A memory advantage for untrustworthy faces

Abstract

Inferences of others’ social traits from their faces can influence how we think and behave towards them, but little is known about how perceptions of people’s traits may affect downstream cognitions, such as memory. Here we explored the relationship between targets’ perceived social traits and how well they were remembered following a single brief perception, focusing primarily on inferences of trustworthiness. In Study 1, participants encoded high-consensus trustworthy and untrustworthy faces, showing significantly better memory for the latter group. Study 2 compared memory for faces rated high and low on a series of traits (dominance, facial maturity, likeability, and trustworthiness), and found that untrustworthy and unlikeable faces were remembered best, with no differences for the other traits. Finally, Study 3 compared information about trustworthiness from facial appearance and from behavioral descriptions. Untrustworthy targets were remembered better than trustworthy targets both from behavior and faces, though the effects were significantly stronger for the latter. Faces perceived as untrustworthy therefore appear to be remembered better than faces perceived as trustworthy. Consistent with ecological theories of perception, cues to trustworthiness from facial appearance may thus guide who is remembered and who is forgotten at first impression.

Introduction

Walking along a busy street or through a crowd, many of the people that pass by slip through our thoughts relatively unnoticed—or, at least, unremembered (Simons & Levin, 1998). But some of the faces in the crowd stick with us. In the present work, we examined why, after brief encounters, some faces are better remembered than others. In doing so, we focused on the role of social traits, particularly trustworthiness. Perceivers quickly and consistently evaluate others’ social traits at first perception (Ambady, Bernieri, & Richeson, 2000). The attribution of a variety of social traits to others based on photos of their faces shows agreement across perceivers after just 50 ms of viewing time (Bar et al., 2006, Rule et al., 2009, Willis and Todorov, 2006). Given the efficient and persistent nature of these judgments at brief perception, it stands to reason that they might influence subsequent processing of the individuals with effects on memory. The question as to what physical qualities support face memory is one that has been explored in some depth. One of the strongest indicators of whether a face is remembered is its distinctiveness (Vokey & Read, 1992). People with atypical appearances are better remembered than those who appear more typical, likely because the former stand out as salient (Light, Kayra-Stuart, & Hollander, 1979).

To date, much of the research examining memory differences based on social traits has utilized methods that direct participants’ attention to particular aspects of the targets. Typically, participants are asked to view a series of individuals and to evaluate aspects of their physical appearance (e.g., facial attractiveness; Barclay & Lalumiere, 2006), or to make specific inferences about their character (e.g., likeability; Bell & Buchner, 2010). For example, in one study, participants were asked to rate a series of targets on a host of traits, including honesty, after which they chose among a number of professions (half of which were criminal occupations; e.g., armed robber) that they believed the target individuals might occupy (Yarmey, 1993); subsequently, the participants were given a memory test. This was not random design, as the researchers were interested in measuring whether trustworthy “good guys” and untrustworthy “bad guys” might be remembered differently; the participants’ judgments foreshadowed this and were expected to influence memory.

Indeed, trustworthiness, a fundamental dimension of person judgment (Oostehrof & Todorov, 2008), has received considerable attention in the study of person memory (e.g., Oda, 1997). For instance, some studies have shown that untrustworthy individuals are remembered better than trustworthy individuals (e.g., Bayliss and Tipper, 2006, Mealey et al., 1996, Mueller et al., 1988, Oda, 1997), even among children (Kinzler & Shutts, 2008). This is often credited to the possibility of an evolved “cheater detection module” in human cognition and behavior (Cosmides, 1989). Other research (Barclay and Lalumiere, 2006, Mehl and Buchner, 2008) has documented some flaws in the methods used in several of these earlier studies and has highlighted the importance of equating the verbal information used to convey trustworthy and untrustworthy behaviors, as well as the importance of the context created by this verbal information (see below; e.g., Bell & Buchner, 2011). Moreover, the bulk of previous work has largely relied upon descriptions of people, rather than the natural signals of trustworthiness inherent in their facial appearances. Yet people show high agreement in their ratings of trustworthiness from faces (e.g., Zebrowitz, Voinescu, & Collins, 1996).

Asking participants to form impressions about trustworthiness and closely related traits (e.g., likeability; see Rule et al., 2010) prior to testing their memory may have limited ecological validity. In everyday life, one typically meets people without conscious goals to evaluate those individuals along a specific trait. Rather, our impressions of others are usually formed quickly, automatically, and without much conscious deliberation (Macrae & Bodenhausen, 2000). This is particularly relevant to cases of eyewitness identifications—an important area of research in memory (e.g., Meissner & Brigham, 2001). Witnessing a crime is generally fast, stressful, and rarely is one expected to be considering specific physical qualities or character traits of the perpetrator. Instead, our exposures to strangers—within and outside of crime scenes—tend to be brief, passing, and unconstrained by evaluative instructions. Thus, in the present work, we sought to explore memory for trustworthy versus untrustworthy targets using an experimental design that might be better-suited to understanding how impressions are formed in a real-world setting. We therefore exposed participants to faces in a passive-viewing task without any contextual information (such as behavioral descriptions; Mehl & Buchner, 2008), complex strategy evaluations (e.g., economic games; Oda & Nakajima, 2010), or instruction to assess specific traits related to the dimension upon which the targets were intended to differ (e.g., criminality, trustworthiness, likeability; Bell & Buchner, 2011); as used in previous studies.

The present work therefore sought to consider the role of facial trustworthiness in memory for individuals in a different context. Principally, we were interested in the role that facial appearance alone plays in memory for individuals seen once and without instruction to process the faces in any particular way; i.e., similar to encountering strangers in everyday life. In addition, we investigated the relative contributions of facial and verbal sources of information about trustworthiness to memory for faces. We therefore adopted an ecological approach to considering memory differences for glimpsed faces differing in trustworthiness.

Drawing from the ecological theory of object perception (Gibson, 1979), ecological theories of social perception posit that percepts have utility and function (McArthur & Baron, 1983; Zebrowitz & Collins, 1997). Facial displays are thought to have relevant inherent value for communicating properties about individuals, particularly their social traits (see Zebrowitz, 1997), and memory for faces might vary according to their functional value (see also Nairne & Pandeirada, 2008). Accordingly, trustworthy faces could signal opportunities for cooperation and affiliation, and untrustworthy faces might signal dangers to be avoided (e.g., Buchner et al., 2009, Cosmides, 1989, Slepian et al., 2012, Suzuki and Suga, 2010). From the perspective of the ecological theory of social perception, we expected untrustworthy faces to be salient to perceivers and therefore better remembered—both because they are considered to be of high ecological value (e.g., Buchner et al., 2009, Suzuki and Suga, 2010) and because they are likely to occur more rarely (e.g., Barclay, 2008), even when they are glimpsed briefly.

Throughout the evolution of research on the relationship between targets’ trustworthiness and how well they are remembered, a number of nuances have emerged. An important theme seems to be that context is highly influential upon these effects. Initial accounts of differences in memory for individuals based on trustworthiness were situated within theories of an evolved capacity for “cheater detection” and reported differences in recognition memory between the trustworthy and untrustworthy targets (e.g., Mealey et al., 1996). Buchner and colleagues revised this perspective to suggest that source memory (the context within which a target is encountered) may be more valuable for evolutionary reasons than recognition memory, and presented empirical evidence in support of this claim (Bell and Buchner, 2010, Bell and Buchner, 2011, Buchner et al., 2009).

Suzuki and Suga (2010) found that trustworthy faces were remembered better than untrustworthy faces but only when they were encoded in an incongruent context (i.e., as behaving in an untrustworthy manner), which participants learned through repeated experiences with the targets in an economic game. Thus, targets who essentially surprised participants with financial punishments were remembered best by the end of the game, suggesting that participants learned to become vigilant about these “wolves in sheep’s clothing.” Thus, the context within which targets were encoded into memory played an important role in each of these effects, and depended critically upon the information that the experimenters supplied to the participants about the targets.

The current work therefore extended this consideration of context to a new domain with high ecological validity: memory for faces encountered briefly, as in first impressions of strangers. That is, similar to making judgments of people encountered in passing on the street, does the relative trustworthiness of those persons affect who is remembered and who is not? A central goal of this work was therefore to assess memory for faces in unconstrained situations where information about targets was captured from their appearances, rather than provided by verbal descriptors or direct interactive experience—such as that obtained in an economic game. Distinct from the majority of previous work in this research area, we therefore implemented an undirected (i.e., passive-viewing) incidental-encoding method of introducing perceivers to the targets and measured how properties of the faces’ inferred traits might relate to whether they are remembered. Given the importance of contextual information in influencing memory, however, we were also concerned with the interaction between perceived (appearance-based) and described (verbally-stated) trustworthy behavior, which might help to illuminate some of the differences observed in previous studies (e.g., Barclay and Lalumiere, 2006, Kinzler and Shutts, 2008, Mealey et al., 1996, Yarmey, 1993). Thus, we sought to consider both sources of information, with particular attention paid to contributions from facial appearance, and examined the relative contributions of facial and verbal information.

Two studies have explored the relationship between the relative trustworthiness of targets’ actual (rather than described) behavior and perceivers’ memory for those faces. Yamagishi, Tanida, Mashima, Shimoma, and Kanazawa (2003) found that participants remembered the faces of people who defected in a prisoner’s dilemma game better than the faces of people who cooperated. This was based only on their facial appearance, as the participants were not given information about who had defected and who had cooperated. Yet, unexpectedly, participants were unable to accurately distinguish the defectors from the cooperators when asked explicitly (as cited in Verplaetse, Vanneste, & Braeckman, 2007). Verplaetse et al. (2007) conducted similar studies in which they found that defectors and cooperators could be distinguished, however, and that their faces affected perceivers’ subsequent recognition memory. The present work therefore aimed to investigate how the perception of social traits might influence individuals’ memory for others’ faces based on a single brief perception. Aside from trustworthiness, there has been some evidence that likeability, a similar and highly-correlated trait (e.g., Rule et al., 2010), affects the recognition of faces. Additional work has shown that other traits, such as dominance (Mueller & Mazur, 1996) and facial maturity (Berry & McArthur, 1985), are particularly important for the social perception of faces and that these traits are especially relevant to how salient a face is to a perceiver (Zebrowitz, 1997). In the current study, then, we examined the relationship between perceivers’ memory for faces and their impressions on these four traits: trustworthiness, likeability, dominance, and facial maturity. Given the reviewed empirical and theoretical work, we predicted that untrustworthy faces would be better remembered than trustworthy ones. Moreover, consistent with the ecological perspective of perception, we hypothesized that facial information might provide a strong contribution to recognition memory, possibly greater than the influence of description-based information, as faces are more perceptually and ecologically salient (Zebrowitz, 1997).

In Study 1, we therefore examined whether faces perceived to be untrustworthy with high consensus would be remembered in a passive-viewing incidental-encoding recognition memory task better than faces consensually perceived to be trustworthy. In Study 2, we expanded this investigation by exploring the potential role that the other traits (likeability, dominance, and facial maturity) might play in both perceivers’ memory for faces, as well as the memorability of individual faces when aggregated across perceivers. Finally, in Study 3, we investigated the interaction between facial cues to trustworthiness and explicit descriptions of the (un)trustworthy character of presented targets.