People are capable of doing horrendous things. Frequent incidents occur in our world where a perpetrator kills, tortures, rapes, or kidnaps a fellow human being. Through modern media, lay people are often confronted with descriptions or pictures of perpetrators who committed such severe injustices. One way to make sense of these perpetrators is to demonize them, that is, to form impressions of them as personifications of pure evilness (e.g., Baumeister, 1997; Darley, 1992). It has been noted that demonizing is psychologically functional to preserve the belief that the world generally is a just place, because perpetrators are classified into an exceptional category that does not qualify as human (Ellard, Miller, Baumle, & Olson, 2002; Hafer & Bègue, 2005). Demonizing can be considered as an extreme form of dehumanization. It is displayed particularly in response to individuals, groups, or institutions that hold attitudes or display behaviors that are considered immoral by the perceiver. As such, demonizing is associated with attributions of extreme, intrinsic, and non-human immorality, and may provide a basis for conflict, destructive behaviors, and severe sanctioning. Although demonizing is a common response to perpetrators, surprisingly little research has been conducted to disentangle the psychological processes that are responsible for this phenomenon (Ellard et al., 2002). The present research sought to examine some of the psychological dynamics underlying demonizing.

Demonizing is rooted in the concept of “evil” (Baumeister, 1997). Although evil may be a rather abstract construct for people, it is likely that people nevertheless have some mental picture of what evildoers might look like. Many religions, folklores, and myths possess anthropomorphized symbols of evil, such as demons, evil spirits, or various representations of the devil—the most well-known being “Lucifer” or “Satan” in Judaism and Christianity. However, contemporary culture also has its non-religious personifications that are considered to represent pure evilness. These personifications may either refer to real people (e.g., Adolph Hitler, Osama Bin Laden), fictitious characters (e.g., Count Dracula, Freddy Krueger), or social groups (e.g., the Ku Klux Klan). The point here is that it seems perfectly normal for people to have personified representations of what evil looks like. Consistent with this observation, social psychologists have assumed that most people posses a prototypical evilness scheme, that is, a coherent mental set of stereotypically diabolical attributes that are associated with evildoers (Baumeister, 1997; Berkowitz, 1999; Ellard et al., 2002). Notably, perceivers tend to believe that evildoers possess a variety of interrelated evilness cues, such as lacking uniquely human emotions (Leyens, Paladino, Rodriguez-Torres, Vaes, Demoulin, & Rodriguez-Perez, 2000), being socially isolated (Baumeister, 1997), and having a reputation of immoral behavior that either holds no regard for others’ well-being, or is even aimed at intentionally hurting others (Berkowitz, 1999). How such evilness cues are attributed to an offender and weighted in demonizing judgments can vary substantially, depending on for instance the type of offense, the offender’s motives, and information about mitigating circumstances. Sometimes, perpetrators may indeed appear to fit a prototypical evilness scheme. For instance, sometimes a socially isolated offender, who committed multiple severe crimes, may come across as cold-blooded, remorseless, or even amused when confronted with the suffering that was caused. In such cases, the behaviors displayed by the perpetrator match expectations of prototypical evilness, rendering it likely that perceivers demonize the perpetrator.

Through various media, however, we often encounter individuals who committed exceptionally heinous crimes, but are not very representative of the prototypical evildoer. An infamous example is the Nazi Adolph Eichmann, one of the main architects of the holocaust, and widely held responsible for sending six million Jewish people to their deaths. As described by Arendt (1963) in her account of the trial, Eichmann was not a prototypical “monster”, nor did he display evidence of a personality disorder. Rather, Eichmann came across as a somewhat bureaucratic, normal, perhaps even boring family man, who seemed motivated primarily by furthering his own career, and by accurately adhering to the laws and ideologies of the Nazi regime. Thus, perpetrators differ on the extent to which they match prototypical expectations of evilness, and this is likely to have implications for the extent to which people are prone to demonize them. In the present contribution, we investigate such influence of prototypical evilness by examining basic cognitive processes that we expect to be associated with demonizing. In particular, by integrating theoretical insights on demonizing with empirical knowledge in related theoretical domains, we suggest that the cognitive resources that perceivers have available is a key factor to predict the influence of prototypical evilness on demonizing. In the following, we present our line of reasoning in more detail, and infer our hypothesis.

Impairment of Cognitive Resources Influences Demonizing

Forming an impression of a perpetrator’s evilness can be a cognitively demanding task. Next to dispositional information pertaining to prototypical evilness, often non-dispositional factors contributed to the occurrence of an offensive act, such as specific features of the crime, mental illness, peer pressure, social-economic factors, and the like. As such, it stands to reason that people need sufficient cognitive resources in order to accurately integrate all relevant pieces of information while establishing the amount of evilness that should be ascribed to the perpetrator. Evidence suggests, however, that people are more prone to rely on prototypical evilness information when evaluating a perpetrator, as the influence of prototypical evilness information on demonizing demands relatively less cognitive effort than consideration of non-dispositional factors. For instance, Miller, Gordon, and Buddie (1999) noted that people are reluctant to take situational factors into account when evaluating a perpetrator, because situational explanations are easily interpreted as attempts to condone the atrocities that were committed. Furthermore, Sargent (2004) found that people who avoid elaborate thought (i.e., low need for cognition) make less complex attributions for a perpetrator’s behavior, which is associated with increased punishment. It is likely that avoiding elaborate thought decreases the extent to which non-dispositional factors are incorporated in a perceivers’ evaluation of a perpetrator’s behavior, increasing the impact of dispositional evilness information. Combining these arguments would suggest that demonizing initially depends on the extent to which a perpetrator matches expectations of prototypical evilness, a first impression for which people can correct if they are able or willing to cognitively elaborate on the event. Hence, prototypical evilness information can be expected to exert a stronger influence on demonizing when perceivers’ cognitive resources are impaired.

This line of reasoning is related to research which revealed that the impairment of cognitive resources can have substantial consequences for person perception. A typical procedure to impair participants’ cognitive resources is to make them remember an eight-digit number during the evaluation of a target person. Such cognitive load has been found to increase the dispositional inferences that perceivers draw from behavior, because people are less able to adjust their inferences for information about situational constraints (Gilbert, Pelham, & Krull, 1988; Trope & Alfieri, 1997). Furthermore, cognitive load increases the application of existing stereotypes in a variety of situations (Gilbert & Hixon, 1991; Van Knippenberg, Dijksterhuis, & Vermeulen, 1999). Although stereotyping is not the same as demonizing—for instance, whereas stereotypes pertain to specific social categories, demonizing may occur in response to any disliked person, group, or institution—these previous findings are relevant to the present purposes by suggesting that impairing cognitive resources increases people’s tendency to rely on existing knowledge structures in their social evaluations.

Extrapolating these arguments to the present purposes, it can be inferred that the extent to which a perpetrator matches prototypical expectations of evilness shapes demonizing responses particularly under conditions of cognitive load. The present research tested this prediction in a series of two experiments. As such, the present experiments were designed to make a novel contribution by illuminating that the impact of prototypical evilness information on demonizing depends on perceivers’ ability to cognitively elaborate on the event. In both the experiments, participants were asked to remember a difficult or an easy telephone number (cognitive load vs. control). After this, they read a description of a severe offense (the brutal murder of a young woman in Experiment 1, and the kidnapping of a young child for ransom in Experiment 2). It was explicitly emphasized that the perpetrator had no previous criminal record. We then manipulated the extent to which the perpetrator’s characteristics were consistent or inconsistent with a prototypical evilness scheme. In the prototypically evil condition, participants read a composite description of the perpetrator suggesting a bad and prototypically evil person. In particular, the perpetrator was described as non-emotional during his arrest (i.e., lacking uniquely human emotions; Leyens et al., 2000), socially isolated (Baumeister, 1997), and with a reputation of behaving antisocially toward his direct social environment as evidenced by various annoying or even hurtful behaviors (Berkowitz, 1999). In the non-prototypically evil condition, the perpetrator was described as distressed during his arrest, as a family man (i.e., not socially isolated), with a reputation of being well-liked by his direct social environment. We predicted that the prototypical evilness manipulation would influence demonizing particularly under conditions of cognitive load (Hypothesis 1).

Experiment 1

Method

Participants and Design

We recruited 80 participants (31 men, 49 women, M age = 22.18, SD = 4.30) at VU University’s campus using flyers. They were assigned randomly to the conditions of a 2 (cognitive load vs. control) × 2 (prototypical evilness: prototypical vs. not prototypical) factorial design. The study lasted approximately 10 min, and participants were paid 2 Euros for participation.

Procedure

Participants were seated in separate cubicles behind computer equipment which was used to present the stimulus information and to register the data. The study was presented as a study on people’s capacities to do multiple tasks simultaneously. In the cognitive load condition, participants were asked to memorize a difficult telephone number: 06-35162974. In the control condition, participants were asked to memorize an easy telephone number: 06-11111111. Participants were informed that they would be asked to reproduce the number at the end of the study.Footnote 1

Participants then read the following description of the brutal murder of a 24-year old female:

Last Friday, a horrible murder was committed in Heemstede. A 24-year old female inhabitant of Heemstede was found dead Friday afternoon, around 3 PM, in an alley of the big market square. Forensic research indicated that the perpetrator first tried to strangle her, and then hit her on the head with a heavy metal object. It is as yet unknown whether the victim and the perpetrator were acquainted. The murder does not seem to be the result of a robbery. Shortly after the crime, the perpetrator was arrested. After it became apparent that the DNA that was found at the victim’s clothes matched the perpetrator’s DNA, the perpetrator, called “Marco G.”, confessed. The perpetrator refrained from commenting on his motives to commit the crime. Marco G. was unknown to the police and has no previous criminal record. He is 40 years old and worked long hours as a truck driver.

The prototypical evilness manipulation was induced by providing participants with background information about the offender. In the prototypically evil condition, participants read the following:

The perpetrator seemed exceptionally calm during his arrest. He was described by people from his neighborhood as a socially isolated individual who rarely left the house. He was considered as annoying by his neighbors, because he ignored continuing requests to reduce noise nuisance. In addition, there were large piles of garbage at his front door on a regular basis. Occasionally, children from the neighborhood who came too close to him were yelled at, and as a result, the children were terrified of him. Recently, one of the children’s football accidentally ended up in Marco G’s garden. When the child tried to get the ball back, Marco G scared the living daylights out of the child by sneaking up on him from behind, and then chasing the child away from his garden while swearing, with a huge smile on his face.

In the non-prototypically evil condition, participants read the following description:

The perpetrator seemed exceptionally upset during his arrest. He was described by people from his neighborhood as a quiet person and a real family man. He is married and has two children, a daughter of 6 years old and a son of 4 years old. He and his wife recently celebrated their 12.5 year wedding anniversary.Footnote 2 Many people from the neighborhood attended the celebration, and expressed how enjoyable this event had been. His neighbors said that, the day before the murder, they had talked to Marco G. in his garden. Marco G. had said how much he was looking forward to take the caravan out of his garage and to go camping outside during the weekends.

We then measured demonizing by assessing participants’ agreement to the following five items (1 = strongly disagree, 7 = strongly agree): “This crime was caused entirely by the offender’s evilness”, “The offender is only motivated to destroy everything that is benevolent”, “The offender seems to enjoy hurting others”, “The offender is immoral”, and “When thinking of the offender, I can only imagine how mean he is”. These items were averaged into a reliable demonizing scale (α = .90). Participants were then asked to reproduce the telephone number. After this, the experiment ended, and participants were debriefed, thanked, and paid.

Results

Telephone Number Recall

The majority of participants were able to accurately recall the telephone number that they were supposed to memorize. In the control condition, 97% accurately recalled the telephone number. In the cognitive load condition, 71% accurately recalled the telephone number, which was significantly less than in the control condition, χ2(1) = 10.18, P < .01. It is further noteworthy that in the cognitive load condition, 8% recalled at least 6 digits and 7% recalled all 8 digits but not in the correct order. These results suggest that participants made a serious effort to hold the telephone number in memory while completing the experiment, and that as intended, the memory task was more difficult in the cognitive load condition than in the control condition.

Demonizing

Means and standard deviations are displayed in Table 1. A 2 (cognitive load) × 2 (prototypical evilness) ANOVA revealed a significant main effect of prototypical evilness, F(1, 76) = 7.74, P < .01, ω 2 = .08, indicating that participants demonized the perpetrator more in the prototypically evil condition (M = 4.23, SD = 1.34) than in the non-prototypically evil condition (M = 3.38, SD = 1.48). This main effect was qualified by the predicted interaction, F(1, 76) = 4.15, P < .05, ω 2 = .04. As expected, the simple main effect of prototypical evilness was significant in the cognitive load condition, F(1, 76) = 12.18, P < .002, ω 2 = .12, but it was non-significant in the control condition, F < 1. These findings support the prediction that cognitive load increases the impact of prototypical evilness information on demonizing.

Table 1 Demonizing as a function of cognitive load and prototypical evilness—Experiment 1

Discussion

The findings of Experiment 1 support the prediction that the impairment of cognitive resources increases the impact of prototypical evilness information on demonizing. As such, the findings of Experiment 1 provide preliminary evidence for the psychological processes that we assumed to contribute to people’s tendency to demonize perpetrators of severe crimes. Although promising, it must be noted that the nature of the offense in Experiment 1 (murder) might provide alternative explanations for the present findings. In particular, perceivers might interpret murder as a rather impulsive act, and as a consequence, they could make different attributions for the prototypically versus non-prototypically evil offender’s motives (e.g., the murder might be interpreted as an act of sadism in the case of the prototypically evil offender, but it might be interpreted as temporary insanity in the case of the non-prototypically evil offender). It was, therefore, important to replicate these results for an offense that had to be deliberately planned in advance. Hence, we conducted a second experiment in which we tested the main hypothesis of the present contribution in the context of kidnapping a 10-year old child for ransom.

Experiment 2

Method

Participants and Design

The experimental design was the same as Experiment 1. We recruited 84 participants by means of flyers in the students’ cafeteria’s of VU University Amsterdam (28 men, 56 women, M age = 20.32, SD = 1.72). The study was preceded and followed by studies of other researchers on topics that were unrelated to the present purposes. Together the studies lasted approximately 45 min, and participants were paid 5 Euros for participation.

Procedure

The study was conducted in the same laboratory as Experiment 1. The manipulation of cognitive load was the same as Experiment 1. After this, participants read the following description of the kidnapping of a 10-year old child for ransom:

A 10-year old girl (the victim) is being kidnapped by Marco G. (the perpetrator). The victim has exceptionally wealthy parents, and Marco G. demands a high sum of money for ransom. The 10-year old victim is held hostage by the kidnapper for two consecutive weeks. During these two weeks, the 10-year old victim is locked up in a very small room. After two weeks, the police finds a clue that leads to the perpetrator. The police performs a raid in the house in which the victim is held hostage. The victim is not harmed physically, but emotionally the young girl is severely traumatized. Marco G., who is caught in the act, confesses to the crime. Marco G. was unknown to the police and has no previous criminal record. He is 40 years old and worked long hours as a truck driver.

The same prototypical evilness manipulation as Experiment 1 was then induced, followed by the same measure of demonizing (α = .83). After this, participants were again asked to recall the telephone number, after which they were debriefed, thanked, and paid for their participation.

Results

Telephone Number Recall

All participants in the control condition (100%) accurately recalled the telephone number. In the cognitive load condition, 74% accurately recalled the telephone number, which was significantly less than in the control condition, χ2(1) = 12.07, P < .01. Furthermore, in the cognitive load condition 14% of the participants recalled at least 6 digits and 5% recalled all 8 digits but not in the correct order. These results correspond to the findings of Experiment 1, and further underscore that the cognitive load manipulation is successful in making participants actively try to memorize the easy number in the control condition and the relatively more difficult number in the cognitive load condition.

Demonizing

Means and standard deviations are displayed in Table 2. A 2 (cognitive load) × 2 (prototypical evilness) ANOVA revealed that participants demonized the perpetrator more in the prototypical condition (M = 4.28, SD = 0.98) than in the non-prototypical condition (M = 3.12, SD = 1.09), as evidenced by a significant prototypical evilness main effect, F(1, 80) = 27.80, P < .001, ω 2 = .24. More important was that the predicted interaction replicated, F(1, 80) = 4.21, P < .05, ω 2 = .04. Again, the effect of prototypical evilness was stronger in the cognitive load condition, F(1, 80) = 26.04, P < .001, ω 2 = .23, than in the control condition, F(1, 80) = 5.20, P < .03, ω 2 = .05. These findings provide further support for the hypothesis.

Table 2 Demonizing as a function of cognitive load and prototypical evilness—Experiment 2

Discussion

The results again corroborated the prediction that the impairment of cognitive resources, by means of a cognitive load manipulation, increases the impact of prototypical versus non-prototypical evilness information on demonizing. As such, it can be concluded that the two experiments presented here provide good support for the psychological processes underlying demonizing that we hypothesized in the introduction.

When comparing the means in Table 1 with the means in Table 2, it is noteworthy that in Experiment 1 the induction of cognitive load was particularly associated with more demonizing in the prototypically evil condition, but in Experiment 2 the induction of cognitive load was particularly associated with less demonizing in the non-prototypically evil condition. In all likelihood, this difference can be explained by considering the attributions that people are likely to make for the two different offenses once given the opportunity to cognitively elaborate on the event (i.e., the control condition). As noted before, the offense in Experiment 1 might be interpreted as a relatively impulsive act. It is likely that this impulsiveness interpretation stimulated perceivers to mitigate the demonizing of a prototypically evil offender if they were provided with the opportunity to cognitively elaborate on the event (e.g., “Perhaps the offender has mental or social problems, leading him to be unable to control his violent impulses”). Such discounting is less likely in the non-prototypically evil condition, where attributions of impulsiveness or temporary insanity only reinforce the first impression of the offender as lacking various evilness cues (e.g., although such attributions do not necessarily reflect favorably on the offender, they do suggest that the offender has typically human emotions).

The offense in Experiment 2, however, had to be carefully planned and executed. This might have led perceivers to exacerbate demonizing of an offender who does not fit the evilness prototype if they were given the opportunity to cognitively elaborate on the event (e.g., “He may not be a ‘typical’ offender, but he did intentionally prepare and conduct the kidnapping of a child”). When under cognitive load people are less capable of making such an elaborated judgment in this condition. Given that the non-prototypically evil description is not entirely “neutral” (i.e., the offender has characteristics that we normally ascribe to a good citizen), people’s first impressions are relatively more likely based on the implicit assumption that the offender is worthy of the benefit of the doubt. These considerations do not apply to the condition where the offender was prototypically evil, as the intentional nature of the crime is consistent with the evilness cues that are ascribed to this offender. Although speculative, these considerations suggest that impairing cognitive resources has the potential of influencing demonizing responses to both prototypically and non-prototypically evil offenders, depending on the level of intentionality that people ascribe to the perpetrator. More research is needed to fully clarify this issue. Importantly, these considerations hold no implications for the main conclusion of the present contribution, which is that variations in the extent to which an offender is congruent with expectations of prototypical evilness influences demonizing particularly when cognitive resources are impaired.

General Discussion

Two experiments support the hypothesis in the context of the murder of a young woman (Experiment 1) and the kidnapping of a 10-year old child (Experiment 2). Building on an integration of theoretical insights on demonizing (Baumeister, 1997; Berkowitz, 1999; Miller et al., 1999) with empirical research on dispositional inference (Gilbert et al., 1988; Trope & Alfieri, 1997) and stereotyping (Gilbert & Hixon, 1991; Van Knippenberg et al., 1999), the present findings are consistent with the theoretical notion that the extent to which a description of a perpetrator is consistent with mental representations of prototypical evilness contributes to demonizing responses, and that these existing knowledge structures exert a particularly strong impact on demonizing if people’s cognitive resources are impaired.

Although the present hypothesis was partly derived from knowledge on dispositional inference and stereotyping, it is important to note that the processes described here are conceptually distinct from these theoretical domains. In particular, previous research indicated that cognitive load increases the tendency to infer traits from behavior (dispositional inference; Gilbert et al., 1988) as well as the tendency to base evaluations on social categories (stereotyping; Van Knippenberg et al., 1999). In the present research, however, both the behaviors that were displayed (murder, kidnapping) as well as the social categories that were associated with these behaviors (criminal offenders) were unambiguously negative in all conditions. Hence, based on a dispositional inference or stereotyping framework alone, one would expect a main effect of cognitive load increasing demonizing compared to the control condition. Our findings indicated no such main effect; instead, cognitive load moderated the impact of prototypical evilness information, which can be explained by the assumption that demonizing responses stem from the extent to which perpetrators are consistent with mental schemes of prototypical evilness. The most plausible account of the present findings is provided by the integrative theoretical framework that we laid out in the introduction, which is focused specifically on the psychological dynamics associated with demonizing.

An additional contribution of the present research is that our demonizing measure may provide researchers with a useful tool to study perceivers’ destructive responses to severe offenses in various ways. First, it stands to reason that high levels of demonizing is not only predictive of high levels of punishment, but also with support for relatively more controversial punitive measures (e.g., the death penalty; torture). Once demonized, an offender is likely to be placed outside of the “scope of justice” by many perceivers, which legitimizes more extreme forms of punishment (Opotow, 1990). Second, research on the psychology of punishment (i.e., retributive justice) typically investigates participants’ responses to moderate offenses, because severe offenses easily lead to ceiling effects when asking for punishment recommendations (Rucker, Polifroni, Tetlock, & Scott, 2004; Van Prooijen, 2006; Van Prooijen & Lam, 2007). This problem is methodological in nature, and may obscure the influence of various social-psychological factors on destructive responses to severe offenses. It is likely that demonizing responses are less sensitive to this ceiling-effect problem, as people may be relatively less inclined to believe that an offender is purely evil than to impose maximum punishment on offenders of severe crimes. Hence, a focus on demonizing may stimulate research on people’s destructive responses to transgressions by broadening the scope of offenses that can be subjected to empirical scrutiny.

The present studies focused specifically on the causal influence of prototypical evilness information on demonizing. In future research, it might also be interesting to examine what factors lead people to make assumptions about the extent to which an offender is consistent with a prototypical evilness scheme. In everyday life, people often do not receive personalized information about evildoers (e.g., the identity of a perpetrator is not always known, given that perpetrators are not always caught), but it is nevertheless likely that people draw a mental picture of what kind of person the perpetrator might be. For instance, one might speculate that relatively more severe crimes lead people to ascribe relatively more “evil” character attributes to the perpetrator (e.g., socially isolated, remorseless). In addition, it might be predicted that people are more likely to ascribe evil character attributes to perpetrators to the extent that they feel emotionally closer to the victim, given that emotional closeness increases sympathy with the victim and stimulates a desire for sense-making of the offender’s behavior (Loewenstein & Small, 2007; see also Van Prooijen, 2010). These ideas suggest that demonizing, in combination with prototypical evilness, may provide a fruitful field for further study that is informative about broader psychological processes that are associated with dehumanization of offenders.

It is important to note that we do not claim that there is one universal evilness prototype that is applied to all offenders. Rather, evilness prototypes are composed of several cues that people tend to associate with evildoers (e.g., social isolation, behaving antisocially in human interaction, and lacking a variety of typically human emotions), but how these cues are weighted in people’s judgments may vary widely across individuals and the type of offense committed. Notably, various crimes may lead people to emphasize different evilness cues. For instance, people may not necessarily perceive a religiously inspired terrorist as socially isolated, but may nevertheless perceive such an offender as very evil due to a belief in this person’s lack of desirable human emotions (e.g., empathy, remorse). At the same time, a mentally retarded criminal may conform to many characteristics that one would normally associate with evilness, and yet people may not necessarily demonize such a person due to a lack of accountability. Hence, there may be multiple evilness cues that have context-specific effects depending on the details of the offense. These more complex dynamics may imply that the effects of cognitive load on demonizing depend on what specific cues are associated with evilness in the context of a specific offense.

The present findings may hold implications for how people make sense of criminal offenders in everyday life. Lay people usually are confronted with criminal offenders only briefly through common media (e.g., TV, internet), and receive only limited information about the offender’s characteristics, actions, and motives. As such, based on the present findings it might be reasoned that a first impression of prototypical (or non-prototypical) evilness is likely to have a strong impact on demonizing, particularly among people who take notice of the offender while simultaneously dealing with other demands and pressures of everyday life. Such demonizing responses may subsequently have an impact on negative emotions, collective action, and public opinion regarding appropriate punishment severity. Having said this, it might also be inferred that prototypical evilness information is much less likely to influence legal professionals who are actively involved in the court trial of a perpetrator. After all, legal professionals are required to carefully examine all factors that are relevant to the perpetrator’s actions, leading them to use all their cognitive resources to investigate the crime. Indeed, it might be the case that these differences in the limited versus full use of cognitive resources may explain why there sometimes is a discrepancy between the general public and legal professionals in justice-based responses to criminal offenders, potentially producing tension in society by means of a public desire for more severe punishment regulations (cf. De Keijser, Van Koppen, & Elffers, 2007). These latter considerations are highly speculative, of course, but they do illuminate that the processes described in the present article may be relevant to understand a variety of human reactions to perpetrators of severe crimes.

To conclude, whereas demonizing is a frequently occurring response to offenders, understanding of this phenomenon hitherto has been a largely theoretical enterprise. The present research sought to shed empirical light on the question which psychological processes produce demonizing of offenders. The findings reported here reveal that demonizing is associated with how well the offender corresponds to a prototypical evilness scheme, and that perceivers are most likely to rely on this evilness scheme when their cognitive resources are impaired. It is yet unclear how these evilness schemes originate in the minds of people; in all likelihood, they are a product of socialization, and are reinforced by both real (e.g., dictators) and fictitious (e.g. Hollywood films) exemplars. Regardless of its origins, people seem to have a relatively concrete idea of the attributes of evildoers, and these evilness schemes may substantially influence the way that they judge specific offenders. The present findings may provide a starting point toward a more solid empirical basis for an integrative theoretical model aimed at explaining why people sometimes perceive offenders as pure evil.