HOW PEOPLE [TRY TO] DETECT LIES IN EVERYDAY LIFE.

• Author: Sanchez, Nuria

1. Introduction

   Lying is considered a reprehensible form of social behavior, and children worldwide are socialized into honesty. An effective way to induce honesty is to foster in children the belief that lying will never go unnoticed, as it will generate specific emotions (e.g. shame) that will be revealed through behavioral cues (see Global Deception Research Team 2006). These practices may explain why people worldwide believe that behavior reveals deceit, and why they share similar stereotypes about deceptive behavior (Global Deception Research Team 2006).

   Similar to lay people, scientists have traditionally believed that deception leaves apparent behavioral traces and have spent decades searching for them. Indeed, the most prominent theoretical perspectives in deception-detection research make specific predictions about 'behavioral cues to deceit' (e.g. Buller and Burgoon 1996, DePaulo et al. 2003, Ekman 2009, Zuckerman et al. 1981).

   However, several meta-analyses cast serious doubt about the prospects of detecting deceit from behavior. First, the association between behavioral cues and deception is weak and under the influence of a host of moderator variables (DePaulo et al. 2003, Luke 2019, Sporer and Schwandt 2006, 2007). Second, across studies, observers' accuracy in judging veracity from behavior is low: 54%, compared to 50% chance accuracy (Bond and DePaulo 2006). Third, training observers to detect deception from behavioral observation alone yields only limited increases in detection accuracy (Hauch et al. 2016).

   These meta-analytical conclusions draw a dim picture concerning humans' prospect to detect deceit. However, one can question their relevance for everyday life lie detection. Most deception studies have been conducted in laboratory settings, and concerns have been raised that laboratory lie-detection experiments fail to mirror real-life deception and its detection (Levine 2018). In particular, there is some evidence that outside the laboratory most lies are detected from non-behavioral information rather than from behavioral cues, and that access to non-behavioral information boosts observers' accuracy in judging veracity. We discuss these topics next.

   1.1. Lie detection in the laboratory as compared to real-world settings

   In a typical laboratory experiment, senders tell inconsequential truths and lies at the request of an experimenter. Observers, who are unacquainted with the senders, have to judge whether each sender is truthful or deceptive. Normally, the senders' communications are either video-recorded, audio-recorded, typed or transcribed; thus, there is no opportunity for interaction. Finally, observers have to make their judgments immediately. Under these conditions, the only information available to observers is the senders' behavior at the time the lie is told. As explained above, behavior is a poor marker of deception; this might explain the poor accuracy rates obtained in laboratory experiments. (1)

   Conversely, in most everyday life circumstances, deception targets know the deceiver (background information) and can question him or her, both at the time the lie is told and later. Also, deception targets can determine whether the sender lied or told the truth long after the lie was told. This provides them with opportunities to carefully search for diagnostic non-behavioral information (such as tangible evidence, information from third parties, etc.). Detectors can also learn about the liar's circumstances, in particular their incentives or motivations to deceive. In short, a number of characteristics of real-life deceptive situations facilitate lie detection by enabling observers to access and use diagnostic non-behavioral information. While laboratory experiments tell us much about lie detection based on behavioral cues, they tell us very little about how people detect deception in real life. (2)

   1.2. In real life, people detect deception from non-behavioral information

   In a seminal study, Park, Levine, McCornack, Morrison and Ferrara (2002) asked college students to recall a lie they had detected in the past and to report how they detected it. The results showed that most everyday life lies are not revealed from behavioral cues, but from non-behavioral information such as third-person information, tangible evidence, inconsistencies with the detectors' knowledge, and the liar's confession. While behavioral cues are visible, vocal, and verbal behaviors [assumed to be] displayed by the sender at the time the lie is told (e.g. fidgeting, pauses, amount of detail in the statement...), non-behavioral information involves some knowledge about the context (i.e., knowledge that goes beyond the sender's behavior during the specific deceptive exchange) that contradicts the deceptive statement (Blair et al. 2012; see also Blair et al.'s [2010] notion of 'content in context'). Examples are pieces of information provided by witnesses or informed others, physical evidence, the detector's specific knowledge about the deception topic, and the liars' ultimate admission that the exchange was deceptive (see Park et al. 2002).

   Besides the predominance of non-behavioral over behavioral indicators, Park et al. (2002) also found that most detected lies had been told by familiar senders (friends, romantic partners, family members...) and had been discovered hours, weeks, or months after being told. Park et al.'s results provide strong evidence that laboratory experiments do not capture the ordinary real-life conditions where deception judgments are made, that empirical results of laboratory experiments cannot be generalized to everyday-life circumstances, and that in everyday life it is non-behavioral information, rather than behavioral cues, that allows people to detect deceit.

   Park et al.'s (2002) findings relative to the prominent role of non-behavioral information (compared to behavioral cues) in detecting real-life deception have been replicated more recently by Levine and Daiku (2019), Masip and Herrero (2015), Novotny et al. (2018), and Park and Lee (cited in Levine 2020). Masip and Sanchez (2019) conducted a mini meta-analysis of the four empirical studies available at the time and found that, across studies, 82% of the indicators reported were non-behavioral, while only 17% were behavioral (but see Sanchez and Masip 2020, for a study failing to find this effect).

   1.3. Non-behavioral information increases observers 'detection accuracy relative to behavioral cues

   Empirical research shows that access to non-behavioral information, which is typically absent in laboratory experiments, facilitates lie detection. Across eight deception-detection experiments, Blair et al. (2010) found that while the accuracy rate of observers with access to behavioral cues only was 57%, the accuracy rate of observers with access to non-behavioral information in addition to behavioral cues was 75%. Similarly, Bond et al. (2013, Experiment 3) found an accuracy rate of 51% among observers with access to visible behavior only, and of 76% among those with access to non-behavioral information (the senders' incentives to lie or tell the truth) in addition to visible behavior. A third group who only had access to non-behavioral information attained a 97% accuracy rate. (3) In short, compared to a behavior-only condition, access to non-behavioral information increases observers' accuracy in judging veracity (see also Blair et al. 2018).

   One reason why non-behavioral information helps people judge veracity is that it allows observers to compare the sender's statement with that [presumably] reliable information (Blair et al. 2010, 2012). This strategy consists of using the so-called correspondence criterion (Blair et al. 2018), which has long been considered by philosophers to be one way of determining whether a belief is true or false (see Dunwoody 2009). It is also consistent with the so-called situational familiarity hypothesis, according to which "in familiar contexts, observers are able to 'visualize' the situation in question and judge the plausibility and validity of verbal content" (Stiff et al. 1989: 560, see also Reinhard et al. 2011, 2012).

   1.4. Summary

   In short, while people worldwide believe that behavioral cues signal deception, the accumulated evidence indicates that behavior is a poor marker of deceit. Laboratory experiments show that people are poor truth/lie detectors, but this might be a consequence of experimental participants typically having access to behavioral cues only. Most everyday life lies are not detected from behavior but from non-behavioral information. Non-behavioral information is normally absent from laboratory experiments and has been shown to increase observers' accuracy in judging veracity. Because laboratory experiments typically fail to capture the characteristics of everyday life deception and its detection, naturalistic studies are needed. We examined some unexplored topics outside the confines of the laboratory.

   1.5. Theoretical background

   Our goal in conducting this study was to examine how lies are detected outside the laboratory, in real-life circumstances. Two recent deception theories helped us make a number of predictions: Levine's (2014, 2020) Truth-Default Theory (TDT) and Street's (2015) Adaptive Lie Detector account (ALIED). However, this research was not conceived to test any specific theory. Nor was it conceived to compare TDT against ALIED. In some respects, these two theories make similar predictions, but they differ in the posited underlying mechanisms. Ascertaining which theory is correct was indeed beyond the scope of this study, which was designed solely to explore how lies are detected in real life.

   Rather than a focused, unitary theory, Levine's TDT is "a collection of quasi-independent mini-theories, models, or effects that are joined by an overarching logic" (Levine 2014: 379). TDT is based on prior research by Levine and his colleagues, and it...