IMPLICIT ASSOCIATION TEST FOR RACE
The Implicit Association Test (IAT) measures human attitudes. Implicit attitudes are considered those judgments or actions of people that are automatic and the person has no idea what causes that judgment or action. The automatic action is a result of influences from societal beliefs. IAT is different from qualitative psychological tests for measuring attitudes because the measurement is of the cognitive processes of memory that cause a person to show a particular attitude automatically. IAT-Race was used to study racial stereotypes in an experiment to determine if the “Obama Effect” was observed. The participants were prompted to go to a web page titled “Implicit Association Test Race.” Several experimental trials were performed to measure their attitudes on Race and if the media coverage about President Obama influenced the reaction times. The hypothesis tested was that when a positive stereotype threat is felt by the participants the participants will perform better than if a negative stereotype or no stereotype was present. The hypothesis was rejected by the experiment’s results.
(experimental psychology, IAT-Race, Implicit Association Test, human attitudes, race stereotypes)
IMPLICIT ASSOCIATION TEST FOR RACE: ATTITUDE’S TO OBAMA
The Implicit Association Test (IAT) was developed in order to measure human attitudes more reliably than questionnaires of participants’ attitudes. Historically psychological testing has been based on qualitative techniques like questionnaires allowing only subjectively achieved values. Questionnaires and self-reporting done by participants in research studies contain the bias of each participant and the subjective opinion of the research interpreting the results. Surveys, questionnaires and self-reporting are examples of qualitative research. Qualitative research only reflects the quality of the perspective and opinion of the participants as well as the interpreter of the data. IAT is different from qualitative psychological tests for measuring attitudes because the measurement is of the cognitive processes of memory that cause a person to show a particular attitude automatically. The IAT is a quantitative research tool. Anthony G. Greenwald, Debbie E. McGhee and Jordan L. K. Schwartz are social psychologists from the University of Washington who have been involved with analyzing attitudes with IAT and comparing the results from their experiments to other researchers’ experiments. “Implicit attitudes are manifest as actions or judgments that are under the control of automatically activated evaluation, without the performer’s awareness of that causation” (Greenwald & Banaji, p. 6-8 as cited by Greenwald, McGhee & Schwartz p. 1464). The IAT has been used to measure many attitudes towards issues in society that are of interest to social psychologists.
An assumption is made for experiments using the IAT that the difference in attitudes between groups will become evident (Greenwald, McGhee, Schwartz, 1998). One example of two groups that are often compared is male and female in order to determine differences in attitudes based on gender. IAT-Race has been used to study racial stereotypes between groups of different ethnicities. David M. Marx from San Diego University, Sei Jin Ko from Northwestern University and Ray A. Friedman (2009) from Vanderbilt University reported the “Obama effect” from an experiment they carried out using a verbal exam methodology. The purpose of the test was to compare the test performance of two groups, Black-Americans and White-Americans. The researchers reported that Black-Americans received better grades on tests after President Obama’s Democratic convention speech and his election to the presidency. The researchers concluded that the performance effects that were measured in the experiment had “a concrete positive influence on Black-American’s academic performance” (Marx, Ko & Friedman, p. 953). The participants had been chosen randomly from a nation-wide sample and none of the participants had any personal interaction with President Obama (Marx, Ko & Friedman, p. 954). The methodology included comparing four separate groups at four different times (Marx, Ko & Friedman, p. 954). Two data collection events were when “Obama’s success was concrete and salient and two data collection events when his success was less salient” (Marx, Ko & Friedman, p. 954). When his success was salient the trait of his success was prominent in the national media, when his success was less-salient his success was not as conspicuous in the national media.
The experiment performed for this report was between Factor 1with two categories, congruent and incongruent, and Factor 2three categories of ethnic groups: African American, Caucasian and Other ethnic groups. The experiment was undertaken to measure the “Obama effect” using a similar methodology to the Marx, Ko and Friedman experiment (2009) described above. The IAT-Race test was designed t show that the ethnic group of the participants
determines the choices they make in test trials. The hypothesis tested was that when a positive stereotype threat is felt by the participants the participants will perform better than if a negative stereotype or no stereotype was present.
Dr. Pamela DeRosse’s students from her Research Methodology class were recruited to participate in order to meet the requirements of the course. In other words we participated in the experiment as a convenience sample because we were class members of Dr. DeRosse’s Experimental Psychology class. Seventy four students participated, 60 students were women and 14 students were male (See table1). The number of females was 81% higher than the number of males based on a total number of students in the class room equal to 74 with 46 more females than males.
The average age of the sample group was 29.19 years with a standard deviation (SD) of 9.8. The average age of the student participants was 29.19 ± 9.8. Therefore the ages ranged from about 19 to 39 years old. An ANOVA statistical analysis demonstrated that the ages between the groups and within the groups were not significantly significant, therefore age was not a confound.
The Chi-Square Tests statistical analysis determined that the proportion of females and males in the 3 racial groups were proportional; very little difference in the proportion of women to men in the three groups existed. The calculations resulted in X2(df=2) = 0.931 when p = 0.628 where X2 = the chi-squared test value, df equals the degrees of freedom and 0.628 is the Pearson Chi-Square; therefore gender was not a confound in the experiment.
The largest ethnic group was Other with 34 of the students self-identifying as other (Middle Eastern, Hispanic or Other). The next largest group was Caucasians with a total of 22 members and the smallest ethnic group was African American with a total of 18. Therefore African Americans made up only 24% of the total sample. The statistical analysis of the type of trial preformed did show within-subjects main effects; this could create a bias in the results due to the Trial Type. The race (ethnicity) of the participants when evaluated with the same statistical analysis did not show any Race x Trial Type interaction. But when a statistical test (Partial Eta Squared) for Race was carried out for between-subjects effects a main effect was calculated so Race could have been a confound of the experiment due to the large number of other and the small number of African Americans.
Each of the students had access to a standard desktop computer connected to the Internet. The participants entered the website provided by the American Psychological Association titled Online Psychology Experiment. After entering the website the student participants were prompted to go to a web page titled “Implicit Association Test Race.” They were asked to provide their Gender, Age, and/or Race. The Timing aspect of the experiment was described to them. Reaction time of the participants was measured. The participants were given to choices in order to describe their attitudes towards white and black Americans. In Figure 1 the same images that were on the web page can be seen. We were asked to correctly classify the race of a person in a photograph, on the left (Figure 1 in the image). Next were to classify words on the basis of 'good' or 'bad' using the image on the right (Figure 2 in the image).
Figure 1. Participants were asked to classify the race from the man’s face
The first choice asked them to tick the preference that most suited their own attitudes: I like both races equally, I strongly prefer Black (or white), or I moderately prefer Black (or White) (See fig. A-1). If the student made the second choice they would report their attitudes using a rating scale where 0 = coldest feelings, 5 = neutral feelings, and 10 = warmest feelings (See fig. A-2). After they had filled in the page of their choice they were shown a button to click in order to submit their data. After the data was submitted they were asked to print out the page proven they had completed the survey.
The computer was also used so the computers could categorize faces and words. For one task screen displayed the words “European-American” on the far right, a white-haired white man’s photo in the bottom-middle and the words “African-American” on the far right of the screen. The next screen only had three words but they were placed in the same configuration as the previous screen: Good at the far left, Joy in slightly larger type at the bottom-middle and Bad on the far left. A second task was to look at a screen with the words “African-American or good’ at the far left, a larger photo of a black man at the bottom-middle and the words “European-American or bad” on the far right (See fig. 2)
Figure 2. Participants were asked to classify the man’s face
The two tasks were to determine incongruent and congruent information. The stereotype congruent trial was designed to determine whether or not the participant has the socially accepted attitude of bad paired with Blacks and good paired with Whites. The stereotype non-congruent test trial took the opposite approach; European-Americans were matched with the word Bad and African-Americans were matched with the word Good. Each task carried out by the participant was timed. The reaction time was interpreted as follows (a) less time was needed to react with automatic implicitly learned social types (b) more time was needed to react with non-automatic reactions linked to implicitly learned social types .If the participant demonstrated a longer reaction time than it was assumed that the participant must have needed more time to sort out an attitude that conflicted with the generally accepted societal attitudes.
The design of the test was a 2x 3 mixed Factorial Design because two different IV trials were performed, one with two levels (congruent and incongruent) and the other with three levels of ethnicity (Caucasian, African-American and Other) The graph (See fig. 3) in the Results section was made using the values from Table 2 using the mean values for the three ethnic groups from the categories of congruent and Incongruent.
The IV1 trial included 2 levels (Factor = 1): incongruent and congruent . The IV1 trial reported data between subjects. The IV2 trial type included 3 three levels (Factor = 2): Caucasian, African-American and Other . The dependent variables (DV) were the recorded reaction times. DV was reported as the Type III Sum of Squares or Mean Square in trial IV1 that considered the relationship of the data between subjects (See table 3). The Type III Sum of Squares or Mean Square for trial IV2 are shown with the statistical results that describe the relationship of the data within the subjects (See table 4).
The stereotype threat on the participants’ task performance was assessed by defining three categories of ethnicity. The three categories participants were allowed to choose were Caucasian, African-American and Other. In other words the participants self-identified their own ethnicity rather than have a researcher assign an ethnicity. The congruent trial was designed to represent the socially acceptable stereotypes of blacks and whites. The incongruent trial was designed to represent a deviance from the socially acceptable attitude (Bad-African-American, Good-European-American. The faster the reaction time in making a selection during the experiment, the closer the attitude of the participant was to socially accepted stereotypes. The slower the reaction in making a selection during the experiment, the attitude of the participant was interpreted as not agreeing with the socially accepted stereotyping of whites and blacks.
The y-axis represents the dependent variables in the experiment (See fig. 2). Note that the Caucasian ethnicity the lines starts at the longest time and then decreases to faster mean reaction times graphed against congruent to incongruent. The timed results of the participants in the African-American and Other ethnicity categories did not vary in the time take to react as much as the Caucasion participants.
Figure 2 used the descriptive statistic data to graph three lines. The x-axis of the graph represents the congruent-incongruent range. The y-axis represents the mean of the reaction times for each of the three ethnic categories. The reaction time means were calculated from the trials, averaged and are recognized as the dependent variables of the experiment.
Figure 3 The range of congruent-incongruent (x-axis) versus the mean reaction time (y-axis)
The experiment rejected the hypothesis that when a positive stereotype threat is felt by the participants the participants will perform faster than if a negative stereotype or no stereotype was present. Statistical analysis demonstrated that the confound was race of the participants and was evident for within-subjects and between-subjects.
There were several limitations to the experiment. The sample was not controlled random sample so the numbers of participants in the three ethnic categories were not well distributed. The poor distribution of the participants race resulted in a confound between and within the subjects. The number of females was much larger than that of males. In society males and females are each about fifty percent of the population. The convenience sample was a limitation because the sample was representative of that Experimental Psychology class but not a good representation of the distribution of ethnicities and gender in society.
IAT is a test that was developed partially for the purpose of finding a measurement that shows validity. In other words, the IAT is has been developed as a tool for measurement of attitudes for the purpose of limiting personal bias in the data and results. A limitation is that often activities that are observed do not match what people self-report (Azjen, 2005).
Conclusions and Future Study
The hypothesis tested was that when a positive stereotype threat is felt by the participants the participants will perform better than if a negative stereotype or no stereotype was present. The experimental results proved the hypothesis. Statistical analysis demonstrated that the confound was the race distribution of the participants and was evident for within-subjects and between-subjects. The “Obama-effect” as reported in the Marx, Mo and Friedman (2009) experiment was observed in the experiment.
More experiments need to be carried out with the researcher controlling the number of members with different ethnicities so that the proportion of whites to blacks in the sample mirrors the same proportion of whites to blacks in the population. An experiment could also include samples with equal numbers of blacks and whites and larger numbers of blacks to whites in order to better understand the relationship of race to stereotype threat.
Ajzen, I. (2005). Attitudes, Personality, and Behavior. Maidenhead, England: Open University Press.
Angeli, E., Wagner, J., Lawrick, E., Moore, K., Anderson, M., Soderlund, L., & Brizee, A. (2010, May 5). General format. Retrieved from http://owl.english.purdue.edu/owl/resource/560/01/
APA (American Psychological Association). (nd). Online Psychology Experiment, APA, available from http://opl.apa.org/Experiments/Start.aspx?EID=5
Aronson, J., Jannone, S., McGlone, M. & Johnson-Campbell, T. (2009). The Obama effect: An experimental test. Journal of Experimental Social Psychology, 45, 957-960.
Greenwald, A. G., McGhee, D. E., & Schwartz, L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 76(6), 1464-1480.
Marx, D. M., Ko, S. J. & Friedman, R. A. (2009). The “Obama Effect”: How a salient role model reduces race-based performance differences. Journal of Experimental Social Psychology, 45, 953-956.
Figure A- 1Online IAT-Race web page for participant self-description (APA, nd)
Figure A- 2 Online IAT-Race web page for second choice for rating groups (APA, nd)