Abstract
Previous research has demonstrated that people remember negative reputational information particularly well. However, most of these experiments manipulated the type of information associated with each face, rather than manipulating the circumstances under which people learn this information. The present experiment examines the effect of the social situation on memory for social-exchange relevant information. Faces were paired with descriptions of cheating, trustworthy, or neutral behavior. In addition, the importance of the social situation was manipulated: Participants had either to decide if they would want to work with the described person on a student project (socially relevant scenario) or if they would want to ask this person what time it is while waiting at an airport (socially irrelevant scenario). A multinomial processing tree model was used to measure old–new item discrimination and source memory. Only in the socially relevant scenario a source memory advantage for cheaters was found.
Navigating through our social world is a complex task. We continuously interact with thousands of people. During these interactions with others, we try to remember information about the people we meet (Hargis & Castel, 2017). In particular, we have to remember whom we can trust, and whom to avoid (Bell, Schain, & Echterhoff, 2014). Indeed, a number of experiments found increased memory for persons with negative reputations (e.g., Buchner, Bell, Mehl, & Musch, 2009; Suzuki & Suga, 2010). From an evolutionary point of view, this memory advantage for norm violators can explain why cooperation is a stable behavior among humans. At first glance, it seems obvious that individuals benefit from mutual cooperation. However, cooperation is also risky because it usually implies fitness costs for the helping individual (Trivers, 1971). Therefore, cooperation can only be a successful strategy when there is the opportunity for reciprocity. According to this concept, cooperation is beneficial for the helping individual when it is reciprocal, that is, when the favor of the helping individual is returned later on (Axelrod & Hamilton, 1981; Tooby & Cosmides, 1992; Trivers, 1971). However, this delayed reciprocity makes huge demands on memory. It requires that an individual is able to recognize others and recall their behavior of previous encounters. Furthermore, to retaliate for unfairly behavior we do not only need a better recognition of the face of defectors, we also have to remember if this person was associated with cheating behavior. Indeed, several studies demonstrated that source memory for faces of cheaters is particularly good (e.g., Buchner et al., 2009; Bell, Buchner, & Musch, 2010).
Tooby and Cosmides (1992, 2005) postulated that humans have evolved specialized brain mechanisms that help them avoid exploitation by cheaters (social contract theory.) Enhanced memory for faces of cheaters has been attributed to a highly specialized cheater detection module (Mealey, Daood, & Krage, 1996). Other theories claim that more general processes are behind these results. Good source memory for cheaters reflects mechanisms that are sensitive to novelty and violations of expectations (Chang & Sanfey, 2009). Our cognitive system may react to this unexpected information with intensified processing (Bell, Mieth, & Buchner, 2015). A cognitive system that relies on schematic processing in normal circumstances but switches to resource-intensive processing when important or unexpected (=schema incongruent) information is encountered can be described as an efficient system (Bell et al., 2015; Sherman, Lee, Bessenoff, & Frost, 1998). Bell, Buchner, Kroneisen, and Giang (2012) concluded that the effects can be explained by a mechanism that is sensitive to emotional expectancy violations. Events that are incongruent with our current positive or negative expectancies characterize situations in which the current approach or avoidance tendencies maybe inappropriate. Good source memory for the emotional incongruent behavior of the other person can help to adjust behavioral tendencies and therefore avoid risky social interactions. In line with this, recent findings showed that participants preferentially remember reputational information that is emotionally incongruent to their expectations (e.g., Bell et al., 2012; Kroneisen & Bell, 2013; Kroneisen, Woehe, & Rausch, 2015).
Furthermore, it seems that motivational factors are also important. Bell et al. (2014) showed that personal costs and benefits influence memory for cheaters, but not memory for trustworthy persons. Their participants remembered faces showing norm violating behavior well when the described behavior of that person involved negative consequences for the participant (e.g. “He lied to you to sell you the useless products of another person.”) but poorly when the described behavior involved personal benefits for the participant (e.g. “He lied to another person to sell that person your useless products.”). Apparently, our categorization of others is based on our judgments about the morality of their behavior. However, it also reflects self-serving biases. Moreover, Kroneisen (2018) demonstrated that personally relevant cheaters are remembered better than personally irrelevant cheaters. Similar to other studies, she compared source memory for faces combined with cheating, trustworthy or neutral behavior descriptions. However, these descriptions were either familiar and had a high likelihood of occurrence (for a student population, e.g. “P. L. sneaks into student parties and pours knockout drops into the drinks of other party guests”) or they were unfamiliar and of low relevance (for a student population, e.g. “P.O. fulfills his military service. Because he has free access to the armory, he steals ammunition and sells it on the black market.”). Results showed that source memory for cheaters was high when likelihood of occurrence and familiarity was high. This indicates that the frequently observed source memory advantage for cheating or immoral behavior is not a general advantage for all behaviors implying some kind of norm violation. It seems to be that the relative importance of defecting influences our memory.
Up until now, most of these past experiments concerning memory for cheaters manipulated the type of information associated with each face, rather than manipulating the circumstances under which people learn this information. Cosmides (1989) already discussed the idea that only situations that contain a social contract will activate a cheater detection algorithm. In line with this, Wagner-Egger (2001) claimed that correct cheater detection should correlate with both, the social costs of cheating and the benefits for the non-cheating counterpart: The higher the costs and the benefits are, the better the cheater detection. However, these discussions were only concerned with the Wason Selection Task (Wason, 1968). Nevertheless, it indicates that it should matter if a situation contains a personally relevant social contract or not.
To summarize, it seems that the relative importance of remembering defectors depends on the circumstances. Based on these results, the question arises if the importance or the significance of the social exchange situation also influences the source memory advantage for cheaters.
The present experiment was designed to test this question. The design of the memory experiment was similar to the design used in previous studies (e.g., Kroneisen et al., 2015; Kroneisen, 2018) with the only exception that the social relevance of the social exchange situation was manipulated. In the socially relevant situation, prior to the learning phase, subjects were asked to imagine that they had to participate in a research project at the university. To complete this project successfully, they had to work together with other students. In the socially irrelevant situation, prior to the learning phase, participants were asked to imagine that they are at the airport because they wanted to visit a relative in Ireland. Currently, they are not sure about the exact time and decided to ask someone about it.1 The basic hypothesis was that source memory should be enhanced for cheating behavior in a social relevant exchange situation in comparison to a socially irrelevant situation.
Experiment
Method
Design
A 3 × 2 design was used with behavioral history (cheating, trustworthy, neutral) as within-subject factor and social relevance (university vs. airport) as between-subjects factor. Old-new recognition, and source memory were the dependent variables.
Given a sample size of N = 63 in the socially relevant condition and N = 64 in the socially irrelevant condition, 72 responses in the source memory test, and α = 0.05, the power to detect a difference between the source memory parameters with an effect size of w = 0.06 (which is in the order of magnitude of the source memory effect observed by Buchner et al., 2009) was = 0.98. The power calculations were conducted using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007).
Participants
Participants were one-hundred and twenty-seven students (108 female) of the University of Koblenz-Landau. All of them received a participation credit. They were randomly assigned to the socially relevant group (63 participants, age range: 18–29 years, M = 22.08; SD = 2.46, 53 female) and the socially irrelevant group (64 participants, age range: 18–37 years, M = 22.02, SD = 2.76, 55 female).
Apparatus and materials
The stimuli consisted of color portrait photographs (640 × 480 pixels) of 72 white male adults from the CVL faces database (Peer, Emersic, Bule, Žganec Gros, & Štruc, 2014).2 Only forward-facing pictures with a neutral facial expression were selected. The faces were randomly assigned to two sets of 36 photographs (set 1 or 2, counterbalanced across participants).
Short descriptions conveyed the behavioral history (cheating, irrelevant to the cheating-trustworthiness dimension [=neutral], trustworthy) of the person shown. Half of the descriptions were taken from Kroneisen (2018, personally relevant descriptions), the other half was newly generated. All descriptions were then rated in a norming study. In this pre-study five students rated 36 behavior descriptions on a scale ranging from −3 [cheating] to +3 [trustworthy]. For instance, “P.L. sneaks into student parties and pours knockout drops into the drinks of other party guests.” would convey a history of cheating. K.R. stays awake after a long student dorm party and helps to clean up and take out the garbage.” would convey trustworthy behavior. “K.I. has a large backpack with several compartments to transport documents safely protected from the rain.” would convey irrelevant behavior that is emotionally neutral. Twelve descriptions were shown in each condition (MCheatingDescriptions = −2.26, SD = 0.22, MTrustworthyDescriptions = 2.63, SD = 0.42, MNeutralDescriptions = 0.28, SD = 0.10) during the encoding phase (for a list of all behavioral descriptions see the Appendix).
Procedure
Overall, the socially relevant group and the socially irrelevant group performed the same experiment. It only differed in the scenario given prior to the learning phase (see Fig. 1 for an overview of the procedure). The socially relevant group was asked to imagine that they had to accomplish a research project at the university. This research project has to be realized in a team of students. The importance of this project for the student was emphasized. In addition, the subjects were told, that their first task would be to select the rest of their team members. After this instruction the subjects should think about this situation for 20 s. The socially irrelevant group was asked to imagine that they are at the airport because they wanted to visit a relative in Ireland. In addition, they were told that they are unsure about the current time and want to ask someone about it (for the exact instructions see the appendix). After this instruction the subjects should think about this situation for 20 s.
Schematic procedure of the experiment
Citation: Culture and Evolution 20, 1; 10.1556/2055.2023.00039
The next steps were very similar to the procedure used by Kroneisen (2018), 12 cheaters, 12 trustworthy, and 12 neutral persons were presented in random order during the encoding phase. Participants were asked to rate if they would choose the person. Each trial started with a headline I would choose this person and a photograph (set 1 or set 2, counterbalanced across participants) shown for 2 s. Next, the photograph disappeared and the behavior description was shown for 5 s. The descriptions of cheating, trustworthy and irrelevant behaviors were randomly assigned to the photographs. Subsequently, the rating scale ranging from 1 (no agreement) to 6 (strong agreement) was shown. Participants made their choice by using the computer mouse and then initiated the next trial. To absorb primacy and recency effects, two buffer photographs and descriptions were included at the beginning and at the end of the encoding phase. After a short break of 20 s, the test phase followed. Here, participants received the instructions for the surprise source memory test. Seventy-two male faces were presented. Half of the faces were old (set 1 or 2, depending on the encoding-phase assignment), and the other half were new. Each trial started with a headline I would choose this person and a photograph. After two seconds a rating scale ranging from 1 (no agreement) to 6 (strong agreement) was shown and participants were asked to rate if they would choose this person. After the rating, participants were asked whether or not they had seen the face during the encoding phase (old-new recognition). If participants indicated that they had seen the face before, they were required to decide whether the face belonged to a cheater, to a trustworthy or to a neutral person (source memory). After selecting one of these answers, the next face was shown. The experiment lasted approximately 35 min.
Measuring source memory
A good measure of source memory should be sensitive to differences in source memory without being sensitive to differences in the ability to remember the items themselves (Murnane & Bayen, 1996). Multinomial processing tree models are stochastic models that allow us to disentangle different cognitive processes underlying task performance. A well-established multinomial source-monitoring model was developed by Bayen, Murnane, and Erdfelder (1996). It can distinguish between old–new recognition, source memory, and various types of guessing biases. An advantage of this model is that its source memory parameters remain unaffected by differences in old-new recognition and by guessing manipulations (Bayen et al., 1996; Bayen & Kuhlmann, 2011; Bayen, Nakamura, Dupuis, & Yang, 2000). Schütz and Bröder (2011) demonstrated that this model is at least as good as an alternative model based on signal detection theory (see also Klauer & Kellen, 2010). An adaptation of the model for the present Experiment is illustrated in Fig. 2 (for a very similar approach see Kroneisen, 2018). The most relevant parameters in the present context are parameter D, representing the probability of recognizing a face as old, and parameter d, representing the conditional probability of remembering the behavior the person showed. All parameter estimations were performed with the program MultiTree (Moshagen, 2010). As a goodness-of-fit measure, the log-likelihood ratio statistic, G2, which is asymptotically chi-square distributed (Hu & Batchelder, 1994) was used.
Bayen et al.'s (1996) source memory model as adapted for the present experiment using the example of socially relevant situation (university). The rectangles on the left side represent the picture shown (male faces that were associated with cheating (−), neutral (n), or trustworthy (+) behavior descriptions or that were new). Letters along the links represent the probabilities with which certain cognitive states occur. Rectangles on the right side represent the answers of the participants (cheating (−), neutral (n), trustworthy (+), or new). The first tree in this figure represents test trials with faces associated with a cheating context high in relevance during the encoding phase. Parameter Dsr- represents the probability of recognizing a face associated to a negative context (cheating behavior) as old. Parameter dsr- represents the conditional probability of also remembering correctly that the face was encountered in a negative context (cheating behavior). If the negative source of the face is not remembered (with probability 1 – dsr-), it is still possible to guess the correct source with probability (a± · a–). Alternatively, it may be incorrectly guessed that the face was associated to a positive context (trustworthy behavior) with probability (a± · (1 – a–)) or to a neutral context with probability (1 – a±). If a cheating face is not recognized as old (with probability of 1 – D– r), it may still be guessed (with probability b) that the face is old. For these faces, the correct behavior history may be guessed with probability (g± · g–). Similarly, it is possible to incorrectly guess that the face was associated to a trustworthy context with probability (g± · (1 – g–)), or to a neutral context with probability (1 – g±). Faces neither recognized as old nor guessed to be old (with probability 1 – b) are incorrectly classified as new. Analogous statements apply to faces associated to neutral behavior (tree 2) or trustworthy behavior (tree 3) and to new faces (tree 4). In addition, the same set of four trees has to be applied for the socially irrelevant situation (airport)
Citation: Culture and Evolution 20, 1; 10.1556/2055.2023.00039
Results3
Encoding-phase ratings
A 3 (behavioral history) ×2 (social situation) repeated measurement ANOVA showed that the ratings differed as a function of the behavioral history variable, F(2, 250) = 591.04, p = 0.02, η2 = 0.83. Contrasts4 indicated that in both situations (socially relevant and socially irrelevant) participants were lower in their agreement to choose cheating persons in comparison to trustworthy persons (socially relevant: t(250) = −26.37, p < 0.001, ηp2 = 0.74; socially irrelevant: t(250) = −19.39, p < 0.001, ηp2 = 0.60), or neutral persons (socially relevant: t(250) = −20.97, p < 0.001, ηp2 = 0.64; socially irrelevant: t(250) = −16.08, p < 0.001, ηp2 = 0.51). There was also an effect of social situation, F(1,125) = 5.50, p = 0.02, ηp2 = 0.04, indicating higher ratings in the socially irrelevant situation. The interaction between behavioral history and context was also significant, F(2,250) = 13.51, p < 0.001, ηp2 = 0.10. Contrasts indicated that the socially relevant and irrelevant situation differed in their cheater ratings, t(285.045) = −4.97, p < 0.001, ηp2 = 0.08. Participants in the socially relevant situation were lower in their agreement to choose cheaters in their team.
Old-new recognition
Old-new recognition is reported in terms of Pr (a sensitivity measure of the two-high threshold model), calculated by subtracting the false alarm rate from the hit rate. Pr as a sensitivity measure was evaluated in validation studies (Snodgrass & Corwin, 1988) and avoids the problem of undefined values arising from the use of d’.
There was a significant effect of social relevance, F(1, 125) = 5.61, p = 0.03, ηp2 = 0.04. Participants in the socially irrelevant condition showed better old-new discrimination in comparison to the socially relevant condition (Msr- = 0.43, SEM = 0.02; Msr+ = 0.42, SEM = 0.02; Msrn = 0.45, SEM = 0.02; Msi- = 0.48, SEM = 0.02; Msi+ = 0.53, SEM = 0.02; Msin = 0.49, SEM = 0.02). Behavioral history did not affect old–new discrimination, F(2, 250) = 0.44, p = 0.64, ηp2 = 0.004. The interaction between both variables was not significant, F(2, 250) = 1.44, p = 0.24, ηp2 = 0.01.
Model-based analysis: old new recognition and source memory (Parameter D and d)
To disentangle the processes involved in old-new recognition, source memory and guessing the multinomial model illustrated in Fig. 2 was used. To analyze if old-new recognition and source memory for cheating or trustworthy persons is different depending on social relevance two sets of the model trees are needed. One set of four trees for the four different types of persons in the socially relevant situation (cheating, trustworthy, neutral, new) and one set of four trees for the four different types of persons in the socially irrelevant situation (cheating, trustworthy, neutral, new). Therefore, we have two sets of parameters in these trees.
As in previous studies, the analysis was started with a base model that builds on the fact that old-new discrimination did not differ as a function of behavioral history. However, the Pr data indicated that there is a significant difference between the university and airport situation. Therefore, in the base model, separately for both conditions, all parameters representing old-new discrimination for the different types of faces were set to be equal. The parameter of representing the probability of detecting new faces as new was set to be equal to the old-new discrimination parameters (a standard assumption of the two-high threshold model of signal detection). Thus, the base model is characterized by the restrictions that Dsociallyrelevant_cheater = Dsociallyrelevant_trustworthy = Dsociallyrelevant_neutral = Dsociallyrelevant_new and Dsociallyirrelevant_cheater = Dsociallyirrelevant_trustworthy = Dsociallyirrelevant_neutral = Dsociallyirrelevant_new. The base model fitted the data (line1, Table 1).
Goodness-of-fit values for the model-based results of source memory (see text for details)
| Model test (Parameter restrictions) | Results |
| Base Model | |
| 1 Dsr– = Dsr+ = Dsrn = Dsrnew; Dsi– = Dsi+ = Dsin = Dsinew | G2(11) = 7.77, p = 0.73 |
| Additional restrictions on the base model | |
| 2 Dsr– = Dsr+ = Dsrn = Dsrnew = Dsi– = Dsi+ = Dsin = Dsinew | ΔG2(1) = 13.96, p < 0.001 |
| 3 dsr– = dsi– | ΔG2(1) = 3.976, p = 0.046 |
| 4 dsr+ = dsi+ | ΔG2(1) = 2.74, p = 0.10 |
| 5 dsr– = dsr+ | ΔG2(1) = 12.49, p < 0.001 |
| 6 dsi- = dsi+ | ΔG2(1) = 0.21, p = 0.65 |
| 7 dsrn = dsin | ΔG2(1) = 0.27, p = 0.60 |
| 8 dsr– = dsrn = dsin | ΔG2(2) = 8.57, p = 0.01 |
| 9 dsr+ = dsrn = dsin | ΔG2(2) = 2.04, p = 0.36 |
| 10 dsi– = dsrn = dsin | ΔG2(2) = 1.19, p = 0.55 |
| 11 dsi+ = dsrn = dsin | ΔG2(2) = 0.43, p = 0.81 |
“sr” and “si” in the indices refer to the factor situation (sr = socially relevant, si = socially irrelevant); “n” refers to the neutral descriptions.
The goodness-of-fit of the base models and all additional restrictions were tested using the goodness-of-fit statistic G2, which is asymptotically χ2 distributed. p values smaller than 0.05 indicate that the additional restrictions are not compatible with the data, as a result of which the hypothesis implied by these restriction must be rejected.
First, it was tested whether the old-new recognition parameter D was indeed higher for the socially irrelevant than for the socially relevant situation as the Pr data indicated. The null hypothesis of no difference can be implemented by imposing the restrictions Dsociallyrelevant = Dsociallyrelevant_trustworthy = Dsociallyrelevant_neutral = Dsociallyrelevant_new = Dsociallyirrelevant_cheater- = Dsociallyirrelevant_trustworthy = Dsociallyirrelevant_neutral = Dsociallyirrelevant_new on the base model. An incompatibility of this assumption with the data should lead to a significant increase in model misfit caused by this restriction as expressed in the goodness-of-fit statistic ΔG2. As can be seen in Table 1 (line2), faces encoded in the socially irrelevant situation were better recognized than faces encoded in the socially relevant situation.
Next, it was examined if the source memory parameter d differed for faces associated with cheating and trustworthy behavior between the socially relevant situation and the socially irrelevant situation (Fig. 3). The null hypothesis of no differences can be implemented by imposing the restrictions dsociallyrelevant_cheater = dsociallyirrelevant_cheater and dsociallyrelevant_trustworthy = dsociallyirrelevant_trustworthy respectively on the base model. An incompatibility of these assumptions with the data should lead to a significant increase in model misfit caused by these restrictions as expressed in the goodness-of-fit statistic ΔG2. As can be seen in line 3 in Table 1 there was a difference in source memory for cheating faces depending on situation: Faces associated with cheating behavior in a socially relevant situation were remembered better than faces associated with cheating behavior in a socially irrelevant situation. However, source memory for trustworthy faces did not differ depending on situation (line 4, Table 1). For the socially relevant condition, a source memory advantage for cheaters in comparison to trustworthy persons could be found (line 5, Table 1). For the socially irrelevant condition, there was no difference in source memory between trustworthy and cheating persons (line 6, Table 1). Source memory for neutral persons did not differ between the situations (line 7, Table 1). In comparison to neutral behavior descriptions, a source memory advantage for cheaters in the socially relevant situation could be detected. However, independently of situation, no differences between neutral and trustworthy persons were found (line 8–11, Table 1).
The results of parameter estimates for the old new recognition parameter D and the source memory parameter d as a function of behavioral history and social relevance. The error bars represent standard errors of the parameters
Citation: Culture and Evolution 20, 1; 10.1556/2055.2023.00039
Test-phase ratings
The choice rating during the test-phase can be seen as an indirect answer to the question if individuals rely on their source memory to select people. Therefore, an ANOVA was conducted with test rating as dependent variable and memory as within-subjects predictor. Memory was split into three categories: no memory (= neither old-new recognition nor source memory), old-new recognition (= correctly remembering a face as old or new), and source memory (=correctly remembering the behavior a face has shown during the encoding phase). A 2 (social relevance: socially relevant vs. socially irrelevant) x 3 (Memory: no memory, old-new recognition, and source memory) x 3 (Behavioral history: cheating vs. trustworthy vs. neutral) ANOVA with test ratings as DV was conducted. For this analysis, participants were only included if they had at least one response in each category of the ANOVA (see Murty, FeldmanHall, Hunter, PhelpsPhelps, & Davachi, 2016, for a similar procedure). Fifty-two participants in the socially irrelevant condition and fifty-three participants in the socially relevant condition remained in the sample.
There was a significant effect of social relevance, F(1, 103) = 4.75, p = 0.03, ηp2 = 0.04. Overall, participants in the socially relevant condition gave lower choice ratings in comparison to the socially irrelevant condition (see Fig. 4). There was no main effect of memory, F(2, 206) = 1.78, p = 0.17, ηp2 = 0.02. However, the ratings differed as a function of the behavioral history variable, F(2, 206) = 46.22, p < 0.001, ηp2 = 0.31. Overall, cheating persons got lower choice ratings in comparison to neutral (t(616.426) = 13.73, p < 0.001) and trustworthy persons (t(616.42) = 11.47, p < 0.001). The interaction between memory and behavioral history was significant, F(4, 412) = 33.61, p < 0.001, ηp2 = 0.25. For the source memory category, contrasts showed that for the socially relevant condition the choice ratings differed between neutral and cheating persons (t(616.42) = 11.81, p < 0.001) and between neutral and trustworthy persons (t(616.42) = 2.52, p = 0.01). Cheaters got lower and trustworthy persons got higher choice ratings. For the socially irrelevant condition choice ratings differed between neutral and cheating persons (t(616.42) = 7.63, p < 0.001), cheaters got lower choice ratings. However, no difference between neutral and trustworthy persons could be detected (t(616.42) = 0.69, p = 0.49).
Mean choice ratings in the test phase as a function of behavioral history, social relevance and memory. The error bars represent standard error of the means
Citation: Culture and Evolution 20, 1; 10.1556/2055.2023.00039
Discussion
The main question of this experiment was whether the social character of an exchange situation also influences how we remember the behavior and actions of others. Social character was manipulated by the descriptions of two situations that differed in their social relevance for the participant. The socially relevant situation involved a clear social contract (working together on a project at the university), the socially irrelevant situation did not involve an important social contract (ask for the time at an airport). In line with the predictions, in the socially relevant situation, source memory was highest for cheaters compared to trustworthy and neutral persons. In the socially irrelevant situation, no differences in source memory depending on behavioral history could be found. Furthermore, participants had better source memory for cheaters in the socially relevant situation (university) in comparison to cheaters in the socially irrelevant situation (airport). However, regardless of the situation, no differences between neutral and trustworthy persons were detected.
Consistent with previous findings (e.g., Bell et al., 2012; Buchner et al., 2009), old-new recognition was not affected by whether faces were associated with negative, positive, or emotionally neutral behavioral descriptions. Interestingly, old-new recognition between the socially relevant and the socially irrelevant situation differed. Overall, the airport condition showed better item memory than the university condition. Here, participants remembered more faces correctly as old. As discussed above, to retaliate for unfairly behavior we need to remember if someone was associated with cheating behavior. Only recognizing the face as old is not enough to avoid or punish cheaters. Moreover, it could be even worse: remembering a face as familiar might increase the risk of being exploited by cheaters since we prefer items high in familiarity (Bornstein, 1989; Buchner et al., 2009; Zajonc, 1968). In line with this idea, many experiments showed no difference in old new recognition between faces associated with unfair, trustworthy or neutral behavior (e.g., Bell et al., 2010, 2012; Kroneisen et al., 2015). The present experiment shows that the social relevance of the situation may influence this effect. On an airport, the probability to meet a person again is very low. Therefore, it is not necessary to remember an association between face and behavior. In line with that, overall source memory for this condition was very low. It is possible that the participants concentrated on the faces alone instead of the combination face and behavior. However, the choice ratings during the test phase indicated that participants had a kind of implicit memory for cheaters. Yet, only participants with at least one response in each category (no memory, old-new recognition, and source memory) were included. Therefore, these results have to be interpreted with caution.
In contrast, working together with other students on a project for the university can be seen as an important social contract situation. Previous behavior of others can influence the outcome of the project. Therefore, it is important to remember persons who showed unfairly behavior before. Furthermore, it is not only important to remember the face but to remember the behavior. Interestingly, the choice ratings during the test phase indicated a kind of implicit memory for cheating and for trustworthy persons. However, as mentioned above, these results should be interpreted with caution.
Kroneisen (2018) showed that the frequently observed source memory advantage for cheating behavior is not generally found. Instead, social contract violations with high probability of occurrence and familiarity (for a student population) were remembered better. This indicates that motivational factors influence source memory. In line with this, Bell et al. (2014) found better memory for immoral persons when they were associated with personal costs but not when they were associated with personal benefits. The personal relevance of the cheating behavior seems to have an impact on source memory. The present experiment extents these results and shows that situations in which there is no clear social contract, such as asking for the time at an airport, have no effect on source memory. These findings do not contradict results of previous studies that used only likeability or attractiveness ratings during the learning phase. The finding that the source memory advantage stays stable even when working on a secondary continuous reaction time task indicates that this mechanism can operate automatically (Mieth, Bell, & Buchner, 2016). However, if it is clearly stated that we will not meet with this person in a social contract situation in the future, the influence of cheating behavior on source memory disappears. It appears that there are specific cues necessary to activate the cognitive mechanisms underlying social cooperation. Interestingly, the influence of context on memory is also a question in another line of research on adaptive memory. In the survival processing paradigm (see, for example, Nairne, Thompson, & Pandeirada, 2007) normally, two conditions are compared: Participants have to imagine being in a situation either relevant or irrelevant to their own survival. After this, a number of words have to be rated according to their relevance for the survival or the control situation. A later unexpected free recall test shows a stable memory advantage for the survival condition only. Based on these results, it is often stated that the fitness relevance of information is determined by context rather than content. Certainly, these two lines of research are not completely comparable. However, there are also some indications that the activation of adaptive mechanisms depend on circumstances.
The experiment reported here also has some limitations. The choice rating in the test phase was done to make the design as similar as possible to the original design of Buchner et al. (2009). Furthermore, it is a measure of implicit memory. However, it is possible that the repeated question may have confused some participants. Nevertheless, it is important to have some kind of implicit memory test. There is some evidence that people use their source memory to plan their later actions (see Murty et al., 2016; Schaper, Mieth, & Bell, 2019). Yet, we are also influenced by other variables, like for example facial trustworthiness. If this variable is added to the experiment, participants tend to rely more on facial trustworthiness than their source memory (Kroneisen, Bott, & Mayer, 2021). If a large discrepancy between the ratings and the source memory test had been found, this would imply that participants may have used variables other than memory for their choice rating (e.g., facial trustworthiness). However, the pictures were counterbalanced throughout the experiment and therefore, it was not likely that facial trustworthiness did influence the results.
Furthermore, again, only male images were shown in this experiment. It is known that female faces are more trustworthy than male faces (Kroneisen & Bell, 2013). However, the inclusion of female images in this case would have added even more variance to the experiment. Therefore, the decision was made to not change this variable in this experiment either and only use male images. Future research should include this variable.
Another critical point concerns the question of generalization. In this experiment, only two scenarios were used. This was done to get as many subjects as possible per scenario to increase the power of the experiment. Also, including many scenarios would also mean to increase the complexity of the experiment. However, this does not easily generalize the results. Future research should use different socially relevant and irrelevant scenarios and examine this more closely.
It can be argued that the two scenarios also differed in terms of complexity. The socially relevant situation required more complex decisions. Also, having a cheater in your group would mean that you could not pass the course, so serious consequences were possible. However, maybe that is what defines a socially relevant situation. If the situation is not socially relevant, there is less information to consider. This raises the question of whether socially relevant situations differ from irrelevant situations on several levels (and this defines those situations).
In summary, this research shows the importance of the social context for memory. Both the content of the information to be memorized and the overall context within the situation are of great importance and should not be neglected in research. Of course, memory research should also be done independently of content in order to understand how it works, but one should not underestimate the importance of the overall context. Memory research must also take into account complex social structures. We are just beginning to understand how exactly these connections work. Therefore, research that clarifies the relations between memory and social contexts is absolutely necessary. Only then is it possible to really understand our memory.
Acknowledgement
I thank Janika Rolf for help with the data acquisition.
References
Axelrod, R., & Hamilton, W. (1981). The evolution of cooperation. Science, 211, 1390–1396. https://doi.org/10.1126/science.7466396.
Bayen, U. J., & Kuhlmann, B. G. (2011). Influences of source-item contingency and schematic knowledge on source monitoring: Tests of the probability-matching account. Journal of Memory and Language, 64, 1–17. https://doi.org/10.1016/j.jml.2010.09.001.
Bayen, U. J., Murnane, K., & Erdfelder, E. (1996). Source discrimination, item detection, and multinomial models of source monitoring. Journal of Experimental Psychology: Learning, Memory & Cognition, 22, 197–215. https://doi.org/10.1037/0278-7393.22.1.197.
Bayen, U. J., Nakamura, G. V., Dupuis, S. E., & Yang, C. L. (2000). The use of schematic knowledge about sources in source monitoring. Memory & Cognition, 28, 480–500. https://doi.org/10.3758/BF03198562.
Bell, R., Buchner, A., Kroneisen, M., & Giang, T. (2012). On the flexibility of social source memory: A test of the emotional incongruity hypothesis. Journal of Experimental Psychology: Learning, Memory, & Cognition, 38, 1512–1529. https://doi.org/10.1037/a0028219.
Bell, R., Buchner, A., & Musch, J. (2010). Enhanced old-new recognition and source memory for faces of cooperators and defectors in a social-dilemma game. Cognition, 117, 261–275. https://doi.org/10.1016/j.cognition.2010.08.020.
Bell, R., Mieth, L., & Buchner, A. (2015). Appearance-based first impressions and person memory. Journal of Experimental Psychology: Learning, Memory, & Cognition, 41, 456–472. https://doi.org/10.1037/xlm0000034.
Bell, R., Schain, C., & Echterhoff, G. (2014). How selfish is memory for cheaters? Evidence for moral and egoistic biases. Cognition, 132, 437–442. https://doi.org/10.1016/j.cognition.2014.05.001.
Bornstein, R. F. (1989). Exposure and affect: Overview and meta-analysis of research, 1968–1987. Psychological Bulletin, 106, 265–289.
Buchner, A., Bell, R., Mehl, B., & Musch, J. (2009). No enhanced recognition memory, but better source memory for faces of cheaters. Evolution and Human Behavior, 30, 212–224. https://doi.org/10.1016/j.evolhumbehav.2009.01.004.
Chang, L. J., & Sanfey, A. G. (2009). Unforgettable ultimatums? Expectation violations promote enhanced social memory following economic bargaining. Frontiers in Behavioral Neuroscience, 3, Article 36. https://doi.org/10.3389/neuro.08.036.2008.
Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition, 31, 187–276. https://doi.org/10.1016/0010-0277(89)90023-1.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175–191. https://doi.org/10.3758/BF03193146.
Hargis, M. B., & Castel (2017). Younger and older adults’ associative memory for social information: The role of information importance. Psychology & Aging, 32, 325–330. https://doi.org/10.1037/pag0000171.
Hu, X., & Batchelder, W. H. (1994). The statistical analysis of general processing tree models with the EM algorithm. Psychometrika, 59, 21–47. https://doi.org/10.1007/BF02294263.
Klauer, K. C., & Kellen, D. (2010). Toward a complete decision model of item and source recognition: A discrete-state approach. Psychonomic Bulletin & Review, 17, 465–478.
Kroneisen, M. (2018). Is he important to me? Source memory advantage for personally relevant cheaters. Psychonomic Bulletin & Review, 25, 1129–1137. https://doi.org/10.3758/s13423-017-1345-1.
Kroneisen, M., & Bell, R. (2013). Sex, cheating, and disgust: Enhanced source memory for trait information that violates gender stereotypes. Memory, 21, 167–181. https://doi.org/10.1080/09658211.2012.713971.
Kroneisen, M., Bott, F. M., & Mayer, M. (2021). Remembering the bad ones: Does the source memory advantage for cheaters influence our later actions positively? Quarterly Journal of Experimental Psychology, 74, 1669–1685. https://doi.org/10.1177/17470218211007822.
Kroneisen, M., Woehe, L., & Rausch, L. S. (2015). Expectancy effects in source memory: How moving to a bad neighborhood can change your memory. Psychonomic Bulletin & Review, 22, 179–189. https://doi.org/10.3758/s13423-014-0655-9.
Mealey, L., Daood, C., & Krage, M. (1996). Enhanced memory for faces of cheaters. Ethology and Sociobiology, 17, 119−128. https://doi.org/10.1016/0162-3095(95)00131-X.
Mieth, L., Bell, R., Buchner, A. (2016). Cognitive load does not affect the behavioral and cognitive foundations of social cooperation. Frontiers in Psychology, 7, 1312. https://doi.org/10.3389/fpsyg.2016.01312.
Moshagen, M. (2010). MultiTree: A computer program for the analysis of multinomial processing tree models. Behavior Research Methods, 42, 42–54. https://doi.org/10.3758/BRM.42.1.42.
Murnane, K., & Bayen, U. J. (1996). An evaluation of empirical measures of source identification. Memory & Cognition, 24, 417–428. https://doi.org/10.3758/BF03200931.
Murty, V. P., FeldmanHall, O., Hunter, L. E., Phelps, Phelps, E. A., & Davachi, L. (2016). Episodic memories predict adaptive value-based decision-making. Journal of Experimental Psychology: General, 145, 548–558. https://doi.org/10.1037/xge0000158.
Nairne, J. S., Thompson, S. R., & Pandeirada, J. N. S. (2007). Adaptive memory: Survival processing enhances retention. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33, 263–273. https://doi.org/10.1037/0278-7393.33.2.263.
Peer, P., Emersic, Ž., Bule, J., Žganec Gros, J., & Štruc, V. (2014). Strategies for exploiting independent cloud, implementations of biometric experts in multibiometric scenarios. Mathematical Problems in Engineering, 1–15. http://www.lrv.fri.uni-lj.si/facedb.html.
Schaper, M. L., Mieth, L., & Bell, R. (2019). Adaptive memory: Source memory is positively associated with adaptive social decision making. Cognition, 186, 7–14. https://doi.org/10.1016/j.cognition.2019.01.014.
Schütz, J., & Bröder, A. (2011). Signal detection and threshold models of source memory. Experimental Psychology, 58, 293–311. https://doi.org/10.1027/1618-3169/a000097.
Sherman, J. W., Lee, A. Y., Bessenoff, G. R., & Frost, L. A. (1998). Stereotype efficiency reconsidered: Encoding flexibility under cognitive load. Journal of Personality and Social Psychology, 75, 589–606. https://doi.org/10.1037/0022-3514.75.3.589.
Snodgrass, J. G., & Corwin, J. (1988). Pragmatics of measuring recognition memory: Applications to dementia and amnesia. Journal of Experimental Psychology: General, 117, 34–50. https://doi.org/10.1037/0096-3445.117.1.34.
Suzuki, A., & Suga, S. (2010). Enhanced memory for the wolf in sheep's clothing: Facial trustworthiness modulates face-trait associative memory. Cognition, 117, 224–229. https://doi.org/10.1016/j.cognition.2010.08.004.
Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. Barkow, L. Cosmides, & J. Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture (pp. 19–136). New York: Oxford University Press.
Tooby, J., & Cosmides, L. (2005). Conceptual foundations of evolutionary psychology. In D. M. Buss (Ed.), The handbook of evolutionary psychology (pp. 5−67). Hoboken, NJ: John Wiley & Sons.
Trivers, R. (1971). The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 35–57. https://doi.org/10.1086/406755.
Wagner-Egger, P. (2001). Costs and benefits in Wason's selection task: The social conditional model. Swiss Journal of Psychology, 60, 117–135. http://doi.org/10.1024//1421-0185.60.3.117.
Wason, P. C. (1968). Reasoning about a rule. The Quarterly Journal of Experimental Psychology, 20, 273–281. https://doi.org/10.1080/14640746808400161.
Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9, 1−27.
Appendix
Behavior Descriptions:
| Trustworthy | S.O. is a member of the university-hiking group. During long marches, he carries the water supplies for everyone. |
| W.W. lives in a deprived area. He gives away necessary school and writing tools to children at his own expense. | |
| W.S. plays guitar in a children home twice a month to entertain the children. | |
| D.P. helps in a homeless shelter alongside his 40-hour week at the university and his side job. | |
| Alongside his studies, B.F. sells homemade clothing on the market and offers unsold goods free of charge to homeless people. | |
| Y.L., a research assistant, donates a part of his pension to an orphanage in his home country because it does not receive financial support from the state. | |
| L. H. lends money to his friends or buys a round in the evening, when they are short of cash. | |
| G.Z. meets with a weaker classmate twice a week and teaches him the complicated lecture material without payment. | |
| D. F. shares his foot with his roommates when they did not have time to cook. | |
| M. F. often queues for the drinks at parties, in order to spare his friends the lining up. | |
| K.R. stays awake after a long student dorm party and helps to clean up and take out the garbage. | |
| S.H. gets up at 3 am to pick up his roommate from the train station and helps him carry the heavy luggage. | |
| Neutral | K.I. has a large backpack with several compartments to transport documents safely and protected from the rain. |
| H.V. buys a huge house because he wants to have a big family later. | |
| R.M. has a radio alarm clock and sets the alarm to the news every morning at seven o'clock. | |
| J.Z. wants to buy a new cell phone. For this, he sells his old books. | |
| U.P. enjoys sport and relaxes with a long run through the forest regularly. | |
| I.R. was born in Italy, likes Italian food very much and travels to Tuscany every summer. | |
| P.M. loves classical music. Every Saturday he watches a symphony concert on TV. | |
| H.T. is a person who loves writing short stories. One time, he has won a regional writing competition. | |
| E.J. studies environmental sciences. Outside of the university, he enjoys working with nature. | |
| N. L. always buys ten stuffed toys from every country he visits to open a museum with them in the near future. | |
| J.L. likes to walk with the dog during his free time and enjoys every weather situation. | |
| G.C. makes seismographic records of earthquakes in California, snowdrifts in Alaska, and hurricane in New Jersey. | |
| Cheating | P.L. sneaks into student parties and pours knockout drops into the drinks of other party guests |
| P.W. is conducting a study with demented persons. He often steals the promised reward from them and claims that they have already received it. | |
| Y.I., a biology student, is cultivating a sensitive plant as part of a student project. He spreads bugs to the plant of his classmate to destroy his results. | |
| M.R. takes over the organization of sports events. He favors the athletes who payed the highest bribe. | |
| P.O. is a coach in the gym. Since he has free access to the equipment locker, he steals parts of the equipment and sells it. | |
| J.L. usually fixes his bike in the backyard. If he is missing a tool, he steals his landlord's tool and claims it is his. | |
| A.W. exploits his demented grandmother. Month after month he educes money from her to squander it with his friends. | |
| F.L. is a research assistant. He secretly sells data of test subjects without their permission in order to enrich himself. | |
| R.Z. wants to buy tickets for a student party. Suddenly, he pushes himself forward in the line and others, who have been waiting for an hour, don't get any tickets. | |
| R.K. works as a bartender in a discotheque alongside his studies. He intentionally returns less change to drunken guests to enrich himself. | |
| F.A. secures a spot in the library early in the morning during the busy exam prep period, but then shows up 3 h later to study. | |
| A.B. eats the food supplies of his roommates during the night while they are asleep. |
Instructions:
Socially relevant situation:
Please try to imagine that you have to accomplish a research project at your university as part of your classes. The project has to be realized in a team of students. The other students are your classmates. This project is important for you as it has to be presented and evaluated at the end of the term. You will have to work with the other students over a longer period of time and meet regularly with your group.
Thus, you should choose your group members well in order to avoid annoyance, additional work or a bad grade.
Socially irrelevant situation:
Please try to imagine that you want to visit your aunt in Ireland. You are standing at the airport and still have a lot of time. Since you have passed through the security check quite quickly, you are now waiting for boarding together with other people in front of your gate. You see lots of people with their suitcases passing by the souvenir shops, ready to set off for the most diverse places in the world. You pull two small but heavy suitcases behind you and can't reach for your mobile phone right now.
Spontaneously, you decide to ask someone for the time.
Instructions for both groups:
We would like you to take a moment to think about this situation. Therefore, this screen will be displayed for 20 s.
The exact instructions can be seen in the appendix and are publicly available via the OSF Framework: https://osf.io/y3m24/?view_only=3623e5872e914f1da8eeb02e06e3fd79
The CVL database contains pictures of 114 persons, there are 7 images for each person, resolution: 640*480 pixels, format: jpeg, done with Sony Digital Mavica under uniform illumination, no flash and with projection screen in the background; persons age: mostly 18).
The data are available via the OSF Framework: https://osf.io/y3m24/?view_only=3623e5872e914f1da8eeb02e06e3fd79
All contrasts were conducted using the emmeans package in R (https://cran.r-project.org/web/packages/emmeans/). Because emmeans results are based on a model that is fitted to all of the data the degrees of freedom differ from a classical t-test.
Holm-method adjustment.
Again, all contrasts were conducted using the emmeans package in R. This package also adjusts the degrees of freedom to the number of tests.
