Seven Clarifications on the Psychology of Dehumanization

Dehumanization Is Perceiving a Person or Group as Lacking Humanness by Degree

Over invokes a “dehumanization hypothesis” that supposedly underpins dehumanization research. It proposes that “victims of intergroup harm are perceived as being similar to nonhuman entities” (p. proof 1) and “as a result, . . . natural inhibitions against causing them harm are eroded” (p. proof 1). This hypothesis is presented in an extremely strong form: It posits that out-groups (and only out-groups) are always dehumanized, that these out-groups are viewed as categorically equivalent to nonhuman entities, that if an entity is perceived as having a single human attribute it cannot have been dehumanized (e.g., Challenge 5), and that dehumanization invariably leads to harm.

Each of these propositions misrepresents contemporary dehumanization research and theory. No singular, shared dehumanization hypothesis exists. Different theoretical approaches to the psychology of dehumanization vary in their accounts of how the phenomenon should be conceptualized and studied. None of the accounts claim that all out-groups are dehumanized, and researchers have painstakingly examined dehumanization’s moderators and boundary conditions (for reviews, see Bain et al., 2014; Haslam & Loughnan, 2014; Vaes et al., 2012). There are no baldly absolutist claims that dehumanization always occurs toward all out-groups at all times, claims that are utterly implausible in our probabilistic and contextualized discipline. In addition, no psychological account presents dehumanization as a categorical or qualitative denial of humanity to humans. Rather, dehumanization is understood and measured as a matter of degree: Some humans are perceived as less than fully human or as less human than others according to ratings of human attributes, implicit associations with nonhuman entities, or other quantitative assessments. Indeed, contemporary dehumanization research has advanced greatly in showing how attributions of humanness vary by degree in subtle and relative ways.

It is also inaccurate to suggest that dehumanization research is guided by the hypothesis that dehumanization plays a causal role in harm. Much of that research examines dehumanization in relation to variables other than harm, such as helping, forgiveness, intergroup contact, moral judgment, stereotyping, and prejudiced attitudes. The fraction of dehumanization research that explicitly addresses harm rarely draws causal inferences, seeking instead to examine whether dehumanizing perceptions are implicated in harm. The claim that, for dehumanization researchers, it is axiomatic that dehumanization causes harm by eroding inhibitions may have been true of some early work (e.g., Bandura et al., 1975), but most recent research does not subscribe to this account, viewing disinhibition as merely one of several potential mechanisms linking dehumanization to harm commission (e.g., moral exclusion, delegitimization, diminished empathy, increased blame). Indeed, work that does address harm sometimes examines how dehumanization justifies it after it has been committed rather than before via disinhibition (e.g., Castano & Giner-Sorolla, 2006). In short, Over’s putative dehumanization hypothesis is a fiction that fails to capture contemporary dehumanization research.

Dehumanization Research Is Grounded in Systematic Studies of Lay Conceptions of Humanness

Over’s Challenge 5 criticizes Haslam’s dual conception of humanness as incomplete. She argues that his conception omits important antisocial attributes such as jealousy, spite, and dishonesty that might define an additional dimension of humanness. We address this issue below in the section Dehumanization Is Not Reducible to Negative Evaluation, arguing that these attributes—and others that differentiate humans from angels—are better conceptualized as indicative of evaluation rather than humanness. Over also criticizes the derivation of Haslam’s two dimensions of humanness, human nature and human uniqueness (Haslam, 2006; Haslam et al., 2005), on the grounds that it involved soliciting judgments of characteristics that were typical of the human category when such judgments depend on the comparison point and are therefore contextually variant.

We are unpersuaded by this criticism. First, the judgment task for assessing human uniqueness did not assess “typicality” and had an explicit comparison point (i.e., “this characteristic is experienced solely by human beings and is not experienced by animals”). The salience of the human–animal contrast in conceptions of humanness has a deep history and underpinned the earlier research on infrahumanization that inspired Haslam’s work (Leyens et al., 2000). Second, the judgments of human nature were not in fact judgments of typicality but of human nature itself (i.e., “this characteristic is an aspect of human nature”). Although this judgment has no explicit comparison point, subsequent research indicated that it aligned closely with the human/machine comparison (Haslam et al., 2008) and that it was remarkably consistent across diverse cultures (Park et al., 2012), suggesting it is not highly contextual. The fact that the two proposed dimensions of humanness correspond closely to the dimensions of mind perception revealed by Gray et al. (2007) also suggests that the dimensions are important, nonarbitrary, and replicable. Finally, Haslam has never argued that his two dimensions of humanness are exhaustive. Judgments of how humans differ from other comparison points may be worth exploring, although we question whether ad hoc contrasts with zombies and dolphins are likely to illuminate basic elements of lay conceptions of humanness.

Dehumanization Is a Psychological Process That Extends Beyond Animal Metaphors

The bulk of Over’s article addresses the consequences and functions of dehumanizing language in hate speech and propaganda—that is, the organized deployment of dehumanizing language—as if this were the primary domain or essence of dehumanization. However, dehumanization research in psychology has mostly examined the psychological processes involved in dehumanization and has only incidentally touched on language use as one among many manifestations. The vast majority of dehumanization research does not address language use directly but examines dehumanization through visual representations (e.g., Kteily et al., 2015) and perceptual (Fincher & Tetlock, 2016), cognitive (Vaes et al., 2020), or neural (Vaes et al., 2019) mechanisms. As a result, much of Over’s critique conflates the specific case of dehumanizing language use (e.g., animal metaphors) with the substantially broader understanding of dehumanization as a set of psychological processes.

Over’s critique of dehumanization research is mistaken even on the specific topic of dehumanizing language use. It suggests that any reference to a person using a nonhuman metaphor is by definition dehumanizing (e.g., in Challenges 1, 3, and 5). No such claim is made in dehumanization research, some of which has explicitly disavowed the view that likening someone to an animal is intrinsically dehumanizing and has documented how animal metaphors vary widely in the extent to which they are positive or negative, inoffensive, or dehumanizing (Haslam et al., 2011). Research has also examined how the functions and consequences of nonhuman metaphors go beyond contributing to a dehumanized perception of groups and individuals (Tipler & Ruscher, 2019). Over’s restricted focus on dehumanizing language in propaganda and hate speech fails to recognize the range and complexities of dehumanization and language use.

Humanness Is Contextual and Dehumanization May Be Subtle

In her critique of the dehumanization hypothesis, Over states that human and nonhuman characteristics or metaphors are often attributed to or associated with the “wrong” targets. In-group members might be associated with nonhuman entities (Challenge 1), whereas out-groups are granted human attributes (Challenges 4 and 5). These observations pose no significant challenge to dehumanization research for three reasons. First, as we discuss elsewhere in this commentary, dehumanization does not entail that out-groups are always and only ascribed nonhuman attributes or that in-groups are never ascribed such attributes. Dehumanization is often a subtle matter of degree, whereby in-group and out-group commonly (but not always) differ in average levels of ascribed humanness. Second, even in contexts in which dehumanization is not subtle, out-groups need not always be dehumanized: Like any other manifestation of intergroup friction, the extent to which dehumanization occurs may depend on moderators such as the presence of threat or conflict. Third, the humanness of an attribute need not be fixed but can depend on the context and target. Paladino and Vaes (2009), for example, showed that the same characteristic (e.g., courage or irrationality) changed in humanness as a function of the group to whom it was attributed. When participants were convinced that the trait was typical of their in-group they judged it more human than when the same characteristic belonged to the out-group. We should therefore not be surprised when in-group members are associated with certain animals or nonuniquely human qualities or when out-group members are ascribed human qualities.

Humanness and Moral Evaluation Are Related but Distinct

Over emphasizes the importance of a set of antisocial attributes that are specific to humans and frequently attributed to out-groups. She presents their omission from dehumanization research as a problem for it (see Challenges 2, 4, and 6), but we disagree. These negative attributes reflect highly moralized judgments that apply especially to humans, but this does not automatically make them suitable for assessing dehumanization. First, dehumanization researchers have been careful to differentiate humanness from evaluation to ensure that dehumanization is conceptually and psychometrically distinct from derogation and prejudice. Failure to do so would make it impossible to disentangle the causes, correlates, and consequences of these phenomena. Terms such as “spiteful” and “disloyal” clearly convey negative evaluation, but only if they also capture salient aspects of humanness would they be suitable for assessing dehumanization. Second, to capture salient aspects of humanness, traits such as these must not only be especially applicable to humans—as most human personality trait terms surely are—they must also be judged particularly indicative of humanness relative to other trait terms. Dehumanization researchers have chosen to develop measures of humanness on the basis of judgments of the extent to which traits or emotions are uniquely or typically human in this comparative fashion, not simply based on a judgment that the traits or emotions tend to apply especially to human targets.

This is not to say that negative evaluation and dehumanization are unrelated. There is ample evidence that moral judgments and dehumanization are closely associated (e.g., Bastian et al., 2011). For example, simply identifying a face as belonging to someone morally culpable leads the person to be perceptually dehumanized (Fincher & Tetlock, 2016). Indeed, immoral actions, even those that are arguably uniquely human (e.g., murder, torture), clearly have dehumanizing implications both for the way their victims (Bastian & Haslam, 2010) and perpetrators (Bastian et al., 2013) are perceived. Nevertheless, evaluation and humanness are conceptually distinct, and judgments of groups on these dimensions are only moderately correlated (Kteily et al., 2015). We agree with Over that evaluative attributes play a major role in social perception but not that their role challenges the existing research on dehumanization.

Dehumanization Is Associated With Many Outcomes Beyond Harm

Over’s Challenge 7 boils down to the observation that some dehumanized targets are not subjected to harm. This observation would undermine dehumanization research only if it claimed that dehumanization invariably leads to harm. It does not, however; nor does any credible psychological claim propose that one variable predicts or causes another without exception. Research has commonly linked dehumanization processes not only to forms of harm but also to diverse phenomena, including protection from burnout among health-care workers (Vaes & Muratore, 2013) and prosocial behavior motivated by a desire for people to reclaim their humanity (Bastian et al., 2013). When research has linked dehumanizing perceptions to harm it has presented them as statistical inferences, not deterministic laws that disallow exceptions.

Challenge 7 also mentions the positive treatment of nonhuman animals. We fail to see how this should be problematic for research on dehumanization. To demonstrate that certain humans are likened to certain nonhumans in a way that is associated with the unpleasant treatment of those humans in no way entails that all animals are treated unpleasantly. The fact that some nonhuman animals are loved and nurtured does not override the fact that some animals are seen as threatening (e.g., apes) or disgusting (e.g., rats), that the broad idea of animality (in relation to humanity) often symbolizes backwardness and barbarity (Haslam et al., 2011), and that human attitudes toward animals can be remarkably inconsistent, combining ostensible loving with actual killing (e.g., the “meat paradox”; Loughnan, Haslam, & Bastian, 2010). It can also be argued that some of the positive treatment of animals derives from processes of anthropomorphism (e.g., Epley et al., 2007), the mirror image of dehumanization.

We believe Over’s inference that the positive treatment of animals invalidates dehumanization research in which people are likened to animals rests on the same misunderstandings of dehumanization discussed above. Animalistic dehumanization—perceiving a person or group as resembling animals in some fashion, as having fewer attributes that differentiate humans from animals than other people, or implicitly associating them more than others with animals—does not amount to believing dehumanized people are categorically or literally nonhuman. It involves perceiving them as less human than other humans along a dimension of degree of likeness to animals, a matter of attenuated humanness rather than banishment from the human category. That some animals are treated well therefore has no relevance to whether people perceived as less human than others, but still categorically human, tend to be mistreated.

Dehumanization Is Not Reducible to Negative Evaluation

Over’s alternative account of the use of animal metaphors is that they are ways to represent targets in a negative fashion. Of course, propagandistic uses of animal metaphors get much of their toxic charge from the negative meanings around disgust, threat, and degradation that these metaphors carry. However, the fact that those virulent meanings are encapsulated in animal terms remains psychologically important over and above their negativity, conveying the additional message that some people are lesser humans than others and do not deserve humane treatment.

More generally, however, dehumanization cannot be reduced to negativity as Over’s account implies. Theorists and researchers have taken care to differentiate humanness from valence and dehumanization from prejudice precisely because they wanted to determine whether dehumanization as a concept goes beyond mere negativity. Infrahumanization research has defined humanness in terms of the species-uniqueness of emotions, a variable orthogonal to emotion valence. Trait sets for assessing Haslam’s dimensions of humanness are likewise orthogonal to trait valence, again so that perceived humanness can be assessed separately from evaluation. Bruneau et al. (2018) showed that dehumanization and negative evaluation are only moderately correlated, independently predict numerous phenomena, and even have distinct neural bases. To reduce the use of animal metaphors or other possible forms of dehumanization to mere negative evaluation is therefore to overlook what is specific to dehumanization.

This is also true in the eye of the target. The Black–ape association goes beyond mere negativity; it is dehumanizing and felt as such by Black targets. Evidence that victims of ostracism self-dehumanize (Bastian & Haslam, 2010), and that minority group members are often aware of being dehumanized (metadehumanization) with important outcomes that go beyond the experience of feeling disliked (reciprocal dehumanization; Kteily & Bruneau, 2017), demonstrates that experienced dehumanization is not reducible to experienced derogation.

Conclusion

Over’s challenges to dehumanization research and theory provide an opportunity to clarify some of its basic assumptions. However, her critique fundamentally misinterprets and misrepresents much of that work. At the root of this misinterpretation, we suspect, is a misunderstanding of what dehumanization researchers in psychology believe the phenomenon to be. To reiterate, dehumanization is not understood as the perception that certain people are literally, qualitatively, or categorically nonhuman—an ontological delusion—but that they are lesser humans. This differentiated, quantitative understanding of dehumanization differs from some understandings in philosophy and genocide studies. When dehumanization researchers’ work is understood correctly, it is not undermined by the fact that some animals are loved or that some humans are hated for their human qualities.