Received: 16 June 2021 Accepted: 22 August 2022 DOI: 10.1002/ejsp.2906 RESEARCH ARTICLE The psychology of hate: Moral concerns differentiate hate from dislike Clara Pretus1 Jennifer L. Ray2 Yael Granot3 William A. Cunningham4 Jay J. Van Bavel5 1 Department of Psychiatry and Forensic Medicine, Universitat Autónoma de Barcelona, Abstract Barcelona, Catalunya, Spain We investigated whether any differences in the psychological conceptualization of 2 Takeda Pharmaceuticals, Cambridge, hate and dislike were simply a matter of degree of negativity (i.e., hate falls on the end Massachusetts, USA 3 Loyola University, Chicago, Illinois, USA of the continuum of dislike) or also morality (i.e., hate is imbued with distinct moral 4 University of Toronto, Toronto, Ontario, components that distinguish it from dislike). In three lab studies in Canada and the Canada United States, participants reported disliked and hated attitude objects and rated each 5 New York University, New York, New York, on dimensions including valence, attitude strength, morality, and emotional content. USA Quantitative and qualitative measures revealed that hated attitude objects were more Correspondence negative than disliked attitude objects and associated with moral beliefs and emotions, Clara Pretus, Department of Psychiatry and Forensic Medicine, Universitat Autónoma de even after adjusting for differences in negativity. In Study 4, we analysed the rhetoric Barcelona, Barcelona, Catalunya, Spain on real hate sites and complaint forums and found that the language used on prominent Email: clara.pretus@uab.cat hate websites contained more words related to morality, but not negativity, relative to Jay J. VanBavel, New York University, complaint forums. Psychology Department, 6 Washington Pl, New York, NY 10003, USA. KEYWORDS Email: jay.vanbavel@nyu.edu attitudes, dislike, emotions, hate, morality 1 INTRODUCTION understanding of the psychological aspects involved in these crimes (Fitzgerald et al., 1997; Meier, 1993). Thus, advancing the understand- The underpinnings of the psychological state of hate have become ing of the psychology of hate can provide valuable insights into how it increasingly relevant as hate-based crimes are on the rise interna- can be best operationalized in the legal system. In the present work, tionally. The UK documented a 30% increase in hate crimes in 2016 we were interested in examining the psychological correlates of our (de Freytas-Tamura, 2017), and reported incidents of hate crimes in shared cultural meaning of “hate”. the United States similarly increased in 2016 compared to the pre- vious year (Berman, 2017). Whereas hate as a legal concept counts on a necessarily specific definition (“intense or extreme dislike, aver- 1.1 The concept of hate sion, loathing, antipathy, enmity or hostility against members of groups or classes of persons identified by protected characteristics”, as per Hate has been generally conceptualized as a negative emotional Brown, 2017), what people perceive and experience as “hate” at a attitude (Allport, 1954; White, 1996) toward persons or groups who psychological level is still not well understood. Legal and psycho- are considered to possess fundamentally negative traits (Allport, logical concepts can be highly interconnected. For instance, legal 1954; Ben-Ze’ev, 2000; Royzman et al., 2006; Sternberg, 2003). Like definitions and procedures around sexual harassment and domestic other affective states (Frijda et al., 1991), hatred can be experienced violence have greatly evolved in the last decades thanks to a greater both as an emotion (“acute hate”) and as a sentiment or emotional This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2022 The Authors. European Journal of Social Psychology published by John Wiley & Sons Ltd. 336 wileyonlinelibrary.com/journal/ejsp Eur J Soc Psychol. 2023;53:336–353. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 337 attitude (“chronic hate”) (Bartlett, 2005; Halperin et al., 2012). Hate There are different ways in which hatred could relate to morality. as an emotion has been described as an acute reaction to a significant Some scholars suggest that hatred as an emotional disposition is linked event characterized by intense feelings, physical symptoms, and desire to negative appraisals evoked by moral transgressions. For instance, for immediate action (Sternberg, 2003, 2005). In this respect, many Allport (1954) conceived of hate as a strong form of dislike that authors have attempted to define hate in contrast to anger, both in escalates in negativity but speculated that hatred cannot exist “unless terms of appraisals and motivational goals (Allport, 1954; Royzman something one values has been violated” (p. 364). Along the same lines, et al., 2006; Sternberg, 2005). Hated targets are viewed as innately evil others have argued that hatred is characterized as the negative evalu- and incapable of change, while anger focuses on the actions attributed ation of a target that is linked with a moral judgment (Royzman et al., to an agent (Ortony et al., 1990), who is perceived as malleable (Fischer 2006), that hate is rooted in seeing the hated target as morally defi- & Roseman, 2007; Halperin, 2011). Thus, hated targets pose an exis- cient or as violating moral norms (Staub, 2004), and that hatred is a tential threat to subjects (Ben-Ze’ev, 2000), who experience low levels direct reaction to protracted harm perceived as deliberate, unjust, and of control (Fitness & Fletcher, 1993) and powerlessness (Fischer et al., stemming from an inner evil character of the hated individual or group 2018). As a result, haters seek to avoid, hurt, or eliminate the object of (Halperin, 2008). These researchers argue that hate is inextricably hate (Allport, 1954; Ben-Ze’ev, 2000; Martínez et al., 2022; Sternberg, linked to morality through negative moral appraisals. 2005), while anger entails motivational goals related to the subject’s Similarly, the Duplex Theory of Hate argues that hate as an emo- desire to change the target (Fischer et al., 2018). tion may be composed of other more basic moral emotions—contempt, When hate is defined as an emotional attitude (vs. hate as an emo- anger, and disgust—which are triggered by moral transgressions tion), it is a more stable disposition towards a hated object that relies (Sternberg, 2003). Particularly, transgressions to communal codes, significantly on cognition (Halperin et al., 2012). In this respect, hatred including hierarchy, have been associated with contempt, transgres- can be conceptually compared to dislike, which can be defined as a sions to personal autonomy or individual rights have been associated preference or negative disposition towards an object that influences with anger, and violations to purity/sanctity have been linked to disgust our behaviour (de Houwer & Hughes, 2020). Hatred and dislike have (see also Rozin, 1999). Therefore, there seems to be some consensus been historically conceptualized as two related concepts. For instance, over the notion that hatred is elicited by moral transgressions, either German poet Johann Wolfgang von Goethe famously speculated that through moral negative appraisals or through moral emotions. Despite “Hatred is active, and envy passive dislike; there is but one step from the numerous accounts of hate as a moral sentiment, no studies to envy to hate” (see Edwards, 1908) and Darwin (1872) argued that our knowledge have empirically tested whether hate is different from “Dislike easily rises into hatred”, suggesting that the two belong on a negative preferences (i.e., dislike) in the moral domain. common conceptual spectrum. From a modern psychological perspective, what we label dislike and hatred both share some characteristics in terms of their dispositional 1.3 Empirical evidence to date nature and negative valence. However, the things that we say that we hate, as opposed to dislike, are transmissible, lead to false attributions, Despite the rich history of theorizing about the nature of hate—and the and motivate violent crimes (Sternberg, 2005). We believe these differ- serious real-world implications of the topic—surprisingly little empir- ences may be understood by taking into account the moral dimension ical research has investigated the psychological conceptualization of of hate. hate. One study found that hate is the sixth most frequently listed exemplar of the concept of emotion (behind happiness, anger, sadness, love, and fear), and therefore, among the most psychologically accessi- 1.2 Hate versus dislike: Does morality matter? ble emotional concepts to people (Fehr & Russell, 1984). Other work exploring hate in interpersonal contexts has found that people write Preferences can transform into values through the process of mor- significantly longer, detailed accounts of their experiences of hate, as alization (Rozin, 1999). As a result, moral values, as compared to compared to experiences of anger, jealousy, or love. However, when preferences, are more central to the self and have a greater ability confronted with hypothetical emotion scenarios, participants were to impact an individual’s life and society at large in powerful ways least accurate at identifying hate (Fitness & Fletcher, 1993). Thus, the (see also Skitka, 2002). In line with this, parent-to-child transmission limited literature on the psychology of hate underscores the every- has been found to be more consistent for moral values than prefer- day importance of the topic, as well as the ways in which hate is little ences (Cavalli-Sforza et al., 1982). Framing an action as moral leads to understood by scientists and laypersons alike. more extreme judgments (Van Bavel et al., 2012) and fosters proso- In attempting to understand the psychological experience of hate, cial behaviour regardless of preferences (Capraro & Rand, 2017). The researchers have investigated what other experiences or emotions willingness to make extreme sacrifices is greater for highly moralized are associated with hate. For example, participants who wrote about values (i.e., sacred values) compared to preferences (Atran et al., 2014). targets of hate often described friends, family, and acquaintances Thus, the distinctive status of hate as “the most destructive affective (56%), but rarely strangers (4%) (Aumer-Ryan & Hatfield, 2007), phenomenon in the history of human nature” (Royzman et al., 2006) suggesting a link between hate and familiarity. More recent work has could well emanate from its moralized nature. found that higher emotional arousal, personal threat perceptions, 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 338 PRETUS ET AL . and attack-oriented behaviours are hate’s distinctive features (see predicted that hated attitude objects would be both more negative and Martínez et al., 2022). Further, people who recalled an experience moral than disliked objects. of hate for their partner reported feeling significantly less in control Across four studies we tested these hypotheses using data from than when experiencing other negative emotions like anger (Fitness controlled laboratory experiments and examined the language used & Fletcher, 1993). Finally, experiencing hate may correspond with online by prominent hate groups. Of note, our lab samples include aversive health consequences, including increased blood pressure Canadian and US students. Therefore, the conceptualizations of hate, and immune system suppression (Dozier, 2002). Consequently, hate’s dislike, and morality discussed in this work are situated within a North inherent link to violence has prompted some researchers to begin American cultural context. Different cultures may have different envi- studying hate as a public health concern (Krug et al., 2002). ronments, social structures, or mental states associated with morality, Concerning the moral dimension of hate, research on hate groups intent, and responsibility that may shape these concepts (see Henrich, suggests that members were particularly likely to cite symbolic, value- 2022). In the lab, we asked participants to generate their own hated and relevant issues as sources for their hate. Moreover, hate groups were disliked attitude objects and rate them on dimensions of valence, emo- also more likely to advocate violence toward the hated group in the tional content, motivation, and morality (Studies 1–3). We also asked face of threats of a moral nature in phone interviews (Green et al., participants to self-reflect on how they understood the difference 1999) and online chat rooms (Glaser et al., 2002). A recent set of stud- between their relationship to their hated and disliked attitude objects ies found that moral values oriented around group preservation are and we coded their qualitative responses (Study 1). The research built predictive of the county-level prevalence of hate groups and associ- upon previous studies of the phenomenology of hate (Aumer-Ryan & ated with the belief that extreme behavioural expressions of prejudice Hatfield, 2007) by including established measures of morality and atti- against marginalized groups are justified (Hoover et al., 2021). As such, tude strength (Skitka et al., 2005) to conduct quantitative tests of our real-world hate groups appear to be motivated by perceived moral key hypotheses. transgressions, even if no actual transgression happened, and this It is important to note that our research largely focuses on lay theo- might foster real-world aggression against vulnerable groups. ries of hate. Thus, any conclusions are limited to the shared semantic content of what people call “hate”. For instance, previous research found that 31% of participants consider hate to be roughly synony- 1.4 Overview mous with extreme dislike (Aumer-Ryan & Hatfield, 2007). In this respect, whereas most theoretical accounts conceptualize hate as an The current research examined the role of morality in the psychology emotional disposition towards persons or groups, people often use of hate. We conducted four studies designed to investigate whether the expression “hate” to refer to objects or concepts. While we did the difference between how people conceptualize hate and dislike is not force a specific meaning upon participants regarding “hate”, we simply a matter of intensity (i.e., hate is merely more negative than dis- separately asked about hate in relation to people/groups versus con- like) or also morality (i.e., hate is imbued with additional psychological cepts/beliefs in Study 3. From our perspective, this lack of constraint ingredients in the moral domain). Do people differentiate hate from dis- in meaning makes any formal test of qualitative differences—over-and- like strictly in terms of intensity, such that dislike is negative and hate above general dislike—more conservative. However, it also introduces is extremely negative on the same continuum? Or do people differen- a constraint on our ability to generalize the results to more ecologically tiate hate from dislike along the moral dimension above and beyond valid contexts. As such, we addressed this limitation in our final study negativity? Most theories focus on differences in negativity or morality by analysing the rhetoric on real hate group websites (vs. complaint without directly testing them concurrently. As such, it is possible that websites). both accounts are correct to some degree. To support the ecological validity of our laboratory research, we To address this question, we tested two primary hypotheses—which conducted a content analysis of websites run by real hate groups, we termed the intensity hypothesis and the morality hypothesis. In line comparing the linguistic content of hate group websites with online with theorizing by Allport (1954) and Ben Ze’ev (2000), hate may best consumer complaint forums and employee complaint forums (Study be conceptualized by people as a negative evaluation. For this reason, 4). Hate group websites were selected to sample expressions of hate, we hypothesized that if the difference between psychological concep- while complaint forums were selected to sample expressions of dislike tualizations of hate and dislike is largely a matter of intensity, then in relation to either products (consumer complaint forums) or peo- hated attitude objects would be rated as more negative than disliked ple/corporations (employee complaint forums). Although hate groups attitude objects (intensity hypothesis). In contrast, following the many are publicly recognized as such, many of these groups adamantly fight theorists who claimed that hate is connected to morality (Allport, 1954; against the hate group label and feel they are merely expressing their Royzman et al., 2006; Staub, 2004), hate may differ in kind from dislike own deeply held values (an issue we discuss in depth below). Thus, the along a moral dimension (morality hypothesis). We hypothesized that content of these websites is more likely to embody the natural occur- hated attitude objects would be associated with more moral emotions, rence of hateful rhetoric and values in the real world. This also expands such as anger, contempt, and disgust (Rozin & Fallon, 1987) and be our research to the domain of intergroup relations since hate groups rated as more connected to core moral beliefs than disliked attitude are defined as organizations that—based on their official statements objects. These hypotheses do not need to be competitive; in fact, we of principles, the statements of their leaders, or their activities—hold 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 339 beliefs or engage in practices that attack or malign entire classes Dependent measures of people, typically for their immutable characteristics (SPLC, 2017). Participants provided both quantitative and qualitative data for their Together, these studies served as an initial exploration into how people six attitude objects. First, participants rated each of their attitude evaluate and represent their own hatred. objects on 13 dimensions by conveying their level agreement with dif- ferent statements on 7-point Likert scales (ranging from 1 = strongly disagree to 7 = strongly agree). These dimensions included valence (i.e., 2 STUDY 1: HATE VERSUS DISLIKE positivity: “Ignoring all my negative feelings, I feel very positive about this attitude object”; negativity: “Ignoring all my positive feelings, I feel Our first study sought to investigate whether differences in peo- very negative about this attitude object”), emotions (i.e., contempt: “I ple’s conceptualization of hate and dislike are simply a matter of feel contempt toward this attitude object”; anger: “I feel anger toward intensity (i.e., hate is more negative than dislike) or also morality (i.e., this attitude object”; disgust: “I feel disgust toward this attitude object”; hate involves additional moral content). To address this question, we fear: “I feel fear toward this attitude object”, and annoyance: “I find the explored what types of attitude objects were associated with hate attitude object annoying/irritating”), attitude strength (i.e., certainty: as compared to dislike, and whether those objects differed as to “I feel very certain about my feelings toward this attitude object”; the valence, emotional content, and motivational and moral concerns extremity: “I feel strongly about this attitude object”, and importance: they promoted. We used analyses of both quantitative ratings and “This attitude object is personally important to me”), motivation (i.e., qualitative reports to compare hate versus dislike on these critical approach: “I am motivated to confront this attitude object”, and avoid- dimensions. ance: “I am motivated to avoid things related to this attitude object”), and morality (i.e., centrality to core moral beliefs).1 Specifically, the morality item was adopted from Skitka et al.’s (Skitka et al., 2005) ques- 2.1 Method tionnaire: “My feelings about this attitude object are connected to my core moral beliefs or convictions”. Finally, participants completed a free The materials, data, and analysis code are available on the Open Sci- response item asking them to describe the similarities and differences ence Framework website (osf.io/5u6my). A power analysis run with the between the three hated and three disliked attitude objects. R package “simr” (Green et al., 2016) showed that, based on 1000 sim- ulations and an alpha = 0.05, recruiting 170 participants would enable Analysis strategy the detection of small effects (e.g., 0.1) for the interaction between Analyses of repeated measures often focus on mean-level differences within-subject attitude object type (hated vs. disliked object) and order in ratings. However, this approach reduces multiple measures to a (hate first vs. dislike first) with a statistical power of 72.6% (95% CI single score for each participant, diminishing power and meaningful [69.72, 75.34]) with six observations per participant. variance. To analyse differences more accurately between hate and dislike, we used multilevel modelling (Hox, 1998). Multilevel mod- 2.1.1 Sample elling allows for the direct analysis of individual ratings and helps overcome violations of independence that occur due to correlated rat- One hundred and seventy-eight University of Toronto students com- ings within participants. When an assumption of independence is not pleted the study in exchange for Introduction to Psychology course satisfied, ignoring dependency among trials can lead to invalid sta- credit. We removed nine participants from our sample who, in gener- tistical conclusions, namely, the underestimation of standard errors ating six attitude objects, did not provide independent ratings for three and the overestimation of the significance of predictors (Cohen et al., disliked attitude objects and three hated attitude objects, resulting in 2002). We therefore created multilevel models with repeated ratings a final sample size of 169 participants. We used repeated measures nested within participants to provide more appropriate estimates of to increase within-subject power. The sample size of each laboratory regression parameters. Multilevel models were implemented in R. We study (Studies 1–3) was determined with the goal to optimize statis- conducted a 2(Type: Hate vs. Dislike) × 2(Order: Hate First vs. Dislike tical power with the constraint of a convenient stopping point (e.g., First) mixed model design, between subjects on the last factor. For com- running as many participants as possible until the subject pool closed pleteness, we also report simple bivariate correlations of negativity, at the end of the semester). We report how all data exclusions (if any), morality, and moral emotions for all three studies in Table 1. all manipulations, and all measures in the study. 2.2 Results 2.1.2 Design and procedure We began by examining the attitude objects participants generated. To Participants were asked to generate three disliked attitude objects visually illustrate these differences in an intuitive fashion, we created and three hated attitude objects, in counterbalanced order. They 1 The current article focuses on the variables directly relevant to the morality and negativity rated these attitude objects on several dimensions and described the hypotheses. Although other variables change from study to study, these variables remained difference between these two sets of attitude objects. consistent. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 340 PRETUS ET AL . TA B L E 1 Bivariate correlations of negativity, morality, and moral Given our mixed-model design, we adjusted for differences in morality emotions (Studies 1 and 2), and bivariate correlations of negativity, in two ways. To begin, we computed a mean morality score for each par- morality, and attitude strength (Study 3). ticipant by averaging each individual’s six morality ratings for all of their Study 1 attitude objects. Then, we created a between-person centred moral- 1 2 3 ity score to reflect how each participant’s mean morality score differed from the average of all the participants’ mean morality scores (i.e., the 1. Morality – grand mean across all subjects). Second, we computed a within-person 2. Negativity 0.20* – centred morality score to reflect how each trial differed from each indi- 3. Moral emotions 0.42* 0.30* – vidual’s mean morality score across all six trials. Then, we re-analysed Study 2 the association between attitude object type (Hated vs. Disliked) and 1 2 3 negativity, adjusting for both between-person and within-person cen- 1. Morality – tred morality scores. As a result, hated attitude objects were rated 2. Negativity 0.16 – as more intensely negative than disliked attitude objects, Mdiff = 0.28, 3. Moral emotions 0.29* 0.23* – 95% CI [0.15, 0.41], t(822) = 4.17, p < .001, even after adjusting for the effects of morality. Negativity ratings remained unaffected by the Study 3 order in which participants generated them (p = .50). The effects of 1 2 3 between-person morality, B = 0.16, 95% CI [0.01, 0.30], t(163) = 2.15, 1. Morality – p = .033, and within-person morality, B = 0.13, 95% CI [0.09, 0.18], 2. Negativity 0.59* – t(821) = 6.07, p < .001, were statistically significant in the model. 3. Attitude strength 0.57* 0.42* – For ease of interpretation, we show the zero-order correlations of the Morality hypothesis dependent variables treating each rating made by participants as indepen- To examine whether hated objects differ from disliked objects in terms dent. of morality, we next tested the morality hypothesis. Consistent with *p < .05. the morality hypothesis, hated objects (M = 5.33, SE = 0.10) were rated as more connected to morality than disliked objects (M = 4.91, “Wordles” or world clouds, to visually represent the frequency with SE = 0.10), Mdiff = 0.42, 95% CI [0.22, 0.63], t(837) = −3.99, p < .001, which each attitude object was listed. The wordles (see Figure 1) give Cohen’s d = 0.24. This effect was qualified by a significant interac- greater prominence to words that are listed with the greatest fre- tion of attitude object type and order, B = 0.44, 95% CI [0.35, 0.92], quency. The most frequently listed disliked attitude objects included t(836) = 2.10, p = .036, such that differences in morality between “groups” and “rude”. The most frequently listed hated objects included hated and disliked objects were significant for participants who were “racism” and “tests”. asked about the hated object first, Mdiff = 0.63, 95% CI [0.35, 0.92], t(835) = 4.34, p < .001, Cohen’s d = 0.34, but non-significant for partic- ipants who were asked about the disliked object first, Mdiff = 0.19, 95% 2.2.1 Quantitative responses CI [−0.11, 0.49], t(836) = 1.26, p = .21, Cohen’s d = 0.14, see Figure 2(a). This suggests that hated attitude objects were perceived as more con- Intensity hypothesis nected to morality than disliked objects, consistent with the morality To examine whether responses to hated attitude objects differ from hypothesis. However, rating hated attitude objects second resulted in disliked attitude objects in terms of degree of negativity, we first tested attenuated differences in morality for hated versus disliked attitude the intensity hypothesis, which predicts that hated objects are associ- objects. ated with greater negativity than dislike. Consistent with the intensity We then examined whether hated attitude objects were associated hypothesis, hated objects (M = 6.07, SE = 0.08) were rated more neg- with moral emotions—contempt, anger, and disgust—to a greater atively than disliked objects (M = 5.74, SE = 0.08), Mdiff = 0.33, 95% CI degree than disliked attitude objects. To do so, we created a moral [0.20, 0.47], t(825) = 4.92, p < .001, Cohen’s d = 0.23. We found no effect emotions scale by averaging ratings of contempt, anger, and disgust for of order (i.e., hate first vs. dislike first), B = −0.15, 95% CI [−0.47, 0.18], t each participant (α = 0.68, see descriptive statistics of each emotion (234) = −0.87, p = .39, nor did we find an interaction between attitude in Table S1). Consistent with the morality hypothesis, hated objects object type and the order in which participants generated their ratings, (M = 4.95, SE = 0.08) were more strongly associated with moral B = 0.03, 95% CI [−0.24, 0.29], t (823) = 0.19, p = .84. This suggests emotions than disliked objects (M = 4.47, SE = 0.08), Mdiff = 0.48, 95% that hated attitude objects are reported as more negative than disliked CI [0.35, 0.61], t(838) = 7.05, p < .001, Cohen’s d = 0.37, see Figure 2(b). objects, consistent with the intensity hypothesis. We also found an effect of order, such that participants who gener- ated hated attitude objects first reported experiencing fewer moral Effects of intensity, adjusting for morality emotions (M = 4.52, SE = 0.10) than participants who generated For a more stringent test of the intensity hypothesis, we also re- disliked attitude objects first (M = 4.89, SE = 0.10), Mdiff = 0.36, 95% CI analysed the effects of intensity adjusting for differences in morality. [0.35, 0.61], t(167) = 2.55, p = .012, Cohen’s d = 0.28. The interaction 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 341 F I G U R E 1 “Wordles” illustrating the frequency with which participants listed each attitude object in Study 1 (Top: Dislike, Bottom: Hate). The larger the word appears, the more frequently it was listed. F I G U R E 2 The effect of hate versus dislike on morality judgments (a) and moral emotions (b) in Study 1. Hated objects were rated as significantly more tied to core moral beliefs and moral emotions than disliked attitude objects. Disliked attitude objects, rated after hated attitude objects, received attenuated morality ratings. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 342 PRETUS ET AL . of attitude object type and order of generation was not statistically t(823) = 9.74, p < .001, remained significant in the model. Again, this significant, B = 0.06, 95% CI [−0.21, 0.32], t(838) = 0.42, p = .67. pattern of findings supports both the intensity hypothesis, as well as This suggests that hated attitude objects were experienced as more the morality hypothesis. closely connected to moral emotions than disliked objects, providing additional support for the morality hypothesis. 2.2.2 Qualitative responses Effects of morality, adjusting for intensity After reporting three disliked and three hated objects, participants Given initial evidence for both the intensity and morality hypotheses, were explicitly asked to compare their hated and disliked attitude we examined the relationship between our main dependent variables. objects. Two independent raters content coded participants’ descrip- We found that morality ratings and the composite measure of moral tions of the similarities and differences between their hated and emotions were highly correlated (r = 0.42, p < .001). This finding disliked attitude objects and a third rater resolved any disagreements demonstrates construct validity by establishing the relation of moral (Cohen’s κ = 0.88). Only 103 participants answered this question; of judgments to affect that is central to morality (see Haidt, 2001). that number, the raters agreed that 17 responses linked hated objects However, we also found that morality ratings and negativity were to morality, while only two responses linked disliked objects to moral- moderately correlated (r = 0.20, p < .001). This presents a potential ity. Compared to chance, where morality should be equally likely to be confound for the intensity hypothesis and morality hypothesis (see linked to either hate or dislike, we found that participants were more Table 1). To help address this issue, we performed a more stringent likely to report that hated objects were more closely linked to morality test of the morality hypothesis by re-analysing the data, adjusting for than disliked objects, χ2(1) = 11.84, p < .001. For instance, one partic- differences in negativity. Following the same strategy as presented ipant wrote, “The hated objects were more moral issues in that they above, we computed a mean negativity score for each participant by represent who I am and what my beliefs are. With those issues, I am averaging each individual’s negativity ratings for all of their attitude prepared to take a stance and argue my point of view with the hope that objects and created a between-person centred negativity score and others will agree. The disliked objects were less moral issues . . . These a within-person centred negativity score. We then re-analysed the represent what appeals to my taste, but I’m not going to go out of degree to which attitude object type was associated with core moral my way to argue out loud or with uptight people.” Another participant beliefs, adjusting for both between-person and within-person centred wrote, “The hated objects are more or less related to my inner beliefs negativity scores. and values of life whereas the other dislike objects are more . . . shal- Providing further evidence for the morality hypothesis, hated atti- low aspects of my life.” These results provide convergent evidence that tude objects (M = 5.27, SE = 0.13) were rated as more connected morality plays a differentiating role between hate and dislike, lending to moral beliefs than disliked attitude objects (M = 4.74, SE = 0.13), further, qualitative support to the morality hypothesis. Mdiff = 0.54, 95% CI [0.25, 0.82], t(822) = 3.72, p < .001, for participants who rated hated attitudinal objects first, even after statistically adjust- ing for the effects of intensity. However, differences in morality ratings 2.3 Discussion were attenuated when hated attitude objects were rated second (M = 4.73, SE = 0.13), Mdiff = 0.08, 95% CI [−0.22, 0.38], t(823) = 0.55, In our first study, we examined whether the conceptual differences p = 0.58. The effect of between-person negativity, B = 0.20, 95% between hate and dislike are simply a matter of degree of negativ- CI [0.04, 0.35], t(169) = 2.43, p = .016, and within-person negativ- ity or also a matter of morality. We found support for the intensity ity, B = 0.32, 95% CI [0.21, 0.42], t(820) = 5.96, p < .001, remained hypothesis—hated objects were viewed as more negative than disliked significant in the model. This pattern of findings provides evidence con- objects—suggesting that the difference between hate and dislike is sistent with both the intensity hypothesis (that the difference between indeed a matter of intensity. However, we also found support for the hate and dislike is the degree of negativity) and the morality hypoth- morality hypothesis—hated attitude objects were rated as more con- esis (such that above and beyond a difference in negativity, morality nected to participants’ core moral beliefs and were associated with differentiates hate from dislike). higher levels of moral emotions (contempt, anger, and disgust) than Similarly, we re-analysed the data on our composite measure of disliked attitude objects—suggesting that the difference between hate moral emotions, adjusting for differences in negativity (i.e., both and dislike may also be a matter of morality. We found convergent between- and within-person negativity). We again found an effect of evidence for this latter hypothesis across quantitative and qualita- hate versus dislike on the moral emotions, with hated attitude objects tive analyses, with self-reports, expressions of moral emotions, and evoking more moral emotions (M = 4.90, SE = 0.08) than disliked spontaneous descriptions. attitude objects (M = 4.53, SE = 0.08), Mdiff = 0.37, 95% CI [0.24, We note that differences in morality were attenuated when par- 0.50], t(824) = 5.60, p < .001, even after adjusting for the effect of ticipants were asked about dislike attitude objects first. We discuss negativity. In this case, the effect of order became non-significant, possible explanations of this order effect below. Importantly, the B = −0.28, 95% CI [−0.58, 0.01], t(244) = −1.87, p = .062. The effect of results supporting the morality hypothesis remained significant even between-person negativity, B = 0.24, 95% CI [0.11, 0.38], t(170) = 3.47, when adjusting for negativity. Above and beyond the effect of nega- p < .001, and within-person negativity, B = 0.32, 95% CI [0.25, 0.39], tivity, both moral concerns and moral emotions explained the variance 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 343 in ratings of hated versus disliked attitude objects. Likewise, partici- tion offered a more stringent test of the hypotheses that hated and pants spontaneously reported that hated objects were more closely disliked objects differentiate in morality by comparing hated objects tied to morality than disliked objects in their qualitative responses. directly with extremely disliked objects. If morality does differentiate These findings provide preliminary evidence that the conceptualiza- between hate and dislike, hated attitude objects should be rated as tion of hate may differ from dislike, and that morality may play a key more connected to moral concerns and as evoking moral emotions to role in explaining this difference. a greater degree than both disliked and extremely disliked attitude We also found unexpected effects due to the order in which partici- objects. We also added manipulation checks for both hate and dislike. pants generated hated versus disliked attitude objects. The main effect Finally, we collected data for Studies 2 and 3 in another country (USA) of hate versus dislike on morality ratings was qualified by an interac- to increase the diversity of our samples and see if the results would tion in which the order of generating the attitude objects moderated generalize outside Canada. morality judgments. In particular, morality ratings were lower for dis- liked attitude objects when they were generated after hated ones. One potential explanation is that when participants thought about hate first 3.1 Method (then dislike), they understood the nature of the conceptual contrast between hate versus dislike and attenuated their responses for their For Study 2, the power analysis based on 1000 simulations and an disliked attitude objects. However, when participants were asked to alpha = 0.05 showed that recruiting 176 participants would enable generate disliked attitude objects first, they may have reported things the detection of small effects (e.g., 0.3) for the interaction between they hate, and only later understood the contrast once they were within-subject attitude object type (hated vs. disliked object) and invited to list hated objects. Indeed, previous research finds that nearly between-subject condition (extreme vs. regular) with statistical power a third of people consider hate to be synonymous with extreme dislike of 81.3% (95% CI [78.74, 83.67]) with two observations per participant. (Aumer-Ryan & Hatfield, 2007). In reviewing participants’ responses in the free response section, we found several comments that support this interpretation of our 3.1.1 Sample data. For instance, one participant remarked: “I was given the ‘dislike’ paper first and then the ‘hate’ paper. Hence, I did not understand that One hundred and eighty-two students at The Ohio State University I had to put emphasis on the word ‘dislike.’ I made a little mess in completed the study in exchange for Introduction to Psychology course that.” Another wrote: “If I had known I was going to be writing about credit. We removed five participants from our sample who entered things I ‘hated’ after the things I ‘disliked’ I would have had some of the only one or two scale rating numbers (of seven possible ratings) for all ‘disliked’ responses be hated responses.” Given that some participants their responses to the dependent measures in the study (suggesting generated hated attitude objects in the dislike condition, our results that they were not paying any attention or engaged in the study) result- probably represent a conservative test of the differences between hate ing in a final sample of 177 participants. Again, we used a within-subject and dislike. Although this speaks to the semantic challenges of study- contrast to increase power. ing this issue, it is important to note that participants reported more moral emotions for hated attitude objects—regardless of order. Thus, any effects of order cannot account for the overall pattern of results. 3.1.2 Design and procedure Participants were asked to generate one hated, one disliked, one liked, 3 STUDY 2: HATE VERSUS EXTREME DISLIKE and one loved attitude object in that order. Participants were randomly assigned to either a regular attitude objects condition (similar to Study The findings in Study 1 provided the first empirical evidence that 1), or to an extreme attitude objects condition in which they generated morality differentiates between hated and disliked attitude objects, attitude objects for hate, extreme dislike, extreme like, and love. even after adjusting for differences in the degree of negativity. How- ever, given that participants attenuated their morality ratings for Dependent measures disliked attitude objects when they rated them second (i.e., after seeing Participants provided quantitative data for their four attitude objects. an explicit contrast between hate and dislike), we sought to address the As in Study, 1, participants rated each of their attitude objects on observed but unpredicted order effects in Study 2. First, participants the same 13 dimensions (i.e., valence, emotions, attitude strength, were made aware in the instructions that they would be generating motivation, and morality). To help disguise our intent, we included both hated and disliked attitude objects. Second, all participants gen- several additional attitudinal (e.g., feel strongly, important to me) and erated their attitude objects in the same order. Third, to help disguise emotional (e.g., distressed, conflicted, excited) dimensions from the our main research question and minimize demand effect, we also had expanded Positive and Negative Affect Schedule (PANAS-X; Watson participants rate liked and loved objects and complete a number of & Clark, 1994). Thus, participants completed a total of 40 ratings per irrelevant emotion ratings. Fourth, we also included a condition in attitude object. We also added two manipulation check items—how which participants compared hate to extreme dislike. This new condi- much participants felt hate and dislike toward their attitude objects— 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 344 PRETUS ET AL . which were mixed in with the PANAS-X items to disguise the intent of 3.2.2 Intensity hypothesis our manipulation checks. As in Study 1, all ratings were made on 7- point Likert scales (ranging from 1 = strongly disagree to 7 = strongly We examined the intensity hypothesis, which predicts that the dif- agree). ference between hated and disliked objects is a matter of degree of negativity. We replicated our effect of hate versus dislike, such that Analysis hated attitude objects were rated more negatively (M = 5.76, SE = 0.13) We focus on responses for the hated and disliked (or extremely dis- than disliked ones (M = 5.44, SE = 0.13), Miffs = 0.32, 95% CI [0.03, 0.61], liked) attitude objects, so no analyses for the liked and loved attitude t(171) = 2.17, p = .03, Cohen’s d = 0.19. We found no effect of extrem- objects are presented. As in Study 1, we created multilevel models with ity condition and no interaction of attitude object type and extremity repeated ratings nested within participants to provide more appropri- condition (ps > .41). This suggests that hated attitude objects are per- ate estimates of regression parameters. We conducted a 2 (Type: Hate ceived to be more negative than disliked ones, replicating our findings vs. Dislike) × 2 (Extremity: Regular dislike vs. Extreme Dislike) mixed from Study 1 supporting the intensity hypothesis. model design, between-subject on the last factor. 3.2.3 Effects of intensity, adjusting for morality 3.2 Results We re-analysed the effects of intensity adjusting for both between- 3.2.1 Manipulation checks person and within-person centred morality scores. Hated attitude objects were rated as more intensely negative than disliked attitude We included ratings of dislike and hate as manipulation check items, objects, Mdiff = 0.34, 95% CI [0.04, 0.63], t(170) = 2.21, p = .03, even mixed in with PANAS-X items. We wanted to confirm that partic- after adjusting for the effects of morality. Negativity ratings remained ipants hated their hated attitude objects more than their disliked unaffected by the extremity condition in which participants gener- attitude objects. Regarding our dislike manipulation check, we were ated them (p = .44). The effects of between-person morality and more agnostic in our predictions. Participants might dislike their hated within-person morality were non-significant in the model (ps < .15). attitude objects more than their disliked ones or they could dislike both attitude objects equally. Hated attitude objects were more disliked (M = 4.21, SE = 0.13) 3.2.4 Morality hypothesis than disliked attitude objects (M = 3.67, SE = 0.13), Mdiff = 0.54, 95% CI [0.24, 0.84], t(171) = 3,55, p < .001, Cohen’s d = 0.42, but only in Replicating the results of Study 1, hated attitude objects were rated the regular condition. In the extreme condition, dislike ratings did not as more connected to core moral beliefs (M = 5.20, SE = 0.15) than differ based on attitude object type, Mdiff = −0.02, 95% CI [−0.33, disliked ones (M = 4.60, SE = 0.15), Mdiff = 0.60, 95% CI [0.26, 0.94], 0.28], t(171) = −0.15, p = .88, Cohen’s d = −0.01. Our dislike manipula- t(171) = 3.45, p < .001, Cohen’s d = 0.31, see Figure 3(a). We found tion check showed that hated attitude objects were more disliked than no effect of extremity condition and no interaction of attitude object disliked attitude objects, but that hated attitude objects were equally type and extremity condition (ps > .19). Further analysis revealed disliked as extremely disliked attitude objects. that the simple effect of hate versus dislike in the extreme condi- As predicted, hated attitude objects were more hated (M = 4.05, tion was not only statistically significant, but, if anything, actually SE = 0.10) than disliked attitude objects (M = 3.12, SE = 0.10), B = 0.92, greater than in the regular condition, Mdiff = 0.83, 95% CI [0.34, 1.31], 95% CI [0.67, 1.17], t(171) = 7.31, p < .001, Cohen’s d = 0.64. This effect t(171) = 3.37, p < .001, Cohen’s d = 0.43. This suggests that hated atti- was qualified by a significant interaction of attitude object type and tude objects are perceived as more connected to morality than disliked extremity condition, B = −0.63, 95% CI [−0.33, 0.28], t(171) = −2.52, attitude objects–-especially when hated attitude objects are compared p = .012. In the regular condition, hate ratings for hated attitude to extremely disliked attitude objects. objects (M = 4.16, SE = 0.15) were greater than for disliked atti- We further tested the morality hypothesis by investigating the rela- tude objects (M = 2.92, SE = 0.15), Mdiff = 1.24, 95% CI [0.89, 1.59], tionship between hate versus dislike on a composite measure of moral t(171) = 6.98, p < .001, Cohen’s d = 0.94. In the extreme condi- emotions (i.e., contempt, anger, and disgust, α = 0.62). Replicating the tion, hate ratings for hated attitude objects (M = 3.93, SE = 0.15) results of Study 1, hated attitude objects evoked a greater degree of were still significantly greater, but relatively less so, than for dis- moral emotions (M = 3.28 SE = 0.08) than disliked attitude objects liked attitude objects (M = 3.33, SE = 0.15), Mdiff = 0.60, 95% CI (M = 2.89, SE = 0.08), Mdiff = 0.39, 95% CI [0.22, 0.56], t(171) = 4.52, [0.25, 0.96], t(171) = 3.38, p < .001, Cohen’s d = 0.37. Our hate p < .001, Cohen’s d = 0.37, see Figure 3(b). Again, we found no effect manipulation check revealed that hated attitude objects were more of extremity condition and no interaction of attitude object type and hated than disliked attitude objects, but especially in the regular dis- extremity condition (ps > .17). This finding provides additional support like condition. Thus, while hate and extreme dislike were equally for the morality hypothesis by showing that hated attitude objects are negative, hate was still rated as conceptually distinct from extreme associated with a greater experience of moral emotions than disliked dislike. attitude objects. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 345 F I G U R E 3 The effect of hate versus dislike on morality judgments (a) and moral emotions (b) in Study 2. Hated objects were rated as significantly more tied to core moral beliefs and moral emotions than disliked attitude objects. 3.2.5 Effects of morality, adjusting for intensity t(170) = 2.85, p = .005, remained significant predictors of morality, adjusting for attitude object type. This pattern of findings similarly sup- Given that we replicated our findings in support of the hypotheses that ports the intensity hypothesis as well as the morality hypothesis, such the difference between hate and dislike is both a matter of intensity that above and beyond a difference in negativity, the experience of and a matter of morality, we tested the intercorrelations of our main moral emotions differentiates hate from dislike. dependent variables. As in Study 1, we found that morality judgments and moral emotions were positively correlated (r = 0.29, p < .01) and that negativity and the moral emotions were moderately correlated 3.3 Discussion (r = 0.23, p < .01), but negativity and morality were not significantly cor- related (r = 0.16, p = 0.075) (see Table 1). To perform a more stringent In Study 2, people were asked to generate hated, disliked, liked, and test of the morality hypothesis, we re-analysed the effects of atti- loved attitude objects. By examining the contrasts of hate versus dislike tude object type and extremity condition on moral judgments and the and hate versus extreme dislike, we were able to explore a more strin- moral emotions, adjusting for both between-person and within-person gent test of our morality hypothesis. Replicating the results of Study 1, negativity. we found additional support for the intensity hypothesis, suggesting that We replicated our Study 1 findings of the effect of hate (M = 5.20, the difference between hate and dislike conceptualizations is indeed a SE = 0.15) versus dislike (M = 4.59, SE = 0.15) on morality, Mdiff = 0.61, matter of degree of negativity. We also found strong support for the 95% CI [0.26, 0.96], t(170) = 3.47, p < .001, even when adjusting for morality hypothesis, finding that attitude objects were rated as more negativity in two ways. We found no effect of extremity condition, connected to peoples’ core beliefs and were associated with moral between-person negativity, or within-person negativity, and no inter- emotions (i.e., contempt, anger, and disgust) to a greater degree than action of attitude object type and extremity condition (ps > .15). In disliked attitude objects. Strikingly, when people contrasted between Study 1, both between and within-person negativity remained signifi- hate versus extreme dislike, they rated their hated attitude objects cant predictors of attitude object type (i.e., hate vs. dislike), adjusting as even more moral than when hate was contrasted with regular dis- for morality. However, in Study 2, attitude object type (i.e., hate vs. like. In addition, the results supporting the morality hypothesis again dislike) was the only significant predictor of morality ratings. This pat- remained significant even when adjusting for between-person and tern of findings supports the morality hypothesis, such that above and within-person negativity. This means that above and beyond the effect beyond a difference in negativity, morality differentiates hate from of negativity, morality (and moral emotions) differentiated between dislike. hated and disliked attitude objects. Similarly, even when adjusting for between-person and within- person negativity, we replicated the effect of hate (M = 3.26, SE = 0.08) versus dislike (M = 2.91, SE = 0.08) on the expression of moral emo- 4 STUDY 3: HATE, MORAL CONVICTION, and tions, Mdiff = 0.35, 95% CI [0.18, 0.52], t(170) = 4.09, p < .001. We found MORAL CONCERN no effect of extremity condition and no interaction of attitude object type and extremity condition (ps > .19). As in Study 1, the effect of Thus far, we have investigated the relationship between hate versus between-person negativity, B = 0.13, 95% CI [0.04, 0.22], t(170) = 2.94, dislike on ratings of negativity, morality, and the moral emotions. How- p = .003, and within-person negativity, B = 0.13, 95% CI [0.04, 0.21], ever, research on moral convictions (or moral mandates) describes 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 346 PRETUS ET AL . further ways in which morality may be defined and measured (see 4.1.2 Design and procedure Skitka, 2002; Skitka et al., 2005; Skitka & Mullen, 2002). Specifically, moral convictions may be distinguished from otherwise strong but Participants were asked to generate one disliked, one extremely dis- non-moral attitudes by the experience of a unique combination of uni- liked, and one hated attitude object in that order. Participants were versalism, factual belief, and justification for action (Skitka et al., 2005). either assigned to a people/groups attitude objects condition or to con- According to this account, moral convictions are experienced as objec- cepts/beliefs attitude objects condition. In each condition, participants tive characteristics of the world (factual belief) that are perceived as were instructed to list either a person/group or concept/belief they absolutes (universalism), disregarding cultural differences or personal hated (e.g., “For this phase of the study we want you to list a PERSON preferences (Haidt et al., 2003), and motivate action (justification for or GROUP you HATE”). action) (Skitka et al., 2005). These aspects of morality have been shown to predict action potential to a greater extent than non-moral attitudes Dependent measures even after controlling for attitude strength (e.g., extremity, importance, Participants provided quantitative data for their three attitude objects. and certainty). For this reason, in Study 3 we included these additional As in Studies 1 and 2, participants rated each of their attitude objects indices of moral conviction (universalism, factual belief, and justifica- on the same 13 dimensions (i.e., valence, emotions, attitude strength, tion for action) to investigate differences in morality between hate and motivation, and morality), as well as rated the other attitudinal (e.g., feel dislike and controlled for differences in attitude strength. strongly, important to me) and emotional (e.g., distressed, conflicted, If differentiating hated from disliked objects is a matter of intensity, excited) dimensions from Study 2. In addition, participants provided hated attitude objects should be rated as more negative than both dis- ratings on items related to morality, including the same item employed liked and extremely disliked attitude objects. Further, if morality does in Studies 1 and 2 (“My feelings about this attitude object are con- play a differentiating role between hated and disliked objects, hated nected to my core moral beliefs or convictions”), and eight additional attitude objects should be rated as more connected to morality (includ- morality items (i.e., “____violates my core moral beliefs”, “____intention- ing universalism, factual belief, and justification for action) than both ally violates my core moral beliefs”, “____is saintly” (reverse-coded), disliked and extremely disliked attitude objects, even after controlling “____is evil”, “____has the right core moral beliefs” (reverse-coded), for attitude strength. “____has the wrong core moral beliefs”, “____has no core moral beliefs”, In addition, there remains a potential alternative explanation for our and “every time I think of ____ my core moral beliefs spring to mind”), pattern of results in Studies 1 and 2: perhaps asking participants to as well as items related to universality (“Any reasonable person would produce examples for hate versus dislike (or extreme dislike) evokes share my feelings about ____”, and “Feelings about ____ are a matter of responses that reflect different classes of attitude objects. To help personal taste” (reverse-coded)), factual belief (“it’s a fact that ____ is ensure this was not the case, Study 3 sought to replicate and extend the wrong” and “it’s a fact that ____ is right” (reverse coded)), and justification differences in morality and intensity observed in Studies 1 and 2 when for action (“I feel morally obligated to do something about____” and “If I assigning participants to generate specific classes of attitudes objects: took action against ____, it would validate my moral beliefs”) for a total either people/groups or concepts/beliefs. of 51 ratings per attitude object. As in the previous studies, all ratings were made on 7-point Likert scales ranging from “1 = strongly disagree” to “7 = strongly agree.” 4.1 Method To test the relationship between hate and morality, we created a morality scale by analysing all morality items (the items employed in For Study 3, a power analysis based on 1000 simulations and an Studies 1 and 2, and the eight additional items described above), as alpha = 0.05 revealed that recruiting 78 participants would enable the well as the items on universalism, factual belief, and justification for detection of small effects (e.g., 0.3) for the interaction between within- action (following Skitka et al., 2005). We completed factor analyses on subject attitude object type (hated vs. extremely disliked vs. disliked these fifteen items and eliminated the items with loadings below 0.299 object) and between-subject attitude object class (concepts/beliefs (“saintly” (reverse-coded) and one of the factual belief items, “feelings vs. people/groups) with a statistical power of 87.1% (95% CI [75.14, about ____ are a matter of personal taste” (reverse-coded)), resulting in 78.88]) with three observations per participant. a 13-item moral conviction scale with very good reliability α = 0.87). Thus, the final moral conviction scale included most of the items related to universality, factual belief, and justification for action. 4.1.1 Sample Analysis Eighty-two students at The Ohio State University completed the study As in Studies 1 and 2, we created multilevel models with repeated rat- in exchange for Introduction to Psychology course credit. We removed ings nested within participants to provide more appropriate estimates four participants from our sample who reported “nothing” or “no one” of regression parameters. Multilevel models were implemented in R. for their hated attitude object (i.e., failed to report something they We conducted a 3 (Type: Hate vs. Extreme Dislike vs. Dislike) × 2 (Atti- hated), resulting in a total sample of 78 participants. Again, we used a tude Object Class: People/Groups vs. Concepts/Beliefs) mixed model within-subject contrast to increase power. design, between-subject on the last factor. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 347 negatively than concepts/beliefs (M = 5.45, SE = 0.13), Mdiff = 0.41, 95% CI [0.03, 0.79], t(76) = 2.14, p = .04, Cohen’s d = 0.32. We found no inter- action of hate versus dislike and attitude object class (ps > .12). This pattern of results suggests that hated attitude objects are perceived to be more negative than disliked ones, replicating our findings from Stud- ies 1 and 2 in support of the intensity hypothesis. Importantly, we did not find that hated attitude objects are perceived to be more negative than extremely disliked attitude objects. 4.2.2 Effects of intensity, adjusting for morality We re-analysed the effects of intensity adjusting for both between- person and within-person centred morality scores. After adjusting for the effects of morality, differences in negativity became non- significant across all contrasts: hated (M = 5.65, SE = 0.13) versus disliked (M = 5.57, SE = 0.13) attitude objects, Mdiff = 0.08, 95% CI [−0.35, 0.52], t(151) = 0.44, p = .90, hated versus extremely disliked (M = 5.72, SE = 0.12) attitude objects, Mdiff = −0.06, 95% CI [−0.25, 0.55], t(151) = −0.41, p = .91, and disliked versus extremely disliked attitude objects, Mdiff = 0.15, 95% CI [0.04, 0.63], t(151) = 0.88, p = .66. The effects of between-person morality, B = 0.72, 95% CI [0.47, 0.96], t(151) = 5.70, p < .001, and within-person morality, B = 0.72, 95% CI [0.55, 0.90], t(75) = 7.99, p < .001, were statistically significant in the model, while object class (concept/belief vs. person/group) and its interaction with attitude object type were non-significant (ps > 12). F I G U R E 4 The effect of hate versus dislike on ratings of (a) negativity, (b) morality and (c) attitude strength in Study 3. Hated Thus, the intensity hypothesis did not survive the inclusion of morality. objects were rated as significantly more negative and were more associated with attitude strength than disliked objects but received similar ratings in negativity and attitude strength as extremely disliked 4.2.3 Morality hypothesis objects. Hated objects were rated as more related to morality than extremely disliked attitude objects and disliked attitude objects. Replicating the results from Studies 1 and 2 with regard to the relation between hate and morality, attitude object type significantly predicted 4.2 Results ratings of morality (Hate vs. Extreme Dislike: B = 0.41, 95% CI [0.13, 0.70], t(154) = 2.86, p = .005, and Hate vs. Dislike: B = 0.97, 95% CI 4.2.1 Intensity hypothesis [0.68, 1.25], t(154) = 6.69, p < .001). Specifically, hated attitude objects (M = 5.25, SE = 0.11) were more related to morality than disliked atti- Replicating the results of Studies 1 and 2, attitude object type had a sig- tude objects (M = 4.27, SE = 0.11), Mdiff = 0.98, 95% CI [0.04, 0.63], nificant effect on negativity, but only for some of the contrasts (Hate vs. t(152) = 6.75, p < .001, Cohen’s d = 0.90, or extremely disliked atti- Dislike: B = 0.77, 95% CI [−0.39, 1.14], t(154) = 4.02, p < .001, but not tude objects (M = 4.82, SE = 0.11), Mdiff = 0.42, 95% CI [0.04, 0.63], Hate vs. Extreme Dislike: B = 0.23, 95% CI [−0.14, 0.61], t(154) = 1.21, t(152) = 2.90, p = .012, Cohen’s d = 0.37, see Figure 4(b). Extremely dis- p = .23). In particular, hated attitude objects (M = 5.99, SE = 0.15) liked attitude objects were also more related to morality than disliked were rated more negatively than disliked attitude objects (M = 5.21, SE ones, Mdiff = 0.56, 95% CI [0.04, 0.63], t(152) = 3.85, p < .001, Cohen’s = 0.15), Mdiff = 0.78, 95% CI [0.33, 1.24], t(152) = 4.10, p < .001, Cohen’s d = 0.49. We found no effect of attitude object class and no interaction d = 0.51, see Figure 4(a). However, hated attitude objects were not sig- of hate versus dislike and attitude object class (ps > .08). These find- nificantly different from extremely disliked attitude objects (M = 5.76, ings further support a distinction between hated and disliked attitudes SE = 0.15), Mdiff = 0.24, 95% CI [−0.22, 0.69], t(152) = 1.23, p = .44, objects rooted in morality.2 Cohen’s d = 0.09. Extremely disliked attitude objects were rated more negatively than disliked attitude objects, Mdiff = 0.55, 95% CI [0.10, 1.00], t(152) = 2.87, p = .013, Cohen’s d = 0.42. 2 In this study, a computer error made it impossible to obtain moral emotion items (i.e., con- tempt, anger, and disgust) as well as several of the PANAS-X items failed for the majority of We also found a main effect of attitude object class on ratings of neg- participants. For this reason, we report analyses for moral concern and moral conviction as ativity, such that people/groups (M = 5.86, SE = 0.14) were rated more opposed to the moral emotion analyses presented in Studies 1 and 2. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 348 PRETUS ET AL . 4.2.4 Effects of morality, adjusting for intensity hated and extremely disliked attitude objects and morality differ- and attitude strength ences between hated and extremely disliked attitude objects became non-significant after statistically adjusting for attitude strength. Consistent with hypotheses that the difference between hated and disliked attitude objects is both a matter of intensity and a matter 4.3 Discussion of morality, we tested the intercorrelations of our main dependent variables. Morality and negativity were highly significantly correlated This study further tested whether the difference between hated and (r = 0.59, p < .001) (see Table 1). To perform a more stringent disliked attitude objects is one of intensity or morality. As in the pre- test of the morality hypothesis, we re-analysed the effects of atti- vious studies, we found that hated objects were rated more negatively tude object type and attitude object class on morality, adjusting for than disliked attitude objects, supporting the intensity hypothesis. How- both between-person and within-person negativity. Even after control- ever, hated objects did not differ from extremely disliked objects in ling for between and within-person negativity, hated attitude objects terms of negativity and differences in negativity did not survive con- (M = 5.10, SE = 0.10) were more related to morality than disliked atti- trolling for morality, contradicting the idea that hate simply falls at the tude objects (M = 4.45, SE = 0.10), Mdiff = 0.65, 95% CI [0.35, 0.96], extreme end of a continuum of negativity. Rather, we found evidence t(151) = 5.09, p < .001, or extremely disliked attitude objects (M = 4.78, that hate is more connected to moral convictions when compared SE = 0.09), Mdiff = 0.32, 95% CI [0.03, 0.61], t(151) = 2.64, p = .025. to disliked attitude objects—even when adjusting for negativity and Extremely disliked attitude objects were more related to morality than attitude strength. Moreover, hated objects were also more related to disliked ones, Mdiff = 0.33, 95% CI [0.03, 0.63], t(151) = 2.64, p = .025. morality than extremely disliked objects after controlling for nega- We found no effect of attitude object class and no interaction of tivity, though these differences were only marginally significant after hate versus dislike and attitude object class (ps > .39). The effect of adjusting for attitude strength. Because attitude strength ratings were between-person negativity, B = 0.42, 95% CI [0.29, 0.57], t(75) = 5.70, similar for hated and extremely disliked attitude objects, attitude p < .001, and within-person negativity, B = 0.41, 95% CI [0.31, 0.51], strength cannot account for differences between hate and extremely t(151) = 7.99, p < .001, also remained significant in the model. disliked attitude objects. Thus, hate appears to be different from dislike Morality and attitude strength were significantly correlated in terms of morality rather than negativity or attitude strength. (r = 0.57, p < .001). Thus, we repeated the same procedure to evaluate Study 3 also examined the types of attitude objects participants the effects of morality after controlling for attitude strength. Hated evaluated: either people and groups or concepts and beliefs. Whereas attitude objects were still more associated to morality (M = 5.10, people and groups were rated more negatively than concepts and SE = 0.10) than disliked attitude objects (M = 4.42, SE = 0.10) after beliefs, they did not differ along the moral dimension, suggesting that controlling for attitude strength, Mdiff = 0.68, 95% CI [0.36, 1.00], all hated attitude objects share moral relevance. This suggests that our t(151) = 4.97, p < .001. However, the contrast between hated and conclusions about differences in morality generalize to multiple targets extremely disliked attitude objects (M = 4.82, SE = 0.10) became only of hate. Although we selected this distinction based on the types of atti- marginally significant, Mdiff = 0.28, 95% CI [−0.03, 0.59], t(151) = 2.14, tude objects generated in the first two studies, it is possible that such p = .085. The effects of between-person attitude strength, B = 0.68, instructions led to the selection of less central, more socially appropri- 95% CI [0.50, 0.87], t (75) = 7.22, p < .001, and within-person attitude ate attitude objects. To test this concern and see if our findings would strength, B = 0.47, 95% CI [0.32, 0.60], t(151) = 6.40, p < .001, were generalize to real world hate groups, we examined differences in the significant. online expression of hate versus dislike. When evaluating differences in attitude strength, hated attitude objects were associated with increased attitude strength (M = 4.96, SE = 0.11) compared to disliked objects (M = 4.32, SE = 0.11), 5 STUDY 4: HATE ONLINE Mdiff = 0.64, 95% CI [0.30, 0.98], t(152) = 4.44, p < .001, Cohen’s d = 0.56, though no differences were found between hated and extremely dis- In the fourth study, we investigated real world-instantiations of hate liked objects in terms of attitude strength (M = 4.66, SE = 0.11), versus dislike expressions on the Internet. Online media have been a Mdiff = 0.30, 95% CI [−0.04, 0.64], t(152) = 2.09, p = .095, Cohen’s prominent platform for transmitting messages of hate—even in 1995, d = 0.28, see Figure 4(c). Moreover, after adjusting for between- in the first days of the commercialized Internet, hate groups were some and within-participants morality, even differences in attitude strength of the earliest to leverage this technology to recruit members, orga- between hated compared to disliked attitude objects became non- nize, and transmit their beliefs. For instance, Stormfront.org is one of significant (p = .39). the world’s oldest hate groups—providing an Internet forum for over Because differences in negativity did not survive adjusting for 300,000 registered users who are affiliated with neo-Nazi and White morality, the findings from Study 3 support the morality hypothesis supremacist groups. This website is run by a former Ku Klux Klan leader but not the intensity hypothesis. Moreover, differences in morality and has been accused of promoting deadly violence, with connections between hated and disliked attitude objects remained significant even to nearly 100 killings (Holpuch, 2014). To date, hate groups flourish when adjusting for between-person and within-person negativity and online and use the Internet and social media to mobilize collective attitude strength. Of note, attitude strength was similar between action (Hoover et al., 2021). 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 349 We analysed the linguistic content of websites belonging to real, and Glassdoor) per subcategory or company to match the word length American hate groups as well as complaint forums (both consumer of hate websites. To compare the use of moral words, we manually complaints and forums expressing employee complaints against corpo- counted usages of the word “moral” and its synonyms.6 For example, rations), in order to determine if the language used on these websites one such usage by the hate group, Gallows Tree Wotansvolk declared: reveals whether they differ in their level of negativity or in their struc- ture as characterized by morality. Because hate groups often do not Votanism teaches lessons of morality and nobility, to walk identify with the label “hate group” themselves, we use a conceptual- as a proud White individual in a world where being White is ization of hate defined by how others see these groups rather than how now considered wrong . . . We understand honor to be one of they see themselves, unlike in Studies 1 to 3. Including an online sam- the foundation blocks which will support our healthy growth ple also allowed us to capture a diversity of perspectives not reflected and advancement. Thus, we wish to fill our ranks with men among our previous undergraduate samples. and women of honor . . . One example of moral word usage by a complaint site was fea- 5.1 Method tured on the Consumer Electronics complaint forum on Complaints- board.com: 5.1.1 Sample First they cheat customers by selling refurbished and used We sampled websites, including 46 sites of known hate groups products in the original packaging. Second, they have made (expressions of hate), 47 threads from different subcategories of a joke out of customers by being unethical and unrespon- an online consumer complaint forum (expressions of dislike towards sive. products/objects), and 51 threads of employee complaints about differ- ent corporations (expressions of dislike towards people/corporations). Finally, to search for differences in word length as well as the emo- Hate groups were selected from the Southern Poverty Law Center’s tional content of words, we employed the Linguistic Inquiry and Word Hate Map and chosen because they were among the most commonly Count programme (Pennebaker, Francis, & Booth, 2001). We compared noted groups among America’s 50 states that had publicly accessi- word frequencies, expressed as a percentage of the total word count, ble websites (and were not specifically noted as “separatist groups” or between hate sites and complaint forums. religious fundamentalist groups3 ). To serve as one natural comparison condition, consumer complaint forums were selected from subcategories of the international com- 5.2 Results plaint site, complaintsboard.com.4 This maintained consistency in the format of the different complaint threads. Employee complaints about We first analysed our data to determine that our hate sites and two corporations were selected from Glassdoor.com, one of the most forms of complaint sites did not fundamentally differ in number of widely used corporate review sites, which verifies posters and screens words. A one-way ANOVA revealed that the text taken from the hate content for accuracy, assuring as much consistency as possible across websites (M = 1279.65 SD = 1731.59) did not significantly differ from types of posts (Associated Press, 2013). We focused on the 51 compa- the consumer complaint forums (M = 1434.89, SD = 125.65) nor the nies with the greatest number of reviews that also had enough 2-star employee complaints (M = 1273.51, SD = 352.96) in terms of the aver- or lower reviews for us to cull to match the word count from the hate age overall word count, F(2,141) = .40, p = .67, η 2 p = .006. As such, group and complaint forum texts.5 Taking them together, we could any linguistic differences noted between the sites were not due to compare sites in which hate is expressed against sites in which dislike differences in composition length. is expressed (against objects or corporate groups). 5.2.1 Intensity hypothesis 5.1.2 Design and procedure We explored whether the different types of websites used different As all hate websites are different, in selecting the text for our sam- frequencies of negative words, as a test of the intensity hypothesis. The ple we tried to maintain as much consistency as possible within and LIWC dictionary examines negative language by comparing the usage between websites. From the hate websites we analysed the text from of approximately 495 different words related to negative emotion, their “About” and “Mission Statement” sections. We then analysed a sufficient number of complaint threads (both from Complaintsboard 6 We searched for the synonyms given by the Merriam Webster Dictionary—conscionable, ethical, honest, honourable, just, principled, scrupulous—in their noun, adjectival, and adver- 3 See osf.io/5u6my for a full list of sites and text files. bial forms. Antonyms (e.g., unethical, injustice) were included in this search; however, alternate 4 At the time of data collection, 17 August 2017, www.complaintsboard.com was the #1 hit meanings were not included (e.g., when “just” is taken to mean exactly or precisely). Addition- upon a Google search of “complaint forums.” ally, spelling errors were considered—when “moral” was meant to represent “morale” it was not 5Data was collected from Glassdoor in September 2019. included, but when “morale” was, by context, intended to mean “moral” it was counted. 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 350 PRETUS ET AL . including “annoyed”, “grief”, and “hurt”, and then provides the percent- controlled lab experiments, this overall pattern of findings suggests age of negative words per hate website (or complaint threads with that morality helps differentiate expressions of hate from expressions matched word length). We specially edited the LIWC dictionary to of dislike. remove synonyms for the moral emotions (e.g., anger, contempt, dis- gust) as well as words related to morality (e.g., steal, guilt, punish), to ensure that we were examining differences in negativity separate from 6 GENERAL DISCUSSION moral concerns. More frequent use of negative emotion words on the part of one type of site would lend support for the intensity hypothesis. In a combination of laboratory studies and a content analysis of real We found that the amount of negative words used differed across sites, online hate and complaint websites, we found initial evidence that dif- 95% confidence inter F(2,141) = 23.72, p < .001, η 2 p = .25. Specifically, ferences in people’s conceptualizations of hate and dislike are not only hate websites (M = 1.31, SD = 0.85) did not use more negative words a matter of negativity but also morality. Morality—via both the expres- than consumer complaint threads (M = 1.40, SD = 0.50), Mdiff = 0.09, sion of moral emotions and moral conviction—differentiates hated 95% CI [-0.22, 0.40], p = 0.76, Cohen’s d = 0.13. However, employee from disliked attitude objects. Individuals rated hated attitude objects complaints featured more negative words (M = 2.12, SD = 0.53) than in the lab as more closely connected to morality than disliked or even both hate sites, Mdiff = 0.81, 95% CI [0.50, 1.12], p < .001, Cohen’s d extremely disliked attitude objects. This distinction still held when = 1.16, and consumer complaints, Mdiff = 0.71, 95% CI [0.41, 1.03], p < adjusting for the relationship between morality and negativity. Further, .001, Cohen’s d = 1.39 . This suggests that hate sites do not evince more real websites known by the United States government to be organized negative emotion words than complaint sites and may even use less hate groups used significantly more moral language in expressing their negative emotion words than employee complaints. Thus, the intensity beliefs as compared with users on complaint forums venting their dis- hypothesis was not supported in this context of real online expressions like. Of note, we found an order effect in Study 1 such that differences of hate and dislike. between hate and dislike were less evident when participants were asked to generate disliked objects first. This suggests that people spon- taneously think about objects that are closer to objects they extremely 5.2.2 Morality hypothesis dislike or hate when asked about dislike without an explicit reference to hate. We next explored the morality hypothesis by comparing the usage Regarding the intensity hypothesis, we found mixed evidence for of the word “moral” and its synonyms across sites. There was a the role of negativity in distinguishing hate expressions from dislike. significant effect of site type predicting moral word usage, 95% con- In Studies 1 and 2, hated attitude objects were rated as more nega- fidence intervals F(2,141) = 11.35, p < .001, η 2 p = .14. Replicating tive than disliked attitude objects, even after controlling for morality, our three lab studies, the rhetoric on hate websites (M = 0.32, suggesting that both morality and negativity independently contribute SD = 0.50) contained significantly more moral words than consumer to hate. These results are aligned with recent work by Martínez et al. complaint forums (M = 0.04, SD = 0.07), Mdiff = 0.27, 95% CI [0.13, (2022), who found increased ratings in 11 self-reported negative emo- 0.42], p < .001, Cohen’s d = 0.78, , and significantly more moral words tions in response to hated compared to disliked targets. However, these than employee complaint forums (M = 0.11, SD = 0.311), Mdiff = 0.21, authors do not explore differences between hated and extremely dis- 95% CI [0.07, 0.35], p = 0.002, Cohen’s d = 0.59. Further, this pat- liked objects. We find this to be a relevant comparison in the light tern of results persisted when accounting for the variability in negative of Study 3, where we find that negativity differences between hated emotion word usage, F(2, 140) = 10.11, p < .001, η 2 p = .13. Thus, our and extremely disliked objects vanished after controlling for morality, analysis of close to 150 online websites and forum threads suggests suggesting that differences in morality accounted for observed dif- that hate groups use more moral language than complaint forums. ferences in negativity. Further, in Study 4, online expressions of hate did not use more negative language than online expressions of dis- like. Thus, whereas hate and dislike seem to differ in both intensity 5.3 Discussion and morality, it is possible that hate and extreme dislike differ mainly in the morality dimension. Future studies would benefit from employ- Consistent with our three laboratory studies, this analysis of real-world ing scales and statistical techniques that allow researchers to obtain hate groups lent further support to the morality hypothesis. The lan- uncorrelated measures of negativity, attitude strength, and morality to guage of hate groups was different, in the moral domain, from that of better assess the independent contribution of each of these constructs complain forums. These results reflect real, uncensored language used in distinguishing hate from dislike. by groups and individuals known to espouse hate or dislike, and often Although it seems easy to recognize expressions of hatred when known to take significant actions in support of these attitudes. Impor- we see them—at Nazi rallies or ethno-cultural genocide—hate is still tantly, the expressions of these groups and individuals are less likely to poorly understood from a scientific perspective. Our studies find that involve their own lay theories about hate or dislike—but rather their morality is a key ingredient that differentiates the conceptualization public positions on these issues. There are, admittedly, several differ- of hate from dislike in the minds of many people. These studies offer ences between hate websites and complaint forums and these data a springboard for empirical research into the psychology of hate. should be treated as preliminary. But taken together with the carefully Centuries of philosophical theory have laid the groundwork for more 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 351 rigorous empirical investigation. For example, our review of the lit- constructed like other complex emotional states rather than a natural erature raised the possibility that hate is motivational. Rempel and kind (see Barrett, 2006). While this is beyond the scope of the present Burris (2005) suggested that we will ignore disliked objects but will work, these are important distinctions that should be examined in wish to harm hated ones. In line with this, people feel more inclined future research on the psychology of hate. to engage in attack-oriented behaviours when they experience hate On a related note, while the conceptualization of anger, contempt, versus dislike (Martínez et al., 2022). Further, while the present lab- and disgust as distinctively moral emotions continues to receive empir- oratory studies manipulated the type of attitude object generated to ical support (see for instance Steiger & Reyna, 2017), other scholars test differences between groups, further research could reverse the have challenged this view, arguing that disgust may have a broader role relationship between our independent and dependent variables. An beyond morality or that anger can be triggered by other moral trans- important test of the connection between hate and morality would be gressions beyond autonomy (see Lomas, 2019). Thus, our results on to determine if experimentally inducing moral emotions could create the differences in moral emotions between hated and disliked attitude hate in a laboratory setting. However, the ethics of doing so must be objects should be treated with caution: whereas these emotions may carefully considered. be necessary for hatred to arise, they may not be sufficient. Whereas One potential alternative explanation is that our instructions to gen- higher ratings in anger, contempt, and disgust were to be expected in erate hated versus disliked attitude objects elicited different classes the hatred versus dislike condition, they should not be taken by them- of attitude objects (e.g., people and groups vs. concepts and beliefs). selves as unequivocal proof of the association between hatred and However, when we explicitly instructed people to generate different morality. classes of attitude objects, we found that the difference between hate The results of the present research might, eventually, be fruitfully versus dislike was robust across these classes. It is also possible that applied to psychological or behaviour interventions against hate. For hated versus disliked attitude objects differed systematically in level instance, work on relations between Israelis and Palestinians suggests of abstraction. Some work has found that people more readily apply that hatred toward the out-group differs from anger in terms of pro- their moral principles to the psychologically distant (Eyal et al., 2008). filing the out-group as evil and intentionally causing harm (Halperin, Perhaps hated attitude objects are more psychologically distant or 2008; see also Parker & Janoff-Bulman, 2013). Yet such conflict may higher in abstraction? Another alternative explanation for our find- be ameliorated and peace proposals more likely to be adopted when ings is that disliked versus hated objects do not need to have an actual the out-group is willing to compromise sacred values—rather than antecedent: whereas people may not know why they dislike something, economic concessions (Ginges et al., 2007). Thus, acknowledging and hatred may be more readily associated with a specific experience, mak- leveraging the moral concerns associated with hatred may provide an ing it easier to link to morality. Future research should address these important avenue for addressing intergroup (as well as interpersonal) possibilities. conflict. We urge research in these areas to continue this line of inquiry An additional reason we believe the differences between hate and in the hopes of designing and testing interventions to alleviate social dislike extend beyond these issues is our study of online hate groups. conflict. The websites we explored did not require users to list attitudes objects Finally, we note that the samples of our first three studies were they hated. In fact, many online hate groups actively disavow their cat- undergraduate students from Canada and the US. This poses a limi- egorization as “hate groups” and the content of their websites often tation in terms of the generalizability of the findings of these studies, focused on their core values (e.g., “. . . teaches lessons of morality and which have been drawn from western educated individuals from nobility, to walk as a proud White individual in a world where being industrialized, rich, democratic societies (WEIRD, see Henrich et al., White is now considered wrong”). Their websites were identified as 2010). We attempted to overcome this limitation in Study 4, where hate groups by third parties. Our analyses nevertheless found much we obtained samples from a more ecological environment (websites higher expressions of morality on these hate websites as compared with English-speaking audiences). Because different cultures could to complaint forums, both about objects and about corporate groups. have different conceptualizations of hate and dislike, future research Together with our lab experiments, this gives us confidence that the should further address this constraint by including cross-national difference between hate and dislike goes beyond simple semantics. representative samples. 6.1 Hate as emotion 7 CONCLUSION The current research relied on self-reports and content coding, which Hate is undeniably a topic of considerable theoretical and practical provides a modest scope for understanding the rich affective experi- interest. One only need open the newspaper or turn on the television ence of hate. At present, it is impossible to determine if hate causes to encounter daily instances of hate from rallies in Charlottesville, a feeling state or if labelling an experience as hate is a consequence Virginia to terrorist attacks in Niger. The rise in hate crimes in the US of an emotional experience (or both). Importantly, our use of the term suggests this issue is as urgent as ever. Although hate is easily recog- hate does not imply that it is a basic emotion. Our belief is that the nized, the empirical literature has largely ignored the topic. Our finding psychological state we colloquially associate with hatred is actively that morality plays a role in differentiating lay theories of hate from 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 352 PRETUS ET AL . dislike helps to address the paucity of research on this important topic. Barrett, L. F. (2006). Are emotions natural kinds? Perspectives in Psychological However, much more work needs to be done before psychologists will Science, 1, 28–58. https://doi.org/10.1111/j.1745-6916.2006.00003.x Bartlett, S. (2005). The pathology of man: A study of human evil. Charles C. have the capacity to address this issue in a meaningful way. Thomas Publisher. Ben-Ze’ev. (2000). The subtlety of emotions. The MIT press. AUTHOR CONTRIBUTION Berman, R. (2017). Hate crimes in the United States increased last Jay J. Van Bavel and William A. Cunningham designed S1-3; Jay J. Van year, the FBI says. The Washington Post, November 13, 2017. https://www.washingtonpost.com/news/post-nation/wp/2017/11/13/ Bavel, Jennifer L. Ray, and Yael Granot designed S4; Jay J. Van Bavel hate-crimes-in-the-united-states-increased-last-year-the-fbi-says/ programmed and conducted S1-3; Yael Granot and Jennifer L. Ray col- Brown, A. (2017). What is hate speech? Part 1: The Myth of Hate. Law lected data and analysed S4; Jay J. Van Bavel, Jennifer L. Ray, Yael and Philosophy, 36(4), 419–468. https://doi.org/10.1007/S10982-017- Granot and Clara Pretus analysed S1-3; Jay J. Van Bavel, Jennifer L. 9297-1 Capraro, V., & Rand, D. G. (2017). Do the right thing: Preferences for moral Ray, and Yael Granot drafted the manuscript; Jay J. Van Bavel, Yael behavior, rather than equity or efficiency per se, drive human proso- Garnot, Jennifer L. Ray, Clara Pretus and William A. Cunningham edited ciality. Judgment and Decision Making, 13(1), 99–111. https://doi.org/10. the manuscript. 2139/ssrn.2965067 Cavalli-Sforza, L. L., Feldman, M. W., Chen, K. H., & Dornbusch, S. M. (1982). ACKNOWLEDGEMENTS Theory and observation in cultural transmission. Science, 218(4567), 19– 27. https://doi.org/10.1126/science.7123211 The authors thank Stacey Fox, Doreen Kinley, Bryan Moeller, Erika Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2002). Applied multiple Price, and Jessica Wilkins for assisting with data collection and coding. regression/correlation analysis for the behavioral sciences. In Applied We also wish to thank members of the NYU Social Perception and Eval- multiple regression/correlation analysis for the behavioral sciences (3rd ed.). uation Lab, UT Computational Affective Neuroscience Lab, and OSU Routledge. https://doi.org/10.4324/9780203774441 de Freytas-Tamura, K. (2017). U.K. reports big rise in hate crime, Group for Attitudes and Persuasion for constructive comments. Parts citing Brexit and terrorist attacks. The New York Times, Octo- of this research were presented by Jennifer Ray to the Association ber 17, 2017. https://www.nytimes.com/2017/10/17/world/europe/uk- for Psychological Science in 2014, to the Attitudes Preconference at hate-crime-brexit-attacks.html the Society for Personality and Social Psychology in 2014, and to the de Houwer, J., & Hughes, S. (2020). Learning to like or dislike: Revealing similarities and differences between evaluative learning effects. Cur- Society for Personality and Social Psychology in 2015. This research rent Directions in Psychological Science, 29(5), 487–49. https://doi.org/10. was partially supported by a SSHRC Canada Graduate Scholarship and 1177/0963721420924752 NSF Award (#1349089) to Jay Van Bavel, a NSF Graduate Research Darwin, C. R. (1872). The expression of the emotions in man and animals. Fellowship to Yael Granot, and SSHRC Standard Research Grant and London: John Murray. 1st edition. Dozier, R. W. (2002). Why we hate: Understanding, curbing, and eliminating Mershon Center for International Security Studies Grant to William hate in ourselves and our world. Contemporary Books. Cunningham. Edwards, T. (1908). A cyclopedia of laconic quotations. F. B. Dickerson Co. Eyal, T., Liberman, N., & Trope, Y. (2008). Judging near and distant virtue and ETHICS STATEMENT vice. Journal of Experimental Social Psychology, 44, 1204–1209. https:// doi.org/10.1016/j.jesp.2008.03.012 This research was approved by the ethical review board of the Uni- Fehr, B., & Russell, J. A. (1984). Concept of emotion viewed from a pro- versity and Toronto and The Ohio State University. All experiments totype perspective. Journal of Experimental Psychology, 113, 464–486. included informed consent. https://doi.org/10.1037/0096-3445.113.3.464 Fischer, A. H., & Roseman, I. J. (2007). Beat them or ban them: The charac- CONFLICT OF INTEREST teristics and social functions of anger and contempt. Journal of Personality and Social Psychology, 93(1), 103–115. https://doi.org/10.1037/0022- The authors declare no conflict of interest. 3514.93.1.103 Fischer, A., Halperin, E., Canetti, D., & Jasini, A. (2018). Why we hate. DATA AVAILABILITY STATEMENT Emotion Review, 10(4), 309–320. https://doi.org/10.1177/17540739177 The data that support the findings of this study are openly available in 51229 Fitness, J., & Fletcher, G. J. O. (1993). Love, hate, anger, and jealousy in close OSF at https://doi.org/10.17605/OSF.IO/5U6MY. relationships: A prototype and cognitive appraisal analysis. Journal of Per- sonality and Social Psychology, 65(5), 942–958. https://doi.org/10.1037/ REFERENCES 0022-3514.65.5.942 Allport, G. (1954). The nature of prejudice. Addison-Wesley. Fitzgerald, L. F., Swan, S., & Magley, V. J. (1997). But was it really sexual Associated Press. (2013). Employees rate their employers, CEOs harassment? Legal, behavioral, and psychological definitions of the work- on Glassdoor. Canadian Broadcasting Company, Retrieved from: place victimization of women. In W. O’Donohue (Ed.), Sexual harassment: https://www.cbc.ca/news/business/employees-rate-their-employers- Theory, research, and treatment (pp. 5–28). Allyn & Bacon. ceos-on-glassdoor-1.1314945 Frijda, N., Mesquita, B., Sonnemans, J., & van Goozen, S. (1991). The dura- Atran, S., Sheikh, H., & Gomez, A. (2014). Devoted actors sacrifice for close tion of affective phenomena or emotions, sentiments and passions. In comrades and sacred cause. Proceedings of the National Academy of Sci- Kenneth T. Strongman (Ed.), International review of studies on emotion (pp. ences of the United States of America, 111(50), 17702–17703. https://doi. 187–225). Wiley. org/10.1073/pnas.1420474111 Ginges, J., Atran, S., Medin, D., & Shikaki, K. (2007). Sacred bounds on Aumer-Ryan, K., & Hatfield, E. (2007). The designs of everyday hate: A qual- rational resolution of violent political conflict. Proceedings of the National itative and quantitative analysis. Interpersona, 1, 143–172. https://doi. Academy of Sciences of the United States of America, 104(18), 7357–7360. org/10.5964/ijpr.v1i2.11 https://doi.org/10.1073/pnas.0701768104 10990992, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2906 by Readcube (Labtiva Inc.), Wiley Online Library on [28/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PSYCHOLOGY OF HATE 353 Glaser, J., Dixit, J., & Green, D. (2002). Studying hate crime with the internet: Pennebaker, J. W., Francis, M. E., & Booth, R. J. (2001). Linguistic inquiry and What makes racists advocate racial violence? Journal of Social Issues, 1, word count: LIWC 2001. Lawrence Erlbaum Associates, 71 (2001). 177–193. https://doi.org/10.1111/1540-4560.00255 Rempel, J. K., & Burris, C. T. (2005). Let me count the ways: An integrative Green, D., Abelson, R., & Garnett, M. (1999). The distinctive political views of theory of love and hate. Personal Relationships, 12(2), 297–313. https:// hate-crime perpetrators and white supremacists. In D. A. Prentice & D. T. doi.org/10.1111/j.1350-4126.2005.00116.x Miller (Eds.), Cultural divides: Understanding and overcoming group conflict Royzman, E. B., McCauley, C., & Rozin, P. (2006). From Plato to Putnam: (pp. 429–464). Russel Sage Foundation. Four ways to think about hate. In R. J. Sternberg (ed.), The psychology of Green, P., MacLeod, C., & Alday, P. (2016). Package ‘simr’. [ Computer hate (pp. 3–35). American Psychological Association. https://doi.org/10. software] https://cran.r-project.org/web/packages/simr/index.html 1037/10930-001 Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist Rozin, P. (1999). The process of moralization. Psychological Science, 10(3), approach to moral judgment. Psychological review, 108(4), 814. 218–221. https://doi.org/10.1111/1467-9280.00139 Haidt, J., Rosenberg, E., & Hom, H. (2003). Differentiating diversities: Moral Rozin, P., & Fallon, A. E. (1987). A perspective on disgust. Psychological diversity is not like other kinds. Journal of Applied Social Psychology, 33(1), Review, 94(1), 23–41. https://doi.org/10.1037/0033-295X.94.1.23 1–36. https://doi.org/10.1111/j.1559-1816.2003.tb02071.x Skitka, L. J. (2002). Do the means always justify the ends or do the ends Halperin, E. (2008). Group-based hatred in intractable conflict in Israel. sometimes justify the means? A value protection model of justice rea- Journal of Conflict Resolution, 52, 713–736. https://doi.org/10.1177/ soning. Personality and Social Psychology Bulletin, 28, 588–559. https:// 0022002708314665 doi.org/10.1177/0146167202288003 Halperin, E. (2011). Intergroup hatred: Psychological dimensions. In D. Skitka, L. J., Bauman, C. W., & Sargis, E. G. (2005). Moral conviction: Another J. Christie (Ed.), The encyclopedia of peace psychology (pp. 557–561). contributor to attitude strength or something more? Journal of Personal- Wiley-Blackwell. https://doi.org/10.1002/9780470672532.wbepp138 ity and Social Psychology, 88(6), 895–917. https://doi.org/10.1037/0022- Halperin, E., Canetti, D., & Kimhi, S. (2012). In love with hatred: rethink- 3514.88.6.895 ing the role hatred plays in shaping political behavior. Journal of Applied Skitka, L. J., & Mullen, E. (2002). The dark side of moral conviction. Analy- Social Psychology, 42(9), 2231–2256. https://doi.org/10.1111/j.1559- ses of Social Issues and Public Policy, 2(1), 35–41. https://doi.org/10.1111/ 1816.2012.00938.x j.1530-2415.2002.00024.x Henrich, J. (2022). Selective cultural processes generate adaptive heuristics. SPLC. (2017). Active Hate Groups 2016. In Intelligence Reports. Southern Science, 376(6588), 31–32. https://doi.org/10.1126/science.abo0713 Poverty Law Center. Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest people in the Staub, E. (2004). The origins and evolution of hate, with notes on prevention. world? Behavioral and Brain Sciences, 33(2–3), 61–83. https://doi.org/10. In Psychology of hate (pp. 51–65). American Psychological Association. 1017/S0140525%D7;0999152X https://doi.org/10.1037/10930-003 Holpuch, A. (2014). Almost 100 hate-crime murders linked to Steiger, R. L., & Reyna, C. (2017). Trait contempt, anger, disgust, and moral single website, report finds. The Guardian, April 18, 2014. foundation values. Personality and Individual Differences, 113, 125–135. https://www.theguardian.com/world/2014/apr/18/hate-crime- https://doi.org/10.1016/j.paid.2017.02.071 murders-website-stormfront-report Sternberg, R. J. (2003). A duplex theory of hate: Development and applica- Hoover, J., Atari, M., Mostafazadeh Davani, A., Kennedy, B., Portillo- tion to terrorism, massacres, and genocide. Review of General Psychology, Wightman, G., Yeh, L., & Dehghani, M. (2021). Investigating the role of 7(3), 299–328. https://doi.org/10.1037/1089-2680.7.3.299 group-based morality in extreme behavioral expressions of prejudice. Sternberg, R. J. (ed.) (2005). The psychology of hate. . American Psychological Nature Communications, 12(1), 1–13. https://doi.org/10.1038/s41467- Association. https://doi.org/10.1037/10930-000 021-24786-2 Van Bavel, J., Packer, D., Haas, I., & Cunningham, W. (2012). The impor- Hox, J. (1998). Multilevel Modeling: When and Why. In: Balderjahn, I., tance of moral construal: Moral versus non-moral construal elicits faster, Mathar, R., Schader, M. (eds) Classification, data analysis, and data high- more extreme, universal evaluations of the same actions. PLoS One, 7(11), ways: Studies in classification, data analysis, and knowledge organization. e4869. https://doi.org/10.1371/journal.pone.0048693 Springer. https://doi.org/10.1007/978-3-642-72087-1_17 Watson, D., & Clark, L. A. (1994). The PANAS-X: Manual for the positive and Krug, E. G., Dahlberg, L. L., Mercy, J. A., Zwi, A. B., & Lozano, R. (2002). negative affect schedule-expanded form. The University of Iowa. https:// World report on violence and health. Lancet, 360(9339), 1083–1088. doi.org/10.17077/48vt-m4t2 https://doi.org/10.1016/S0140-6736(02)11133-0 White, R. K. (1996). Why the Serbs fought: Motives and misperceptions. Lomas, T. (2019). Anger as a moral emotion: A “bird’s eye” systematic review. Peace and Conflict: Journal of Peace Psychology, 2(2), 109–128. https://doi. Counselling Psychology Quarterly, 32(3–4), 341–395. https://doi.org/10. org/10.1207/s15327949pac0202_2 1080/09515070.2019.1589421 Martínez, C. A., van Prooijen, J. W., & Lange, P. A. M. V. (2022). Hate: Toward understanding its distinctive features across interpersonal and intergroup targets. Emotion, 22(1), 46–63. https://doi.org/10.1037/ SUPPORTING INFORMATION EMO0001056 Additional supporting information can be found online in the Support- Meier, J. S. (1993). Notes from the underground: integrating psycho- ing Information section at the end of this article. logical and legal perspectives on domestic violence in theory and practice. Hofstra Law Review, 21(4), 1295–1366. Available at: http:// scholarlycommons.law.hofstra.edu/hlr/vol21/iss4/4 Ortony, A., Clore, G., & Collins, A. (1990). The cognitive structure of How to cite this article: Pretus, C., Ray, J. L., Granot, Y., emotions. Cambridge University Press. https://doi.org/10.1017/ CBO9780511571299 Cunningham, W. A., & Van Bavel, J. J. (2023). The psychology of Parker, M. T., & Janoff-Bulman, R. (2013). Lessons from morality-based hate: Moral concerns differentiate hate from dislike. European social identity: The power of outgroup “hate,” not just ingroup “love.” Journal of Social Psychology, 53, 336–353. Social Justice Research, 26(1), 81–96. https://doi.org/10.1007/s11211- https://doi.org/10.1002/ejsp.2906 012-0175-6