The Stability of Implicit Racial Bias in Police Officers

• Published date: 01 March 2018
• DOI: 10.1177/1098611117732974
• Date: 01 March 2018
• Subject Matter: Articles

Article
Police Quarterly
2018, Vol. 21(1) 30–52
The Stability of Implicit
! The Author(s) 2017
Reprints and permissions:
Racial Bias in Police
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1098611117732974
Officers
journals.sagepub.com/home/pqx
Lois James1
Abstract
Research on police officers has found that they tend to associate African Americans
with threat. Little is known however about the stability of implicit racial bias in police
officers, whose attitudes could be expected to fluctuate based on their day-to-day
encounters or from internal stressors such as fatigue. To investigate, this study tested
80 police officers using the Weapons Implicit Association Test (IAT) on four separate
occasions. Officers’ sleep was also monitored using wrist actigraphy. Officers’ IAT
scores varied significantly across the testing days (f ¼ 2.36; df ¼ 1.468; p differences in IAT scores were associated with officers’ sleep (f ¼ 6.49; df ¼ 1.468;
p and that when officers slept less prior to testing they demonstrated stronger asso-
ciation between Black Americans and weapons. The implications of these findings
within the current climate of police–citizen unrest are discussed.
Keywords
implicit bias, fatigue, racial tensions, police legitimacy
The policing profession has undergone what many have termed a ‘‘crisis of
legitimacy’’ over the past 3 years (L. James, Fridell, & Straub, 2016; Nix,
Campbell, Byers, & Alpert, 2017; Paoline, Gau, & Terrill, 2016; Todak, 2017).
Although racial tensions have consistently permeated the history of American
policing, the 2014 shooting of Michael Brown in Ferguson brought allegations
1Sleep and Performance Research Center, Washington State University College of Nursing, Spokane, WA,
USA
Corresponding Author:
Lois James, Sleep and Performance Research Center, Washington State University College of Nursing,
P.O. Box 1495, Room 422C, Spokane, WA 99210-1495, USA.
Email: lois_james@wsu.edu

---

James
31
of racial bias to the forefront of the national conversation about police legitim-
acy. At about the same time, the law enforcement community began to pay more
attention to implicit bias, and in particular the notion that of‌f‌icers may not
display explicit stereotypes but can still be inf‌luenced by biases in their behavior,
judgments, and decisions (Dasgupta, 2013; L. James, Klinger, & Vila, 2014).
In response to broad concerns about racially motivated policing, and amidst
growing evidence that police disproportionately target African Americans
across a range of police activities (Fridell & Lim, 2016; Fryer, 2016), implicit
bias training is becoming a staple among many police departments.
There is debate, however, surrounding implicit attitudes, beliefs, and stereo-
types (Blair, 2002; Dasgupta, 2009; Devine, 1989; Gawronski & Bodenhausen,
2006; Wilson, Lindsey, & Schooler, 2000). Specif‌ically, are these cognitive
processes stable (like traits) or are they variable and malleable (like states)?
The answer to this question has important implications for police training.
First, if implicit bias is stable and impervious to change (Wilson et al.,
2000), then we must ask whether ef‌forts to educate of‌f‌icers on the science of
bias will actually cause them to modify their behavior. Alternatively, if implicit
bias is malleable then can repeated exposure to counter stereotypes reduce
bias? (Plant & Peruche, 2005). Relatedly, if implicit bias is inf‌luenced by
of‌f‌icer-level variables such as stress and fatigue (Ma et al., 2013), would train-
ing ef‌forts targeted at reducing these stressors be more ef‌fective than implicit
bias training? To date, however, the stability of implicit bias in police of‌f‌icers
has been largely ignored in the empirical literature. As such, the current
study tackles this issue by measuring implicit racial bias in a sample of
police of‌f‌icers, exploring the variability of bias by testing the same of‌f‌icers
across four time points, assessing whether variation in of‌f‌icer sleep produces
changes in implicit bias, and discussing the implications of the f‌indings for
police training.
Measurement of Implicit Bias
Implicit bias is def‌ined as subtle and largely unconscious or semiconscious atti-
tudes that inf‌luence behavior (Dasgupta, 2013). It is very dif‌ferent from explicit
bias, which is made up of attitudes, beliefs, and stereotypes that an individual
claims as part of him or herself. Whereas explicit bias is typically measured by
asking someone to identify their beliefs, implicit bias is measured in a more
nuanced way. The Implicit Association Test (IAT) is one of the primary ways
of measuring implicit bias, which it does by analyzing how strongly a person
mentally associates two separate concepts, such as race and weapons (Project
Implicit, 2011). Participants are presented with stimulus words or images

---

32
Police Quarterly 21(1)
and asked to categorize the stimulus by selecting one of two computer keys. The
dif‌ference in time to respond between strongly associated items and less strongly
associated items is called the IAT ef‌fect (with a larger time dif‌ferential signifying
a larger ef‌fect). For example, respondents are typically faster to respond when a
f‌lower is paired with a desirable attribute than when a bug is paired with a
desirable attribute (Greenwald, McGhee, & Schwartz, 1998), presumably
because it takes longer to complete the task when the pairs are cognitively less
compatible.
The underlying assumption of the IAT test is that all individuals have cog-
nitive processes that are outside of their conscious awareness (Greenwald &
Banaji, 1995). As such, implicit measures of attitudes improve upon explicit
measures in several important ways. First, explicit measures are unable to cap-
ture cognitive processes that are outside of our awareness (Greenwald & Banaji,
1995), whereas the IAT test is designed to reveal these unconscious attitudes
through automatic response. Second, by measuring automatic responses, impli-
cit measures of prejudice are considered impervious to ef‌forts to artif‌icially
conceal attitudes (Greenwald et al., 1998; Greenwald, Poehlman, Uhlmann, &
Banaji, 2009; Kim, 2003), while explicit measures allow the test taker time to
consider the social repercussions before selecting their response. Research has
found associations between implicit measures, explicit measures, and externa-
lized behaviors toward members of social groups (Hofmann, Gawronski,
Gschwendner, Le, & Schmitt, 2005; McConnell & Leibold, 2001). However, a
meta-analysis of 122 studies found the predictive validity of IAT score on beha-
vior was stronger and less variable, especially for socially sensitive topics, com-
pared with explicit measures (Greenwald et al., 2009).
Researchers have largely used the IAT to measure how strongly people associ-
ate attributes with certain social groups (e.g., African Americans, homosexuals,
or smokers). For example, Nosek, Banaji, and Greenwald (2002) found partici-
pants more strongly associated men with science and women with liberal arts.
There has been considerable debate surrounding the interpretation of these
f‌indings—namely, while the creators of the IAT have argued that associations
indicate an af‌fective preference or valence (i.e., positive or negative evaluations
of certain social groups compared with others; Dasgupta, Greenwald, & Banaji,
2003), others counter that the associations ref‌lected in the IAT merely indicate
familiarity or exposure to certain stereotypes (Kinoshita & Peek-O’Leary, 2005;
Rothermund, Wentura, & De Houwer, 2005). In other words, dif‌ferences ref‌lect
passively acquired attitudes, not personal feelings. For example, a police of‌f‌icer
may personally believe that suf‌fering from a mental illness does not indicate
weakness on the part of the of‌f‌icer but may nevertheless implicitly associate
mental illness and weakness because of the widespread stigma against mental
illness endemic in police culture (Baker & Baker, 1996; Church & Robertson,
1999; He, Zhao, & Archbold, 2002).

---

James
33
The Stability of Implicit Bias
Decades of research on implicit attitudes and beliefs promoted the idea that
implicit bias is a stable phenomenon (Devine, 1989; Wilson et al., 2000). From
this line of research came the idea that although explicit biases could be changed,
implicit biases were ‘‘trait like,’’ and at best, one could hope to understand and
ignore them (Bargh, 1999). A competing line of research however has emerged
which claims the opposite—that implicit attitudes are in fact quite malleable and
can be readily changed by altering one’s environment (Blair, 2002; Dasgupta,
2009; Gawronski & Bodenhausen, 2006). For example, if a police of‌f‌icer who is
implicitly biased against Latinos moves from an impoverished and high crime
Latino neighborhood to an af‌f‌luent and low crime Latino neighbourhood, they
are likely to experience changes in their attitudes about Latinos. This could
consequently alter their beliefs, stereotypes, and even behaviors toward
Latinos (Dasgupta, 2013). This is the concept of counter stereotyping, or expos-
ing individuals to information that goes against their established attitudes in an
attempt to alter them, and is a foundation of law enforcement implicit bias
training (L. James et al., 2016).
Of course, one would also expect the reverse would be true, a point that is
especially salient for police of‌f‌icers. If an of‌f‌icer is consistently exposed to infor-
mation that conf‌irms their attitudes, then those attitudes (and subsequent inf‌lu-
ences on behavior and decisions) will likely get stronger. For...