Reactions to offenders: Psychological differences between beliefs versus punishment.

In the present research, we examined a discrepancy between people's beliefs about, versus punitive reactions towards, offenders. Particularly, appraisals of offenders along the dimension of communion (i.e., being friendly or trustworthy) should primarily affect people's beliefs about them, as reflected in demonizing and conspiracy theories, and to a lesser extent observers' punitive reactions. However, actual evidence of transgression should (more strongly than beliefs) influence observers' punitive reactions. In two studies, we manipulated communion and transgression ambiguity in the context of financial offences. The transgression was presented as either an observable and clear-cut immoral case (non-ambiguous transgression) or as a case that involves a grey area between what is legal or illegal (ambiguous transgression). Study 1 revealed that viewing an offender as low (as opposed to high) in communion predominantly influenced demonization and conspiracy beliefs involving the offender. The transgression manipulation, however, mostly influenced observers' punitive reactions and their underlying punitive motives. Similar findings were obtained in Study 2. We conclude that although beliefs about offenders and punitive reactions are often strongly related, they are actually grounded in different psychological processes.

Background

People are often confronted with descriptions of offenders who committed criminal offences. Such acts prompt a range of negative beliefs and behavioural responses among observers. Common negative beliefs towards these offenders include demonization, that is, impressions of them as pure evil (Baumeister, 1997; Darley, 1992). Furthermore, people may believe conspiracy theories that assume the offender to be part of a larger corrupt network (Douglas, Sutton, & Cichocka, 2017; Van Prooijen & Van Vugt, 2018). A common behavioural response to criminal offenders, however, is to punish them (e.g., Carlsmith, Darley, & Robinson, 2002). What drives these beliefs versus punitive responses? In the present contribution, we propose that while these various responses towards offenders often appear closely aligned, they are actually rooted in related but distinct psychological antecedents. Specifically, we argue that people's beliefs about offenders are predominantly driven by their expectations (along the person perception dimension of communion; Abele, 2003; Abele & Wojciszke, 2007; Abele & Bruckmüller, 2011; Abele et al., 2016; Wojciszke, 2005), while punitive behaviours towards offenders are predominantly driven by the offender's observable rule‐breaking behaviour.

Various studies indeed support a relationship between beliefs about offenders and punitive action. For instance, providing descriptions of an offender as evil influences subsequent punitive judgements, suggesting a link between demonizing and punishment (Van Prooijen, Coffeng, & Vermeer, 2014). Likewise, conspiracy beliefs often involve allegations of corruption and other criminal behaviour that is subject to legal punishment (e.g., Van Prooijen & Acker, 2015). But do these relationships imply that beliefs about offenders and punishment of offenders are grounded in similar psychological processes? People organize their beliefs about others based on their expectations of them (Abele et al., 2016; Brambilla, Rusconi, Sacchi, & Cherubini, 2011; Fiske, Cuddy, & Glick, 2007; Wojciszke, 2005). Seeing one as cold, unfriendly, untrustworthy, or impolite might shape people's expectations or subjective assumptions of covert activities of a target person. Viewing possible offenders as low in these ‘communion’ traits therefore would affect people's beliefs about them as being inherently evil (Baumeister, 1997) and assumptions of conspiring in secret with like‐minded others to commit crimes. But independent of such negative beliefs, we argue that for punitive action people first require hard evidence of actual rule‐breaking. Apart from punitive intentions, we also explore observers’ motives for punishing offenders, which can provide information about the goals that people wish to achieve through their punitive actions. In the following, we illuminate the different antecedents of beliefs about, versus punitive action towards, offenders.

Beliefs about offenders

When people form an overall impression about others, they are often confronted with a plethora of information comprising people traits and behaviours (Neuberg & Fiske, 1987; Peeters, 1983; Reeder & Brewer, 1979; Skowronski & Carlston, 1987). The research literature distinguishes two core dimensions that underlie person perception, namely agency and communion (also referred to as warmth and competence or social vs. intellectual goodness–badness; Bakan, 1966; Cuddy, Fiske, & Glick, 2008; Fiske, Cuddy, Glick, & Xu, 2002; Rosenberg, Nelson, & Vivekananthan, 1968). Agency, on the one hand, refers to task functioning, is related to ‘getting ahead’, and involves traits such as ambition, confidence, efficiency, and capability. Communion, on the other hand, pertains to functioning in social relations and to ‘getting along’ and involves qualities such as friendliness, kindness, trustworthiness good intention, honesty, and fairness (Abele, 2003; Abele & Bruckmüller, 2011; Abele & Wojciszke, 2007; Abele et al., 2016; Wojciszke, 2005). The communion dimension is of primary importance in person perception, is weighted more heavily in social judgements, is processed in earlier stages of information processing, and leads to faster recognition of others’ intentions than the agency dimension (Abele & Bruckmüller, 2011; Wojciszke, Baryla, Parzuchowski, Szymkow, & Abele, 2011; Ybarra, Chan, & Park, 2001). Such primacy of the communion dimension occurs because it includes other‐profitable traits (i.e., traits that have a direct bearing on the well‐being of other people), whereas agency includes self‐profitable traits and have a direct bearing for the trait possessors themselves (Peeters, 1983, 2001; Peeters & Czapinski, 1990).

We propose that the dimension of communion predicts a range of beliefs that people typically have of offenders. Specifically, low communion of an offender would increase demonizing of offenders, and conspiracy beliefs. People scoring low in communion are viewed as malevolent, mal‐intentioned, incapable of reform, but also distant, cold, and unaffectionate (Giner‐Sorolla, Leidner, & Castano, 2011). These appraisals fuel impressions of the offender as evil non‐human beings (Leyens et al., 2000; Li, Leidner, & Castano, 2014) with anti‐social intentions. Demonizing (Berkowitz, 1999) is a special, dehumanizing kind of moral characterization that identifies others as pure evil (see Reicher, Haslam, & Rath, 2008). Demonizing specifically makes assumptions of an interrelated set of typically evil character traits, such as lacking human emotions, being socially isolated, and taking pleasure into hurting others. As such, demonizing ascribes diabolical attributes to target people or groups (Van Prooijen & Van de Veer, 2010).

Closely associated with demonizing is conspiracy beliefs, defined as explanatory beliefs that involve a number of actors who join together in secret agreement and try to achieve hidden goals that are perceived as unlawful or malevolent (Zonis & Joseph, 1994). Put differently, perceiving a target person as evil increases the likelihood of perceiving this target person as part of an evil group that secretly colludes to commit harm. Conspiracy beliefs are moderately related with, but conceptually distinct from, distrust (Goertzel, 1994; Van Prooijen & De Vries, 2016). Although distrust refers to an abstract aversive feeling towards a person or group, conspiracy beliefs constitute a specific allegation of misconduct enacted by a different, often powerful group.

While empirical research on conspiracy theories mostly focuses on political or corporate conspiracy theories (for overviews, see Douglas et al., 2017; Van Prooijen & Van Vugt, 2018), it stands to reason that people also perceive suspected offenders as part of conspiracies if they are perceived as low in communion. The role of communion is underscored by research that connects conspiracy beliefs with a sense of threat (Newheiser, Farias, & Tausch, 2011; Whitson & Galinsky, 2008) and that shows a causal effect of perceived immorality on conspiracy beliefs (Van Prooijen & Jostmann, 2013). In sum, we hypothesize that information that a target offender lacks communion traits increases conspiracy beliefs that include the offender as part of a corrupt network (Hypothesis 1). Furthermore, we expect that the relationship between communion and conspiracy beliefs is mediated by demonizing of the offender (Hypothesis 2; see Figure 1 for the mediation model).

Figure 1

Diagram of the mediation model.

We further argue, however, that the effects of communion are more prominent on these beliefs about offenders than on their desire to actually punish them. Certainly, information about communion reveals one's intentions (see Abele et al., 2016) and has diagnostic value for one's future actions towards society (Peeters, 1983, 2001), and perceivers might harm targets low in communion or help targets high in communion (Wojciszke, 2005, see also Cuddy et al., 2008 for a review). While we therefore do not exclude the possibility that communion appraisals increase punitive reactions, such appraisals rest on assumptions of other's motives, intentions, and possible future actions. Communion appraisals lack the concrete evidence of rule‐breaking necessary to justify punishment. Put differently, communion exerts stronger effects on beliefs about offenders (i.e., demonizing and conspiracy beliefs) and weaker on punitive action; but likewise, in the following we argue that transgression ambiguity exerts stronger effects on punitive action than on beliefs about offenders.

Transgressions and motives for punishment

Although information about communion might trigger punitive reactions under certain conditions, a common community norm associated with punishment is that no person should be punished unless proven guilty. Indeed, evidence suggests that the most impactful factor influencing punishment is evidence of actual rule‐breaking (Dworkin & Blumenfeld, 1966; Taylor & Hosch, 2004; for an overview, see Van Prooijen, 2018). In the present research, we will compare conditions where a protagonist commits morally dubious behaviour that is either illegal or arguably a legal grey area, and test whether this manipulation shapes punishment independent of communion. Moreover, we examine how rule‐breaking behaviour influences different (i.e., retributive vs. utilitarian) motives for punishment.

The legal system of most nations is committed to the notion that only acts that are illegal should be punished (Green, 2014). By labelling certain conduct as ‘criminal’ and punishing accordingly, our legal system regulates behaviour and stimulates compliance to the law. In many cases, however, there is considerable ambiguity if a person's behaviour was actually unlawful (e.g., fraud vs. creative accounting, tax evasion vs. tax avoidance, and bribe vs. campaign contribution; Braithwaite, 1991; Friedrichs, 2010; Sutherland, 1940; Shapiro, 1990). Although such ambiguous transgressions are not encouraged and often are frowned upon by the public, our system sanctions transgressions only if a clearly articulated rationale can be provided (Green, 2014). In this study, we therefore hypothesize that observers will display stronger intent to punish offenders who carried out a legally unambiguous as opposed to ambiguous transgression. Of importance, we expect this effect to be stronger than, and occur independent of any possible effect of communion (Hypothesis 3).

Besides punishment intentions, we also considered the differential motives that punishers display when assigning sanctions. The empirical and philosophical literature distinguishes between two main types of motives for punishment: utilitarian motives (Bentham, 1789), and retributive or just deserts motives (Kant, 1797). Each of these motives serves different goals and illustrates whether people aim to prevent future harm‐doing or punish past harm‐doing.

Utilitarian motives for punishment aim to reduce the likelihood of future offences, and thus maximize happiness and minimize suffering (see Carlsmith & Darley, 2008; Nagin, 1998; Van Prooijen, 2018). The purpose of utilitarian punishments is to control the offenders’ behaviour through deterrence of future crimes, incapacitation of a known liability to society, or restoration/rehabilitation of offenders (Carlsmith & Darley, 2008). In contrast to utilitarian motives, retributive/just deserts punishments are based on the moral philosophy of deontology according to which punishment must be proportionate to the harm inflicted. Retributive punishment motives are heavily fuelled by moral emotions (e.g., anger, disgust), and its objective is not preventing future offences per se, but retaliating for perpetrators’ past behaviour (Goldberg, Lerner, & Tetlock, 1999; see also Van Prooijen, 2018). We expected that independent of communion information, people would display both utilitarian and retributive punishment against an offender who was involved in an observable and clear‐cut (vs. ambiguous) offence. Specifically, perceivers may justify their punishment intention against an offender through both the moral imperative to sanction the already committed offence (retributive motives) but also by preventing such transgressions in the future (utilitarian motives).

In sum, the current study seeks to test the broader idea that beliefs about, versus punishments of offenders – while often related in practice – have different antecedents, such that beliefs about offenders primarily depend on the person perception dimension of communion (and to a lower extent on transgression ambiguity), while punishments and punitive motives primarily depend on transgression ambiguity (and to a lower extent on the communion dimension). One may speculate that information about an offender's agency (which may reveal an offender's capability to plan and perform a malevolent course of action) might also affect people's judgements of, and punitive reactions against, the offender. For explorative reasons, and in order to establish a complete design that takes both person perception dimensions into account, we also manipulated agency in Study 1.

We conducted two experimental studies to test our hypotheses. Study 1 examined the effect of communion, agency, and transgression (non‐ambiguous vs. ambiguous) on observers’ (1) beliefs about the offenders (demonization and conspiracy beliefs) and (2) punitive reactions. Study 2 aimed to replicate Study 1, but focused only on communion given that as predicted, it was a more potent predictor of beliefs about offenders than agency (e.g., Abele et al., 2016; Cuddy et al., 2008; De Bruin & Van Lange, 2000; Vonk, 1996; Wojciszke, 2005; Wojciszke & Abele, 2008; Wojciszke, Bazinska, & Jaworski, 1998).

STUDY 1

Method

Participants

A total of 211 British participants (150 females, 60 males, one unknown; M age = 37.7, SD = 11.96) living in the United Kingdom took part in this study. Participants came from various education levels (21.9% high school, 60.5% bachelor degree, 14.8% master's degree, 2.4% PhD, 0.5% other). The majority of participants were occupied as employees (59.5%). 23.8% were unemployed and the rest were students (8.6%) or worked in managerial positions (8.1%). A sensitivity analysis (G*power) for a fixed‐effect ANOVA with eight conditions reveals that the current sample gives 95% power to detect a medium effect size (Cohen's f = 0.25). Participants were recruited via Prolific academic and were paid £ 0.90 (€1.00) for their participation. The studies reported in this article were programmed in Qualtrics such that each IP address could participate only once.

Experimental design and procedure

We designed a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) × 2 (offender's agency: high/low) between‐subjects experiment. Participants were asked to read a vignette (in English), which presented a case where an entrepreneur carried out a financial offence (tax evasion). The offence was either ambiguous (‘…After investigation the case was found to involve a grey line between legal and illegal taxation activities…’) or non‐ambiguous (‘…After investigation the case was a clearly tax‐evasion activity…’). In order to keep the group identity of the entrepreneur identical with the one of participants (British), the offence was presented as having taken place in the United Kingdom.

Respondents also received information about both facets of communion and agency of the entrepreneur based on the description of these facets in Abele et al. (2016). High versus low communion information: ‘…Many people [do not] see the entrepreneur as a well‐intentioned and [or] trustworthy person’ ‘…They would describe him as a very warm and generous person who likes socializing with others [They would describe him as very cold, selfish, and distant’]. High versus low agency information: ‘….The entrepreneur is presented by his surroundings as [not] a very competent…’… and [or] confident person. He is [not] capable of planning and [or] implementing very difficult tasks [difficult tasks]. His intelligence and skills are remarkable [questionable]. The selection of communion and agency characteristics was based on Abele et al. (2016) description of the ‘Big two’ person perception dimensions and on the scale of stereotype content (Fiske et al., 2002). All materials are in the Appendix S1. Respondents were assigned randomly to the conditions.

Three items served as manipulation checks for ambiguity of transgression: ‘The entrepreneur was involved in a clearly illegal tax evasion activity’, ‘there is a grey line whether the activity the entrepreneur was involved in is a legal or illegal activity’ (recoded), and ‘it is uncertain whether the entrepreneur's actions are legal’ (recoded) (1 = absolutely disagree, 7 = absolutely agree; α = .88). Two items served as manipulation checks for the entrepreneur's communion and agency, respectively: ‘How warm do people find the entrepreneur as a person?’ and ‘How competent/intelligent do people find the entrepreneur?’ (1 = not at all, 7 = very much). The measurement of conspiracy beliefs, demonization, and punishment variables followed. In order to avoid priming the outcome variable (conspiracy beliefs) with the demonization items (mediator), we measured conspiracy beliefs first. The study was anonymous and participation was voluntary. Participants were debriefed and thanked after filling in the questionnaire.

Dependent variables

Conspiracy beliefs

We developed a 5‐item scale of conspiracy beliefs about the offender, which included the following items: ‘The entrepreneur did not act alone; he colluded with key‐persons in evading to pay taxes’, ‘the British company has a hidden agenda that aims to help its powerful members evade paying taxes’, ‘the entrepreneur conspired with other powerful people in evading to pay taxes’, ‘a secret conspiracy helped the entrepreneur carry out this act’, and ‘the entrepreneur designed the tax evasion plan in secret together with persons in key‐positions in the British government’ (α = .84).

Demonization

We used the 5‐item demonization scale by Van Prooijen and Van de Veer (2010) for the assessment of demonization of the offender. We adjusted the scale to the specifics of this study (e.g., ‘this act was caused entirely by the entrepreneur's evilness’, ‘the entrepreneur seems to enjoy committing fraudulent behaviors’; α = .85).

Punishment intentions

A 3‐item scale assessed observers’ intentions to punish the offender (‘the entrepreneur deserves to be punished’, ‘the entrepreneur should receive a sanction for his act’, and ‘it is right that the entrepreneur be sanctioned by the responsible authority’; α = .91).

Motives for punishment

We developed a 16‐item scale to assess the various motives for punishment, including (1) utilitarian motives and its sub‐dimensions, namely, deterrence (e.g., ‘a very severe punishment should be assigned to the entrepreneur so that he doesn't repeat this behavior in the future’), incapacitation (e.g., ‘we must protect society from tax evasion; the entrepreneur must be imposed several financial restrictions’), and restorative/rehabilitative motives for punishment (e.g., ‘the best way to address this sort of behavior is to offer the entrepreneur the chance to learn from his mistakes and reintegrate into society’); and (2) retributive motives (e.g., ‘the entrepreneur should be assigned a punishment equivalent to the magnitude of his transgression’). This scale was based on Carlsmith (2008) and Kugler et al. (2013) while a similar version of this scale was recently used in Fousiani and Demoulin (in press) and in Fousiani, Yzerbyt, Kteily, and Demoulin (2019). Cronbach's alpha was .91 for utilitarian motives, and .90 for retributive motives for punishment. The scale can be found in the Appendix S1. All dependent variables were measured on a 7‐point scale (1 = absolutely disagree, 7 = absolutely agree).

Results

Correlations between the study variables are presented in Table 1. These correlations indicate that beliefs (demonizing, conspiracy beliefs) and punitive intentions and motivations indeed are significantly correlated. Moreover, perceived warmth of the offender (manipulation check item) was moderately correlated with demonizing and conspiracy beliefs, suggesting related yet distinct construct. To further establish their conceptual distinction, we also conducted a principal component analysis with Varimax rotation on the items of these three constructs. The three factors that were extracted (perceived warmth, conspiracy beliefs, and demonization) explained 43.5%, 14.7%, and 9.8% of the variance, respectively, and all had an Eigenvalue higher than 1. With one exception,1 the items had high loadings (|f ij| > .50) on the predicted factor with no cross‐loadings. The results of the factor analysis confirm that the three constructs are distinct. The factor loading matrix is presented in Appendix S2. In the following, we will test whether beliefs and punishment are largely driven by different antecedents.

Table 1

Pearson correlation coefficients between study variables

	1	2	3	4	5	6	7	M (SD)
1. Conspiracy beliefs	1	.50***	.38***	.46***	.32***	−.19**	.08	3.64 (1.24)
2. Demonization		1	.46***	.58***	.44***	−.32***	.13	3.29 (1.25)
3. Punishment intentions			1	.66***	.70***	−.14*	.08	5.08 (1.47)
4. Utilitarian motives				1	72***	−.15*	.05	4.41 (1.18)
5. Retributive motives					1	−.15*	.17*	5.13 (1.32)
6. Offender's communion						1	.17*	3.92 (2.54)
7. Offender's agency							1	4.42 (2.30)

*p < .05; **p < .01; ***p < .001.

Manipulation checks

We analysed the manipulation checks of the perceived transgression ambiguity, communion, and agency manipulations by means of 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) × 2 (offender's competence: high/low) univariate analyses of variance (ANOVA). These analyses yielded a main effect of the transgression manipulation on perceived ambiguity of transgression, F(1,202) = 382.86, p < .001, η2 = .66. Participants perceived the transgression as clearer and less ambiguous in the non‐ambiguous condition (M = 5.56, SD = 1.39) than in the ambiguous condition (M = 2.29, SD = 1.02). The effects of both communion and agency manipulations on perceived ambiguity of transgression were non‐significant (Fs < 10). We also found a main effect of the communion manipulation on perceived communion as intended, F(1,202) = 529.40, p < .001, η2 = .72. As expected, participants perceived the offender as higher in communion in the high communion condition (M = 6.13, SD = 1.20) than in the low communion condition (M = 1.80, SD = 1.43). The effects of either transgression or agency manipulations on perceived communion were not significant (Fs < 1) as expected. Finally, the effect of agency manipulation on perceived agency also came out significant, as intended F(1,202) = 325.10, p < .001, η2 = .62. As expected, participants perceived the offender as higher in agency in the high agency condition (M = 6.20, SD = 1.23) than in the low agency condition (M = 2.62, SD = 1.63). As intended, the effects of neither communion nor transgression manipulations on perceived agency were significant (Fs < 1).

Unexpectedly, the interaction effect of transgression and agency on perceived agency of the offender was significant F(1,202) = 3.99, p = .047, η2 = .02. Participants in the high agency – non‐ambiguous condition perceived the offender as higher in agency (M = 6.32, SD = 1.16) than in the high agency – ambiguous condition (M = 6.11, SD = 1.28). In contrast, people in the low agency – non‐ambiguous condition (M = 2.24, SD = 1.55) perceived the offender as lower in agency than in the low agency – non‐ambiguous condition (M = 2.90, SD = 1.64). This unexpected effect was small, however, and no other significant main effects or interaction effects occurred (Fs < 1). In sum, the manipulation of transgression only affected perceived transgression, the manipulation of communion only affected perceived warmth, and finally, the manipulation of agency affected perceived agency. It can be concluded that participants perceived the manipulations as intended.

Conspiracy beliefs

Participants’ scores were submitted to a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) × 2 (offender's competence: high/low) univariate ANOVA with conspiracy beliefs as the main dependent variable. The main effect of communion on conspiracy beliefs was significant F(1,203) = 4.69, p = .03, η2 = .023. Observers reported stronger conspiracy beliefs when the offender was viewed as cold and immoral (low in communion) rather than warm and moral (high in communion) (Means and standard deviations in Table 2). The effects of agency F(1,203) = .17, p = .68, η2 = .001 or transgression ambiguity F(1,203) = .67, p = .42, η2 = .003 were not significant. Furthermore, all the interaction effects were not significant (Fs < 1). These results support Hypothesis 1.

Table 2

Means and standard deviations for the study variables across experimental conditions

	Non‐ambiguous transgression		Ambiguous transgression		High communion offender		Low communion offender		High agency offender		Low agency offender
	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
Conspiracy beliefs	3.58	1.27	3.69	1.21	3.45	1.27	3.83	1.18	3.60	1.24	3.69	1.25
Demonization	3.36	1.19	3.23	1.30	2.95	1.16	3.62	1.25	3.35	1.28	3.22	1.22
Punishment intentions	5.77	1.06	4.56	1.51	4.87	1.44	5.28	1.48	5.16	1.61	5.00	1.31
Utilitarian motives	4.80	1.06	4.13	1.18	4.29	1.21	4.53	1.13	4.35	1.25	4.48	1.10
Retributive motives	5.13	1.32	4.75	1.33	5.00	1.39	5.26	1.23	5.26	1.32	5.00	1.30

All ratings were on 7‐point scales ranging from 1 = absolutely disagree to 7 = absolutely agree.

Demonization

Similarly, we conducted a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) × 2 (offender's competence: high/low) univariate ANOVA with demonization as dependent variable. The main effect of communion on demonization of the offender was also significant F(1,203) = 17.14, p < .001, η2 = .08. Observers demonized offenders more when they are viewed as low rather than high in communion (Means and standard deviations in Table 2). The effects of transgression ambiguity F(1,203) = .61, p = .44, η2 = .003 or agency F(1,203) = 1.68, p = .20, η2 = .008 were not significant. Furthermore, all the interaction effects were not significant (Fs < 1). These findings underscore that the effects of communion on beliefs about offenders emerged independent of transgression ambiguity.

As a next step, we conducted a mediation analysis through bootstrapping with communion traits of the offender as the independent variable (effect‐coded −1: low, 1: high in communion), conspiracy beliefs as the dependent variable, and demonization as mediator. The total effect of communion on conspiracy beliefs was negative and significant: Observers reported stronger conspiracy beliefs when the offender was low as opposed to high in communion. When demonization of the offender was added as a mediator in the model, the indirect effect was significant but the direct effect was not, resulting in full mediation (see Table 3 for the relevant statistics). These results support Hypothesis 2.

Table 3

Mediation results with demonization as mediator

Effects of warmth on	Total effect	Direct effect (c΄)	Unstandardized paths		Indirect effect		Ratio of indirect to total effect
			a	b	Estimate	BCA CI
Conspiracy beliefs (Study 1)	−.20 (.08)*	−.03 (.08)	−.34 (.08) ***	.49 (.06) ***	−.16 (.05)	−.27 to .08	.86
Conspiracy beliefs (Study 2)	−.29 (.09) **	−.23 (.09) **	−.22 (.08) **	.30 (.07) ***	−.06 (.03)	−.13 to .01	.20

Standard errors in parentheses (bootstrap standard errors for the indirect effect estimate); BCA CI: bias‐corrected and accelerated bootstrap confidence interval; paths a and b correspond to the prediction coefficients of the independent variable to the mediator (path a) and of the mediator to the dependent variable (path b), see Figure 1.

*p < .05; **p < .01; ***p < .001.

Punishment intentions

A 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) × 2 (offender's agency: high/low) ANOVA was conducted with punishment intentions as the main dependent variable. As expected, the effect of type of transgression ambiguity on punishment intention was significant F(1,203) = 43.36, p < .001, η2 = .18. In line with our hypotheses, people have a stronger intent to punish offenders who committed non‐ambiguous as compared to ambiguous transgressions.

The effect of communion on punishment intention was also significant F(1,203) = 6.26, p = .01, η2 = .03 showing that people have a stronger intention to punish offenders that are low rather than high in communion. In order to further investigate the magnitude of the effect of communion on conspiracy beliefs and punishment intention, we also conducted an ANOVA with conspiracy beliefs and punishment intention as repeated measures. The interaction between the measure (conspiracy beliefs and punishment intention) and offender's communion was not significant (F < 1). Accordingly, we ran an ANOVA with repeated measures to test the effect of transgression ambiguity on both conspiracy beliefs and punishment intention. The interaction between transgression ambiguity and measure (conspiracy beliefs vs. punishment intention) came out significant F(1,203) = 48.89, p < .001, η2 = .19. Consistent with our line of reasoning, results revealed a stronger effect of transgression ambiguity on punishment intention than on conspiracy beliefs (see Table 2).

The effect of agency on punishment intention was not significant (F < 1). Finally, the interaction between transgression ambiguity and agency was marginally significant F(1,203) = 3.99, p = .047, η2 = .02. People have a stronger intention to punish an offender who has committed a non‐ambiguous transgression and is high rather than low in agency (M high = 6.07, SD = .98, M low = 5.49, SD = 1.07). The mean difference for high and low agentic offenders was not significant in the ambiguous transgression condition (M high = 4.50, SD = 1.65, M low = 4.61, SD = 1.36). No other interaction effects were significant (F < 1).

Punishment motives

We also conducted a multivariate analysis of variance with motives for punishment (utilitarian and retributive) as dependent variables. The multivariate effect of transgression ambiguity on this set of variables was significant F(2,202) = 13.46, p < .001, η2 = .12. In line with our predictions, observers display both utilitarian F(1,203) = 16.42, p < .001, η2 = .08 and retributive F(1,203) = 26.56, p < .001, η2 = .12 motives for punishing offenders who have committed non‐ambiguous offences. Furthermore, there was a marginally significant effect of agency on retributive motives for punishment F(1,203) = 3.91, p = .05, η2 = .02 showing that people assign more retributive punishments to high agentic offenders as compared to low agentic ones. The effect of agency on utilitarian motives F(1,203) = .17, p = .69, η2 = .001 was not significant (Means and standard deviations in Table 2).

Moreover, on punitive motives, neither the multivariate effects of communion (Fs < 1) nor the communion by offender ambiguity interaction were significant (Fs < 1) suggesting that the effects of transgression ambiguity on punitivity indeed are independent from communion, as predicted. These results are in line with Hypothesis 3.

Finally, the multivariate interaction effect between communion and agency on our punishment motive variables was marginally significant F(2,202) = 3.05, p = .05, η2 = .03. The univariate interaction effect showed that people assign more retributive punishments F(1,203) = 6.10, p = .014, η2 = .03 to offenders who are high in communion but also high (as opposed to low) in agency (M highagency = 5.34, SD = 1.28, M lowagency = 4.59, SD = 1.43). The mean difference for low in communion offenders was not significant (M highagency = 5.18, SD = 1.38; M lowagency = 5.34, SD = 1.09). The univariate effects of the communion by agency interaction on utilitarian motives were not significant (Fs < 1). No other interaction effects were significant (Fs < 1).

Discussion

Consistent with Hypothesis 1, results revealed that expectations of offenders – notably information about communion – strongly influenced observers’ beliefs about the offenders. Specifically, observers endorsed conspiracy beliefs when offenders were presented as low rather than high in communion. Consistent with Hypothesis 2, demonization of offenders mediated this relationship. These effects emerged independent from transgression ambiguity, which instead predicted punishment intentions and motivations, as predicted in Hypothesis 3. We also observed an effect of communion, but only on punitive intent (not on punitive motivations), independent from transgression ambiguity (as indicated by the non‐significant interaction), and with a weaker effect size than the transgression ambiguity main effect. From additional analyses, we found that although communion exerted an equally strong effect on conspiracy beliefs as on punitive intentions, transgression ambiguity exerted a stronger effect on punishment intention than on conspiracy beliefs. Taken together, these findings support the idea that beliefs about offenders are predominantly driven by communion perceptions, and punitivity is predominantly driven by transgression ambiguity.

In a more exploratory fashion, we also examined the effects of agency, which influenced punitive reactions of observers. This analysis revealed that people have a stronger intention to punish an offender who has committed a non‐ambiguous transgression and is high rather than low in agency. Moreover, people have stronger retributive motives to punish offenders who are both high in communion and in agency. However, the effects of agency were exclusive to retributive motivations and did not generalize to punitive intent or utilitarian motives.

A limitation of Study 1 is that we only included one single item for the manipulation check of communion, referring to the trait of warmth. According to Abele et al. (2016), communion is a two‐faceted dimension, including traits related to warmth and traits related to morality. While experimentally distinguishing between these two facets is beyond the purposes of the present paper, in Study 2 we address this issue by including one more item to assess the perceived morality of the offender. In Study 2, we sought to replicate and extend our main findings. Moreover, in Study 2, we focus exclusively on communion rather than agency, given that communion is a stronger predictor of impression formation process and overall perception of others (e.g., De Bruin & Van Lange, 2000; Vonk, 1996; Wojciszke, 2005; Wojciszke & Abele, 2008; Wojciszke et al., 1998), and was a key factor in stimulating beliefs of offenders.

STUDY 2

Methods

A total of 229 British participants (148 females, 81 males, one unknown; M age = 37.02, SD = 11.90) living in the United Kingdom took part in this study. Participants came from various education levels (24% high school, 62.4% bachelor degree, 10% master's degree, 2.2% PhD, 1.3% other). Moreover, the majority of participants were occupied as employees (62.4%), 18.8% were unemployed, and the rest were students (8.7%) or had managerial positions (10%). According to a sensitivity analysis (G*power) for a fixed‐effect ANOVA with four conditions, the present sample gives 95% power to detect a medium effect size (Cohen's f = 0.24). Participants were recruited via Prolific academic and were paid £ 0.90 (€1.00) for their participation.

Participants were randomly assigned to the conditions of a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) between‐subjects experiment. Similar to Study 1, respondents were asked to read a vignette (in English), which presented a case where an entrepreneur carried out a financial offence (money laundering). The offence was again either non‐ambiguous (‘…After investigation it became clear that the entrepreneur indeed “cleaned” money that was derived from illegal activities…’) or ambiguous (‘…After investigation it remains uncertain whether the entrepreneur actually “cleaned” his money as there is no evidence that this money was derived from illegal activities…’). As in Study 1, in order to keep the group identity of the entrepreneur identical with the one of participants (British), the offence was presented as having taken place in the United Kingdom. Participants were also given information about the entrepreneur's perceived communion, similar as in Study 1. All materials are in the Appendix S1. Three items served as manipulation checks of ambiguity of the transgression: ‘The entrepreneur clearly was involved in a money laundering activity’, ‘it is unclear whether the entrepreneur was involved in any illegal activities’ (recoded), and ‘it is uncertain whether the entrepreneur's action is illegal’ (recoded) (1 = absolutely disagree, 7 = absolutely agree; α = .94). Two items served as manipulation checks for entrepreneur's communion: ‘How warm do people find the entrepreneur as a person?’ and ‘How well‐intentioned do people find the entrepreneur as a person?’ (1 = not at all, 7 = very much; α = .96). The study was anonymous and participation was voluntary. Participants were debriefed and thanked after filling in the questionnaire.

For the assessment of conspiracy beliefs (α = .80), demonization (α = .85), punishment intentions (α = .96), utilitarian motives (α = .91), and retributive motives (α = .90), we used identical scales to the ones we used in Study 1. All scales were adjusted to the specifics of this study.

Results and discussion

Correlations between the variables are presented in Table 4. These correlations yielded a similar pattern as in Study 1, suggesting that our key constructs (e.g., communion, demonizing, conspiracy beliefs) are related but distinct constructs. We again conducted a principal component analysis with Varimax rotation. Three factors were extracted (perceived communion, conspiracy beliefs, and demonization; Eigenvalues >1) which explained 45%, 15%, and 10.7% of the variance, respectively. The items had high loadings (|f ij| > .50) on the predicted factors and no cross‐loadings, with one exception.2 The results of the factor analysis confirm that the three constructs are distinct. The factor loading matrix is presented in Appendix S3.

Table 4

Pearson correlation coefficients between study variables

	1	2	3	4	5	6	M (SD)
1. Conspiracy beliefs	1	.26***	.16**	.23***	.21**	−.25***	4.35 (1.36)
2. Demonization		1	.62***	.62***	.48***	−.24***	3.61 (1.19)
3. Punishment intentions			1	.70***	.66***	.03	4.98 (1.70)
4. Utilitarian motives				1	74***	−.19**	4.53 (1.21)
5. Retributive motives					1	−.15*	5.22 (1.40)
6. Offender's communion						1	4.07 (2.42)

*p < .05; **p < .01; ***p < .001.

We analysed the manipulation checks by means of 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) univariate ANOVAs with perceived transgression ambiguity and perceived communion as the main dependent variables. These analyses yielded a main effect of the transgression manipulation on perceived transgression ambiguity, F(1,225) = 256.42, p < .001, η2 = .53 and a main effect of the communion manipulation on perceived communion, F(1,225) = 772.32, p < .001, η2 = .77. Participants perceived the transgression as clearer and less ambiguous in the non‐ambiguous condition (M = 5.18, SD = 1.46) than in the ambiguous condition (M = 2.24, SD = 1.32). Moreover, participants perceived the offender as warmer and more moral in the high communion condition (M = 5.98, SD = 1.16) than in the low communion condition (M = 1.71, SD = 1.15). Neither the effect of communion manipulation on perceived transgression F(1,225) = 3.63, p > .05, η2 = .02 nor the effect of transgression manipulation on perceived communion were significant F(1,225) = 3.44, p < .05, η2 = .02. Finally, no significant interaction effects occurred (Fs < 1) ensuring that the manipulation of transgression only affected perceived transgression and the manipulation of communion only affected perceived communion of the offender. It can be concluded that participants perceived the manipulations as intended.

Participants’ scores were submitted to a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) ANOVA with conspiracy beliefs as the main dependent variable. The univariate effect of communion on conspiracy beliefs was significant F(1,225) = 10.46, p = .001, η2 = .04. Observers reported stronger conspiracy beliefs when the offender was perceived as low rather than high in communion. As expected, the effect of transgression ambiguity on conspiracy beliefs was not significant F(1,225) = .73, p = .39, η2 = .003 (Means and standard deviations in Table 5). Finally, no significant interactions between the variables were observed (Fs < 1), suggesting that the effects of communion on beliefs were independent from transgression. These findings provided full support for Hypothesis 1.

Table 5

Means and standard deviations for the study variables across experimental conditions

	Non‐ambiguous transgression		Ambiguous transgression		High communion offender		Low communion offender
	M	SD	M	SD	M	SD	M	SD
Conspiracy beliefs	4.26	1.35	4.47	1.37	4.09	1.27	4.67	1.40
Demonization	4.00	1.04	3.15	1.25	3.41	1.18	3.85	1.17
Punishment intentions	5.78	1.22	3.83	1.64	4.97	1.80	4.99	1.57
Utilitarian motives	4.93	.98	3.96	1.31	4.39	1.28	4.70	1.12
Retributive motives	5.61	1.14	4.66	1.55	5.10	1.48	5.38	1.28

All ratings were on 7‐point scales ranging from 1 = absolutely disagree to 7 = absolutely agree.

We then ran an ANOVA with demonization of the offender as the main dependent variable. The effect of communion on demonization was significant, F(1,225) = 12.80, p < .001, η2 = .05. Observers demonized offenders more strongly if they were viewed as low rather than high in communion. The main effect of transgression ambiguity on demonization was also significant F(1,225) = 31.53, p < .001, η2 = .12 (Means and standard deviations in Table 5). Finally, the interaction between the variables was not significant (Fs < 1), suggesting that the effects of communion on demonization were independent from transgression.

Next, we conducted a mediation analysis with communion trait of the offender as the independent variable (effect‐coded −1: low, 1: high), conspiracy beliefs as the dependent variable, and demonization as mediator. The total effect of communion on conspiracy beliefs was negative and significant: Observers reported stronger conspiracy beliefs when the offender was low as opposed to high in communion. When demonization of the offender was added as a mediator in the model, both the direct and indirect effects were significant, resulting in partial mediation (see Table 3 for the relevant statistics). In sum, the mediation hypothesis (Hypothesis 2) was again supported.

Participants’ scores were submitted to a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's warmth: high/low) ANOVA with punishment intentions as the main dependent variable. In line with our predictions, people have a stronger intention to punish offenders who have committed non‐ambiguous than ambiguous offences F(1,225) = 106.87, p < .001, η2 = .32. As expected, the effect of communion on punishment intent was not significant F(1,225) = 1.83, p = .18, η2 = .008 (Means and standard deviations in Table 5). Finally, the communion by transgression ambiguity interaction effect was not significant (Fs < 1).

Participants’ scores were submitted to a 2 (transgression ambiguity: ambiguous/non‐ambiguous) × 2 (offender's communion: high/low) MANOVA with motives for punishing the offender (utilitarian and retributive) as the main dependent variables. As expected, the multivariate effect of transgression ambiguity on punishment motive was significant F(2,224) = 23.41, p < .001, η2 = .17. In line with our predictions, people display stronger utilitarian F(1,225) = 45.62, p < .001, η2 = .17 and retributive F(1,225) = 30.61, p < .001, η2 = .12 motives for punishing offenders who have committed non‐ambiguous offences. The multivariate effect of communion on this set of variables was also significant F(2,224) = 4.43, p = .01, η2 = .04. Results revealed that people assign stronger utilitarian F(1,225) = 8.70, p = .004, η2 = .04 and retributive F(1,225) = 5.54, p = .02, η2 = .02 punishments to low as opposed to high in communion offenders (Means and standard deviations in Table 5). Finally, the communion by transgression ambiguity interaction effect was not significant (Fs < 1).

In order to disentangle the magnitude of the unexpected effects of communion on punishment motives and of transgression ambiguity on demonization, we conducted an ANOVA with demonization and punishment motives as repeated measures. The interaction between the measure (demonization and punishment motives) and offender's communion was not significant (F < 1). The interaction between the measure and transgression ambiguity was not significant either (F < 1).

These findings replicate Study 1 and further support our predictions. Expectations of offenders in terms of communion predominantly influence people's beliefs, that is demonizing and endorsement of conspiracy beliefs, while observable transgression notably influences punitive reactions. Unlike Study 1, we also observed an effect of transgression ambiguity on demonizing, and an effect of offender's perceived communion on people's punitive motives. We return to these findings below.

GENERAL DISCUSSION

The results obtained in two studies supported the hypothesis that information about communion predominantly influence observers’ beliefs about offenders, and only to a lower extent punitive reactions. Specifically, communion‐related information about offenders increased observers’ endorsement of conspiracy beliefs, while demonization of offenders mediated this relationship. Furthermore, in line with our predictions, people's punitive reactions depended more strongly on offender’ observable rule‐breaking behaviour than on communion. In fact, people predominantly punished offenders who clearly committed an illegal act, a finding that is in line with prior research (Dworkin & Blumenfeld, 1966; Green, 2014). Moreover, our manipulation check findings suggest that communion and transgression ambiguity were independent (i.e., a clear transgression did not make a protagonist seem colder or more immoral (low in communion) nor did low communion make a protagonist appear as an offender). These findings suggest related, yet different processes for beliefs and punishments of offenders, such that beliefs mostly depend on expectations along the communion dimension, and punishments mostly depend on transgression ambiguity.

In this study, we found that communion is the main dimension that influences individuals’ beliefs about offenders. More specifically, we showed that when people expect malevolent intentions of low in communion targets they try to make sense of these intentions by demonizing them (Amaral, 2002; Brambilla et al., 2011; Engell, Haxby, & Todorov, 2007; Willis & Todorov, 2006; Winston, Strange, O'Doherty, & Dolan, 2002) and seeing them in turn, as evil conspirators. This finding corroborates and extends prior work according to which conspiracy beliefs are the result of a sense‐making process that is triggered by subjective threat experiences (Corneille, Yzerbyt, Rogier, & Buidin, 2001; Mashuri & Zaduqisti, 2015; Newheiser et al., 2011). The present findings suggest that while conspiracy theories have been predominantly investigated in the sphere of political or corporate conspiracy theories (Douglas et al., 2017; Van Prooijen & Van Vugt, 2018), they are also relevant in people's perceptions of offenders. More specifically, this research extends prior work on the relationship between perceived immorality and conspiracy beliefs (Van Prooijen & Jostmann, 2013) and is the first to show that people also perceive suspected offenders as part of conspiracies depending on appraisals along the communion dimension of person perception.

In line with our hypotheses, punitive reactions mostly emerge through a different antecedent, that is, observable violations of formal rules, and to a lesser extent through expectations of offender's intentions. In both our studies, people were confronted with a financial offence which involved high or low degrees of ambiguity about its lawfulness. Findings revealed that for punitivity, observers primarily require evidence that the protagonist's behaviour indeed violated formal rules. Instead, people can be relatively lenient in their punishment for behaviours that may be morally questionable yet not clearly illegal. In addition, besides people's punishment intentions, transgression ambiguity influenced observers’ underlying punitive motives as well. More specifically, when people have evidence about an offender's transgression, they display stronger utilitarian and retributive motives for punishment. Clearly, individuals, when confronted with observable and unquestionable immoralities, strive to justify their punishment intention by indicating both retributive (i.e., past‐oriented proportionate) and utilitarian (prevention‐oriented) motives for punishment. These findings are in line with the notion that while social factors such a perceived communion may sometimes influence punitivity, the main driver of punishment is strength of evidence (Taylor & Hosch, 2004).

It is important to note that although the effects of transgression ambiguity on observers’ punitive intentions and motives are straightforward throughout both studies, this is not always the case for the effects of communion on these variables. In Study 1, we observed an effect of communion manipulation on observers’ punitive intentions (but not on punitive motives), and in Study 2, we found an effect of communion manipulation on punitive motives (but not on punitive intentions). These effects were independent from transgression ambiguity (as indicated by the non‐significant interaction) and emerged inconsistently among Studies 1 and 2. It should be noted that communion, because it is crucial in global evaluations of others (Wojciszke, 2005; Wojciszke et al., 1998) may create a relatively urgent need among observers to behaviourally react (e.g., punitive reaction in this study) against unfriendly, cold, and untrustworthy targets (see Cuddy et al., 2008 for a review). Future research should further investigate the effects of expectations of offenders along the communion dimension on perceiver's behavioural reactions.

One limitation of our study is that the protagonist who was involved in the offence was in both our studies a powerful entrepreneur. There is evidence that people endorse conspiracy theories particularly about powerful people or groups. Specifically, conspiracy belief has been linked to prejudice against powerful groups (Imhoff & Bruder, 2014), while low‐status people are more likely to endorse conspiracy theories (Crocker, Luhtanen, Broadnax, & Blaine, 1999; Uscinski & Parent, 2014). Based on the present findings, we cannot conclude with certainty if communion perceptions also increase conspiracy theories about low‐power offenders (e.g., a burglar). In addition, one might speculate that perceived power differences between the offender and the observer influence the process through which people form their beliefs or wish to assign punishments. Future research should investigate the role of offenders’ but also observers’ power in beliefs and punitive reactions. Another limitation of this study is that we particularly focused on one specific type of offence, that is, a financial offence, which might differ in perceived severity or emotional reaction it may generate to observers as compared to other types of offence (e.g., occupational crime and transnational crime). Based on the current findings, we cannot conclude with certainty if the effects observed here generalize across different types of offence. Further, it should be noted that this study focused on observers’ punitive intention rather than punitive behaviour against the offenders. Future research should use behavioural measures in order to replicate these effects.

Finally, these studies we manipulated actor's communion as one single concept in line with the classical perspective of the person perception theory (Abele, 2003; Bakan, 1966). However, communion is a bi‐faceted construct including ‘warmth’ and ‘morality’ (see Abele et al., 2016). There is evidence that the facet of morality is of primary importance in the impression formation process because it has increased functionalist value (Brambilla, Sacchi, Rusconi, Cherubini, & Yzerbyt, 2012; Brambilla et al., 2011). More specifically, moral traits are more relevant than warmth (or agency traits) in defining whether someone represents an opportunity or a threat (Willis & Todorov, 2006) and thus are judged as highly informative and valuable (Van Prooijen & Ellemers, 2015). Future research should go one step forward and distinguish between the two facets of communion (e.g., warmth vs. morality) and investigate the unique effect that each facet along the communion dimension may have on perceiver's beliefs and punishing reactions.

A methodological strength of our study is that we manipulated ambiguity of transgression in different ways among our studies. In Study 1, ambiguous offence involved a grey line between legal and illegal activities, whereas in Study 2, ambiguous offence involved uncertainty and absence of evidence about the immoral act. This suggests that the findings observed here generalize across different types of transgression ambiguity.

Conclusion

These results reveal that observers’ beliefs about offenders versus their punitive reactions are grounded in related yet distinct psychological processes. Individuals, when confronted with offenders who are seen as low in communion, try to make sense of the offenders by demonizing them, and by seeing them, in turn, as evil conspirators. Punitive reactions, however, are particularly driven by the offender's observable rule‐breaking behaviour, even when appraisals of the actor (i.e., perceived communion) can make punitive action more likely under some circumstances. Taken together, it can be concluded that although beliefs about offenders and punitive reactions are often strongly related, they have different antecedents: Beliefs primarily depend on communion (i.e., person perception), while punishments and punitive motives primarily depend on transgression ambiguity (i.e., evidence for rule‐breaking).

Conflict of interest

Author Kyriaki Fousiani declares that she has no conflict of interest. Author Jan‐Willem van Prooijen declares that he has no conflict of interest.

Compliance with ethical standards

This research involves human participants. All procedures performed in this study were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Data availability

Data and Appendix S1 are available from the Open Science Framework at https://osf.io/u29p8/?view_only=e0f9ca49982f412899e3f66a7b61e6f1.

Appendix S1. Vignettes and Scales—Studies 1 & 2.

Click here for additional data file.

Appendix S2. Study 1—Factor analysis (PCA with Varimax rotation) of perceived communion of the offender, conspiracy beliefs, and demonization.

Appendix S3. Study 2—Factor analysis (PCA with Varimax rotation) of perceived communion of the offender, conspiracy beliefs, and demonization.

Click here for additional data file.