Transcript Picture - Liz Redford

Implicit Preference for White People over Black People
Decreases with Repeated Implicit Association Tests (IATs)
Emma Grisham, Dylan Musselman, Taylor Barnette, Melissa Powers, Gorana Gonzalez, John Conway, Rick Klein, Liz Redford
University of Florida
Discussion
Introduction
The purpose of this study is to investigate whether
familiarity with Implicit Attitude Tests (IATs) poses a
threat to the methodological validity of the measure and
reliability of the data. It is important to continuously
analyze the validity of a measure to recognize and
understand its limitations. We hypothesize that implicit
biases decrease with experience taking implicit attitude
measures. Since it’s invention, several concerns have
arisen regarding the methodological validity of the IAT
and practice effects, or effects on performance due to
repeated exposure to a task. Nosek, Banaji, and
Greenwald (2002) rationalized that individuals who
participant in multiple IATs have little to no significant
effect on observed results. Even so, research is geared
towards improving the IAT, specifically the inclusion of
practice trials. Research by Greenwald, Nosek, and
Banaji (2003) has shown that inclusion of practice trials
results in a better interpretation of implicit attitudes.
However, these same findings show evidence that with
practice, implicit biases decrease, both within
the session and with past IAT experience. A current
investigation was a 2011 study showing
that participants can be taught to practice strategies to
fake IAT scores (Rohner, Schroder-Abe, & Schutz). Are
practice effects a more serious issue?
Methods
We used data from a 2013 sample from the
demonstration website of Project Implicit, a website where
volunteers can choose from a variety of tests to measure
their unintended biases (implicit.harvard.edu). See Figure
1 and Figure 2. Participants in this study completed the
Race (Black faces vs. White faces) Implicit Association
Test (IAT). In addition to the standard 5 trial IAT,
participants were asked sets of explicit measures on racial
attitude, sets of personality and political opinion questions,
and demographic questions. Participants were mostly
female (61.5%) and White (69.7%). 12.5% were Black or
African American, and the remaining (17.8%) were Native
American/Alaskan, East or South Asian, Pacific Islander,
Multiracial, or other. Participants took the IAT and were
then given their IAT score.
Examples of the
Race Implicit
Association Test. The
IAT uses reaction
times to measure
implicit attitudes.
Figure 1. A picture of the race IAT, matching
“Black People” with “Bad”
Figure 2. A picture of the race IAT, matching
“White People” with “Good”
Results
Figure 3.
Results from the Race
Implicit Association
Test. A significant
difference was present
in all except for the last
two conditions (taking
the IAT 3-5 times or 6+
times previously).
We used an ANOVA to analyze mean differences in IAT scores between five groups based on the number of IATs
participants had previously taken: 0, 1, 2, 3-5, 6+. Levene’s test was significant, (4, 119022) = 3.34, p = 0.01, so we
report Welch’s F. The omnibus test of mean differences was significant, F (4, 6195.90)= 151.30, p < .0001. Tukey’s
post-hoc comparisons revealed that, as previous experience with IATs increased, IAT D score decreased. Each
increase in IAT experience was associated with a significant decrease in IAT score (all ps < .04; see Figure 3) except
between the two most experienced groups (p = .40). The greatest difference was observed between those with no
prior experience, M = .34, SD = .45, and those with the most, M = .21, SD = .44, d = .30, a medium effect size.
Our purpose was to test the hypothesis that
people who take the IAT more would have lower
mean IAT scores. Our hypothesis was supported,
and the overall effect size was very large. Practice
effects may explain this result. There is a chance
that people learn how to respond faster, thus
reducing their measured bias each time they take
the IAT. They then appear to have less bias than
they really do; in other words, they beat the system.
Alternatively, it could be that people are motivated to
become less prejudiced the more that they take the
IAT. This presents us with a third variable problem:
people who are motivated to reduce their prejudice
may be the people who are more likely to take the
IAT repeatedly to assess their progress. If the score
decrease is due to practice effects, as we are
proposing, then the IAT may be less valid after
multiple uses. Our main limitation is that we cannot
make causal conclusions because our data is
correlational. Future studies should test whether the
same results obtain with IATs measuring attitudes
toward social categories other than race, such as
gender or sexual orientation.
References
Greenwald, A. G, Nosek, B. A., & Banaji, M. R.
(2003). Understanding and using the Implicit
Association Test: I. An improved scoring algorithm.
Journal of Personality and Social Psychology, 85,
197-216.
Nosek, B. A., Banaji, M. R., & Greenwald, A. G.
(2002). Harvesting implicit group attitudes and
beliefs from a demonstration website. Group
Dynamics, 6, 101-115.
Rohner, J., Schroder-Abe, M., Schutz, A. (2011).
Exaggeration is harder than understatement, but
practice makes perfect!: Faking success in the IAT.
Experimental Psychology, 58, 464-472.