The perception of having been taken advantage of is painful and aversive (Vohs et al., 2007). Not only does being “duped” result in a disadvantageous outcome for the victimized person, but it also implies that one has been a fool for not having seen it coming. For this reason, the feeling of being duped is often accompanied by self-blame and other self-conscious emotions like shame, embarrassment, and regret. Consequently, people are strongly motivated to avoid feeling duped, and when they anticipate or expect exploitation, they often behave in a pre-emptively defensive way.

For example, such pre-emptive defensiveness or pre-emptive selfishness has been discussed as one explanation for social loafing, which describes the phenomenon that people tend to decrease effort when working in groups compared to when working alone (Latané et al., 1979). One reason why people loaf in groups appears to be that they expect others to do the same: They work less hard because they are afraid that their group members would otherwise “free ride” on their efforts. Stated differently, the feeling of being duped is so aversive that one would rather slack off and put up with a worse outcome than being the “sucker” who carries the rest of the group (Kerr, 1983).

Importantly, there are stable inter-individual differences in the motivation to avoid exploitation. While some people hate the idea of falling prey to other people’s malicious intentions, others simply care less about potentially being duped. The personality trait capturing such a latent fear of exploitation is called “victim sensitivity” (VS). VS has been shown to predict uncooperative and even immoral behavior in socially uncertain interdependence situations, when the risk of being duped is (considered to be) high. For instance, Gerlach et al. (2012) found victim-sensitive people to be less forgiving and more revengeful following a relational transgression committed by a romantic partner or close friend. Similarly, Faccenda and colleagues (2009) showed that after being fouled, soccer players high in victim sensitivity are more inclined to engage in unsportsmanlike, transgressive behaviors themselves. Additionally, victim-sensitive individuals also tend to make more egoistic choices in social dilemma games and other enticing situations (Fetchenhauer & Huang, 2004; Gollwitzer & Rothmund, 2011; Gollwitzer et al., 2005, 2009; Maltese et al., 2016; Rothmund et al., 2011) and more likely engage in counterproductive work behavior when feeling treated unfairly in organizational contexts (Lavelle et al., 2018).

A theoretical explanation for these findings is provided by the Sensitivity to Mean Intentions (SeMI) Model (Gollwitzer & Rothmund, 2009; Gollwitzer et al., 2013). According to this model, victim-sensitive individuals become more easily suspicious and in turn often behave selfishly and uncooperatively. More specifically, people high in VS are assumed to be particularly sensitive toward contextual cues indicating untrustworthiness. When such cues are present, a suspicious mindset (consisting of an attributional bias regarding others’ malevolence, legitimizing cognitions regarding one’s own selfish behaviors, and an avoidance-related motivational state) is activated in victim-sensitive people, and in this state of suspiciousness, there is less adherence to social norms and fairness principles. Thus, the suspicious mindset is theoretically conceptualized as the mechanism explaining the effect of VS on uncooperative behavior. In technical terms, the SeMI model describes a moderated mediation: When cues of untrustworthiness are present (“moderator”), VS activates a suspicious mindset (“mediator”), which then results in the withdrawal of trust and cooperation in social exchanges.

Empirical Evidence for the Suspicious Mindset

In situations that require trust, victim-sensitive individuals’ strong fear of exploitation leads them to react defensively and uncooperatively in order to protect themselves from possible victimization. In other words, victim-sensitive people are strongly motivated to avoid being duped. While a number of studies have looked at the “total effect” of VS (in the presence of untrustworthiness cues) on uncooperative behavior, the precise cognitive and motivational processes underlying such defensiveness have received less attention. As a consequence, evidence for the suspicious mindset and its three components (i.e., hostile interpretations, legitimizing cognitions, avoidance motivation) has been so far indirect.

The few studies that investigated the suspicious mindset have mostly used scenario-based approaches. For example, Gerlach et al., (2012, Study 3) applied the SeMI model to research on interpersonal forgiveness and instructed their participants to imagine that a close friend had harmed them (for instance, by disclosing a secret). After imagining this scenario, participants were asked how they interpreted the friend’s intentions, that is, whether they thought the friend’s post-transgression behavior was indicative of ulterior motives (attributional bias), and whether or not unforgiving reactions on their behalf would be justified (legitimization). In line with the SeMI model’s predictions, victim sensitivity was negatively related to participants’ willingness to forgive, and this relationship was (partially) mediated by hostile interpretations and self-protective legitimizing cognitions.

Following these findings, Maltese et al., (2016; Study 1) used ambiguous scenarios to show that heightened expectancies of injustice can (in part) account for victim-sensitive people’s uncooperative behavior in trust games. As expected by the authors, participants high in VS were more inclined to endorse and anticipate unjust outcomes when reading the scenarios (at least when a suspicious mindset had been activated before), and these anticipations mediated the negative effect of VS on trust and cooperation.

Lastly, Rothmund et al. (2011) were able to demonstrate that experiencing virtual aggression from a victim’s perspective resulted in reduced trust expectations as well. In their study, participants first played a video game sequence in which they were either betrayed by a video game character or not. Afterward, they took part in a common goods dilemma game in the real world. Here, participants had to indicate (1) how much they wanted to contribute, and (2) how much they expected other players to invest in this situation. Again, a significant indirect effect was found: When confronted with virtual untrustworthiness, participants were more reluctant to cooperate in the subsequent dilemma game, and this was mediated by heightened mistrust expectancies with regard to others’ investments.

In sum, previous research has attempted to elucidate the suspicious mindset and its components. Although these studies have shed some light on the cognitive dynamics of people high in VS, they have focused mainly on one element (namely the attributional bias), oftentimes by using scenario-based approaches with low ecological validity. Therefore, more research is needed in this context, which looks for stronger and more direct support regarding all three elements of the suspicious mindset. This is what Study 1 of the present research aims to do. In Study 2, we go a step further by illuminating the motivational processes operating behind the self-protective concerns of victim-sensitive people.

Motivational Processes Underlying the Pre-Emptive Defensiveness

What exactly motivates people to avoid exploitation and to react pre-emptively selfish when the risk for being duped is considered to be high? One of the few considerations of this question is given by Vohs et al. (2007). As briefly discussed above, these authors argue that people are motivated to avoid exploitation because the feeling of being duped is an aversive emotional state related to self-blame and other negative self-conscious emotions. More specifically, Vohs et al. (2007) posit that feeling duped arises from the perception of having been taken advantage of in a social context after one chose to trust the other person. Therefore, the victim is partly to blame for its situation: The unfavorable outcome could have been avoided if a different decision had been made (i.e., not trusting the other party), and for this reason, the experience of victimization results in self-recrimination. Stated differently, being deceived by another person has threatening implications for the self; it suggests that one is socially incompetent or stupid. Because people are reluctant to adopt such negative self-views, they will hence be motivated to avoid being the “sucker.” Thus, according to Vohs et al. (2007), individuals with a high fear of exploitation (i.e., people high in VS) are defensive because the anticipation of being exploited threatens their positive self-image.

Challenging this theorizing, we argue that victim-sensitive people’s defensiveness is not so much about avoiding self-image threats, but more about maintaining a sense of control. As a vast amount of research shows, people generally prefer to have (perceived) control over themselves, their environment, and the outcome of social situations (e.g., Fiske, 2003; Kay et al., 2008; Rothbaum et al., 1982; Skinner, 1995, 1996). Victim-sensitive individuals, however, might be particularly motivated to be in control: As Schmitt et al. (1995) demonstrated, a higher sensitivity to own unjust disadvantages was associated with a higher need for control. In our view, socially uncertain interdependence situations are particularly threatening to this need for control. In these situations, untrustworthiness cues act as a warning signal for vulnerability and victimization—an aversive outcome imposed by others. The individual should hence feel at the mercy of untrustworthy forces and, as a consequence, should be motivated to (re)establish a sense of control by preventing the expected exploitation (e.g., Rothbaum et al., 1982; Skinner, 1996). Thus, we argue that the defensiveness shown by people high in VS reflects a strategy to cope with the anticipated loss of control over the situation.

Study 2 was specifically designed to test these two competing hypotheses (i.e., anticipated exploitation threatens victim-sensitive people’s positive self-image vs. their need for control) directly against each other. Notably, if we were able to explain why people high in VS tend to react with self-protective and uncooperative tendencies in socially uncertain situations, then we could design strategies in order to attenuate these reactions. Understanding the motivation behind victim-sensitive individuals’ defensiveness becomes especially relevant when considering that a high fear of exploitation crucially impairs social interactions.

The Present Research

The aim of the present research is to further illuminate the defensiveness mechanism underlying victim sensitivity. More specifically, we pursue two goals: In Study 1, we simultaneously investigate all three elements of the suspicious mindset by using immersive virtual reality technology. A virtual environment has the advantage that it allows to study participants’ behavior in an ecologically realistic way while maintaining a high degree of experimental control (McCall & Blascovich, 2009). Thus, Study 1 goes far beyond what previous research on the suspicious mindset has achieved. In addition, we introduce social distancing as a behavioral measure of the mostly neglected third component of the mindset, avoidance motivation. Accordingly, two hypotheses are being tested in this study: First, we expect to replicate the established effect of VS on trust and cooperation. Second, we hypothesize that this relation is mediated by hostile interpretations, legitimizing cognitions, and participants’ approach-avoidance behavior in a virtual world.

In Study 2, we go a step further by exploring why people high in VS actually show such pre-emptive defensiveness. As discussed above, two hypotheses are plausible in this context: On the one hand, it may be that victim-sensitive individuals are defensive because they are afraid that being duped threatens their self-image (Vohs et al., 2007). On the other hand, persons high in VS might be defensive because they want to maintain a sense of control over what is happening in their social environment.

Study 1

Study 1 was designed to investigate the suspicious mindset in a more immersive and direct way than has been done in past research. More specifically, the focus of Study 1 was on participants’ avoidance motivation; a component of the suspicious mindset that has been neglected in previous studies. As a behavioral measure for this avoidance-related motivational state, social distancing was used. We expected that in a state of suspiciousness, individuals high in VS would (physically) distance themselves from others because in trust-relevant situations, any interaction partner represents a source of potential exploitation (Gollwitzer & Rothmund, 2009; Gollwitzer et al., 2013).

To activate such a state of suspiciousness in our participants, we confronted them with untrustworthiness cues in the beginning (see below). Then, they had to solve easy tasks in a fully immersive virtual world together with an ostensible partner while we tracked the distance between both players. The use of an immersive virtual environment allowed us to accurately, continuously, and unobtrusively measure participants’ distancing behavior (Bailenson et al., 2003; Kane et al., 2012) and, at the same time, to standardize the behavior of the ostensible partner (all of the confederate’s movements were pre-programmed). After completing the tasks in the virtual world, participants played a modified trust game (Berg et al., 1995). Victim sensitivity and other personality traits were measured several days or weeks prior to the laboratory appointment in a separate online questionnaire.

In sum, Study 1 tested the mediation described in the SeMI model (Gollwitzer et al., 2013). Thus, we predicted that after being confronted with untrustworthiness cues, people high in VS would keep more distance toward their partner’s avatar (M1), would interpret the other person’s intentions and behavior in a more distrustful, hostile way (M2), and would more strongly endorse legitimizing cognitions regarding their own uncooperative behavior (M3). We further assumed that these mechanisms would explain (i.e., mediate) the negative effect of VS on participants’ cooperation in a trust game (DV).


Taking into account the efforts that come with the use of the VR laboratory, our goal was to recruit as many participants as feasible within our six-month time frame (November 2018 to May 2019). The study, which was ostensibly about “coordination in a virtual world,” was advertised via university mailing lists, flyers, and an advertisement in a local newspaper. The study consisted of two parts: an online questionnaire (which was completed by 269 participants) and a subsequent laboratory appointment several days or weeks later (here, 92 subjects took part). Participants whose datasets could not be unequivocally matched were excluded. In addition, three participants had to be excluded because they were suspicious with regard to the confederate (n = 2) or because of technical issues in the virtual reality laboratory (n = 1). Thus, the final sample consisted of 84 participants (48% male, 52% female). The majority (93%) were undergraduate students from a wide variety of disciplines; ages ranged from 18 to 50 years (M = 23.6, SD = 4.6 years). Participants received monetary compensation for their time, the exact amount (4€-10€) was dependent upon their decisions in the trust game.

Materials and Measures

As mentioned above, the study consisted of an online questionnaire in which personality traits were measured, and a laboratory appointment in which we assessed social distancing, hostile interpretations, legitimizing cognitions, and trusting behavior. Data from these different parts were matched via a personalized code that participants created on their own and that ensured (pseudo-)anonymity. For privacy reasons, this code was later deleted and replaced by a numerical code in the dataset.

Personality Traits

At the beginning of the online questionnaire, participants were informed that the study consisted of two parts and that it would take approximately one hour to complete. After giving informed consent, participants provided demographic information and then completed several trait measures, including victim sensitivity (VS), observer sensitivity (OS), general trust, and the Big Five. Victim and observer sensitivity were assessed with 10 items each (Justice Sensitivity Inventory; Schmitt et al., 2010), which were answered on a Likert scale from 0 (“totally disagree”) to 5 (“totally agree”). Example items are: “It bothers me when others receive something that ought to be mine” (VS) and “I am upset when someone does not get a reward that he/she has earned” (OS). The scales have good internal consistency (in the present study: Cronbach’s αs = 0.84 and 0.89, respectively). General trust (α = 0.81) was measured with a German version of the six-item General Trust Scale by Yamagishi and Yamagishi (1994) and was answered on a six-point Likert scale as well. Finally, neuroticism (α = 0.80), extraversion (α = 0.85), and openness to experience (α = 0.58) were measured with the BFI-10 (Rammstedt & John, 2007); here, response scales ranged from 1 to 5. Several other scales were assessed as well but because they are not relevant for the present research, they will not be referred to any further. At the end, participants had the possibility to sign up for the second part of the study, which was scheduled several days or weeks later. In addition, they were informed that the laboratory appointment would take place with a second participant.

Virtual Reality

In the laboratory, participants were informed that their partner (i.e., a female confederate) would be a couple minutes late but that they would start preparations without her. The confederate being unpunctual had two reasons: First, because it should raise doubts about her dependability and trustworthiness, and second, because participants should not see the other person in the “real” world (to avoid any positive or negative first impressions). For this reason, participants were also told not to talk to the other person once she arrived.

Next, participants received an information sheet, which informed them that they would first solve different coordination tasks in a virtual world and that they would play a (trust) game with their partner afterward. They were also given a brief summary of the trust game instructions to make the possibility of a later exploitation salient. After participants gave their informed consent, they put on a head mounted display (HMD) and were led to their starting position for the virtual tasks (while wearing the HMD, participants were unable to see any part of the “real” world). Next, the experimenter secretly informed the confederate via SMS or WhatsApp, who arrived at the laboratory approximately two minutes later. The confederate signed the informed consent as well and then (ostensibly) put on a second HMD.Footnote 1

Next, the experimenter explained that both players would be randomly assigned to an avatar (i.e., a human representation in the virtual world), but that they would only see the avatar of their counterpart. In fact, participants did not see any virtual representation of their own body. The confederate, however, was always represented by a female avatar with a neutral facial expression (see Fig. 1). A neutral appearance was chosen because, as previous research has demonstrated, victim-sensitive individuals perceive neutral faces as less trustworthy (Gollwitzer et al., 2012; Study 1). A neutral facial expression, albeit a subtle cue, should thus activate a suspicious mindset in individuals high in VS, especially if combined with other untrustworthiness cues (like the tardiness of the confederate and the possibility of later exploitation in the trust game). Next, the experimenter started the virtual environment. Here, participants saw a spacious plaza with a fountain, plants, and some shops and restaurants (see Fig. 1). In addition, a blue and a red cross on the ground indicated the starting points for the respective players (participant and confederate were facing each other when standing on their starting positions). We did not present participants with auditory input in the virtual world because they had to listen to the experimenter’s instructions.

Fig. 1
figure 1

Virtual environment and female avatar used in Study 1. Both the plaza and the avatar were taken from built-in packages of the Vizard virtual reality software (WorldViz). Printed by permission

After familiarization with the virtual world in a warm-up task, participants’ approach-avoidance behavior was assessed in three proxemic tasks. Here, participants were asked to (1) walk around their partner to read a combination of numbers and letters from their back (this task was validated in previous research, see Bailenson et al., 2003), (2) to avoid additional agents (i.e., virtual humans controlled by a computer algorithm) who walked through the virtual place, and (3) to approach their partner up to the point where the closeness still felt natural and comfortable for them (e.g., Sommer, 2002; Uzzell & Horne, 2006).

Importantly, the two players took turns in performing these tasks; thus, only the participant or avatar were moving at any given time (in Tasks 1 and 3 the participant went first, in Task 2 the confederate). In all proxemic tasks, the position and orientation of participant and avatar were continuously tracked and these data were used to compute the minimum distance for each task.Footnote 2 We chose minimum distance as a measure for social distancing because it has been successfully used in both traditional proxemic research and studies investigating interpersonal distance in virtual environments (e.g., Bailenson et al., 2001, 2003).

Trust Game

After leaving the virtual world, participants were escorted to another room in which they were seated in front of a computer screen. They were informed that they would now play a game with their partner and that the points they won in this game would be added to their 2€ compensation fee: For every 10 points they gained, they would be payed 1€ extra. The experimenter then left the room, ostensibly to tend to the other person who still waited in the virtual reality laboratory,Footnote 3 while participants familiarized themselves with the instructions of the game (adapted from Berg et al., 1995). The rules were explained as follows: Both players would make two decisions, one in the role of “sender” (i.e., trustor) and one in the role of “receiver” (i.e., trustee), but they would not be informed about the decisions of their partner until the end of the game. As senders, they would receive 20 points, which they could either keep (those points would be added to their payment) or send to the receiver. More specifically, participants could send 0, 5, 10, 15, or 20 points that were then quadrupled by the experimenter. The receiver, on the other hand, could then decide to either keep all or to share any amount of the quadrupled points with the sender.

After participants read these instructions, they first made their decision as senders (which was our focal DV in this study) and then as receiver. However, because they were not informed about the decisions of their alleged partner during the game, participants had to make the second decision for all possible outcomes (strategy method). After each decision, participants answered a short follow-up survey about their feelings and thoughts during the respective decision (e.g., “I am satisfied with my decision” and “I was afraid of being taken advantage of by the other player”; see Table 4 ).

SeMI Components

After the trust game, participants answered a series of questions about their perception of the other person and her behavior in the virtual world. Four of these items assessed hostile interpretations (e.g., “Do you think the other person is trustworthy?” (reversed), “Do you think the other person has hostile motives?”; Cronbachs α = 0.75) and two items measured legitimizing cognitions (i.e., “Do you think it is justified not wanting to work with the other person?”, “Suppose you were to work on a project with the other person (e.g., for a seminar). Do you find it justified to refuse?”; Cronbachs α = 0.67). Ten additional items (e.g., “I had the impression that the other person kept some distance to me in the virtual world”) were included for exploratory purposes. All of these follow-up items (see also Appendix A; Tables 5 and 6) were answered on six-point Likert scales. Finally, participants were probed for suspicion, fully debriefed, and received their compensation fee.Footnote 4

Results and Discussion

To test our hypotheses, we correlated victim sensitivity (and other relevant personality traits) with our three mediator variables and our DV (i.e., trust behavior) in a first step. These correlations, as well as means and standard deviations of the respective variables, can be found in Table 1. Correlations among personality measures resembled earlier findings (e.g., Schmitt et al., 2005, 2010): For example, victim sensitivity was positively related to observer sensitivity and neuroticism, but uncorrelated with extraversion or openness.

Table 1 Means, standard deviations, and correlations between measured variables (Study 1)

As can be seen in Table 1, minimum distance,Footnote 5 that is, the shortest distance that participants maintained between themselves and the avatar of the other person, was uncorrelated with victim sensitivity (r = 0.11, p = 0.31). Interestingly, though, exploratory analyses of the follow-up items revealed that whereas participants high in VS showed no avoidance tendencies themselves, they still perceived the other person to be more distant and reserved. Thus, despite the fact that avatar movements were pre-programmed, victim-sensitive participants more strongly had the impression that the other person kept some distance to them in the virtual world (r = 0.25, p = 0.02). Unexpectedly, we also found observer sensitivity and openness to experience to be positively correlated with the social distance measure, which implies that participants high in OS and openness kept more distance to the avatar than participants low in these traits.

With regard to the other two suspicious mindset components, we found that VS was positively correlated with hostile perceptions of the other player (r = 0.24, p = 0.03), but uncorrelated with legitimizing cognitions (r = 0.14, p = 0.21). Thus, participants high in VS interpreted the other person’s intentions and behavior in a more negative, mistrustful way, but they did not consider uncooperativeness toward the other person as more justifiable. However, this may have not been necessary because participants high in VS showed no uncooperative behavior in our study anyway: Contrary to our hypothesis and to the results of prior research, we found no correlation between victim sensitivity and trust behavior (r = 0.09, p = 0.40). In fact, the decision to send points was unrelated to any of the personality traits measured, as well as to the suspicious mindset variables (see Table 1).

Mediation Analysis

Despite the insignificant total effect of VS on trust behavior, we conducted our multiple mediation analysis as planned. This was based on the reasoning that indirect effects can exist even though a total effect is not statistically significant (e.g., Hayes, 2018), for instance, because different indirect effects may suppress each other.

More specifically, we included all three mediator variables (i.e., hostile interpretations, legitimizing cognitions, social distancing) simultaneously into the model while controlling for observer sensitivity, general trust, neuroticism, and extraversion.Footnote 6 An overview of the results can be found in Fig. 2. Overall, only 10% of the variance in trust behavior was explained by our model. Contrary to our hypothesis, victim sensitivity had no (total or direct) effect on trust. With regard to the mediators, we found VS to predict hostile interpretations (β = 0.27, p = 0.04), but no significant effects on legitimizing cognitions or social distancing were found (p’s = 0.14 and 0.78, respectively). In accordance with the correlation analyses, none of these mediator variables significantly influenced the amount of points sent in the trust game (although the positive effect of social distancing reached marginal significance). For this reason, it is not surprising that no indirect effect of VS on trust was found: Bootstrapping results showed that the 95% confidence intervals always contained zero. Thus, we were unable to replicate the established negative relationship between victim sensitivity and trust, and we also did find no support for the assumption that social distancing, hostile interpretations, and legitimizing cognitions (as elements of a suspicious mindset) mediated this hypothesized relation.

Fig. 2
figure 2

Multiple mediation of the effect of VS on trust via social distancing, hostile interpretations, and legitimizing cognitions. Standardized regression coefficients are depicted. *p < .05 (two-tailed)

In sum, victim sensitivity was not (directly or indirectly) related to distrust in our findings. However, when taking a closer look at our DV it became clear that a ceiling effect (which reduces variation in trusting behavior) might have contributed to this lack of findings: While more than 50% of participants decided to send all of their 20 points to their partner, only one person sent nothing.Footnote 7 Importantly, exploratory analyses of the follow-up items revealed that high-VS participants were less satisfied with their decision (“I am satisfied with my decision”; r =  − 0.29, p = 0.01) and had a greater fear of exploitation in the trust game than low-VS participants (“I was afraid of being taken advantage of by the other player”; r = 0.24, p = 0.03). In other words, participants high in VS were afraid that the other person might take advantage of them, but they did not act upon this fear. One reason for this could be that the untrustworthiness cues that we provided participants with were not strong enough to trigger self-protective behavior. Another explanation is given by Dunning et al. (2014): These authors argue that humans sometimes show trust in others (especially strangers) because they believe that norms of respect force them to do so. Thus, in many situations, individuals do not really expect benevolence from a cooperation partner, but feel obliged to show trusting behavior nonetheless. It is therefore possible that victim-sensitive participants in our study sent points to their partner because this was what they felt they should do, although it was not what they really wanted to do.

Although we were unable to find support for our main hypotheses, the results of Study 1 still provided some important insights into the defensiveness mechanism underlying victim sensitivity. For instance, our results show that whereas VS did not predict social distancing, participants high in VS perceived the confederate to be physically distant—despite the fact that the partner’s behavior was standardized. Together with the finding that victim sensitivity promoted hostile interpretations of the other person’s behavior and intentions in general, this lends further support for the “inaccuracy hypothesis” (Gollwitzer et al., 2012, 2013): According to this notion, VS implies an asymmetrical sensitivity toward untrustworthiness cues, which results in less accurate predictions of other people’s cooperativeness (i.e., a bias in social judgments). In addition, it is also noteworthy that VS predicted hostile interpretations even though our cues of untrustworthiness were rather subtle in this study.

In conclusion, we were able to demonstrate that victim sensitivity was related to more unfavorable perceptions of the other player as well as to fear of exploitation, but we did not find any effects on a behavioral level (i.e., participants high in VS showed no social distancing and no withdrawal of trust and cooperation). Hence, while these results still suggest that victim-sensitive individuals’ information processing is characterized by a pronounced suspiciousness in socially uncertain situations, no conclusions about the underlying motivational processes could be drawn. Therefore, Study 2 was conducted to investigate why individuals high in VS are more likely than others to act pre-emptively defensive and how their self-protective strikes may be attenuated.

Study 2

Study 2 was designed to test the hypothesis that restoring a sense of control over what happens in one’s social environment can effectively alleviate victim-sensitive individuals’ tendency to distrust others. Similar to Study 1, participants played a trust game with an unknown fellow participant and participants’ behavior as senders (or “trustors”) was used as our focal DV. To manipulate the activation of a suspicious mindset, half of the participants were confronted with cues of untrustworthiness (i.e., a respective facial expression and a message; see below).

To restore a sense of control, we used an established control-affirmation procedure adapted from Kay et al. (2008), in which participants are prompted to recall a situation in which they had recently experienced a sense of control (see Methods section). Although this procedure has been successfully used in prior research, we figured that recalling a situation in which participants had control might not only affirm their momentary sense of control, but may also and inadvertently boost their self-regard. For instance, recalling a situation in which one has shown moral courage and protected another person from being bullied may effectively reinstall a sense of control, but may also (re)affirm one’s moral virtues and self-regard (e.g., Monin & Miller, 2001). Therefore, in order to test whether restoring a sense of control is specifically responsible for alleviating the effect of suspiciousness on distrust among victim-sensitive individuals, we also implemented an experimental condition in which participants were self-affirmed (using a procedure adapted from Monin et al., 2008).

Thus, Study 2 used a 2 (suspicious mindset, no suspiciousness) × 3 (control-affirmation, self-affirmation, no affirmation) between-subjects design with victim sensitivity as a continuous moderator variable. We predicted that when a suspicious mindset has been activated, victim-sensitive individuals would be less trustful toward a fellow participant when no affirmation occurred and that this effect was alleviated specifically after restoring participants’ sense of control. Thus, we hypothesized a suspiciousness × affirmation × victim sensitivity three-way interaction effect.


Undergraduate students from a mid-size German university were invited via email to participate in an online study about “social interactions, self-views, and emotions.” In exchange for participation, five online book vouchers worth 50€ each were raffled among all participants. The invitation contained a link to the survey, which was active for two months. We aimed at recruiting as many participants as possible within this time frame. Detecting a small- to medium-size three-way interaction effect (i.e., f2 = 0.03, as in Gollwitzer & Rothmund, 2011; Study 2), with a power of 80% would have required a sample size of 325 if α = 5% (calculated with G*Power 3.1; Faul et al., 2009).Footnote 8

At the end of the data collection phase, 321 participants had started the study; 291 completed it. Participants who needed less than 1 min (n = 12) or more than 2 h to complete the online survey (n = 5) were discarded from further analyses. One additional case was deleted because this participant gave an invalid response (i.e., 12) to the trust game measure (values could only range between 0 and 10, see below). Thus, the final sample consisted of 273 participants (173 female; 100 male). Ages ranged between 18 and 64 years (M = 25, SD = 7.1 years). Students were enrolled in a broad range of undergraduate programs, including psychology (n = 18). A sensitivity power analysis revealed that, with 273 valid cases, the population effect would have to be equal or larger than f2 = 0.036 in order to be detected with a probability (i.e., power) of 80% on a significance level of 5%. We considered this to be at the upper limit of a plausible effect, given that especially ordinal interactions are often smaller than expected (e.g., Lakens & Caldwell, 2021). Therefore, we decided to increase our significance level to α = 10%, which we considered justifiable given that our hypothesis is directional. This increased our power to detect our pre-specified population effect of f2 = 0.03 with 273 participants to 82%.

Materials and Measures

Trust Game

After participants gave their informed consent, they were introduced to the rules of the trust game. They were told that they would be paired with a second (female)Footnote 9 participant from another university in the south of Germany, with which we would be cooperating on a research project. Participants were informed that they would play two rounds with their partner: In Round 1, participants would be the “sender” (i.e., trustor) and their partners would be “receivers” (i.e., the trustee). In Round 2, roles would be switched. Unbeknownst to participants, there was no second round, and there also was no other participant. Participants received 10 lottery tickets, of which they could send any number (between 0 and 10) to their partner. The number of tickets sent would be tripled by the experimenter so that receivers would now have their own 10 tickets plus the tripled number of received tickets from their partner (i.e., the real participant). At that point, receivers would have to decide whether or not they want to split the total number of tickets equally between both players (this version of the trust game has been described by Kuwabara, 2005; for an application in research on the SeMI model, see Gollwitzer & Rothmund, 2011; Study 2). Participants were also told that the number of lottery tickets they had in the end would determine their chances of winning the 50€ book voucher: The more tickets, the higher the likelihood of winning one of these vouchers.

Suspiciousness Manipulation

Next, participants saw a picture that their respective partner has allegedly taken of herself during their experimental session. These pictures were selected from the Amsterdam Dynamic Facial Expression Set (ADFES; van der Schalk et al., 2011), a stimulus set containing 648 emotional (facial) expressions displayed by 22 models (for a validation, see Wingenbach et al., 2016). For the present purpose, two pictures were selected (see Fig. 3), one displaying a neutral facial expression (left panel) and the other displaying joy (right panel). More precisely, the neutral facial expression was used to induce a “suspicious mindset” among participants, whereas the joyful facial expression was used to prevent the induction of suspiciousness. Using similar stimuli, Gollwitzer et al., (2012; Study 1) showed that a neutral facial expression is indeed sufficient to induce a suspicious mindset among victim-sensitive individuals.

Fig. 3
figure 3

Stimulus material used in Study 2 to induce (left panel) vs. not induce (right panel) suspiciousness. Stimuli were selected from the Amsterdam Dynamic Facial Expression Set (ADFES; van der Schalk et al., 2011). Reprinted by permission

To make our manipulation even stronger, participants received a (fake) message from their partner, which she allegedly wrote to introduce herself. These messages were designed to induce a suspicious mindset (“Hi, my name is Heike. I’m 25 and studying business administration in Tübingen. It’s awesome that I have the chance to win something here.”) vs. to reduce the likelihood of suspicions being raised (“Hi, my name is Sarah. I’m 25 and studying social work in Tübingen. It’s awesome that we both have the chance to win something here.”). Prior research showed that a message in which one’s partner focuses on maximizing their own profit (as in the suspicious mindset condition) vs. on maximizing the team’s collective profit (as in the no suspiciousness condition) can effectively amplify the effect of distrust on uncooperative responses (Parks et al., 1996).

Affirmation Manipulation

Next, participants were randomly assigned to either a control-affirmation, a self-affirmation, or a no-affirmation (control) condition. Participants in the control-affirmation condition were instructed to recall a positive event from the last couple months for which they were personally responsible, in which they experienced a “sense of control over what had happened” (adapted from Kay et al., 2008). Participants in the self-affirmation condition were prompted to recall a positive event from the last couple months in which they were able to display a virtue or express a value that was personally relevant for them, such as a sense of humor, social competence, and musical talent (taken from Monin et al., 2008). Participants in the no-affirmation (control) condition were asked to list all the things they do on a typical weekday. In all three conditions, participants were asked to use the free-response field provided to write a short text (100 words max). These texts were screened and coded for validity and credibility. All participants followed the instructions and wrote a principally credible text, which was in line with the respective prompt.

After writing these texts, participants were re-familiarized with the rules of the trust game. Then, they actually played the trust game and decided how many lottery tickets they would want to transfer to their partner. This measure was our dependent variable.

Personality Traits

Finally, participants completed a battery of personality trait measures, including the 10-item Victim Sensitivity Scale (α = 0.83), the 10-item Observer Sensitivity Scale (α = 0.84), both taken from Schmitt et al. (2010), and the 6-item General Trust Scale (α = 0.82) taken from Yamagishi and Yamagishi (1994). Response scales on all measured variables ranged between 1 (“totally disagree”) to 6 (“totally agree”). After responding to a number of control questions (e.g., “What do you think this study was about?”), participants were informed about their final amount of lottery tickets, thanked, and debriefed.

Results and Discussion

Looking at our central DV, participants’ decisions in the trust game, many of them (n = 119, i.e., 44%) decided to transfer all of their lottery tickets (i.e., 10) to their partner. Only 7 participants (2.6%) decided to keep their 10 tickets for themselves. The median was 8, the mean was 7.57 (SD = 2.7), and the skewness index was − 0.83 (SE = 0.15), suggesting a moderate to high negative skewness and a significant deviation from the normal distribution (Shapiro–Wilks’ W = 0.832, p < 0.01). Therefore, we used a procedure proposed by Box and Cox (1964) to find the optimal transformation for our variable,Footnote 10 which, in our case, was

$$Y_{{{\text{trans}}}} = \frac{{\left( {Y + 0.5} \right)^{\lambda } - 1}}{\lambda }$$

with Y denoting the trust measure in its original metric, and Ytrans denoting the transformed variable (see also Fox & Weisberg, 2011; Velilla, 1993). Means, standard deviations, and correlations between the transformed trust variable and the three personality traits measured are displayed in Table 2.

Table 2 Means, standard deviations, and correlations between measured variables (Study 2)

As expected, victim sensitivity was positively related to observer sensitivity and negatively related to general trust, which is why we used observer sensitivity and general trust as covariates in the regression model reported below. In addition, the transformed trust variable was negatively related to victim sensitivity and positively related to general trust across all experimental conditions.

We predicted that when a suspicious mindset has been activated, victim-sensitive individuals would be less trustful toward a fellow participant when no affirmation occurred, and that this effect would be alleviated specifically by a control-affirmation (vs. a self-affirmation) procedure. To test the hypothesized suspiciousness × affirmation × victim sensitivity three-way interaction effect, we contrast-coded the suspiciousness conditions (no suspiciousness = –1, suspicious mindset =  + 1) as well as the affirmation conditions so that Contrast 1 reflected an effect of control-affirmation vs. baseline (control-affirmation =  + 1, self-affirmation = 0, no-affirmation = –1) and Contrast 2 reflected an effect of self-affirmation vs. baseline (control-affirmation = 0, self-affirmation =  + 1, no-affirmation = –1). Victim sensitivity was standardized to facilitate the interpretation of regression weights. Observer sensitivity and general trust were included as covariates.

The estimated regression coefficients are reported in Table 3. The two focal predictors in this model are the three-way interaction terms in the two bottom rows of the table. Adding these two terms to the model increased the explained variance by 1.3% (from R2 = 0.093 without the three-way interaction terms to R2 = 0.106 including them). Inspecting their regression weights suggests that a control-affirmation procedure (but not a self-affirmation procedure) was able to make victim-sensitive individuals just as trustful as their victim-insensitive counterparts, as indicated by the VS × Suspiciousness × Contrast_1 interaction effect, B = 0.913, SE(B) = 0.505, p = 0.072. This interaction effect is also graphically displayed in Fig. 4. Simple slope analyses confirmed that while victim sensitivity was negatively related to trust in the suspicious mindset condition when no affirmation occurred (B =  − 1.834, p = 0.015), this effect disappeared in the control-affirmation condition (B =  − 0.459, p = 0.596). Thus, although participants high in VS tend to distrust others in socially uncertain situations, this pre-emptive defensiveness can actually be attenuated if a sense of control is (re-)established.

Table 3 Results from the moderated regression model (Study 2)
Fig. 4
figure 4

Predicted values visualizing the suspiciousness × affirmation interaction effect for participants low vs. high in victim sensitivity (± 1 SD around the sample mean; Study 2)

General Discussion

Being duped is a highly aversive, painful experience which people are motivated to avoid. For this reason, especially people with a high fear of exploitation (i.e., people high in victim sensitivity) tend to act pre-emptively selfish and hostile in situations in which the risk for being “suckered” is considered to be high. However, although past research has repeatedly investigated the relation between VS and uncooperativeness in socially uncertain situations (e.g., Faccenda et al., 2009; Fetchenhauer & Huang, 2004; Gollwitzer et al., 2005; Lavelle et al., 2018), the assumed mechanism (i.e., the suspicious mindset) has received far less attention (for exceptions, see Gerlach et al., 2012; Maltese et al., 2016; Rothmund et al., 2011). To close this research gap, we conducted two studies investigating the cognitive and motivational processes underlying such pre-emptive defensiveness. Together, the findings crucially expand our knowledge of the defensive motivational system in victim-sensitive individuals.

The Suspicious Mindset

To better understand how victim sensitivity translates into selfish and uncooperative behavior, we examined possible mediating processes in Study 1. More precisely, we used virtual reality technology to create an immersive environment in which participants interacted with another person to unobtrusively measure social distancing (as a form of self-protective, avoidant behavior). Although our main focus was on this approach-avoidance behavior, we also assessed hostile interpretations of the other person’s behavior and intentions, and legitimizing cognitions with regard to own uncooperative reactions in self-report (i.e., the other two components that are assumed to constitute the suspicious mindset). Finally, participants played a trust game with their alleged partner over real money.

Notably, the results did not support our hypotheses for the most part. First, we were unable to show the expected correlation between victim sensitivity and social distancing. More specifically, VS only predicted hostile and mistrustful perceptions of the other person, but not legitimizing cognitions or avoidance behavior. However, although participants high in VS showed no distancing tendencies themselves, they still perceived the other person to keep their distance, which was by design a purely subjective impression. Thus, our findings corroborate the notion that VS promotes a bias in social judgments: Victim-sensitive individuals focus strongly on untrustworthiness cues, which distorts their perceptions of their social environment (Gollwitzer et al., 2012, 2013).

Second, trust rates were overall very high, yet unrelated to any of the measured variables. In contrast to previous studies, victim-sensitive participants did not send fewer points to their partner in the trust game, even though they reported a higher fear of exploitation than victim-insensitive participants. Thus, we were unable to replicate the established effect of victim sensitivity on distrust, and, as a consequence, we were also unable to show that the suspicious mindset mediated this relation. However, there are a number of possible reasons that might explain why the expected direct and indirect effects of VS on trust were absent in our findings. For instance, the sample size may have simply been too small to ensure adequate statistical power for the detection of the hypothesized effects. Monte Carlo simulations suggest that several hundred participants are required if small to medium sized effects are tested in a (multiple) mediation model (Fritz & MacKinnon, 2007; Ma & Zeng, 2014; Thoemmes et al., 2010). For this reason, future studies should aim at recruiting a much larger sample than the one used in the present study.

In addition, the SeMI model and previous research (e.g., Gollwitzer & Rothmund, 2011; Rothmund et al., 2011; Süssenbach & Gollwitzer, 2015) suggest that VS predicts uncooperative and distrustful behavior only if there are sufficient indications that an interaction partner might be untrustworthy. The untrustworthiness cues that participants encountered in Study 1, though, were relatively subtle. In addition, participants interacted with the other person for quite some time before playing the trust game, and, as previous research demonstrates, prior contact with a partner can increase trustworthiness perceptions (Glaeser et al., 2000). That said, the fact that victim-sensitive participants reported more fear of exploitation and less satisfaction with their own decision in the trust game corroborates the notion that our manipulation was (at least, in part) successful: Participants high in VS were more fearful and suspicious of their partner’s intentions, although they did not act on it.

In conclusion, victim sensitivity was exclusively related to an attributional bias regarding others’ malevolence in Study 1, which confirms and complements past findings on the suspicious mindset in victim-sensitive individuals (Gerlach et al., 2012; Maltese et al., 2016; Rothmund et al., 2011). However, although VS predicted more unfavorable perceptions of the other player, we were not able to demonstrate that this bias actually mediated behavioral trust and cooperation.

Motivational Basis of Defensiveness

To examine why exactly victim-sensitive people tend to react defensively in socially uncertain situations and how such pre-emptively selfish behavior may be alleviated, a second study was conducted. More precisely, Study 2 was designed to test whether victim-sensitive individuals are defensive because (1) they are afraid that being exploited threatens their self-image (Vohs et al., 2007) or because (2) they want to maintain a sense of control over their social environment. To compare these two possible motivations directly against each other, we employed an experimental design in which participants were either self-affirmed or control-affirmed (or not affirmed at all) after the activation of a suspicious mindset (vs. after no induction of suspiciousness), and then instructed to play a trust game with an ostensible partner.

Our results suggest that a control affirmation was able to reduce the negative effect of VS on trust behavior when a suspicious mindset was evoked, although the effect was small and only significant on a 10% level. Affirming a positive self-image, on the other hand, did not significantly alleviate the negative effect of VS on trust behavior (see Fig. 4). These findings suggest that affirming a sense of control tends to be more relevant for high-VS individuals than affirming a positive self-image. In this context, it should be noted that, in contrast to Study 1, we activated a relatively strong suspicious mindset in Study 2 to test our hypotheses as strictly and conservatively as possible. The fact that affirming a sense of control made victim-sensitive individuals just as trustful as their victim-insensitive counterparts suggests that restoring control may be an effective strategy to mitigate suspiciousness. Following up on these insights is, in our view, an important avenue for future research.


The present research further corroborates the SeMI model’s predictions and expands our knowledge about the pre-emptive defensiveness shown by victim-sensitive individuals, but, of course, there are also some limitations. First and foremost, the sample size of Study 1 was rather small, and the effect found in Study 2 (i.e., the VS × Suspiciousness × Contrast_1 interaction effect, which was significant on a 10% level, see Table 3) was also smaller than expected. For this reason, the suspicious mindset and its components should be investigated in more highly powered studies in the future, especially if more complex designs are applied. This will allow to draw conclusive inferences about the underlying defensiveness mechanism.

Another constraint lies in the fact that the minimum distance measures were uncorrelated across proxemic tasks in Study 1, although all of these tasks were designed to assess participants’ distancing behavior. However, they differed in their demands: While Task 1 (reading the number from the other person’s back) and Task 3 (moving toward the other person and stopping at a comfortable distance) specifically asked participants to approach their partner, Task 2 focused on the avoidance of additional virtual humans (therefore increasing the possibility that participants were too distracted to pay attention to the confederate or to their physical proximity). Here, we focused on the minimum distance assessed in the first proxemic task, which should represent the most valid measure for social distancing in our study.Footnote 11 Nevertheless, future research may want to use more than one measure of approach-avoidance behavior (or even additional operationalizations of avoidance motivation such as the avoidance of eye contact).


A large body of empirical findings shows that a high fear of exploitation (i.e., high victim sensitivity) results in uncooperative and hostile behaviors in socially uncertain situations. Importantly, this pre-emptive defensiveness implies a focus on others and their behavior: Instead of worrying about themselves and the impression they make, people high in VS are more concerned about what others are up to. In other words, defensiveness is an inherently social phenomenon, not a merely self-centered phenomenon. Investigating this social sensitivity is important not only from a theoretical but also from an applied perspective: If we understand which cognitive and motivational processes are at work in victim-sensitive individuals, we may be able to develop effective interventions capable of preventing or attenuating adverse effects of VS on trust and cooperation. The present research suggests that boosting a sense of control helps victim-sensitive people to overcome their habitual suspiciousness. Stated differently, providing people with a sense of control might prove to be an effective tool to improve social interactions and to break vicious cycles of non-cooperation.