do they cohere as a set? We believe that both IMT2 and TDT pass muster on this criterion.

Second, theory needs to make specific, testable, and falsifiable predictions. Examples of precise predictions include the fact that Quantity violations will be the most frequent form of deception (IMT2 Proposition IM2), and the linear base-rate effects in TDT. Again, and at the risk of sounding militant, we entreat our fellow scholars to rigorously assess extant deception theories in terms of this criterion. If no findings are possible that would be inconsistent with a “theory,” then the “theory” isn’t a theory.

Third, and perhaps most critically, theory-based predictions need to be data consistent. We know that this sounds silly to suggest. Do we really have to overtly state the importance of sustaining only those theoretical propositions that are data-supported? Unfortunately, yes. Why? Because arguably the bulk of current guiding presumptions in the deception literature simply don’t match the data.

Beyond these three, we also believe that useful theory needs to be generative. By generative, we don’t mean simply that the theory needs to spawn research, although we hope our theories do that in abundance. But—akin to our view on theory in general—we reject the value of research-for-research’s sake. Rather than just generating research, good theory needs to make novel predictions, lead to new findings, and take research in directions that would not have been taken absent the theory. Theory should tell us how to design our research and which variables to prioritize. And all the while, and most important, theory should propel our knowledge of the domain forward in powerful, profound ways.

What Is Deception?

One of the first notable differences between our new theories and the status quo is how deception is defined. The vast majority of deception theory and research is based on the idea that deception is, by definition, intentional. Like the original IMT, IMT2 and TDT adopt functional definitions of deception and do not require conscious intent. Our theories do not preclude intentional, and planned, deception; but we do not limit the scope of deception to the communicative result of conscious intent.

Cue Theories: A Path Not Taken

Many prior deception theories can be characterized as “cue theories.” Examples include Ekman and Friesen’s (1969) original leakage theory and its newer variants, Zuckerman, DePaulo, and Rosenthal’s (1981) four factor theory, Interpersonal Deception Theory (Buller & Burgoon, 1996), and Vrij’s cognitive load paradigm (Vrij & Granhag, 2012; Vrij, Granhag, & Porter, 2010). What each of these theories has in common are the predictions that honest people behave differently from liars; that (consequently) cues exist that can usefully distinguish truths from lies; and that deception cues are explainable in terms of various psychological mediating states such as emotions, anxiety, arousal, cognitive effort, and/or strategic efforts to appear honest.

IMT2 and TDT, in contrast, might be characterized as noncue theories. Neither deception cues nor their mediating processes play a role in either theory. Neither theory specifies that truths and lies necessarily produce systematic differences in emotional states, arousal, cognitive effort, or the desire to be believed. As a consequence, neither theory is concerned with the nonverbal behaviors associated with felt emotions, arousal, cognitive effort, and the desire to appear honest. Both IMT2 and TDT place the theoretical and empirical focus elsewhere.

Stakes?

A widely accepted assertion in deception research is that high-stakes lies are different than small, everyday lies (e.g., Harwood, 2014). High stakes are those in which a liar has much to gain from successful deception and much to lose if the lie is detected. The theoretical basis for the importance of stakes goes back to Ekman and Friesen (1969). In their view, leakage cues were tied to strongly felt emotions, and stakes were tied to emotional intensity. When the stakes were low, little difference in emotions between honest and deceptive communication would exist. Correspondingly, leakage and deception cues were expected only when stakes were sufficient high.

As research evolved, it became increasingly clear that deception cues were weak and inconsistent. The idea of stakes came to be used in a circular fashion to save theory from the data. When predicted findings did not obtain, it was reasoned that the stakes were not sufficiently high. It was known that the stakes were not high enough because the predicted findings were not obtained. Although stakes is often held to be a critical consideration in deception research, surprisingly little evidence suggests that stakes make a substantial difference. According to meta-analyses, the extent of sender motivation exerts relatively little influence on either observable deception cues or deception detection accuracy (Bond & DePaulo, 2006; DePaulo et al., 2003). In DePaulo et al.’s meta-analysis of deception cues, only one cue (vocal pitch; 158 cues were examined) was found that both exhibited a substantial difference (i.e., d > .2) between honesty and deception and that was significantly moderated by sender motivation. In terms of accuracy, Bond and DePaulo report average accuracy of 53.3% for motivated lies compared to 53.4% for lies classified as unmotivated. Thus, although stakes is held by prior theory to be a crucial consideration, consistent empirical evidence for this view has failed to emerge.

Defenders of the primacy of high-stakes lies view will be quick to dismiss the lack of affirmative evidence for their predictions as a result of weak induction of stakes. “If only the stakes were higher,” they opine, “then deception cues would have been observed”; and “if only the induction of stakes were stronger, then the predicted differences between high and low stakes would surely obtain!” But such arguments, although coherent from within the theoretical view that produced them, are counterfactual. Furthermore, as mentioned previously, they lead to circular reasoning.

Motivation is critical to both IMT2 and TDT in the same way. Both theories hold that people deceive for a reason, and if honesty is sufficient for efficient goal attainment, people will be honest. But, the magnitude of stakes is not a key variable in either theory. The key issue for both of the new theories is the match between what is desired and the truth, not the consequences of potential detection.

Cognitive Load

We can think of perhaps no other issue within deception that currently evokes more zealotry and dogmatism than cognitive load. Many scholars are adamant that deception always evokes more load than truthtelling, regardless of context. To even suggest that certain conditions may exist under which deception is more efficient than truthtelling is treated as heresy.

Such zeal is understandable. Some scholars have built their reputations and staked their careers—in terms of publications and funding—on the validity of this premise. But the fact that people are ardently invested in a belief set doesn’t mean that it’s true. Anyone who has read Festinger, Riecken, and Schachter’s (1956) brilliant When Prophecy Fails knows this to be the case. And with regard to cognitive load, even the simplest of anecdotes defeats this premise. Ask any 10-year-old why she chose to lie to a friend, and she’ll tell you: because it was easier than telling the truth. We sometimes feel like deception researchers are the only people on the planet who maintain that lying is always harder than truthtelling.

Consider the issue of memory—a scholarly domain we feel is central to understanding deception. If we ask you, the reader, to “tell the truth” about what you were doing on the evening of March 2, 2012, at precisely 7:42 PM EST—or, alternatively, lie about it—which production task is easier? Sifting through the maze of long-term memory to find metaphorically cobwebbed bits of now-disassociated information and string them together into a coherent narrative? Or immediately and seamlessly stream false information directly from working memory to speech production? The answer is so obvious it’s absurd: the latter. Yet some scholars continue to maintain that truthtelling is always more efficient than lying—even when false information is readily deployable from working memory and the truth isn’t; the truth is difficult to retrieve from long-term memory; and/or the truthful information is impossible to contextually package in a face- and relationship-maintaining fashion.

Anecdotes and arguments aside, when we look at the data, we find that cues linked with cognitive effort fare no better than those linked with arousal or emotions in meta-analyses. If lying were, in fact, consistently more cognitively effortful than truthtelling, we would expect longer response latencies and less fluent speech to be associated with deception. DePaulo et al. (2003) show that this is not the case. The only cue in DePaulo et al. that was both mentioned as an indication of cognitive effort and that showed a substantial difference between truths and lies was pupil dilatation (d = .39), and pupil dilation is also an indication of arousal.

It would be one thing if such beliefs were benignly ensconced. But the cognitive load viewpoint is aggressively advanced and defended by adherents. We, and our colleagues (including some of the commentators in this issue), have regularly (over the last decade) attempted to publish studies documenting the contextual conditions under which truthtelling evokes greater load than lying. Yet such manuscripts routinely have been blockaded by reviewers and editors who have no reason for rejecting the data other than that they disagree. Such dogmatism has no place in science.

Communication Medium and Channel Issues

A long mainstay of deception detection research has been the search for medium or channel effects. Early research predicted differences between text, audio-only, and audiovisual presentation for truths and lies. Cue theories all predict such differences because various cues are more or less available in certain communication formats and media. With the advent of Interpersonal Deception Theory, focus was shifted to the difference between face-to-face and mediated communication. More recently still, much research involves a search for differences between various forms of new media (texts, email, social networks, online dating, etc.) or is predicated on the existence of differences by new and old media.

Meta-analysis, however, clearly shows that deception detection accuracy is remarkably consistent across mediums and channels. The across-study average accuracy is between 53% and 54% (Bond & DePaulo, 2006). In comparison to this grand mean, average accuracy is 54.0% for audiovisual-mediated communication, 53.8% for audio only, 52.6% when there is no interaction at all, 52.8% with face-to-face interaction, and 54.0% with observation of interaction with a third party. Mere media does not seem to make much difference.

Media is not a variable in either of our theories. The focus of both theories is more on face-to-face interaction, but we would expect the predictions of both theories to hold across new and old media.

The Primacy of Information and Contextual Constraints

For decades, deception research has been mired in models positing individual differences (e.g., self-monitoring, Machiavellianism, message design logics, lie acceptability, etc.) as causing deception. Both IMT2 and TDT suggest simpler antecedents. The strongest determinant of whether someone will deceive or tell the truth is the nature of the information they possess in working and long-term memory, and the particular contextual constraints they face in disclosing the truthful information. In situations where little personal, relational, or professional costs are attached to disclosing information, people will tell the truth. When people possess information they deem too problematic to disclose, they will deceive. When contextual demands “put them on the spot” to disclose unpleasant truths—and no potential falsehoods are readily accessible within working or long-term memory—they will confess. In all of these cases, the driving force behind such behaviors is the practical, contextual goodness-of-fit of the possessed information, and the contextual constraints that determine that goodness-of-fit.

When one embraces such a “situated” perspective, several implications immediately arise—implications that precisely match extant data. For examples, most people will tell the truth most of the time (Serota & Levine, 2014; Serota, Levine, & Boster, 2010). Why? Because much of daily discourse doesn’t involve problematic information. Similarly, almost everyone will lie, if the conditions are right (e.g., the truthful information they possess is sufficiently problematic; false information is readily available within working- or long-term memory; Levine, Kim, & Hamel, 2010). And—an exceptionally important practical implication—we should be able to get people to confess the truth, if we ask them the right questions (Levine, Clare, et al., 2014; Levine, Shulman, Carpenter, & DeAndrea, 2013). What are “the right questions”? Those that allow easy access to truthful disclosure within memory but difficult access to alternative false information.

Contextualized Verbal Content, Not Cues, Lead to Detection Accuracy

Contextual goodness-of-fit is also critical for lie detection. To the extent that our theories, especially TDT, have verisimilitude, the way to achieve higher accuracy is not through the observation or even active prompting of deception cues. Instead, verbal content is key and verbal content must be understood in the context it was produced. High accuracies can be obtained by questioning designed to yield answers that can be either fact checked or at least judged for plausibility, and by questioning designed to get liars to be honest and confess their lies (Blair, Levine, & Shaw, 2010; Levine, Blair, & Clare, 2014, Levine, Clare, et al., 2014). This is, after all, how lies are detected outside the lab: either by comparing what is said to some type of evidence or by confession (Park, Levine, McCornack, Morrison, & Ferrerra, 2002).

Conclusion

We are honored and gratified to have our new theories presented in this special issue of Journal of Language and Social Psychology. We hope readers find our articles and the commentaries informative and thought provoking. We hope too that this special issue provides a critical turning point in the academic study of deception away from hopeful myths and toward verisimilitude.

Our vision of the future of deception theory and research is one in which serious attention is given to message production. Deception is seen as functional, but not necessarily consciously intentional. Preoccupation with cues is replaced with attention is situation-specific message content. Variables such as stakes, media, and channel are not so important. This shift of focus will enable deception detection research to break out of the slightly-better-than-chance quagmire and obtain and replicate higher levels of accuracy. Mostly importantly, we envision a future in which the simplest of empirical goals are attained: theoretical predictions are borne out in the data, and replicate.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Timothy R. Levine is a professor in the School of Media and Communication, Korea University, Seoul, South Korea. He has published extensively on the topics of deception, credibility assessment, and interpersonal communication.

Steven A. McCornack (PhD, 1990, University of Illinois at Urbana-Champaign) is an associate professor in the Department of Communication at Michigan State University. His primary research area is deceptive discourse production.

References

Blair

J. P.

Levine

T. R.

Shaw

A. S.

(2010). Content in context improves deception detection accuracy. Human Communication Research, 36, 420-439.

Bond

C. F.

Jr. DePaulo

B. M.

(2006). Accuracy of deception judgments. Personality and Social Psychology Review, 10, 214-234.

Bond

C. F.

Jr. Levine

T. R.

Hartwig

(2014). New findings in nonverbal lie detection. In Granhag

P. A.

Vrij

Vershuere

(Eds.), Deception detection: Current challenges and new directions. Chichester, England: Wiley.

Buller

D. B.

Burgoon

J. K.

(1996). Interpersonal deception theory. Communication Theory, 6, 203-242.

Cole

(2014). Reconsidering the role of intentionality in deceptive communication: A commentary on IMT2 and TDT from a relational science point of view. Journal of Language and Social Psychology, 33(4) 393-397.

DePaulo

B. M.

Lindsay

J. J.

Malone

B. E.

Muhlenbruck

Charlton

Cooper

(2003). Cues to Deception. Psychological Bulletin, 129, 74-118.

Ekman

Friesen

W. V.

(1969). Nonverbal leakage and clues to deception. Psychiatry, 32, 88-106.

Fiedler

Walka

(2006). Traing lie detectors to use nonverbal cues instead of global heurisitcs. Human Communication Research, 20, 199-223.

Festinger

Riecken

H. W.

Schachter

(1956). When prophecy fails: A social and psychological study of a modern group that predicted the destruction of the world. Minneapolis: University of Minnesota Press.

10.

Greene

J. O.

(2014). Just looking around. Journal of Language and Social Psychology, 33(4) 398-404.

11.

Harwood

(2014). Easy lies. Journal of Language and Social Psychology, 33(4) 405-410.

12.

Levine

T. R.

(2014). Truth-default theory (TDT): A theory of human deception and deception detection. Journal of Language and Social Psychology, 33(4) 378-392.

13.

Levine

T. R.

Blair

J. P.

Clare

D. D.

(2014). Diagnostic utility: Experimental demonstrations and replications of powerful question effects in high stakes deception detection. Human Communication Research, 40, 262-289.

14.

Levine

T. R.

Clare

D. D.

Blair

J. P.

McCornack

S. A.

Morrison

Park

H. S.

(2014). Expertise in deception detection involves actively prompting diagnostic information rather than passive behavioral observation. Human Communication Research (published online July 9).

15.

Levine

T. R.

Feeley

McCornack

S. A.

Harms

Hughes

(2005). Testing the effects of nonverbal training on deception detection accuracy with the inclusion of a bogus train control group. Submitted to Western Journal of Communication, 69, 203-218.

16.

Levine

T. R.

Kim

R. K.

Hamel

L. M.

(2010). People lie for a reason: An experimental test of the principle of veracity. Communication Research Reports, 27, 271-285.

17.

Levine

T. R.

Park

H. S.

McCornack

S. A.

(1999). Accuracy in detecting truths and lies: Documenting the “veracity effect.” Communication Monographs, 66, 125-144.

18.

Levine

T. R.

Shulman

Carpenter

DeAndrea

D. C.

(2013). The impact of accusatory, non-accusatory, bait, and false evidence questioning on deception detection. Communication Research Reports, 30, 169-174.

19.

McCornack

S. A.

(1997). The generation of deceptive messages: Laying the groundwork for a viable theory of interpersonal deception. In Greene

J. O.

(Ed.), Message production: Advances in communication theory (pp. 91-126). Mahwah, NJ: Lawrence Erlbaum.

20.

McCornack

S. A.

Morrison

Paik

J. E.

Wisner

Zhu

(2014). Information Manipulation Theory 2 (IMT2): A propositional theory of deceptive discourse production. Journal of Language and Social Psychology, 33(4) 348–377.

21.

Park

H. S.

Levine

T. R.

McCornack

S. A.

Morrison

Ferrerra

(2002). How people really detect lies. Communication Monographs, 69, 144-157.

22.

Serota

K. B.

Levine

T. R.

(2014). A few prolific liars: Variation in the prevalence of lying. Journal of Language and Social Psychology. doi: 10.1177/0261927X14528804.

23.

Serota

K. B.

Levine

T. R.

Boster

F. J.

(2010). The prevalence of lying in America: Three studies of reported deception. Human Communication Research, 36, 1-24.

24.

Van Swol

L. M.

(2014). Questioning the assumptions of deception research. Journal of Language and Social Psychology, 33(4) 411–416.

25.

Verschuere

Shalvi

(2014). The truth comes naturally! Does it? Journal of Language and Social Psychology, 33(4) 417–423.

26.

Vrij

Granhag

P. A.

(2012). Eliciting cues to deception and truth: What matters are the questions asked. Journal of Applied Research in Memory and Cognition, 1, 110-117.

27.

Vrij

Granhag

P. A.

Porter

S. B.

(2010). Pitfalls and opportunities in nonverbal and verbal lie detection. Psychological Science in the Public Interest, 11, 89-121.

28.

Walczyk

J. J.

(2014). A commentary on Information Manipulation Theory 2: Its place in deception research and suggestions for elaboration. Journal of Language and Social Psychology, 33(4) 424–430.

29.

Zuckerman

DePaulo

B. M.

Rosenthal

(1981). Verbal and nonverbal communication of deception. In Berkowitz

(Ed.), Advances in experimental social psychology (Vol. 14, pp. 1-59). New York, NY: Academic Press.