Original Research Reports Bleeding-Heart Liberals and Hard-Hearted Conservatives: Subtle Political Dehumanization Through Differential Attributions of Human Nature and Human Uniqueness Traits

This research demonstrated that human nature (HN) and human uniqueness (HU) traits capture the content of Americans' stereotypes about liberals and conservatives, respectively. Consistent with expectations derived from dehumanization theory, people more strongly associated HN traits with liberals than with conservatives, and more strongly associated HU traits with conservatives than with liberals. A trait × target ideology × perceiver ideology × trait valence interaction suggested that both liberals and conservatives more strongly associated their ingroup with stereotype-consistent positive traits, and their outgroup with stereotype-consistent negative traits. Mediation analyses revealed that outgroup antipathy, but not ingroup liking, explained the relationship between ideology and political outgroup dehumanization. Finally, humanness traits captured subtle differences in political stereotype content not captured with the warmth and competence dimensions derived from the stereotype content model. Together, these results indicate that differential attributions of HN and HU traits capture political stereotype content and function to subtly dehumanize one's political opponents. which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. " The man who is not a socialist at twenty has no heart, but if he is still a socialist at forty he has no head. " Aristide Briand (1862-1932) Both liberals and conservatives express prejudice and intolerance towards each other and like other forms of prejudice, these attitudes are likely linked to stereotyped beliefs about these political groups (Fiske & Taylor, 2008). It is unsurprising then that American political discourse is saturated with epithets steeped in stereotypes: latte liberals, fat cats, bleeding hearts, and rednecks. In this paper, we attempt to understand the content of political stereotypes in the U. S., and how these stereotypes function to dehumanize one's political opponents.

sual across both men and women, and Australian and Chinese participants.Importantly, these differential attributions do not necessarily involve hostility towards the target group (Bain et al., 2009).Indeed, many of these comparisons are in the absence of strong intergroup hostilities (e.g., women and men; psychology and medical students), which may explain why these beliefs are oftentimes shared across groups (e.g., Bain et al., 2006;Graham et al., 2012).
Further, instead of universally attributing outgroups with less humanness, these subtle forms of dehumanization can also involve attributing an excess of certain types of humanness to outgroups.For example, Bain et al. (2009) found that whereas Australians attributed greater HU to Chinese targets, Chinese attributed greater HN to Australian targets.Thus, subtle mechanistic and animalistic dehumanization processes need not necessarily involve universally attributing less humanness to outgroups-instead, some outgroups may be attributed excessive levels of certain traits as a way to perpetuate group stereotypes and maintain positive group distinctiveness (Bain et al., 2009;Bain, Haslam, DeSouza, & Kashima, 2008).
We suggest that human nature and human uniqueness traits also differentially capture the content of liberal and conservative stereotypes, respectively.Common liberal stereotypes such as the "bleeding heart", as well as many of the qualities identified in empirical studies of liberal stereotypes in the U.S. (e.g., young, emotional, compassionate, sensitive, quick-acting; Graham et al., 2012;Udolf, 1973), reflect human nature qualities (e.g., passionate, trusting).In contrast, popular notions of conservatives as "rigid" or "self-interested", as well as qualities identified by empirical studies in the U. S. (e.g., conventional, strict, rigid;Carney, Jost, Gosling, & Potter, 2008;Udolf, 1973) reflect qualities unique to humans (e.g., cold, organized).Moreover, theoretical perspectives that characterize liberals and conservatives as respectively applying nurturant versus strict parenting styles (Lakoff, 2002) and approach versus inhibition/avoidance motives (Janoff-Bulman, 2009) to their political attitudes are also generally consistent both with the "bleeding heart" liberal and "hard-hearted" conservative stereotypes and with HN and HU traits, respectively.Thus, based on popular notions as well as theory and research in political psychology, we expect Americans to associate liberals with HN traits and conservatives with HU traits.
As mentioned above, people subtly dehumanize groups even towards whom they feel little hostility.This is clearly not the case across the ideological divide-liberals and conservatives readily express prejudice and intolerance towards each other (see Brandt, Reyna, Chambers, Crawford, & Wetherell, in press for a review).Thus, in addition to observing political stereotypes shared across the political spectrum, we might also expect antipathy towards one's political opponents to motivate subtle dehumanization tendencies.We therefore expected that conservatives' tendency to attribute HN traits to liberals would be motivated by antipathy towards liberals, and that liberals' tendency to attribute HU traits to conservatives would be motivated by antipathy towards conservatives.Dehumanization effects are hypothesized to occur independent of trait valence (e.g., Bain et al., 2009).Of course, given the intense hostility between liberals and conservatives in the United States and deepening political polarization (Layman & Carsey, 2002), however, we might anticipate valence effects on liberals' and conservatives' trait ratings of themselves and each other.We therefore varied the valence of the assessed traits to determine whether valence moderated the expected interaction between target ideology (i.e., liberal vs. conservative) and humanness type (i.e., HN vs. HU).
Finally, any analysis of stereotype content should recognize the warmth and competence dimensions that research on the stereotype content model (SCM; Fiske et al., 2002) has shown are fundamental to social perceptions.According to the SCM, the warmth dimension captures traits associated with the perceived intent of a group, whereas the competence dimension captures traits associated with the perceived ability of a group (Fiske, Cuddy, & Glick, 2007).Haslam et al. (2009) argue that the warmth and competence dimensions are not redundant with human nature and human uniqueness dimensions, respectively.For instance, whereas warmth and competence generally reflect positive traits, HN and HU reflect both positive and negative traits.Furthermore, whereas Agreeableness corresponds to warmth and Conscientiousness to competence, HN encapsulates Openness, Extraversion and Neuroticism, whereas HU reflects both Conscientiousness and Agreeableness (Haslam, Bain, Douge, Lee, & Bastian, 2005).Thus, for the purposes of contrasting the expected differential attribution of HN and HU traits with a well-validated model of stereotype content, we also examined people's differential attributions of warmth and competence to liberals and conservatives.

Materials and Procedure
Participants evaluated liberals and conservatives on 10 HU and 10 HN traits drawn from three sources in the dehumanization literature (specifically, Haslam et al., 2005;Loughnan et al., 2009;Loughnan & Haslam, 2007).We intentionally avoided choosing stimuli that had clear political implications (i.e., "conservative").HU and HN traits were balanced for valence (see Table 1).At the same time, participants also rated liberals and conservatives on four warmth traits (warm, honest, well-intentioned, friendly) and four competence traits (competent, intelligent, skillful, capable) drawn from Fiske et al. (2002).For all 28 traits, participants were asked, "To what extent do you think each of the following traits characterize liberals [conservatives]", and responded to a 6-point item (1 = Strongly disagree; 6 = Strongly agree).Target order and trait order were randomized across participants.
On a separate page, participants were then again presented with the list of HN and HU traits and selected the five they believed best defined "humanness".This procedure determines whether participants generally define humanness more in terms of HN or HU, as recent evidence suggests cultural variation in defining humanness (e.g., Chinese and Australians define humanness more in terms of HU and HN, respectively;Bain, Vaes, Kashima, Haslam, & Guan, 2012).Participants then reported their political ideology

Overview of the Analyses
We first report descriptive statistics for and bivariate correlations among the variables, followed by analyses of which traits (HN vs. HU) people more frequently used to define humanness.We then test our primary hypotheses regarding the differential attributions of HN and HU traits to liberals and conservatives using a 3 (Perceiver Ideology: liberal, moderate, conservative) × 2 (Target Ideology: liberal, conservative) × 2 (Trait: HN, HU) × 2 (Valence: positive, negative) mixed model ANOVA, with Perceiver Ideology as a between-subjects factor and Target Ideology, Trait, and Valence as within-subjects factors.
We treated the continuous ideology variable as a categorical variable in mixed model ANOVA for two primary reasons.First, the U.S. political system is a deeply polarized two-party electoral system, with the two major political parties, Democrats and Republicans, generally endorsing liberal/left-wing versus conservative/right-wing political philosophies, respectively (Delli Carpini & Keeter, 1993).Thus, whereas people certainly vary in the strength of their liberal or conservative political identifications, at least in the U.S. context, the liberal and conservative labels historically predict endorsement of candidates from the two major parties (Jost, 2006), and strongly influence people's perceptions of political candidates, even above party membership information (Crawford, Jussim, Madon, Cain, & Stevens, 2011).Second, given a design with three within-subjects variables and one between-subjects variable, the results are easier to illustrate with ANOVA than with multiple regression.We used scores on the 7-point political ideology item to categorize participants below the midpoint as liberals (N = 59), at the midpoint as moderates i (N = 34), and above the midpoint as conservatives (N = 25).
Following this analysis, we use tests of multiple mediation to determine whether outgroup antipathy and ingroup liking explained the relationship between political ideology and the animalistic and mechanistic dehumanization of political groups.Finally, we examined the value of a stereotype content model approach to political stereotypes with a 3 (Perceiver Ideology: liberal, moderate, conservative) × 2 (Target ideology: liberal, conservatives) × 2 (Dimension: warmth, competence) mixed model ANOVA, with Perceiver Ideology as a between-subjects factor and Target ideology and Dimension as within-subjects factors.

Correlations and Descriptive Statistics
Table 2 reports the correlations among and descriptive statistics for each of the variables.Regardless of trait type (i.e., HN or HU), conservatism positively correlated with attributions of positive traits to conservatives and negative traits to liberals, and negatively correlated with attributions of negative traits to conservatives and positive traits to liberals.These results suggest strong ingroup favoritism and outgroup derogation in this sample, consistent with other research on political hostility (Brandt et al., in press).

Defining Humanness
A 3 (Perceiver Ideology: liberal, moderate, conservative) × 2 (Humanness: HN, HU) mixed model ANOVA on humanness definitions with Perceiver Ideology as a between-subjects factor and Humanness as a within-subjects factor revealed a Humanness main effect, such that people chose more HN words (M = 3.41, SD = 1.12) than HU words (M = 1.59,SD = 1.12) to define humanness, F(1, 115) = 65.53,p < .001,η p 2 = .36.The Perceiver Ideology × Humanness interaction was not significant, F(2,115) = .15,p = .858.Moreover, a single-sample t-test indicated that the average number of HN words chosen (3.41) was significantly above the 2.50 midpoint, t(117) = 8.80, p < .001,d = .63.Thus, like Australian participants in Bain et al. (2012), Americans across the political spectrum defined humanness more in terms of HN than HU.We address the implications of these findings for political dehumanization in the Discussion section.
ii Table 4 presents the results of simple effects analyses for all target comparisons.
Liberals attributed more positive traits to liberals than to conservatives, and more negative traits to conservatives than to liberals, clearly reflecting political hostility.That said, comparisons of the effect sizes suggest that liberals perceived greater differences between liberals and conservatives on positive HN traits than on positive HU traits, and on negative HU traits than on negative HN traits.These findings imply that liberals sought to see themselves as distinct from conservatives on positive traits associated with liberals (HN traits) and on negative traits associated with conservatives (HU traits).Moderates' perceptions clearly reflect this stereotype content: they attributed more positive HN traits to liberals and more negative HU traits to conservatives, but did not perceive ideological differences on negative HN and positive HU traits.Like liberals, conservatives also sought positive distinctiveness in differential trait attribution: they attributed more positive HU traits to themselves and more negative HN traits to liberals, but did not perceive ideological differences on positive HN or negative HU traits.
Simple effects on trait comparisons, presented in Table 5, provide an alternative view of these findings.Regardless of perceiver ideology or trait valence, liberals were rated higher in HN than HU traits.Beliefs about conservatives, however, were more nuanced.Moderates tended to rate conservatives higher in positive HU traits than positive HN traits and higher in negative HU traits than negative HN traits, although these effects did not reach conventional levels of significance.While liberals rated conservatives equally low in positive HN and HU traits, they rated them Note.df = 1,101; HN = human nature traits; HU = human uniqueness traits.
far higher in negative HU traits than negative HN traits.However, conservatives rated themselves equally high in HN and HU traits, regardless of trait valence.

Mediation by Outgroup Antipathy
We next examined whether antipathy towards conservatives explained the relationship between liberalism and the mechanistic dehumanization of conservatives, and whether antipathy towards liberals explained the relationship between conservatism and the animalistic dehumanization of liberals.To operationalize animalistic dehumanization, we computed difference scores reflecting greater attribution of HN than HU traits to liberals by trait valence (i.e., positive HN -positive HU; negative HN -negative HU).To operationalize mechanistic dehumanization, we computed difference scores reflecting greater attribution of HU than HN traits to conservatives by trait valence (i.e., positive HU -positive HN; negative HU -negative HN). iii We first examined bivariate correlations between ideology and these difference scores; ideology was negatively correlated with mechanistic dehumanization of conservatives on negative traits, r(112) = -.24,p < .001,indicating that liberals were more likely than conservatives to associate more negative HU than HN traits with conservatives.
Ideology was also positively correlated with animalistic dehumanization of liberals on positive traits, r(113) = .20,p = .030,indicating that conservatives were more likely than liberals to associate more positive HN than HU traits with liberals.Ideology was uncorrelated with the difference score for liberals on negative traits, r(110) = -.07,p = .495,and the difference score for conservatives on positive traits, r(113) = .10,p = .247.
We therefore conducted two separate multiple mediation analyses on the two difference scores that produced significant correlations with ideology, using Preacher and Hayes' (2008) INDIRECT macro.In each model, we specified ideology as the independent variable and feeling thermometer ratings towards liberals and conservatives as separate mediator variables.
iv Figure 1 displays the analysis of mechanistic dehumanization of conservatives on negative traits.Ideology predicted both antipathy towards conservatives and liking of liberals.However, antipathy towards conservatives, but not liking of liberals, predicted mechanistic dehumanization of conservatives.Including these two mediators in the model reduced the relationship between ideology and mechanistic dehumanization to non-significance, and the indirect effect of antipathy towards conservatives was significant (see Figure 1 note).
Reversing the arrangement of outgroup antipathy and mechanistic dehumanization suggested that mechanistic dehumanization did not mediate the relationship between liberalism and outgroup antipathy, as liberalism had significant direct effects on outgroup antipathy when controlling for mechanistic dehumanization, b = -10.01,SE = 1.16, p < .001.

Discussion
This study provides the first evidence that differential attributions of human nature and human uniqueness traits capture the content of Americans' stereotypes about liberals and conservatives, respectively.A robust Target Ideology × Trait interaction indicated that as predicted, people assigned more human nature (HN) traits to liberals than to conservatives, and more human uniqueness (HU) traits to conservatives than to liberals.Alternatively, people assigned more HN than HU traits to liberals, and more HU than HN traits to conservatives.This interaction was not moderated by perceiver ideology, indicating that across the political spectrum, HN and HU traits generally reflect the stereotype content of the U.S. political left and right, respectively.
That said, an unexpected Perceiver Ideology × Target Ideology × Trait × Valence interaction indicated that the valence of the trait mattered when liberals and conservatives made ingroup and outgroup trait associations.
Generally speaking, both liberals and conservatives more strongly associated their ingroup with stereotype-consistent positive traits, and their outgroup with stereotype-consistent negative traits.Whereas liberals more strongly associated positive traits with themselves than with conservatives regardless of humanness type, this difference was especially pronounced on positive HN traits.Similarly, conservatives more strongly associated positive HU traits with themselves than with liberals, but did not make differential attributions on (counter-stereotypical) positive HN traits.Thus, consistent with expectations from social identity theory (Tajfel & Turner, 1986), both liberals and conservatives appear to engage in social creativity strategies towards positive distinctiveness, seeing their own group as more strongly embodying traits central to the group identity than their outgroup.Such stereotype-consistent trait differentiation also occurred on negative traits.Specifically, whereas liberals more strongly associated negative traits with conservatives than with liberals regardless of humanness type, this difference was especially pronounced on negative HU traits.Similarly, conservatives more strongly associated negative HN traits with liberals than with conservatives, but did not make differential attributions on (counter-stereotypical) negative HU traits.
Examining trait attribution comparisons, people across the political spectrum more strongly attributed HN than HU traits to liberals, regardless of trait valence.It is unsurprising that liberals assigned more HN than HU traits to themselves, given that they (along with the entire sample) defined humanness more in terms of human nature than human uniqueness.The greater attribution of HU than HN traits to conservatives depended on the ideology of the perceiver and trait valence.Whereas liberals rated conservatives equally low in positive HU and HN traits, they rated them far higher in negative HU than HN traits; thus, these findings suggest that liberals believe conservatives have an excess of negative HU traits, both overall and compared to themselves.Conservatives, however, were reluctant to attribute greater HU than HN traits to their own group, regardless of valence, likely because they defined humanness more in terms of HN than HU.
The mediation analyses suggest that outgroup derogration, not ingroup favoritism, motivates subtle forms of dehumanization among both liberals and conservatives.Outgroup antipathy (but not ingroup liking) mediated the relationship between liberalism and the mechanistic dehumanization of conservatives, operationalized as the tendency to assign more negative HU than HN traits to conservatives.Given that liberals define humanness more so in terms of HN than HU traits, this greater attribution of negative HU than HN traits to conservatives appears especially malicious.
Further, outgroup antipathy (but not ingroup liking) mediated the relationship between conservatism and the animalistic dehumanization of liberals.One of the more interesting findings in this study is that animalistic dehumanization of liberals occurred through more strongly associating liberals with positive, not negative, HN than HU traits.It does not appear that conservatives believed liberals have a deficit in positive HU traits, as their average rating of liberals on positive HU traits (M = 3.68, SD = .85)is above the midpoint of the 6-point scale.Instead, this finding suggests that conservatives believe liberals have an excess of positive HN traits.Indeed, as seen in Table 4, conservatives rate liberals as slightly but not significantly higher in positive HN traits than they do themselves.
To investigate this idea further, we asked a separate sample of thirty-seven self-identified conservatives (recruited from MTurk) to describe the types of liberals they think are "childlike and emotional" and "cold and rigid".Interest-ingly, these conservatives had little trouble listing a host of childlike and emotional liberals (e.g., young people; environmentalists; supporters of generous welfare policies), and several volunteered that "all liberals" were childlike and emotional.In contrast, while some conservatives indicated that liberal elites or bureaucrats were cold and rigid, several volunteered that liberals are actually not cold and rigid, and if anything, are "warm to a fault".These qualitative responses are consistent with an interpretation of our findings suggesting that conservatives think liberals suffer from an excess of positive HN traits, and that prejudice against liberals drives this belief (see Figure 2).
These findings also suggest that HN and HU traits reflect stereotype content in a different manner than the warmth and competence dimensions more frequently used in stereotype content research (Fiske et al., 2002).If warmth and competence are simply redundant with HN and HU, respectively, then people should more strongly associate liberals with warmth and conservatives with competence.This was not the case, however-whereas liberals were rated higher in warmth than were conservatives, liberals and conservatives were rated equally high in competence (and alternatively, conservatives were rated higher in competence than warmth, but liberals were rated equally high in warmth and competence).Thus, the warmth and competence dimensions fail to capture the distinct stereotype portraits of U.S. liberals and conservatives that were reflected in differential attributions of human nature and human uniqueness traits, respectively.This study therefore provides additional evidence (see also Haslam et al., 2009) that humanness ratings provide value for understanding stereotype content beyond warmth-competence ratings that should be explored in future research.The fact that outgroup antipathy motivated differential attributions of humanness traits further suggests the importance of exploring HN and HU trait associations in other intergroup contexts.
Contextual factors such as threat may lead ingroup members to dehumanize outgroup members, which allows the perpetration of inhumane and violent actions against outgroup members without moral compunction (see Goldenberg et al., 2009;McAlister, Bandura, & Owen, 2006;Motyl & Pyszczynski, 2009;Sternberg, 2003).Our results suggest that liberals and conservatives dehumanize one another, and that distinctive political stereotypes lie at the heart of such dehumanization.This political dehumanization may explain the apparent rise in demonizing rhetoric and hostility across the political divide (e.g., claims that President Obama is the "Anti-Christ", Haidt, 2012;Pyszczynski, Henthorn, Motyl, & Gerow, 2010).Future studies could explore this possibility by examining whether activation of "bleeding heart" liberal and "hard-hearted" conservative stereotypes increases partisan polarization.
Of course, one goal of future research would be to determine how political hostility and stereotyping could be reduced.For example, individual differences in identification with all of humanity (e.g., McFarland, Webb, & Brown, 2012) may mitigate political dehumanization effects.Hostility-reduction interventions may use reminders of the humanness of outgroup members more generally (e.g., Motyl et al., 2011), or more specifically by emphasizing the (counter-stereotypical) human nature qualities of conservatives and uniquely human qualities of liberals.That said, research in other contexts suggests that humanness reminders may backfire, for example by increasing perpetrators' expectations of forgiveness for wrongdoings and decreasing victims' collective action motivations (Greenaway, Louis, & Wohl, 2012;Greenaway, Quinn, & Louis, 2011).Thus, it is possible that while common humanity reminders may reduce stereotyping and prejudice among liberals and conservatives, they may still be unwilling to cooperate with each other.Future research should explore these important questions.
These results also have implications for framing effects in political communication.For example, given the acceptance of the conservative stereotype by both liberals and moderates, it may be easier and more effective for

Political Dehumanization 98
Democratic campaigns to frame Republican opponents as unfeeling automatons than as intellectual lightweights lacking seriousness.Indeed, President Obama's re-election campaign successfully cast his opponent Mitt Romney as an "out-of-touch fat cat", and Obama reportedly joked that Romney was not "human enough" to get elected President (Martin & Thrush, 2012).By contrast, Republican campaigns may find it easier to frame Democratic opponents as trivial or affable to a fault (as Senator John McCain's campaign did when it likened then-Senator Barack Obama to a celebrity during the 2008 Presidential election) rather than as cold, unfeeling, or impersonal.
The fact that moderates and conservatives alike shared this liberal stereotype may make this framing strategy all the more effective.

Limitations
Although there are many benefits to using MTurk samples over and above convenience samples of college students as outlined above, there are also several known limitations that should be noted.First, MTurk samples are not as representative as national probability samples such as those used in the American National Election Studies, which tend to be older and more politically conservative than MTurk samples (Berinsky et al., 2012).It would therefore be optimal to replicate these findings in nationally representative samples.Second, there is often less experimenter control in internet samples such as MTurk relative to lab samples, which may impact the validity of some research findings.For example, participants may be more distracted or inattentive (Oppenheimer, Meyvis, & Davidenko, 2009); they may be more likely to misrepresent themselves given the greater anonymity provided in internet samples (Kahan, 2013); and they may have greater knowledge of the research protocol, either because of past participation in related studies (Chandler, Mueller, & Paolacci, 2013), through online sharing of experiment information in MTurk participant forums (Mason & Suri, 2012), or through online retrieval of answers to questions in a given protocol (Kahan, 2013).
Several strategies have been suggested for mitigating against some of these validity threats, including removing participants who fail embedded attention checks (e.g., Oppenheimer et al., 2009) and prescreening participants for their knowledge of or previous experience with a particular research protocol (Chandler et al., 2013).Of course, removing inattentive participants should only increase statistical power (Oppenheimer et al., 2009), and the present study was relatively novel and did not require "accurate" responding to particular questions; thus, it is unlikely that the above issues posed serious validity threats to our study.Still, it is important for social and political psychologists to consider both the benefits and limitations of the relatively new recruitment strategy of MTurk.
There is a rich dehumanization literature (see Haslam et al., 2009 for a review), and this study employed only one particular methodological approach to examining political dehumanization.For example, dehumanization may occur through differential associations of targets with particular traits, as we found in the current study, but also with particular non-human entities (e.g., animals and machines for animalistic and mechanistic dehumanization, respectively; see, e.g., Loughnan & Haslam, 2007).Further, while we only examined explicit trait associations, other dehumanization research has employed methodologies that assess implicit associations as well (e.g., Loughnan & Haslam, 2007;Saminaden, Loughnan, & Haslam, 2010).Finally, whereas scores of traits have been used to assess HN and HU characteristics, we relied on a relatively small set of traits in the present study (i.e., 10 traits for each humanness dimension).Thus, future research could replicate and extend our work by sampling a richer variety of traits, and examining whether liberals and conservatives are more strongly explicitly and implicitly associated with non-human entities such as animals and machines, respectively.Using a greater variety of methods will increase our understanding of the generalizability of the present findings.
It is also worth noting that these findings are specific to the U.S. political context alone, and future research should explore their generalizability to and implications for other political contexts.For example, such differential trait attributions observed in this U.S. sample may be limited to similar political systems marked by only two major parties and hyperpolarization.Further, in nations with more than two competitive political parties, these differential trait associations may only hold for far-left and far-right (but not center-left or center-right) parties and individuals.Future research should explore these possibilities.

Conclusion
Consistent with other recent findings (Chambers et al., 2013;Crawford & Pilanski, 2013;Wetherell et al., 2013), this study makes clear that liberals and conservatives do not like each other very much: liberals clearly attributed more negative than positive traits to conservatives, and conservatives attributed more negative than positive traits to liberals.That said, people do not simply attribute more negative traits to their ingroup than to their outgroup.
Instead, ideologues use differential attributions of human nature and human uniqueness traits to stereotype and dehumanize their political opponents.
ii) To examine whether these findings depended on the operationalization of perceiver ideology as a categorical or continuous variable, we also conducted a mixed GLM with trait, valence and target ideology as within-subjects variables and the continuous perceived ideology variable as a covariate (Tabachnick & Fidell, 2013).As in the analyses reported in the text, the Trait × Target and Perceiver Ideology × Target Ideology × Trait × Valence interactions were significant (Fs > 16.01, ps < .001),while the Perceiver Ideology × Target Ideology × Trait was not (F < 1).
iii) An alternative operationalization of dehumanization is to create difference scores reflecting a comparison between liberals and conservatives on the different trait types (e.g., greater attribution of positive HN traits to liberals than to conservatives; greater attribution of negative HU traits to conservatives than to liberals).However, these difference scores only reflect the extent to which positive traits and negative traits are associated with one group over the other, and therefore confound subtle dehumanization with the strong political hostility observed in this sample (see Table 2).Unsurprisingly, mediation analyses using target comparison difference scores revealed significant indirect effects of ideology on these target comparison difference scores through both outgroup antipathy and ingroup liking (results are available from the corresponding author on request).Thus, we present the trait comparisons (e.g., positive HN vs. positive HU traits for liberals), which capture subtle political dehumanization while also eliminating political hostility as an alternative explanation for any correlations with political ideology.iv) Feeling thermometer ratings were transformed by dividing each by 10 in order to enhance the interpretability of the unstandardized regression coefficients in the model.They were also reverse-scored so that higher values indicated greater antipathy.

Figure 1 .
Figure 1.Antipathy towards conservatives mediates the relationship between ideology and the mechanistic dehumanization of conservatives on negative traits.Note.Path labels are based on Baron and Kenny's (1986) nomenclature.Path values represent unstandardized regression coefficients, with standard errors in parentheses.Higher scores on ideology represent greater conservatism.Higher scores on antipathy represent greater dislike of the target.Mechanistic dehumanization of conservatives on negative traits is the difference score of negative HU traits -negative HN traits for conservatives.Adjusted R 2 = .17,p < .001.Confidence intervals with 5,000 bootstrap samples for the indirect effect through antipathy towards conservatives, Lower = -.29,Upper = -.09.The indirect effect through antipathy towards liberals was not significant, Lower = -.11,Upper = .10.A contrast of the indirect effects was significant, Lower = .02,Upper = .33. **p < .01. ***p < .001

Figure 2 .
Figure 2. Antipathy towards liberals mediates the relationship between ideology and the animalistic dehumanization of liberals on positive traits.Note.Path labels are based on Baron and Kenny's (1986) nomenclature.Path values represent unstandardized regression coefficients, with standard errors in parentheses.Higher scores on ideology represent greater conservatism.Higher scores on antipathy represent greater dislike of the target.Animalistic dehumanization of liberals on positive traits is the difference score of positive HN traits -positive HU traits for liberals.Adjusted R 2 = .05,p < .05.Confidence intervals with 5,000 bootstrap samples for the indirect effect through antipathy towards liberals, Lower = .001,Upper = .16.The indirect effect through antipathy towards conservatives was not significant, Lower = -.09,Upper = .06.A contrast of the indirect effects was not significant, Lower = -.02,Upper = .23. *p < .05. ***p < .001 Journal of Social and Political Psychology 2013, Vol.1(1), 86-104 doi:10.5964/jspp.v1i1.184

Table 2
Descriptive Statistics for and Correlations among VariablesNote.dfs ranged from 111 to 116.FT = feeling thermometer; HN = human nature traits; HU = human uniqueness traits.

Table 5
Simple Effects Analyses for Trait Comparisons Reflecting the Perceiver Ideology × Target Ideology × Trait × Valence Interaction