A - Dehumanization Understanding The Relationship Between Attribute-Based and Metaphor-Based

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Group Processes & Intergroup

Relations
http://gpi.sagepub.com

Understanding the Relationship between Attribute-Based and Metaphor-Based


Dehumanization
Steve Loughnan, Nick Haslam and Yoshihisa Kashima
Group Processes Intergroup Relations 2009; 12; 747
DOI: 10.1177/1368430209347726

The online version of this article can be found at:


http://gpi.sagepub.com/cgi/content/abstract/12/6/747

Published by:

http://www.sagepublications.com

Additional services and information for Group Processes & Intergroup Relations can be found at:

Email Alerts: http://gpi.sagepub.com/cgi/alerts

Subscriptions: http://gpi.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.co.uk/journalsPermissions.nav

Citations http://gpi.sagepub.com/cgi/content/refs/12/6/747

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes &
Intergroup Relations
2009 Vol 12(6) 747–762

Understanding the Relationship


between Attribute-Based and
Metaphor-Based Dehumanization
Steve Loughnan, Nick Haslam and Yoshihisa Kashima
University of Melbourne

Previous research has adopted two distinct approaches to the study of dehumanization. One has
focused on the denial of human attributes to groups (attribute-based dehumanization) and the
other on the likening of group members to nonhumans (metaphor-based dehumanization).
The relationship between these two approaches has yet to be examined. The current studies
seek to clarify this relationship by integrating the two approaches. Using Haslam and colleagues’
(Haslam, 2006; Haslam, Loughnan, Kashima & Bain, 2008) model of dehumanization, we
examined whether attribute-based dehumanization leads to metaphor-based dehumanization,
and vice versa. In Study 1 participants read about a novel group that was described either as
lacking one type of humanness or as being like a nonhuman. In Study 2 a concrete learning
task taught participants that a novel group lacked a specific type of humanness. In both studies,
participants explicitly learned to dehumanize the group and inferred the corresponding type
of attribute- or metaphor-based perception (e.g. perceived a group as animal-like after learning
that it lacked uniquely human attributes, and vice versa). Implicitly, however, participants were
able to directly learn but not robustly infer the corresponding type of dehumanization. We
suggest that the relationship between the two types of dehumanization can be understood using
cognitive models of metaphor-making.

keywords dehumanization, infrahumanization, metaphor

The study of dehumanization1 is experiencing attributes to others or on the likening of others


rapid growth within social psychology. Since to nonhumans, and the present paper aims to
the emergence of infrahumanization theory bridge these attribute- and metaphor-based
less than a decade ago (Leyens et al., 2001), approaches.
the field has expanded to encompass a wide Most previous research has examined attribute-
variety of phenomena in group and person per- based dehumanization. Research on infra-
ception, and a rich assortment of research humanization exemplifies this focus on denials
methods and theoretical perspectives. At the
same time, approaches to the field have subtly Author’s note
diverged. The purpose of this paper is to address Address correspondence to Steve Loughnan,
one central aspect of this divergence and attempt Department of Psychology, University of
to integrate it. Dehumanization research has Melbourne, Parkville VIC 3010, Australia.
tended to focus either on the denial of human [email: lost@unimelb.edu.au]

Copyright © The Author(s), 2009


Reprints and permissions: http://www.sagepub.co.uk/journalsPermissions.nav
12:6; 747–762;
Downloaded from http://gpi.sagepub.com at University of Melbourne DOI:21,10.1177/1368430209347726
Library on October 2009
Group Processes & Intergroup Relations 12(6)

of humanness, specifically the denial of unique- (2007) found that although some occupational
ly human emotions to outgroup members groups (i.e. artists) are likened to animals, others
(Boccato, Cortes, Demoulin, & Leyens, 2007; (i.e. business people) are likened to robots.
Leyens et al., 2001; Paladino et al., 2002). Under The distinction between attribute- and metaphor-
this paradigm, researchers first established based dehumanization captures most previous
which characteristics are unique to humans dehumanization research, but their relationship
(i.e. secondary emotions), and then examined remains theoretically and empirically uncertain.
the differential attribution of these character- Infrahumanization theorists have argued that
istics to ingroup and outgroup. The denial of because HU attributes define the boundary be-
uniquely human emotions to outgroup mem- tween humans and animals, their denial tacitly
bers was taken to indicate that they are subtly animalizes outgroup members (Leyens et al.,
viewed as less human than the ingroup (Leyens, 2007). Indeed, Leyens et al. (2007, p. 140) state
Demoulin, Vaes, Gaunt, & Paladino, 2007; Leyens that ‘infrahumanization in a process by which
et al., 2001). people consider their ingroup as fully human
A focus on the denial of humanness-defining and outgroups as less human and more animal-
attributes also characterizes the dehumanization like’. Similarly, Haslam claimed that ‘if people
research of Haslam and colleagues. Haslam are perceived as lacking what distinguishes
(2006) proposed that humanness may be under- humans from animals, they should be seen im-
stood as attributes that are unique to humans plicitly or explicitly as animal-like’ (Haslam,
(human uniqueness; HU) and those that are 2006, p. 258). However, the claim that the denial
essential to being human (human nature; HN). of HU leads to animalization has not received
In interpersonal comparisons, people tend to empirical scrutiny. With regards to a denial of
deny HN to others relative to the self (Haslam & human nature, Haslam (2006, p. 258) claims
Bain, 2007; Haslam, Bain, Douge, Lee, & Bastian, that ‘the shared, typical, or core properties of
2005), and deny both HN and HU to particular humanness are also those that distinguish us
groups in intergroup contexts (Loughnan & from automata’. Again however, the claim than
Haslam, 2007; Saminaden, Loughnan, & the denial of HN leads to mechanization has
Haslam, in press). Similar to infrahumanization not been investigated. Although attribute- and
research, studies in this tradition examine metaphor-based types of dehumanization have
dehumanization as the subtle denial of particular occasionally been examined simultaneously
human attributes. (e.g. Loughnan & Haslam, 2007; Saminaden
A second and largely independent stream of et al., in press), whether denying human at-
research has examined dehumanization as the tributes to a group and seeing its members as
association of outgroup members with non- nonhuman are causally linked has not been
human entities. This metaphor-based approach examined. Thus, although there are theoretical
has typically examined the likening of out- links between denials of HN and robot or
groups to animals. For instance, Viki et al. (2006) machine metaphors, and between denials of HU
investigated the perceived likeness of ingroup and animal metaphors (Haslam, 2006; Leyens
and outgroup members to animals and found a et al., 2007), there is currently no evidence for
tendency to consider outgroup members more these associations. This represents an important
animal-like than ingroup members. A similar gap in our understanding of dehumanization.
effect was also found when comparing tradi- The current research aimed to address this
tional and modern people (Saminaden et al., gap by directly examining the relationship be-
in press). Other research has established that tween attribute- and metaphor-based dehuman-
ethnic outgroups can be likened to apes relative ization. It builds on recent advances in our
to the ingroup (Boccato, Capozza, Falvo, & understanding of the formation of implicit
Durante, 2008; Goff, Eberhardt, Williams, & and explicit representations of social groups.
Jackson, 2008). Metaphor-based dehumaniza- Social cognitive research on implicit and explicit
tion research has not remained limited to the attitude formation has recently examined the
study of animalization. Loughnan and Haslam mechanisms through which people come to

748

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

hold evaluative beliefs about others. One tech- attributes lead to metaphors. This model tacitly
nique widely employed in this research has been underlies much of the existing theorizing in both
the ‘novel group’ paradigm, whereby partici- the infrahumanization and dehumanization
pants are required to form an attitude about literatures. Although these frameworks differ
a completely fictitious group (Gregg, Seibt, & in their respective types of humanness, both
Banaji, 2006; Ranganath & Nosek, 2008; Rydell, argue that a particular set of human attributes
McConnell, & Mackie, 2008; Rydell, McConnell, is required to be considered fully human.
Mackie, & Strain, 2006; Rydell, McConnell, Denying another group these attributes shifts
Strain, Claypool, & Hugenberg, 2007). Under the them away from the human and towards the
novel group technique, participants are typically nonhuman. Stated otherwise, attribute-based
provided with positive or negative information dehumanization leads to metaphor-based de-
about the group before indicating their implicit humanization. Finding support for this model
and explicit evaluations. would validate existing theory. Alternatively, a
In the current research we adapted this metaphorical association between the outgroup
approach to the study of dehumanization. In and a particular nonhuman might lead to the
two studies participants were given descrip- inference that the group lacks a specific sense
tions that dehumanized a novel group in either of humanness. Put simply, metaphors lead to
an attribute-based (i.e. low in HU or HN) or attributes. In contrast to the preceding model,
metaphor-based manner (i.e. being animal- the second model suggests that metaphors
like or robot-like). We then assessed i) whether cause attribute-based dehumanization.
people were able to directly learn metaphor- The third model is the reciprocity model.
and attribute-based dehumanization at both an This model combines the two preceding models
implicit and explicit level, and ii) whether they by positing two routes to full metaphor and at-
were further able to infer the corresponding tribute dehumanization. The outgroup can be
type of dehumanization at both an implicit viewed as lacking a specific sense of humanness
and explicit level. If attribute- and metaphor- which leads to the inference that they are like a
based dehumanization are related in the man- nonhuman and being likened to a specific non-
ner suggested by infrahumanization theory human causes people to infer that the group
(Leyens et al., 2007) and dehumanization theory lacks a specific sense of humanness. The reci-
(Haslam, 2006), then learning that members of procity model allows both previous accounts
a group lack a particular sense of humanness to be accurate; attributes lead to metaphors
(e.g. HU) should lead people to infer that they and metaphors lead to attributes. If this model
are like a particular nonhuman (e.g. animals). is correct we can provide an integrated, bi-
If the effect is bi-directional, then learning that directional account of the relationship between
group members are like a particular nonhuman these two types of dehumanization. Under
(e.g. robots) should lead people to infer that they this model neither type of dehumanization
lack a specific sense of humanness (e.g. HN). is primary, rather both types may lead to and
This paper examines the relationship between be inferred from the other. This set of models
both types of dehumanization with regards to is not intended to be exhaustive. Indeed, it is
the directionality of inference (attributes to possible that the two types of dehumanization
metaphors/metaphors to attributes) and level are unrelated or that a third variable can explain
of acquisition and inference (implicit/explicit). their co-occurrence. However, this set is sufficient
Using this design we can foresee several different for an initial investigation, and seems to cover
patterns of results. With regard the directionality the most plausible relationships.
of inference, at least three types of relationship These studies additionally allow us to ex-
are possible. plore whether dehumanization is learned and
One possibility is that being denied a specific inferred spontaneously (and thus appears at an
sense of humanness might lead to a group being implicit level) or non-spontaneously (and thus
viewed as like a specific nonhuman. That is, appears only at an explicit level). Many social

749

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

inferences and aspects of impression formation hypothesized that the denial of specific attrib-
occur spontaneously, whereas others require utes would lead directly to learning attribute-
deliberative processing (for a recent review see based dehumanization and indirectly to inferring
Uleman, Saribay, & Gonzalez, 2008). If the ac- metaphor-based dehumanization. Likewise, we
quisition of dehumanization occurs rapidly at predicted that the metaphorical likening of a
both levels it would suggest that, like attitudes social group to nonhumans would lead directly
(Gregg et al., 2006; Rydell et al., 2007), people to learning metaphor-based and indirectly to
can readily form dehumanizing impressions inferring attribute-based dehumanization.
of others. By contrast, if people are only able These measures of direct learning acted addi-
to acquire dehumanizing representations ex- tionally as manipulation checks to ensure that
plicitly, this would suggest that the process of we were able to induce different types of de-
learning to dehumanize a group is cognitively humanization. We measured dehumanization
demanding. The same can be said for dehuman- at both an implicit and explicit level.
izing inferences. If they appear at both an
implicit and explicit level then we have evidence Method
that people can rapidly expand their sense of Participants One hundred and fourteen under-
dehumanization (for a related discussion on graduates (89 female, mean age 20.56 years,
the rapid generalization of implicit attitudes SD = 4.69) participated in partial fulfillment of
see Ranganath & Nosek, 2008). In contrast, if course requirements. Participants completed
dehumanization occurs at an explicit but not four tasks: a categorization task, a scenario read-
an implicit level, there is evidence that the in- ing task, an implicit dehumanization measure,
ference process is cognitively demanding. It and an explicit dehumanization measure. All
may be the case that learning that a group is four tasks involved participants working with two
less than fully human is more complex than novel groups: the Hebians and the Nopoes.
learning that they are positive or negative, and
accordingly requires the investment of consider- Materials A categorization task was used to
able cognitive resources. Finally, it is possible that ensure participants were able to quickly and
dehumanization occurs at an implicit but not accurately discriminate between names repre-
an explicit level. This pattern of results would senting two novel groups: Hebians and Nopoes.
suggest that people are perhaps hesitant to To ease the categorization of a name as either
report dehumanization (even of novel groups), Hebian or Nopoe, all Hebian names contained
despite these beliefs forming easily. In sum, this the letter string ‘h-e-b’ and all Nopoe names
study design allows us to distinguish multiple contained the string ‘o-p-o’ (see Gregg et al.,
different accounts of the relationship between 2006). The categorization task involved the
these two aspects of dehumanization. presentation of single names from both groups.
Participants were required to respond to Hebian
names by striking the ‘A’ key and Nopoes
Study 1
names by striking the ‘L’ key. Each response
In the first study, we examined both attribute- was followed by immediate accuracy feedback,
and metaphor-based dehumanization. At the with a 700 ms timeout.
attribute level, we examined the denial of Participants were presented with one of four
HN and HU. At the metaphor level, the group stories depicting a novel group. All stories started
was likened to either animals or robots. We with a short description of the novel group as
adopted Gregg et al.’s (2006) approach for an ethnic group living in a fictitious nation
inducing attitudes towards a novel group by (Hebia or Nopoia). The two Hebian stories de-
having participants read a short description scribed the ‘typical Hebian’ as possessing certain
of the group. Four dehumanizing stories attributes. The aim of these stories was for par-
were presented: two attribute-based and two ticipants to form an impression of the group as
metaphor-based. Consistent with theory we lacking one sense of humanness. In the low-HU

750

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

condition, the group was described as posses- concepts (two targets, two distractors) with each
sing attributes low in human uniqueness. term repeated five times, resulting in 60 targets
Conversely, in the low-HN condition, the group and 60 distractors. All terms were matched with
was described as possessing attributes low in their paired distractor for familiarity, frequency,
human nature. The humanness attributes were and length using a lexical database (Baayens,
drawn from previous research where they were Piepenbrock, & van Rijn, 1995). Participants’
extensively validated and matched from fre- task was to distinguish targets (go response)
quency, familiarity, and word length (Haslam & from distractors (no go response). Given the
Bain, 2007; Loughnan & Haslam, 2007; ease of identifying Hebian or Nopoe names, a
Saminaden et al., in press). The Nopoes’ stories response deadline of 600 ms was imposed.
presented the group as like a particular type of To measure explicit attribute-based and
nonhuman using nonhuman descriptors and metaphor-based dehumanization a question-
behaviors. One story animalized the Nopoes naire was administered. The attribute-based
directly by labeling them as animal-like and dehumanization questions required participants
indirectly by reference to their animal-like be- to rate the degree to which a series of HN and
haviors and appearance. The other story mech- HU traits were typical of the group on a six-point
anized the Nopoes directly by labeling them scale (1 = not at all typical, 6 = highly typical).
as robot-like and indirectly through reference The 20 traits were drawn from previous re-
to robotic behaviors. Behaviors were used to search (Haslam et al., 2005; Loughnan &
bolster the direct metaphor, and were selected Haslam, 2007) and included the 12 traits used
to reflect those highly characteristic of animals in the GNAT task with the addition of four
and robots based on previous research (Haslam, HN traits (i.e. nervous, trusting, sociable, and
Kashima, Loughnan, Shi, & Suitner, 2008) (see impatient) and four HU traits (i.e. humble,
Appendix for all study stimuli). broadminded, stingy, and shallow). For par-
To assess implicit dehumanization partici- ticipants in the Hebian condition this ques-
pants completed a Go No-go Association Task tionnaire served as an explicit measure of
(Nosek & Banaji, 2001). The GNAT is an implicit direct learning. For participants in the Nopoe
association measure which, similar to the SC-IAT condition, it assessed the inference of attribute-
(Karpinski & Stienman, 2006), measures a single based dehumanization. The metaphor-based
association between two target categories. The dehumanization task required participants to
GNAT measures the strength of association indicate the perceived likeness of the novel
between two target concepts, indexed using a group to particular nonhumans. Participants
d-prime (d′) statistic, and has been previously em- indicated how much the group was like animals
ployed to examine dehumanization (Capozza, and robots using a six-point scale (1 = very unlike,
Andrighetto, Falvo, & Trifiletti, 2006; Loughnan 6 = very like). For participants in the Nopoe
& Haslam, 2007). Participants were presented condition this questionnaire was an explicit
with four individually randomized blocks of measure of direct learning. For participants in
GNAT trials. These blocks tested the associations the Hebian condition, it assessed the inference
between the novel group and both i) the two of metaphor-based dehumanization.
senses of humanness and ii) the two nonhumans
(animals and robots). For the animal stimuli we Procedure The experiment was administered
decided to focus specifically on apes. Previous to groups of one to eight participants, who were
research has shown that social groups are randomly assigned to one of the four stories.
subtly likened to apes (Boccato et al., 2008; Goff All participants completed the categorization
et al., 2008), and that this can have particularly task, reading task, GNAT, and questionnaire in
harmful consequences for the group (Goff order. Following the completion of the ques-
et al., 2008). Each concept was represented by tionnaire participants were thanked and fully
six terms. Each GNAT block contained four debriefed.

751

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

Results participants in the animalized condition viewed


The results of the name discrimination task indi- the group as possessing less HU than HN,
cated that participants were able to distinguish F(1, 24) = 9.22, p = 0.006 (see Table 1). These results
the Hebian and Nopoe names at high speed indicate that learning that a group is like a spe-
with over 80% accuracy, well above chance cific type of nonhuman leads to the inference that
level responding. An analysis of discrimination it lacks particular psychological attributes.
accuracy and average d′ scores in the GNAT Explicit metaphor-based dehumanization
revealed that 14 participants performed at was measured by asking participants how much
or below chance and they were subsequently the group was like ‘animals’ and ‘robots’. A 4
excluded from further analysis. The exclusion (story condition) × 2 (nonhuman; animal, robot)
of participants who respond below chance is mixed model ANOVA yielded the predicted
recommended when using d′ values (Green & interaction between condition and nonhuman,
Swets, 1966). F(3, 95) = 54.30, p < 0.001. Consistent with ex-
pectations, participants rated the animalized
Explicit dehumanization The questionnaire group as significantly more animal-like than
required participants to rate the group on HN robot-like, F(1, 24) = 52.80, p < 0.001, and the
and HU trait terms. Both HN and HU terms mechanized group as more robot-like than
demonstrated good reliability (Cronbach’s animal-like, F(1, 23) = 15.13, p = 0.001. This
α = 0.89, 0.89). To examine the effect of story finding indicated that the stories were successful
condition on ratings of humanness, a 4 (Between: in manipulating overt representations of the
story condition) × 2 (Within: humanness type; group as like certain nonhumans. Participants
HN, HU) mixed model ANOVA was conducted. in the low-HU condition also viewed the group
There was no significant main effect of condi- as significantly more animal-like than robot-like,
tion, F(1, 95) = 3.33, p = 0.077, indicating no F(1, 26) = 43.20, p < 0.001, and participants
general tendency to rate the group higher on in the low-HN condition viewed the group as
either sense of humanness, but a significant significantly more robot-like than animal-like,
interaction between humanness and condition, F(1, 23) = 15.13, p = 0.001 (see Table 1). Thus
F(1, 95) = 118.46, p < 0.001. As hypothesized, participants who learned that a group lacked
participants in the low-HU condition rated the certain humanness-defining attributes inferred
group as possessing less HU than HN traits, that it was like particular types of nonhuman.
F(1, 25) = 179.28, p < 0.001, whereas participants In sum, participants both directly learned and
in the low-HN condition rated the group as inferred dehumanization at an explicit level.
possessing less HN than HU, F(1, 24) = 122.17,
p < 0.001. This result indicates that the manipu- Implicit dehumanization Response accur-
lation of humanness attributes was successful at acies for each participant in each GNAT block
an explicit level. Participants in the mechanized were converted into d′ values by subtracting the
condition—who read that Nopoes were robot- standardized probability of a false alarm from
like—viewed the group as possessing less HN the standardized probability of a hit (Nosek &
than HU, F(1, 22) = 120.13, p < 0.001, whereas Banaji, 2001). Higher d′ values indicate a stronger

Table 1. Explicit dehumanization learning and inference

Condition HN HU Animals Automata


Low-HN 2.32a 4.73b 2.54a 4.20c
Mechanization 2.30a 3.94b 2.13a 5.04c
Low-HU 4.82a 2.35b 3.66c 1.59d
Animalization 3.34a 2.55b 4.36c 1.60d
Note: Values with different subscripts are significantly different across rows.

752

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

association. The attribute-based (Hebian) and revealed no significant difference between


metaphor-based (Nopoe) story conditions were d′ values for the Animal and Robot blocks,
analyzed separately. F(1, 24) < 1, ns.
To examine the effect of attribute-based story The two metaphor-based (Nopoes) story
manipulation on implicit associations, a 2 conditions were examined in the same manner.
(Between: low-HN, low-HU) × 4 (Within: HN, In the ‘animalized’ condition the planned
HU, animal, robot) mixed model ANOVA was comparison indicated that participants did not
conducted. The results indicated a significant implicitly associate Nopoes with animals more
interaction between block-d′ and story type, than robots, F(3, 24) = 1.09, p = 0.20, or HU
F(3, 46) = 5.54, p = 0.002. Age and gender had attributes less than HN attributes, F(3, 24) < 1,
no effects in any of the study analyses. To investi- ns. These fi ndings suggest that reading the
gate the effect of the low-HU story on implicit description of Nopoes as animal-like did not
associations between the groups, humanness lead to the formation of implicit associations
attributes, and types of nonhuman, two planned with either nonhumans or perceived humanness.
comparisons were conducted. The first compared In the ‘mechanized’ condition, the planned
d′ values for the HN block with the HU block, comparisons indicated that Nopoes were
revealing a significant difference between the associated more with Robots than Animals, F(3,
two attribute types, F(1, 26) = 6.34, p = 0.018. 22) = 11.92, p = 0.002, and marginally less with
Participants who read that Hebians were low HN attributes than HU attributes, F(3, 22) = 3.78,
in HU associated them with HU attributes less p = 0.065 (see Table 2).
than with HN attributes, indicating that the story
manipulation influenced perceived attribute- Discussion
based humanness at an implicit level. The second The findings of Study 1 are twofold. First,
planned comparison contrasted the d′ values they indicate that people can learn and infer
for the Animal and Robot blocks, revealing that dehumanizing perceptions of groups. Second,
participants associated the Hebians with animals although direct learning occurs at an implicit
more than robots, F(1, 26) = 4.86, p = 0.037. and explicit level, inference robustly occurs
This result suggests that participants inferred only explicitly.
an implicit relationship between the group Participants were able to learn to explicitly
and a particular type of nonhuman based on and implicitly dehumanize the novel group
the type of humanness that the group lacked in both attribute-based and metaphor-based
(see Table 2). ways. When a novel group is dehumanized,
The pattern of results observed in the low- people learn it rapidly regardless of the type of
HU condition was expected to reverse for the dehumanization. This direct learning indicates
low-HN condition. Indeed, participants who the speed with which dehumanizing stereotypes
read that Hebians lacked HN attributes im- can develop. Further, the robustness with which
plicitly associated them with HU attributes participants were able to learn both implicit
more than HN attributes, F(1, 24) = 7.13, and explicit dehumanization indicates that
p = 0.013. In contrast, the planned comparison explicitly presented dehumanizing material

Table 2. Implicit dehumanization learning and inference

Condition HN d′ HU d′ Animals d′ Automata d′


a b c
Low-HN 1.32 1.60 2.34 2.26c
Mechanization 1.11a 1.47b 1.99c 2.24d
Low-HU 1.69a 1.47b 2.63c 2.37d
Animalization 1.48a 1.37a 1.92b 1.99b
Note: Values with different subscripts are significantly different across rows.

753

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

can lead to implicit learning (for similar results The scenario-based learning paradigm may
in the attitude domain see Gregg et al., 2006; have activated the explicit processing system,
Ranganath & Nosek, 2008). leading to robust explicit dehumanization, while
At an explicit level, this dehumanization providing only limited learning opportunities
spreads from attributes to metaphors, and vice to the implicit learning system, resulting in
versa. Importantly, the pattern of inference weaker implicit dehumanization. Alternative
is entirely consistent with the predictions of learning paradigms that have been previously
dehumanization (Haslam, 2006) and infra- demonstrated to lead to strong implicit and
humanization (Leyens et al., 2001) theory; the explicit representations might yield stronger
denial of HU attributes leads to likening to implicit results. This possibility is examined in
animals, and denial of HN attributes leads Study 2. In sum, Study 1 indicates that people
to likening to robots. This study is the first to pro- can learn to dehumanize at both an implicit and
vide direct evidence for a reciprocal causal link explicit level. Moreover, at an explicit level, a
between attribute- and metaphor-based forms theoretically consistent form of dehumanization
of dehumanization. Whereas previous research can be inferred.
has demonstrated the co-occurrence of attribute-
based and metaphor-based dehumanization
Study 2
(Loughnan & Haslam, 2007; Saminaden et al.,
in press), this study indicates that denying one The aim of Study 2 was to replicate the results
type of humanness leads people to infer the of Study 1 and to investigate the possibility that
corresponding dehumanizing metaphor. an alternative learning paradigm may lead to
Interestingly, this movement from one type robust inference of implicit dehumanization.
of dehumanization to the other was only This study focused on the learning of attribute-
robust at an explicit level. At an implicit level, based dehumanization and the subsequent
although participants were generally able to inference of metaphor-based dehumanization.
learn dehumanizing attribute- or metaphor- We elected to focus on this particular path of
based perceptions directly, they were unable to dehumanization for several reasons. A majority
consistently infer one type of dehumanization of previous dehumanization and infrahuman-
from the other. People were able to infer low ization research has worked from an attribute-
HN from mechanistic dehumanization and based approach (Boccato et al., 2007; Haslam &
animalization from low HU. However, the Bain, 2007; Haslam et al., 2005; Leyens et al.,
former result only reached marginal signifi- 2001; Paladino et al., 2002). Thus, focusing on
cance and the latter failed to replicate in the this approach aligns this study with previous
next study. These findings imply that people research. Importantly, previous social cognitive
neither spontaneously infer dehumanizing research investigating attitude acquisition has
attributes from dehumanizing metaphors, nor focused on how people move from more basic
spontaneously use dehumanizing metaphors (e.g. behavioral) representations to more com-
when a target group is ascribed dehumanizing plex (e.g. trait, evaluative) representations (e.g.
attributes. They move between attribute-based Ranganath & Nosek, 2008). Choosing to examine
and metaphor-based dehumanization when they the movement from more basic humanness
are required to do so. attributes to more abstract metaphors aligns
Nonetheless, it is possible that this result this study more closely with previous research.
arose due to the type of learning paradigm em- Finally, to broaden the scope of the findings, we
ployed (i.e. scenario-based learning). Dual-route examined associations with animals in general
attitude models have recently dissociated a fast, rather than apes in particular.
reflective, and cognitively expensive explicit One possible explanation of the lack of spon-
learning system and a slower, unconscious, less taneous dehumanization in Study 1 was the
cognitively demanding implicit learning sys- choice of learning paradigm (i.e. scenario-
tem (Gawronski & Bodenhausen, 2006, 2007). based). Recent implicit attitude research has

754

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

employed the concrete-learning paradigm to Learning trials consisted of the presentation of


investigate the acquisition of both implicit a series of trait terms and behavioral descriptors.
and explicit attitudes. This paradigm provides Participants responded by guessing whether
participants with multiple single instances of be- the attribute was typical of the group. The traits
haviors or traits descriptive of the target group were the 20 adjectives employed in Study 1
or individual. Participants are required to guess and previous research (Loughnan & Haslam,
whether the description is accurate, forming an 2007). These were supplemented by 10 brief
impression of the target over multiple trials. behavioral descriptors previously shown to
Although Gregg et al. (2006) demonstrated align with HN or HU (Wilson, 2008); five
that this technique generates similar results indicative of HN (e.g. ‘maintaining rela-
to the scenario-based approach, Rydell et al. tionships’) and five of HU (e.g. ‘reading’) (see
(2006) have recently shown that concrete- Appendix).
learning can lead to rapid implicit and explicit Attributes were presented individually on
attitude formation. Further, although explicit at- a computer monitor and participants were
titudes are sensitive to small amounts of counter- required to guess whether the attribute was
attitudinal information, implicit attitudes characteristic of the Luupites (press ‘A’) or
gained using this technique are highly resistant not (press ‘L’). Participants were informed
to change (Gregg et al., 2006; Rydell et al., whether their guess was correct (e.g. ‘Correct:
2007). Finally, in a recent study investigating Luupites are Friendly’) or incorrect (e.g. ‘Incor-
the generalization of implicit attitudes between rect: Luupites are not Friendly’). For half the
members of the same outgroup, Ranganath and participants HN terms were uncharacteristic
Nosek (2008) demonstrated that implicit at- (low-HN), and for the remaining participants
titudes acquired in this fashion quickly spread HU terms were uncharacteristic (low-HU).
amongst members of the same group. In short,
concrete-learning induces strong implicit and Name differentiation Participants completed
explicit attitudes, and therefore may lead to a name differentiation task identical to that
strong implicit and explicit dehumanization. used in Study 1. This task enabled participants
Further, this technique has been demonstrated to practice identifying names representing
to induce attitude spread across groups and Luupites from distractors (i.e. Niffite names)
therefore may also lead to inferences of other at high speeds. As in Study 1, all Luupite names
associations. contained the character string ‘l-u-p’ (see Gregg
et al., 2006).
Method
Participants Eighty-six undergraduates (54 fe- The GNAT Participants completed a four-
male, 32 male) with a mean age of 19.15 years block GNAT, pairing Luupites with HN, HU,
(SD = 2.10), participated for partial fulfillment Animals, and Robots. Each block consisted
of course requirement or $10 payment. of 120 trials, including 60 targets comprising
three repetitions of two sets of 10 stimulus
Materials Participants were required to com- items (see Appendix). The HN and HU items
plete several tasks in a set order: a concrete were identical to the trait terms employed in
learning task, a name differentiation task, the concrete learning task, to which the partici-
a four-block GNAT, and an attribute-based pants had prior exposure, and did not include
dehumanization and metaphor-based dehuman- the behavioral descriptors. Animal and robot
ization measure. concepts were represented by exemplars of those
categories. Rather than focusing specifically on
Concrete learning Participants were ran- likening to apes (Study 1), which may narrow
domly assigned to learn that a novel group, the scope of the findings, a more general ani-
the Luupites, lacked human nature (low-HN) mal category was drawn from prior research
or human uniqueness (low-HU) attributes. (Loughnan & Haslam, 2007). The distractors

755

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

formed the complementary category to the of Niffite names as Luupite names). In both
targets. For example, HN targets were coupled the low-HN (M = 2.67) and low-HU conditions
with HU distractors (Nosek & Banaji, 2001). (M = 2.73) participants rapidly and accurately
All terms were matched with their paired dis- dissociated the names, performing well above
tractor for familiarity, frequency, and length chance, ts(85) > 28.0, p < 0.001. Independent-
using a lexical database (Baayens et al., 1995). samples t-tests revealed that name learning did
Trial- and block-order was randomized between not vary by condition, t(85) = 0.50, p > 0.05. In
participants. All other features of the GNAT short, participants performed with a high level
were identical to that of the previous study. of accuracy, indicating that they successfully
learned the attribute-based dehumanization and
Dehumanization questionnaire Participants the capacity to identify Luupite names.
completed the metaphor-based and attribute-
based dehumanization questionnaire for the Explicit dehumanization The two sets of
Luupites. This questionnaire was identical to humanness items demonstrated excellent
that presented in Study 1 but was presented reliability despite containing both positive and
electronically. The 22 questions appeared negative trait terms (Cronbach’s α = 0.96, 0.89)
individually in the center of the screen and par- and so were averaged to create HN and HU
ticipants responded using a six-point scale scores for each participant. A 2 (condition;
ranging from 1 (not at all typical, very unlike) to low-HU or low-HN) × 2 (humanness type; HU
6 (highly typical, very like). and HN) mixed model ANOVA yielded the
predicted two-way interaction between condi-
Results tion and humanness, F(1, 82) = 395.57, p < 0.001.
Preliminary analysis Participants in both con- As expected, people in the low-HN condition
ditions completed two sets of concrete learning rated the group lower on HN terms than people
trials designed to teach them that the group in the low-HU condition, F(1, 82) = 357.12,
lacked one set of humanness attributes. As p < 0.001, and people in the low-HU condition
participants were forced to guess whether each rated the group lower on HU than those in the
attribute was typical of the Luupites or not, low-HN condition, F(1, 82) = 197.47, p < 0.001
d′ sensitivity scores were calculated for each (see Figure 1).
participant. A hit refers to correctly attributing To investigate whether learning attribute-
a Luupite characteristic to the Luupites (e.g. a based dehumanization leads to metaphor-based
HN trait in the low-HU condition) and a false dehumanization a 2 (condition: low-HN or low-
alarm to incorrectly attributing a non-Luupite HU) × 2 (nonhuman: animal and robot) mixed
characteristic to the Luupites (e.g. a HU trait model ANOVA was conducted. The predicted
in the low-HU condition). Average d′ scores
were significantly above chance in the low-HU
(M = 1.36) and low-HN (M = 1.20) conditions,
ts(85) > 18.20, ps < 0.001. Independent-
samples t-tests revealed that the differences
between low-HN and low-HU d′ scores did
not reach significance, t(85) = 1.79, p = 0.079.
Two participants consistently demonstrated d′
values more than two standard deviations below
the mean, and thus were excluded from all
further analysis. Participants also performed well
at the name dissociation task. Again, d′ scores Figure 1. Explicit ratings of novel group on
were calculated using the standardized number humanness attributes and likeness to nonhumans
of hits (correct identifications of the Luupite in the low-HU (dark bars) and low-HN (light bars)
name) and false alarms (incorrect classifications conditions, Study 2.

756

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

interaction emerged with participants rating the low-HN or low-HU) × 2 (nonhuman: animal
low-HU Luupites more animal-like than robot- and robot) mixed model ANOVA was conducted.
like, and the low-HN Luupites more robot-like This analysis did not yield the expected two-way
than animal-like, F(1, 82) = 31.59, p < 0.001 interaction between learning condition and
(Figure 1). Simple effects analysis indicated that nonhuman, F(1, 81) = 2.19, p = 0.14. There was a
this difference was significant for both the low- large main effect of nonhuman, F(1, 81) = 14.93,
HU condition, F(1, 42) = 11.00, p = 0.002, and p < 0.001, indicating that people were more
the low-HN condition, F(1, 40) = 21.76, p < 0.001. sensitive when responding to animals compared
In short, participants are able to learn and infer to robots (Figure 2). Simple effects analysis
explicit attribute-based and metaphor-based were conducted for each of the two conditions,
dehumanization. This finding is consistent with revealing that participants associated the
predictions and with the findings of Study 1. Luupites more with robots than animals in the
low-HN condition, F(1,39) = 13.00, p = 0.001,
Implicit dehumanization In the analysis of the with no difference in the low-HU condition,
implicit association (GNAT) data a 2 (condition: F(1,42) = 3.11, p = 0.085. In sum, people have
low-HN or low-HU) × 2 (humanness: HN and difficulty learning and inferring dehumaniza-
HU) mixed model ANOVA failed to yield the tion implicitly.
predicted two-way interaction between condition
and humanness, although the effect trended in Discussion
the expected direction F(1, 81) = 2.89, p = 0.09. The results of Study 2 replicated the findings
Participants in the low-HU condition more of Study 1. Consistent with the scenario-based
readily associated the group with HN traits than learning of the previous study, concrete learning
with HU traits, whereas those in the low-HN yielded explicit dehumanization and mixed
condition more readily associated the group with evidence for implicit dehumanization. This study
HU than with HN traits (see Figure 2). Planned employed a new learning paradigm previously
comparisons revealed that this difference demonstrated to induce both implicit and ex-
failed to reach significance in both the low-HN plicit attitudes (Ranganath & Nosek, 2008; Rydell
condition, F(1, 81) = 0.30, p = 0.58, and the low- et al., 2006). The technique proved effective
HU condition, F(1, 81) = 1.59, p = 0.21. at teaching people that a novel group lacked
To investigate whether learning about the certain types of humanness. Further, participants
humanness characteristics possessed by a group moved beyond the supplied attribute-based
might lead to the inference of implicit metaphor- information to perceive a group that lacked
based dehumanization a second 2 (condition: humanness attributes to be like nonhumans,
although this inference occurred only at an ex-
plicit level. In short, concrete learning proved
an effective technique for inducing explicit
dehumanization and lead to the direct learning
of implicit dehumanization.

General discussion
This study aimed to investigate the relation-
ship between the two predominant approaches
in the study of dehumanization: attribute-
based and metaphor-based dehumanization.
Figure 2. Implicit associations (d′) between novel We taught participants to dehumanize groups
group, humanness attributes and types of nonhuman using one type of dehumanization and examined
in the low-HU (dark bars) and low-HN (light bars) whether they inferred the second type. Across
conditions, Study 2. two studies we robustly demonstrated that

757

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

people can learn to dehumanize explicitly and of a particular experimental design. Rather, it
implicitly, whereas inference is far stronger at may suggest something about the process by
an explicit level than at an implicit level. which people come to acquire elaborate, multi-
level representations of others. We suggest that
Dehumanization theory Both infrahumaniza- this may reflect a specific instance of a more
tion and dehumanization theory have posited general metaphor-making process.
that a denial of specific attributes leads to people Previous cognitive psychology research has dem-
being subtly likened to various nonhuman onstrated two processes at work in laypeople’s
entities. The denial of HN is theorized to lead use of metaphors (Bowdle & Gentner, 2005;
to likening to robots, machines or objects Gentner & Kurtz, 2005; Gentner & Markman,
(Haslam, 2006) and the denial of HU to liken- 1997; Markman & Gentner, 1993). These pro-
ing to animals (Haslam, 2006; Leyens et al., cesses involve representing the specific attrib-
2001, 2007). Previous research has provided utes possessed by two entities and assessing the
indirect evidence for this claim, demonstrating degree of alignment or concordance between
that groups denied HN tend to be mechanized them. For instance, people may think that
and those denied HU animalized (Loughnan & members of a specific group are uncouth,
Haslam, 2007). The current research provides stupid, and lack linguistic competence (low-HU
evidence not only that this is the case, but that attribute-based dehumanization). They may
the effect is bi-directional; the denial of human- also think that this is true of animals. Through
ness attributes can be inferred from metaphor- the process of attribute alignment, the high
based dehumanization and vice versa. degree of concordance between the attributes
This fi nding is important for at least two possessed by the group and animals leads
reasons. First, it can help explain the regularity people to form this dehumanizing metaphor.
with which the denial of one type of humanness ‘Candidate inference’ (Gentner & Markman,
is paired with a specific nonhuman metaphor. 1997) inverts this process to explain how people
The co-occurrence of low-HU and animalization can move from a higher-level metaphor-based
has been observed in multiple studies (Loughnan representation to infer the attributes possessed
& Haslam, 2007; Saminaden et al., in press) as by the entities. This process starts from the ex-
has low-HN and mechanization (Loughnan & istence of a working metaphor (e.g. those people
Haslam, 2007). This co-occurrence may arise due are robots). People possess lay theories about
to the inference process demonstrated in our the psychological characteristics of robots
studies. Second, the finding helps to explain (Haslam et al., 2008), specifically that they are
how people can come to hold a metaphor-based cold, unfeeling, and inflexible. People can use
dehumanization of groups without directly an established metaphor to infer attributes
learning those metaphors. If people already view possessed by the target (i.e. the group) from
a group as lacking a specific type of humanness the well-known base (i.e. the nonhuman entity).
(e.g. HU) then this research indicates they can Thus, a group likened to robots (metaphor-
infer that the group resembles a specific kind based dehumanization) can be inferred to be
of nonhuman (e.g. animals). In sum, this re- cold, unfeeling, and inflexible (attribute-based
search provides direct evidence for a previously dehumanization).
untested theoretical claim. Employing this metaphor-making account
to explain the relationship between attribute-
Why do people infer at an explicit but not an based and metaphor-based dehumanization is
implicit level? The second key finding of this useful for several reasons. Cognitive psychol-
research is that people are able to infer the the- ogists have already demonstrated that the pro-
oretically consistent type of dehumanization cesses of attribute alignment and candidate
only at an explicit level. The use of different inference require considerable cognitive invest-
learning paradigms and both broad and specific ment (Gentner & Kurtz, 2005). Given these
conceptualizations of the nonhuman category constraints, this clarifies why dehumanization
suggest that this effect is not simply an artifact might be readily learned and inferred at an

758

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

explicit but not implicit level. When given the Being gracious and being impersonal are both
time and resources to engage in effortful pro- considered low in human nature, but the latter
cessing, people are able to engage attribute is arguably more likely to lead to a mechanistic
alignment and candidate inference processes inference than the former. The capacity
that lead to the elaboration of dehumanizing for specific attributes to lead to nonhuman
representations. In contrast, when time and metaphors has yet to be investigated. These
cognitive resources are constrained through specifics are beyond the scope of the present
the use of an implicit technique (De Houwer studies, but nevertheless could be accommodated
& Moors, 2007), people are unable to employ within the model above.
these processes and thus only direct learning
effects are found. Previous dehumanization
Conclusion
researchers have suggested that the movement
from subtle to blatant dehumanization might be a This study aimed to examine an important
motivated processes (Kofta & Miroslawska, under but previously untested component of current
review), and this research may help shed light on theory; that denials of specific types of human-
the mechanism through which it is achieved. defining attributes lead to a likening of others
The second benefit of adopting this model to nonhumans. Incorporating two distinct ap-
is that it allows us to understand how people proaches to the study of dehumanization in
might come to hold dehumanizing metaphors the literature, attribute- and metaphor-based
of groups without direct exposure to those dehumanization, we examined people’s ability
metaphors. Although such metaphors have to move between these two forms. Across two
historically been attached to certain groups studies our results indicate that people can learn
(Jahoda, 1999; Lott, 1999) and continue to be to dehumanize and will infer the correspond-
employed in the discourse of extremists, main- ing type of dehumanization in a theoretically
stream, everyday dehumanization may spring consistent manner. Importantly, this latter effect
from a different source. These groups may was much stronger explicitly than implicitly.
be stereotyped in ways that are dehumanizing We conclude that this difference between the
(e.g. lacking HU), and this may lead to the in- implicit and explicit may be understood as a
ference of dehumanizing metaphors (e.g. being special instance of a more general metaphor
animals). This possibility is consistent with the making ability.
finding of Goff et al. (2008) that contempor-
ary Americans who claimed that they had not
Note
been exposed to the African-ape metaphor
and were not aware of the history of metaphor- 1. We use the term ‘dehumanization’ to
based dehumanization nevertheless associated refer to the denial of humanness. Both
African-Americans with apes. In short, there infrahumanization theory and Haslam’s
may be multiple routes to the acquisition of (2006) model of dehumanization fall under
this framework.
dehumanizing perceptions of groups.
Finally, it should be noted that these studies
have dealt with broad attribute-metaphor asso- References
ciations, as proposed by infrahumanization and
Baayens, R. S., Piepenbrock, R., & van Rijn, H.
dehumanization theory. It is possible that some (1995). The CELEX lexical database (Version 2).
specific instances of dehumanization rest on Philidephia, PA.
highly differentiated trait-metaphor sets. For Boccato, G., Capozza, D., Falvo, R., & Durante, F.
instance, saying that members of a group are (2008). The missing link: Ingroup, outgroup, and
‘leeches’ more readily leads to the inference the human species. Social Cognition, 26(2), 224–234.
that they are parasitic than that they lack linguis- Boccato, G., Cortes, B., Demoulin, S., & Leyens, J.
tic competence. Further, some traits may privy (2007). The automaticity of infrahumanization.
dehumanizing inferences more than others. European Journal of Social Psychology, 37(8), 987–999.

759

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

Bowdle, B., & Gentner, D. (2005). The career of non-humans in three cultures. Social Cognition,
metaphor. Psychological Review, 112(1), 193–216. 26(2), 248–258.
Capozza, D., Andrighetto, L., Falvo, R., & Trifiletti, E. Jahoda, G. (1999). Images of Savages: Ancient roots
(2006). The use of GNAT to study intergroup of modern prejudice in Western culture. London:
dehumanisation. Testing, Psychometrics, Routledge.
Methodology, in Applied Psychology, 13(2), 295–304. Karpinski, A., & Stienman, R. (2006). The single
De Houwer, J., & Moors, A. (2007). How to define category implicit association test as a measure of
and examine the implicitness of implicit implicit social cognition. Journal of Personality and
measures. In B. Wittenbrink & N. Schwarz Social Psychology, 91(1), 16–32.
(Eds.), Implicit measures of attitudes (pp. 179–194). Kofta, M., & Miroslawska, M. (under review).
New York: Guilford Press. Infrahumanization and ingroup bias in the minimal
Gawronski, B., & Bodenhausen, G. (2006). group paradigm: How do they relate to each other?
Associative and propositional processes in Unpublished manuscript.
evaluation: Conceptual, empirical, and Leyens, J.-Ph., Demoulin, S., Vaes, J., Gaunt, R., &
meta-theoretical issues. Reply to Albarrasin, Paladino, M.-P. (2007). Infrahumanization: The
Hart, and McCulloch (2006), Kruglanski and wall of group differences. Social Issues and Policy
Dechesne (2006), and Petty and Brinol (2006). Review, 1(1), 139–172.
Psychological Bulletin, 132, 745–750. Leyens, J.-Ph., Rodríguez-Pérez, A.,
Gawronski, B., & Bodenhausen, G. (2007). Rodríguez-Torres, R., Gaunt, R., Paladino, M.-P.,
Unraveling the processes underlying evaluation: Vaes, J., et al. (2001). Psychological essentialism
Attitudes from the perspective of the APE model. and the differential attribution of uniquely
Social Cognition, 25(5), 687–717. human emotions to ingroups and outgroups.
Gentner, D., & Kurtz, K. (2005). Relations, objects, European Journal of Social Psychology, 31, 395–411.
and the composition of analogies. Cognitive Lott, T. (1999). The Invention of Race: Black culture
Science, 30(5), 609–642. and the politics of representation. Oxford: Blackwell
Gentner, D., & Markman, A. (1997). Structure Publishers.
mapping in analogy making and similarity. Loughnan, S., & Haslam, N. (2007). Animals and
American Psychologist, 52(1), 45–56. androids: Implicit associations between social
Goff, P., Eberhardt, J., Williams, M., & Jackson, M. groups and nonhumans. Psychological Science,
(2008). Not yet human: Implicit knowledge, 18(2), 116–121.
historical dehumanization, and contemporary Markman, A., & Gentner, D. (1993). Structural
consequences. Journal of Personality and Social alignment during similarity comparisons.
Psychology, 94(2), 292–306. Cognitive Psychology, 25, 431–467.
Green, D. M., & Swets, J. A. (1966). Signal Detection Nosek, B., & Banaji, M. (2001). The Go/No-go
Theory and Psychophysics. New York: Wiley. association task. Social Cognition, 19(6), 625–666.
Gregg, A., Seibt, B., & Banaji, M. (2006). Paladino, M.-P., Leyens, J.-Ph., Rodríguez-Torres, R.,
Easier done than undone: Asymmetry in the Rodríguez-Pérez, A., Gaunt, R., & Demoulin, S.
malleability of implicit preferences. Journal of (2002). Differential association of uniquely and
Personality and Social Psychology, 90(1), 1–20. non-uniquely human emotions with the ingroup
Haslam, N. (2006). Dehumanization: An integrative and the outgroup. Group Processes & Intergroup
review. Personality and Social Psychology Review, Relations, 5(2), 105–117.
10(3), 252–264. Ranganath, K., & Nosek, B. (2008). Implicit
Haslam, N., & Bain, P. (2007). Attributing greater attitude generalization occurs immediately,
humanness to the self: Moderators and the explicit attitude generalization takes time.
attribution of lesser humanness to others. Psychological Science, 19(3), 249–255.
Personality and Social Psychology Bulletin, 33(1), Rydell, R., McConnell, A., & Mackie, D. (2008).
57–68. Consquences of discrepent explicit and implicit
Haslam, N., Bain, P., Douge, L., Lee, M., & Bastian, B. attitudes: Cognitive dissonance and increased
(2005). More human than you: Attributing information processing. Journal of Experimental
humanness to self and others. Journal of Social Psychology, 44(7), 1526–1532.
Personality and Social Psychology, 89, 937–950 Rydell, R., McConnell, A., Mackie, D., & Strain, L.
Haslam, N., Kashima, Y., Loughnan, S., Shi, J., & (2006). Of two minds: Forming and changing
Suitner, C. (2008). Subhuman, inhuman, valence-inconsistent implicit and explicit
and superhuman: Contrasting humans with attitudes. Psychological Science, 17(11), 954–958.

760

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Loughnan et al. attribute and metaphor dehumanization

Rydell, R., McConnell, A., Strain, L., Claypool, H., & and implicit theories. Annual Review of Psychology,
Hugenberg, K. (2007). Implicit and explicit 59, 329–360.
attitudes respond differently to increasing Viki, T., Winchester, L., Titshall, L., Chisango, T.,
amounts of counterattitudinal information. Pina, A., & Russell, R. (2006). Beyond secondary
European Journal of Social Psychology, 37(6), emotions: The infrahumanisation of outgroups
867–878. using human-related and animal-related words.
Saminaden, A., Loughnan, S., & Haslam, N. Social Cognition, 24(6), 723–752.
(in press). Afterimages of savages: Implicit Wilson, S. (2008). Perceptions of the humanness of
associations between “primitives”, animals and behaviours. Unpublished data, The University of
children. British Journal of Social Psychology. Melbourne.
Uleman, J., Saribay, S., & Gonzalez, C. (2008).
Spontaneous inferences, implicit impressions,

Appendix
Study 1
Story Stimuli
Low-HN Low-HU Animals Robots
Gracious Warm Considered animals Considered robots
Civil Vivacious Live outside Live in cities
Refined Nervous Excellent perception Excellent precision
Intelligent Fickle Irrational Highly rational
Hard-hearted Intolerant No language Complex language
Conforming Trusting Spontaneous Rigid
Conservative Distracted Creative Inflexible
Shallow Outgoing Tough hides Methodical
Polite Curious Thick hair Strict routine
Moral Aggressive Athletic Programmed
Cold Exuberant Totemism Computer use
Organized Emotional Animalistic games Computerized games
Rude Friendly
Humble Jealous
Thorough Sociality
Broadminded Fun
Impersonal Open

Study 2
Learning Task Stimuli
Human Nature Human Uniqueness
Traits Traits
Friendly Sociable Humble Broadminded
Fun-loving Trusting Thorough Stingy
Curious Aggressive Cold Polite
Jealous Impatient Shallow Hard-hearted
Nervous Distractible Organized Conservative
Behaviors Behaviors
Maintaining relationships Writing
Having sex Reading
Verbal communication Studying
Snoozing Performing calculations
Forming groups Numerical reasoning

761

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009


Group Processes & Intergroup Relations 12(6)

GNAT Stimuli
Animals Animals
HN HU Automata (Study 1) (Study 2)
Study 1 Friendly Conservative Robot Chimp Alligator
Fun-loving Cold Software Baboon Animals
Curious Hard-hearted Android Gorilla Beast
Aggressive Thorough Cyborg Monkey Cattle
Distractible Organized Mechanism Orangutan Chimpanzee
Jealous Polite Automata Gibbon Elephant
Study 2 Sociable Humble Machine – Kangaroo
Additional Trusting Shallow Computer – Mammals
Stimuli Impatient Broadminded Synthetic – Platypus
Nervous Stingy Mechanical – Primates

Biographical notes include dehumanization, psychological


essentialism, and prejudice.
steve loughnan received his PhD from the
University of Melbourne. His research interests yoshihisa kashima is Professor of Psychology at
include dehumanization and metaphors. the University of Melbourne. He has research
interests in stereotyping, communication
nick haslam is Professor of Psychology at the processes, essentialism, cross-cultural
University of Melbourne. His research interests psychology, and connectionist modeling.

762

Downloaded from http://gpi.sagepub.com at University of Melbourne Library on October 21, 2009

You might also like