How Adaptive Is Memory for Cheaters?

Enhanced Memory for Cheaters as an Adaptation for Reciprocal Altruism?

The selection pressure exerted by the benefits and challenges of living in social groups has most likely been an important force driving human evolution (Dunbar, 2003). Therefore, social interaction is assumed to be supported by a variety of cognitive modules for social cooperation, speech, face processing, and theory of mind, among other things (Geary, 2007). Social contract theory (Cosmides & Tooby, 1989, 2005) posits that the human mind contains a specific cheater-detection module that supports social cooperation among unrelated individuals.

Human cooperation has fascinated evolutionists since Darwin (1871) because cooperation often entails accepting costs in order to help others, and thus does not result in immediate benefits for the cooperative individual. From an evolutionary viewpoint, it seems mysterious that cooperative strategies could outcompete more egoistic strategies that seem to have clear advantages at first glance. One solution to this problem is reciprocal cooperation (Axelrod & Hamilton, 1981; Trivers, 1971). Cooperation can be successful in the long run if it is based on reciprocity (i.e., if people return the favors they have been receiving). However, the problem remains that groups of reciprocal cooperators may be infiltrated by cheaters who try to exploit these cooperative tendencies by reaping the benefits of cooperation while shying away from the costs of helping others. Therefore, cooperation can evolve only in conjunction with mechanisms that help people detect and avoid cheaters (Cosmides & Tooby, 1989, 2005).

On the basis of these premises, Cosmides and Tooby have postulated that humans, as members of a highly cooperative species, have a cheater-detection mechanism that facilitates reasoning about social exchanges. The interpretation of the results that purportedly provide evidence for such a domain-specific reasoning mechanism is, however, quite controversial, and more general theories have been proposed to explain why people are so efficient at detecting cheaters (Sperber, Cara, & Girotto, 1995).

Given that reciprocity is often delayed across repeated interactions, reciprocal cooperation makes huge demands on memory. For instance, a “tit for tat” strategy—which is considered one of the most successful reciprocal strategies in social cooperation (Axelrod & Hamilton, 1981)—consists of cooperating with strangers and familiar cooperators, and retaliating against cheaters. Obviously, this strategy requires the ability to recognize cheaters. Therefore, it seems natural to generalize the predictions of social contract theory to memory. Indeed, it has been argued that people should be especially good at remembering cheaters because doing so would help them avoid social exploitation (Mealey, Daood, & Krage, 1996; Oda, 1997).

Memory for cheaters was first examined from an evolutionary perspective by Mealey et al. (1996), who presented people with pictures of faces paired with short statements describing the pictured persons as cheaters, neutral, or cooperative (Table 1). Subsequently, in a surprise face-recognition test, the faces were presented together with new faces, and participants had to indicate which faces they recognized from the initial learning phase. Faces of cheaters were better recognized than faces of trustworthy persons, albeit only when the faces were associated with professions indicating low social status. This was interpreted as evidence for a highly specific cheater-recognition mechanism. However, several follow-up studies failed to replicate the face-recognition advantage for cheaters (e.g., Mehl & Buchner, 2008).

OPEN IN VIEWER

Table 1.

Example Behavior Descriptions From Buchner, Bell, Mehl, and Musch (2009)

Condition	Description
Cheater	K. S. is a secondhand-car dealer. Regularly, he sells restored crash cars as supposedly accident-free and conceals serious defects from the customers.
Neutral	J. L. is a gardener. He is extremely interested in orchids and owns a collection of some very rare and expensive exemplars.
Cooperator	N. G. is a mechanic. He is always eager to provide spare parts as cheap as possible for his clients and to fulfill his jobs efficiently.

The absence of a face-recognition advantage for cheaters is not unexpected from an evolutionary perspective. Simply recognizing a face as familiar (item memory) and thus being able to classify it as “old” is not enough to avoid cheaters—it is also necessary to remember that the face belongs to a cheater. In other words, it is important to remember the cheating context in which the face was encountered (source memory).

To examine whether source memory for cheaters is indeed enhanced, we and our colleagues (Buchner, Bell, Mehl, & Musch, 2009) extended the paradigm used by Mealey et al. (1996) to include a face-recognition test in which participants were asked to remember whether the faces belonged to new, cheating, neutral, or cooperative persons.

We measured source memory using the multinomial source-monitoring model of Bayen, Murnane, and Erdfelder (1996). This model yields a source-memory parameter that represents the conditional probability of remembering the context of an item correctly recognized as old. This measure is known to be unaffected by variables that influence accuracy in old-new recognition and guessing, and can therefore be considered a “process-pure” measure of source memory. With this source-memory measure, a highly consistent pattern of data emerged across a number of experiments (Fig. 1). Consistent with evolutionary theories of reciprocal cooperation, source memory was much better for faces of cheaters than for faces of cooperators or control faces, regardless of whether the faces were remembered immediately or after a 1-week delay. In other words, participants were better at remembering that a face was associated with cheating than that a face was associated with neutral or trustworthy behavior.

OPEN IN VIEWER

Fig. 1.

Source memory (i.e., memory for the type of context in which a face was encountered) as a function of stimulus-character condition across four experiments in Buchner, Bell, Mehl, and Musch (2009). Participants were presented with pictures of people’s faces, each of which was accompanied by a description that characterized the person as a cheater, neutral (control condition), or cooperative. The source-memory parameter was obtained using Bayen, Murnane, and Erdfelder’s (1996) source-monitoring model and represents the conditional probability of remembering the source of an item correctly recognized as old. Error bars represent the 95% confidence intervals.

This pattern of results could be taken as evidence of a cognitive module that supports the avoidance of cheaters. However, such a conclusion should not be accepted unless the “cheater specificity” of the source-memory advantage is supported by empirical tests. Therefore, the research program described in the following sections has focused on the question of whether the source-memory advantage can be attributed to well-established, general effects of variables such as distinctiveness (Schmidt, 1991), expectancy violation (Ranganath & Rainer, 2003; van Kesteren, Ruiter, Fernandez, & Henson, 2012), and emotion (Kensinger, 2007) that have been invoked to explain many phenomena in the memory literature.

How “Cheater Specific” Is the Source-Memory Advantage for Cheaters?

One particularly promising account is based on the assumption that cheating elicits negative emotions, which may in turn cause the context eliciting these emotions to be particularly well attended and, hence, encoded. Given the bulk of literature showing that negative, threatening information is better remembered than neutral or positive information is (e.g., Kensinger, 2007), it might be more parsimonious to attribute the source-memory advantage for cheaters to negativity than to a domain-specific module. We (Bell & Buchner, 2010) tested this negativity hypothesis by examining whether the source-memory advantage generalizes to threatening information that is irrelevant for social exchange: disgusting information. A source-memory advantage for faces associated with disgusting behavior (see Table 2) was observed (Fig. 2). When we compared memory for cheating contexts and memory for disgusting contexts directly, we found that the two types of contexts were remembered equally well (Bell, Giang, & Buchner, in press). Consistent with the assumption that the source-memory advantage for faces of cheaters might be mediated by negativity, the memory advantage was abolished when the valence of the cheating behavior was only mildly negative (Bell & Buchner, 2011).

OPEN IN VIEWER

Table 2.

Example Behavior Descriptions From Bell and Buchner (2010)

Condition	Description
Disgust	J. F. is a shop assistant. He never has a handkerchief, so he always wipes his nose with his shirt sleeve, which is already stained and sticky.
Neutral	T. U. is a dyer. He owns a vacuum cleaner from a British manufacturer, which is equipped with an extremely effective, exchangeable carbon filter.
Pleasantness	O. H. is a crane operator. He has an inspiring enthusiasm and his charm and wit cheer up even the most miserable persons.

OPEN IN VIEWER

Fig. 2.

Source memory (i.e., memory for the type of context in which a face was encountered) as a function of context in the two experiments reported in Bell and Buchner (2010). Participants were presented with pictures of people’s faces, each of which was accompanied by a description that characterized the person’s behavior as disgusting, neutral (control condition), or pleasant. The source-memory parameter was obtained using Bayen, Murnane, and Erdfelder’s (1996) source-monitoring model and represents the conditional probability of remembering the source of an item correctly recognized as old. Error bars represent 95% confidence intervals.

The most general hypothesis that may be proposed to explain these results is that source memory is generally enhanced for negative information. Unfortunately, the case is not so simple. General theories based on valence or arousal cannot explain why the source-memory advantage does not generalize to victims of cheating (Bell & Buchner, 2011). When stimulus persons were described either as perpetrators or as victims of cheating, there was a clear source-memory advantage for faces of perpetrators, but faces of victims were poorly remembered even though the descriptions of perpetrators and victims were equally negatively valenced and arousing. Furthermore, faces associated with aggressive behaviors toward others were better remembered than faces associated with self-directed aggressive behaviors, even though the descriptions of both types of behaviors were equally negative and arousing (Bell, Giang, & Buchner, in press).

These findings suggest that negativity or arousal per se do not automatically enhance source memory. Rather, the information has to be threatening (i.e., associated with negative consequences for other people) to be especially well remembered. This conclusion is also supported by an earlier finding of Bliss-Moreau, Barrett, and Wright (2008) that faces are more efficiently associated with negative emotion if they are paired with a description implying that the person is causing harm to others than if they are paired with an equally negative but nonthreatening (nonsocial) description.

How Adaptive Is Adaptive Memory?

Although the findings reviewed above show that the source-memory advantage is more general than predicted by evolutionary theories, it is still consistent with the assumption that the human mind is biased toward the processing of threatening social information, the remembering of which might help people to avoid harmful encounters with others. However, recent studies have demonstrated that people do not always show a processing advantage for negative over positive information. Rather, they seem to focus on information that violates their positive or negative expectations, which is consistent with findings showing enhanced source memory for information that is emotionally incongruent with participants’ expectancies (Cook, Marsh, & Hicks, 2003).

Given that expectation congruity and incongruity are key concepts involved in explaining many memory phenomena (Ranganath & Rainer, 2003; van Kesteren et al., 2012), the incongruity hypothesis represents an account of another domain-general mechanism that might contribute to source memory for faces of cheaters. When participants played cooperation games in which most interaction partners were cooperative, cheating was better remembered than cooperation, a result consistent with predictions based on the evolutionary account. However, when most interaction partners were cheaters, this pattern was reversed (Barclay, 2008; Bell, Buchner, & Musch, 2010; Volstorf, Rieskamp, & Stevens, 2011). Furthermore, the source-memory advantage for cheaters was particularly pronounced for smiling and trustworthy-looking faces (Bell, Buchner, Kroneisen, & Giang, in press; Suzuki & Suga, 2010; Fig. 3).

OPEN IN VIEWER

Fig. 3.

Source memory for faces (i.e., memory for the type of context in which a face was encountered) as a function of the character (cheater or cooperator) and facial expression (smiling or angry) of stimulus persons in Experiment 2 of Bell, Buchner, Kroneisen, and Giang (in press). The source-memory parameter was obtained using Bayen, Murnane, and Erdfelder’s (1996) source-monitoring model and represents the conditional probability of remembering the source of an item correctly recognized as old. Error bars represent 95% confidence intervals.

Ironically, it can be argued that this more general and more flexible mechanism is more beneficial for cooperative individuals than a mechanism that focuses only on cheating would be. First, the expectation-incongruity mechanism is efficient, because it uses cognitive resources to encode information about the character of a person only when the appropriate approach or avoidance response cannot be generated by less resource-intensive mechanisms (e.g., guessing). Furthermore, by focusing only on cheaters, people may risk becoming “locked up” in a negative motivational-emotional state (cf. Rothermund, Voss, & Wentura, 2008), which may lead to isolation in social situations in which the degree of cooperation is low. A more flexible mechanism would allow an individual to adapt to the changing demands of different situations, and to escape from aversive social situations by focusing on trustworthy individuals with whom he or she might form beneficial alliances against the hostile social environment.

However, although memory was enhanced for both positive and negative expectancy violations in most experiments (Barclay, 2008; Bell et al., 2010; Volstorf et al., 2011), results from several experiments were consistent with the assumption that the weighting of negative and positive expectancy violations is asymmetrical (Bell, Buchner, et al., in press; Suzuki & Suga, 2010). People seem to remember negative events that violate positive expectations better than they remember positive events that violate negative expectations. Thus, in order to explain the whole pattern of results, it may be necessary to postulate both an expectancy-violation mechanism and an asymmetrical weighting of unexpected negative and positive information.

Conclusion

Across a number of experiments, we found that source memory for faces of cheaters was enhanced in comparison to source memory for other types of faces, results consistent with predictions derived from evolutionary models of reciprocal cooperation. This finding has been interpreted as evidence for specialized memory modules that have evolved to facilitate reciprocal altruism within human groups and societies. However, this hypothesis should not be accepted unless it can be shown that the module is activated only in situations involving social exchange.

The research program described here has shown that a small set of long-known, general memory mechanisms is well suited to explain the memory advantage for cheaters obtained by us (Buchner et al., 2009). First, human memory may be conceived of as adaptive in the general sense that people preferentially attend to, and remember, threatening information. Second, memory for information that is incongruent with currently activated approach or avoidance tendencies seems to be generally enhanced. Thus, two general mechanisms may act in concert to lead to improved memory for relevant information not only when cheating may occur, but in a wide variety of situations. These more general mechanisms may be more adaptive than a fixed focus on cheating would be, because they allow for the flexible adjustment of cognitive processing and behavior in different environments. Furthermore, these mechanisms might be useful not only for avoiding cheaters in social exchange, but also for avoiding a range of other harmful stimuli. In conclusion, the results suggest that cheating is indeed particularly well remembered in most situations because it stimulates strong negative emotions and violates social expectancies; however, it is not necessary to postulate specific cognitive modules to explain this memory advantage.

Recommended Reading

Cosmides, L., & Tooby, J. (2005). (See References). A review of the evidence in favor of a cheater-detection module.

Erdfelder, E., Auer, T.-S., Hilbig, B. E., Aßfalg, A., Moshagen, M., & Nadarevic, L. (2009). Multinomial processing tree models. A review of the literature. Zeitschrift für Psychologie/Journal of Psychology, 217, 108-124. A comprehensive review of multinomial modeling.

Nairne, J. S., & Pandeirada, J. N. S. (2010). Adaptive memory: Nature’s criterion and the functionalist agenda. The American Journal of Psychology, 123, 381-390. A plea for a functional perspective on memory.