855638

research-article2019
PPSXXX10.1177/1745691619855638De HouwerImplicit Bias Is Behavior

ASSOCIATION FOR
PSYCHOLOGICAL SCIENCE

Perspectives on Psychological Science

Implicit Bias Is Behavior: A Functional- 2019, Vol. 14(5) 835­–840

© The Author(s) 2019
Article reuse guidelines:
Cognitive Perspective on Implicit Bias sagepub.com/journals-permissions
DOI: 10.1177/1745691619855638
https://doi.org/10.1177/1745691619855638
www.psychologicalscience.org/PPS

Jan De Houwer
Department of Experimental-Clinical and Health Psychology, Ghent University

Abstract
Implicit bias is often viewed as a hidden force inside people that makes them perform inappropriate actions. This
perspective can induce resistance against the idea that people are implicitly biased and complicates research on
implicit bias. I put forward an alternative perspective that views implicit bias as a behavioral phenomenon. more
specifically, it is seen as behavior that is automatically influenced by cues indicative of the social group to which
others belong. This behavioral perspective is less likely to evoke resistance because implicit bias is seen as something
that people do rather than possess and because it clearly separates the behavioral phenomenon from its normative
implications. Moreover, performance on experimental tasks such as the Implicit Association Test is seen an instance
of implicitly biased behavior rather than a proxy of hidden mental biases. Because these tasks allow for experimental
control, they provide ideal tools for studying the automatic impact of social cues on behavior, for predicting other
instances of biased behavior, and for educating people about implicitly biased behavior. The behavioral perspective
not only changes the way we think about implicit bias but also shifts the aims of research on implicit bias and reveals
links with other behavioral approaches such as network modeling.

Keywords
implicit bias, functional psychology, racism

In this brief article, I argue that there is merit in think- & Stewart, 2005).1 These measures are then used to pre-
ing about implicit bias as a behavioral phenomenon. I dict the behaviors that are assumed to be driven by an
first discuss the more widespread perspective that implicit bias (see Fig. 1). From this perspective, implicit
views implicit bias as a latent mental construct and bias thus qualifies as a latent mental construct.
highlight two important downsides of this perspective. This popular perspective on implicit bias has a num-
Next, I explain the alternative, behavioral perspective ber of downsides. First, it fosters a quite disturbing view
and discuss the potential advantages, limitations, and on implicit bias: It is an uncontrollable, hidden force
implications of that perspective. inside people that makes them perform inappropriate
actions. Being told that we are implicitly biased can
therefore threaten core beliefs about who we think we
Implicit Bias as a Latent Mental Construct are and aspire to be (e.g., Sukhera, Milne, Teunissen,
Implicit bias is often thought of as an unobservable Lingard, & Watling, 2018). Although this disturbing view
structure in the mind of an individual (e.g., an associa- might not be an inevitable consequence of defining
tion in memory) that drives behavior in an unconscious implicit bias as a hidden mental structure, it is likely to
manner (e.g., Amodio & Mendoza, 2010). Because it be an important source of the defensive reactions that
cannot be observed directly, measures are needed that many people display when being told that they are
index the strength and nature of implicit bias. Several implicitly biased (e.g., Howell, Gaither, & Ratliff, 2015).
experimental tasks have been adopted for this purpose,
such as the Implicit Association Test (IAT; e.g., Greenwald,
Corresponding Author:
McGhee, & Schwartz, 1998), evaluative priming tasks Jan De Houwer, Ghent University, Henri Dunantlaan 2, B-9000 Ghent,
(e.g., Fazio, Jackson, Dunton, & Williams, 1995), and the Belgium
affect-misattribution procedure (Payne, Cheng, Govorun, E-mail: [email protected]
836 De Houwer

implicit bias. The cornerstone of this perspective is the

idea that implicit bias is a behavioral phenomenon
rather than a mental structure. In other words, implicit
B1 B2 B3 B4 Measure bias is something that people do rather than something
Physical that people possess. More specifically, implicit bias can
be defined as implicit group-based behavior, which is
Mental behavior that is influenced in an implicit manner by
cues that function as an indicator of the social group
Bias to which others belong. For instance, saying that some-
one is racially biased means that part of what that
person does (e.g., whether the person smiles at some-
Fig. 1. Schematic representation of implicit bias as a latent mental one, shakes hands with someone, hires someone for a
construct that causally influences physical behavior (B1 . . . B4) and
can be indexed using a measure that predicts behavior.
job) is influenced by cues indicative of the racial group
of others (e.g., skin color). The influence of these social
cues can be labeled as implicit when it occurs quickly,
Thus, the latent-mental-construct perspective has effortlessly, unintentionally, unconsciously, or in a way
undoubtedly contributed to the controversy that sur- that is difficult to control (see Moors, 2016, and De
rounds the notion of implicit bias. It is also likely to Houwer et al., 2009, for more details). 2 For instance,
hamper attempts to reduce implicit bias in society, not someone shows an implicit racial bias when he or she
only because of the controversy it instills but also quickly and unintentionally responds fearfully to the
because the metaphor of a hidden mental structure presence of another person because of the skin color
encourages the idea that implicit bias is a stable entity of that person. Implicit group-based behavior can be
that is difficult to change and control (e.g., Sukhera referred to as biased in that behavior is influenced (by
et al., 2018). social cues) in a systematic way. Note that the behav-
Second, a latent-mental-construct perspective com- ioral perspective is thus amoral in the sense that it does
plicates the measurement of implicit bias and thus not require a judgment about whether the impact of
research on this topic. As indicated by the thick line in social cues on behavior is inappropriate according to
Figure 1, measures of implicit bias are assumed to some norm. It allows for moral debates but separates
directly tap into the unobservable structure and hence them from the behavioral phenomenon in itself. It also
to provide a proxy of this structure. The fact that these does not assign blame for biased behavior but simply
measures are used to predict biased behaviors is based implies that the behavior is a function of social cues in
on the assumption that variations in the measurement the environment. In the remainder of this section, I
outcome reflects variations in the underlying implicit discuss potential advantages, downsides, and implica-
bias. It is, however, notoriously difficult to validate this tions of a behavioral perspective on implicit bias.
assumption (e.g., De Houwer, Teige-Mocigemba,
Spruyt, & Moors, 2009). In fact, because implicit mea-
Accepting implicit bias
sures are likely to reflect multiple mental constructs
and processes, it is highly unlikely that differences in Compared with implicit bias as seen from a latent-
measurement outcomes reflect differences in one spe- mental-construct perspective, people might be more
cific mental construct. This is not a problem if the willing to entertain the possibility that their behavior
measure does exactly what it is expected to do (e.g., it is implicitly biased in a behavioral sense. The behav-
adequately predicts biased behavior). But in those (fre- ioral perspective implies only that social cues can have
quent) cases in which the measure does not conform automatic effects on behavior. It does not require peo-
with expectations, it is very difficult to determine ple to accept a specific theory for why their behavior
whether this is due to a problem with the measure (e.g., might sometimes be biased (e.g., the idea that behavior
it does not capture implicit bias adequately) or with can be controlled by hidden entities within a person)
the theory about the construct (i.e., ideas about when or to accept the idea that biased behavior necessarily
and how implicit bias influences behavior; for a discus- violates some norm. Moreover, implicit bias as behavior
sion of this issue in the context of attitudes research; is probably seen as more malleable than implicit bias
see De Houwer, Gawronski, & Barnes-Holmes, 2013). as a hidden mental structure and might therefore
heighten the belief that the problem of implicit bias can
be remedied. Of course, people will invest effort into
Implicit Bias as a Behavioral Phenomenon
trying to prevent or counteract implicit bias only if they
These important downsides can be circumvented by perceive it to be potentially inappropriate in some
adopting an alternative, behavioral perspective on respect. As noted earlier, a behavioral perspective on
Implicit Bias Is Behavior 837

implicit bias allows for debates on these normative group-based behavior that is observed under well-
issues but separates them from debates about whether controlled conditions.
behavior can be influenced automatically by social cues. Although a behavioral perspective on implicit bias
Thus, adopting a behavioral perspective is likely to help strips performance on experimental tasks from its privi-
dampen some of the controversy that surrounds the leged position as a proxy of hidden mental structures,
notion of implicit bias (e.g., Jost et al., 2009). it continues to assign a prominent role to these tasks.
As noted by a reviewer, many people also might Most importantly, the tasks provide ideal tools to exam-
resist the idea that they act in biased ways. They might ine whether and when behavior is implicitly biased.
even resist the idea that biased behavior can be From a behavioral perspective, the presence of implicit
changed. Although this is certainly possible, a behav- bias can be established by showing an impact of social
ioral perspective is likely to engender less resistance cues on behavior under conditions of automaticity.
than a latent-mental-construct perspective. Both per- Because experimental tasks allow researchers to (a)
spectives entail the possibility of implicitly biased vary the presence of social cues while controlling for
behavior, but only the latter requires assumptions about possible confounds, (b) establish conditions of auto-
hidden mental causes of biased behavior. Both perspec- maticity, and (c) carefully register changes in behavior,
tives allow for the idea that when one is vigilant, implic- they are ideally suited to study implicit bias. Although
itly biased behavior can be prevented or counteracted, it can be challenging to demonstrate the automatic
but the latent-mental-construct perspective puts the impact of social cues on behavior even in experimental
spotlight on stable mental structures rather than on tasks, a behavioral perspective frees research on implicit
malleable behavior. Moreover, the behavioral perspec- bias from the need for proxies of hidden mental con-
tive allows one to clearly separate implicit bias as a structs, thereby side-stepping debates about whether
behavioral phenomenon from the often heated debates and when variations in task performance reflect varia-
about the appropriateness of biased behavior. Finally, tions in hidden mental structures. The focus of research
as will be argued in the next section, experimental tasks is no longer on the hidden mental structures but on the
can be used to objectively reveal (changes in) biased behavior itself.3
behavior and thus to weaken resistance against the idea On the basis of the plausible assumption that differ-
of (changes in) biased behavior. Ultimately, empirical ent instances of a particular type of implicit bias (e.g.,
research will have to determine whether a behavioral racial bias, gender bias) are related to each other (e.g.,
perspective increases acceptance of the idea of implicit persons likely to emit certain types of racial behavior
bias. However, by putting forward the behavioral per- are also more likely to emit other types of racial behav-
spective as a promising alternative for the latent-mental- ior), one can also continue to explore the potential of
construct perspective, the current article already sets experimental tasks as tools to predict real-life biased
the stage for this research. behavior. From a behavioral perspective, the predictive
utility of performance in these tasks will depend on the
extent to which performance is influenced by the same
Examining implicit bias cues under the same conditions as real-life biased
The idea that implicit bias is a behavioral phenomenon behavior (see Ciarrochi et al., 2016). One could say that
also removes the need for measures that tap into a performance in experimental tasks such as the IAT is
latent mental construct. Instead, so-called measures of related to real-life instances of implicit bias as driving
implicit bias can simply be viewed as instances of in a driving simulator is related to driving in real life.
implicit bias, that is, as specific examples of behavior Because the two elements of each pair are instances of
that is influenced automatically by cues indicative of the same phenomenon that occur under different cir-
the social group of others (e.g., see Ciarrochi et al., cumstances, it is likely that the relation between both
2016, and De Houwer et al., 2013 for related ideas). For elements will depend on how similar those circum-
instance, saying that performance on a race-related stances are. Just as the relation between driving in a
version of the IAT is an instance of implicit racial bias simulator and real-life driving can be examined without
implies that IAT performance (more specifically, dif- having to assume that driving in a simulator taps into
ferences in the speed of responding in the different the latent mental construct that determines real-life
blocks of the IAT) is influenced automatically by the driving, so too can the relation between performance
racial cues of the stimuli on the screen. From this per- in experimental tasks and real-life biased behavior be
spective, IAT performance, as well as performance on examined without the need to assume that the former
several other experimental tasks (e.g., the evaluative is a measure of a hidden mental structure that deter-
priming task; Fazio et al., 1995), is simply implicit mines real-life biased behavior (see Fig. 1).
838 De Houwer

Finally, a behavioral perspective on implicit bias also to establish and study whether and when conscious
facilitates the use of experimental tasks for educational feelings and thoughts are influenced automatically by
purposes. For instance, rather than having to interpret social cues such as skin color. From a behavioral per-
a race IAT score as an index of a hidden mental struc- spective, such feelings and thoughts would be seen as
ture that biases behavior, it can simply be described as instances of implicitly biased behavior, which, like
an example of how behavior (i.e., response times) can other instances of biased behavior, are things that peo-
be influenced by race-related cues even when people ple do rather than possess. Studying implicitly biased
do not have the intention to be influenced by those feelings and thoughts does not require a definition of
cues. As noted above, such a behavioral framing of implicit bias as a latent mental construct.
performance in experimental tasks is likely to engender What about unconscious feelings and thoughts?
less defensive reactions than a framing in terms of hid- Because it is not possible to observe or intervene in
den mental structures. Because experimental tasks can these feelings and thoughts directly, they are typically
be used to provide objective information about actual considered to fall outside the scope of behavioral
differences in performance, they are well suited for research. However, adopting a behavioral perspective
demonstrating implicit group-based behavior. on implicit bias does not deny the possibility of adopting
Moreover, because the aim is to illustrate a type of a cognitive perspective. Whereas behavioral research on
behavior rather than to measure a latent mental con- implicit bias can be seen as directed at documenting the
struct, education on implicit bias no longer needs to environmental conditions under which implicit group-
refer to assumptions about hidden mental structures based behavior occurs (i.e., the moderators of implicit
and how those structures relate to task performance bias), cognitive research on implicit bias can be seen as
and real-life biased behavior. The claim that an indi- focused on documenting the mental processes that medi-
vidual or group of individuals is displaying implicit ate implicit group-based behavior (i.e., the mental medi-
group-based behavior can be based solely on perfor- ators of implicit bias). Because the two approaches have
mance during the experimental task. In principle, any different aims, they do not compete but complement
automatic effect of any social cue on any type of behav- each other. I thus subscribe to a functional-cognitive
ior would suffice to demonstrate implicit bias. Because framework for research on implicit bias that allows one to
of the amoral nature of the behavioral perspective, it reconcile behavioral and cognitive research (De Houwer,
is not necessary also to argue that performance on the 2011; Hughes, De Houwer, & Perugini, 2016). In fact, it
experimental task is inappropriate in some normative can be argued that cognitive research on implicit bias will
sense. Of course, the extent to which a demonstration benefit from clearly separating the phenomenon to be
of implicit bias will have an impact is likely to depend explained (i.e., implicit group-based behavior) from
on the extent to which additional evidence shows that mental theories of that phenomenon (e.g., the idea that
within the global population or a section of the popula- associations in memory are responsible for implicit
tion, biased performance in the experimental task is group-based behavior; for more details, see De Houwer
related to important and normatively inappropriate real- et al., 2013; De Houwer, Hughes, & Barnes-Holmes,
life instances of biased behavior. Accumulating this 2017; Hughes et al., 2016).
additional evidence will require effort, but this work Rather than denying the possibility of a cognitive
needs to be done anyway, regardless of the perspective perspective on implicit bias, a behavioral perspective
on implicit bias that one takes. In sum, a behavioral on implicit bias firmly shifts the focus from the mental
perspective facilitates education on implicit bias by level to the behavioral level. I believe that such a shift
shifting the focus toward actual behavior. in focus is appropriate not only because of the practical
implications described elsewhere in this article but also
because it brings the concept of implicit bias in line
What about feelings and thoughts?
with the aims of many researchers who study implicit
Some readers might worry that a behavioral perspective bias. From a behavioral perspective, the problem of
misses out on the fact that feelings and thoughts are implicit bias in society is ultimately a behavioral prob-
crucial in implicit bias. When considering conscious lem. It is all about (changing) what people do. It is not
feelings and thoughts, this problem can easily be cir- primarily about the mental causes of biased behavior.
cumvented by adopting a broad definition of behavior. Having theories about the mental causes of biased
Behavior is not necessarily limited to the movement of behavior can be useful, not in the least because theories
muscles and glands. In addition, conscious feelings and can generate ideas about how to change the biased
thoughts can be observed (be it only by the person behavior, but ultimately, the phenomenon itself is a
who has them) and thus treated as instances of (covert) behavioral one. Just consider the fact that most (applied)
behavior (Skinner, 1953). Hence, it is possible researchers would be happy with finding ways to
Implicit Bias Is Behavior 839

reduce biased behavior even if they do not have a good Conclusion

cognitive theory of implicit bias (for a discussion of the
idea that applied topics in psychology are always pri- As Machado and Silva (2007) elegantly argued, science
marily about behavior in a broad sense, see De Houwer in general and psychology in particular have much to
et al., 2017). gain from conceptual analyses. In this article, I put
forward a new perspective on a psychological concept
that had a huge impact on society during the past
Implications decade: implicit bias. More specifically, I put forward
Although it is difficult to foresee how exactly the field the idea that implicit bias can be thought of as implicit
of implicit bias research will change as the result of group-based behavior. Adopting this behavioral per-
adopting a behavioral perspective, the impact is bound spective implies not only a shift in thinking about
to be substantial. First, it will change the way we com- implicit bias but also has important implications for the
municate about implicit bias. As noted above, there are aims and practices of research on implicit bias.
good reasons to assume that this shift will weaken the
controversy surrounding the concept of implicit bias Action Editor
and thus increase willingness to tackle the problem of Laura A. King served as action editor for this article.
implicit bias, which, at its core, is a behavioral problem.
Second, the behavioral perspective will facilitate ORCID iD
research on implicit bias by eliminating the need for Jan De Houwer https://orcid.org/0000-0003-0488-5224
measures that can function as proxies of hidden mental
biases. Experimental tasks will still be used to examine Acknowledgments
implicit bias, predict real-life biased behaviors, and I thank Sean Hughes and Pieter Van Dessel for their helpful
educate people about implicit bias, but the focus of this comments on earlier drafts of the manuscript. An outline of
research will be shifted to actual behavior, more specifi- this article was presented at the NSF Implicit Bias Conference,
cally, the conditions under which behavior is influenced Alexandria, Virginia, on September 28, 2017.
by cues indicative of social groups as well as the condi-
tions under which different biased behaviors are related. Declaration of Conflicting Interests
Adopting a behavioral perspective on implicit bias The author(s) declared that there were no conflicts of interest
will also bring research on implicit bias into contact with respect to the authorship or the publication of this
with other approaches in science that focus on behavior article.
rather than latent mental constructs. One particularly
interesting approach is offered by networks models. In Funding
their most basic form, network models are graphic ways The preparation of this article was made possible by Methu-
of describing covariations or causal relations between salem Grant BOF16/MET_V/002 of Ghent University
different environmental events, including different
behaviors (Borsboom & Cramer, 2013; Robins, 2013). Notes
For instance, a network model of depression reveals
1. I use the term experimental task in a broad sense that applies
how different symptoms of depression relate to each to any task that allows researchers to exert control over the situ-
other and change over time into stable states (Cramer ation that people are exposed to and to register behavior in a
et al., 2016). Such a network perspective provides ways standardized manner.
to talk about and examine depression that do not 2. I thus equate the term implicit with the term automatic and
require one to assume a common latent mental cause define automatic as an umbrella term that refers to different
of all symptoms. Just as network models have revolu- automaticity features that do not necessarily overlap.
tionized research on psychopathology (e.g., Borsboom 3. Critics of the behavioral view could emphasize the challenges
& Cramer, 2013; Borsboom, Cramer, & Kalis, 2019), they that it raises. Note, however, that a latent-mental-construct per-
have the potential to change the face of research on spective raises the same challenges because performance on
experimental tasks can be a proxy of hidden mental biases
implicit bias. Most importantly, network models of
only if it is related to social cues in the environment. Hence, the
implicit bias do not require the assumption that all
use of proxies not only requires demonstrating the automatic
instances of biased behavior depend on a common impact of social cues on behavior (e.g., that IAT performance
latent mental cause. Instead, they can reveal how is influenced by skin color rather than familiarity of the faces
instances of (racial) behavior causally influence each presented on a computer screen) but also raises the additional
other (e.g., whether and when frequency of contact challenge of showing that this effect is related in a particular
with outgroups leads to racial bias in hiring). way to hidden mental biases.
840 De Houwer

References measure of racial attitudes: A bona fide pipeline? Journal

of Personality and Social Psychology, 69, 1013–1027.
Amodio, D. M., & Mendoza, S. A. (2010). Implicit intergroup Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998).
bias: Cognitive, affective, and motivational underpinnings. Measuring individual differences in implicit cognition:
In B. Gawronski & B. K. Payne (Eds.), Handbook of implicit The implicit association test. Journal of Personality and
social cognition (pp. 353–374). New York, NY: Guilford. Social Psychology, 74, 1464–1480.
Borsboom, D., Cramer, A., & Kalis, A. (2019). Brain disorders? Howell, J. L., Gaither, S. E., & Ratliff, K. A. (2015). Caught
Not really: Why network structures block reductionism in in the middle: Defensive responses to IAT feedback
psychopathology research. Behavioral & Brain Sciences, among Whites, Blacks, and Biracial Black/Whites. Social
42, Article e2. doi:10.1017/S0140525X17002266 Psychological & Personality Science, 6, 373–381.
Borsboom, D., & Cramer, A. O. J. (2013). Networks: An inte- Hughes, S., De Houwer, J., & Perugini, M. (2016). The
grative approach to the structure of psychopathology. functional-cognitive framework for psychological research:
Annual Review of Clinical Psychology, 9, 91–121. Controversies and resolutions. International Journal of
Ciarrochi, J., Brockman, R., Duguid, J., Parker, P., Sahdra, B., Psychology, 51, 4–14.
& Kashdan, T. (2016). Measures that make a difference: Jost, J. T., Rudman, L. A., Blair, I. V., Carney, D. R., Dasgupta, N.,
Optimizing psychological measurement to promote well- Glaser, J., & Hardin, C. D. (2009). The existence of implicit
being and reduce suffering. In R. Zettle, S. Hayes, T. bias is beyond reasonable doubt: A refutation of ideologi-
Biglan, & D. Barnes-Holmes (Eds.), Handbook of con- cal and methodological objections and executive summary
textual behavioral science (pp. 320–347). Malden, MA: of ten studies that no manager should ignore. Research
John Wiley. in Organizational Behavior, 29, 39–69. doi:10.1016/j.riob
Cramer, A. O. J., van Borkulo, C. D., Giltay, E. J., van der .2009.10.001
Maas, H. L. J., Kendler, K. S., Scheffer, M., & Borsboom, D. Machado, A., & Silva, F. J. (2007). Toward a richer view of the
(2016). Major depression as a complex dynamic system. PLOS scientific method: The role of conceptual analysis. American
ONE, 11(12), Article e0167490. doi:10.1371/journal.pone Psychologist, 62, 671–681. doi:10.1037/0003-066X.62.7.671
.0167490 Moors, A. (2016). Automaticity: Componential, causal, and
De Houwer, J. (2011). Why the cognitive approach in psy- mechanistic explanations. Annual Review of Psychology,
chology would profit from a functional approach and vice 67, 263–287.
versa. Perspectives on Psychological Science, 6, 202–209. Payne, B. K., Cheng, C. M., Govorun, O., & Stewart, B. D.
De Houwer, J., Gawronski, B., & Barnes-Holmes, D. (2013). (2005). An inkblot for attitudes: Affect misattribution as
A functional-cognitive framework for attitude research. implicit measurement. Journal of Personality and Social
European Review of Social Psychology, 24, 252–287. Psychology, 89, 277–293.
De Houwer, J., Hughes, S., & Barnes-Holmes, D. (2017). Robins, G. (2013). A tutorial on methods for the modeling and
Psychological engineering: A functional-cognitive per- analysis of social network data. Journal of Mathematical
spective on applied psychology. Journal of Applied Psychology, 57, 261–274.
Research in Memory and Cognition, 6, 1–13. Skinner, B. F. (1953). Science and human behavior. New
De Houwer, J., Teige-Mocigemba, S., Spruyt, A., & Moors, A. York, NY: MacMillan.
(2009). Implicit measures: A normative analysis and Sukhera, J., Milne, A., Teunissen, P. W., Lingard, L., & Watling, C.
review. Psychological Bulletin, 135, 347–368. (2018). The actual versus idealized self: Exploring res­
Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. ponses to feedback about implicit bias in health profes-
(1995). Variability in automatic activation as an unobtrusive sionals. Academic Medicine, 93, 623–629.