Introduction
As human beings, our minds are continuously active in gathering information from our social environments; creating and reinforcing associations and attitudes about the stimuli that we encounter. Our minds are so prolific at forming such associations that they may do so unconsciously, and even if they oppose our own fundamental value systems (Smith & Nosek, 2010). Greenwald and Banaji (1995) attributed the term ‘implicit social cognition’ to the cognitive processes that occur outside of our conscious awareness or control in relation to social psychological constructs such as outlooks, stereotypes and self-concepts. The term ‘implicit’ is in this context applied to a variety of cognitive processes that are not actively controlled or self-reported by individuals. An implicit measure can therefore assess associations without an individual even requiring an awareness of the relation between the response and the measured content. (Nosek & Greenwald, 2009)
As Smith and Nosek (2010) defined it, the Implicit Association Test (IAT) is an experimental sorting task designed to measure the strength of the implicit associations that we accumulate daily through interactions with our environments. The IAT intends to capture the strength of such associations irrespective of whether or not individuals believe that they actively possess them, and whether or not they are believed to be held as valid or true. The IAT is predicated on the assumption that it is easier to make the same response to two things and thus associate them with one another when they are related than when they are unrelated. The IAT offers a window into capturing both positive and negative associations about a multitude of concepts without the need for introspective consultation; subsequently making the test essential for clarifying the mechanisms underlying our social behaviour.
Although self-reported attitudes are important predictors of behaviour, the IAT appears to be a better predictor for topics that are socially sensitive such as racial and cultural prejudices. A meta-analysis by Greenwald et al (2009) of 184 predictive validity studies using the IAT found positive predictive validity across all 51 evaluated domains. IAT results were shown to predict variation in behaviour that was not accounted for by explicit measures. According to Fazio (1990) and Nosek (2005) people can have an implicit reaction toward a topic but choose not to report it because they do not agree with their own reaction, or do agree, but refrain from expressing it due to it’s controversial nature. For example, implicit preferences for White people compared to Black people predicted voting for John McCain versus Barack Obama in the 2008 US presidential election (Greenwald et al, 1998; Greenwald et al, 2009; Payne et al, 2010). Clearly, IAT tests are integral measures in assessing the cognitive processes that determine people’s attitudes and subsequent actions.
Over the past few decades it has become the general consensus that people are increasingly egalitarian in their attitudes towards race, however IAT analyses reveal that individuals still display systematic ethnocentric differences in their implicit attitudes. Payne B.K et al (2005) found this through flashing prime stimuli such as Black and White faces briefly to participants before presenting them with less familiar photographs of Chinese faces. Individuals were told to ignore the prime images, and instead instructed to rate whether the unfamiliar pictographs were more or less aesthetically pleasant than the average photograph. The results from this study indicated that the presence of the primes still initiated a racial preference in individuals towards images of their own race. The methodology used however is slightly problematic as the dependent variable measured is a decisive explicit rating from participants. Albeit, it can be argued that the rating is secondary because the assessment relies on whether or not implicit racial associations indirectly influence these explicit ratings. Implicit measures typically assess and predict aspects of human cognition that are not readily revealed by means of self-report (Nosek, Hawkins & Frazier, 2011). Empirical evidence for the factors influencing the predictive validity of implicit measures are clearly still in their infantile phases, meaning further research is necessary in this field in order to draw any valid definitive conclusions. (Nosek et al 2011)
As with many psychological studies, replications of IAT experiments do not always give results that indicate an effect. (Maxwell, Lau & Howard, 2015). A replication of Ranganath and Nosek’s (2008) experiment by Cohn (2014) aimed to replicate the original findings. However, no immediate generalisation of attitudes towards stimuli in participants was found, suggesting lack of reliability for the racial IAT findings in modern society. It is evident that there is a gap in literature in this domain, suggesting the need for validation that the IAT can still measure implicit racism in 2017 irrespective of extraneous influences that may interfere with its effectiveness. Smith and Nosek (2010) likened the IAT to measures of blood pressure “by showing some consistency over time but fluctuations from moment to moment.”
the present state of knowledge provides a foundation for the next age of implicit social cognition: clarification of the mechanisms underlying implicit measurement and how the measured constructs influence behavior. There is still discrepancy as to whether or not implicit associations that differ to an individuals pressed beliefs can be attributed to them, or the culture that they are integrated in. (Smith and Nosek, 2010).
Therefore the purpose of this experiment is to establish IAT’s continuing relevancy in an increasingly global and diverse population, and to investigate the prevalence of black/white racial stereotypes in the context of a multiracial environment.
Our hypothesis was that participants, when asked to classify a random sequence of black faces and bad words and white faces with good words (a congruent trial) would have a lower mean reaction time compared to when asked to categorize black faces with good words and white faces with bad words (an incongruent trial). Such a result would therefore reflect the presence of implicit attitudes in overseas cultures.
Method
Participants
The sample of participants consisted of 171 students from University College London; composing of 143 Women, 27 Men and 1 N/A. The individuals that took part in the research ranged in age from 17-36, with a mean age of 18.97. All participants were recruited using an opportunity sample, and were not reimbursed for their time as their involvement was elementary to either/or the PSYCH1107 and PSYCH1103 modules that they were derived from. Participants took part in both trials of the experiment and were assigned randomly to counter-balanced conditions.
The UCL Department of Psychology Ethics Committee granted full ethical approval prior to the conducting of the experiment. Consent was obtained from each individual prior to their participation, and they were fully debriefed upon completion.
Design
We intended to measure participants ability to classify two sets of stimuli with the same two buttons corresponding to options in the top right or left-hand corners of the screen. The independent variable (IV) that we manipulated was whether the stimuli presented to the participants were ‘congruent’ or ‘incongruent’. A congruent trial would require a participant to classify a random sequence of White faces and good words with one key, and a different sequence of Black faces and bad words with the other designated key. The incongruent trial alternatively required people to categorise and classify the sequence of White faces and bad words with one key, and the sequence of Black faces and good words another. The participants mean reaction times (meanRT) for both the congruent and incongruent trials were calculated as the dependent variable (DV).
We carried out the study using a repeated measures, within subjects design. Counterbalancing was achieved through randomisation of the order in which each participant was exposed to the congruent and incongruent trials. We decided to do so to eliminate any potential order effects associated with a repeated measures design, ensuring that the data only signified participants implicit associations.
Materials
We employed the implicit attitudes test (IAT) in order to measure participants ability to classify two sets of stimuli with the same two buttons. We then evaluated individual mean reaction times to the stimuli introduced in the experiment.
The stimuli presented to each participant consisted of 12 images; 6 of which were pictures of White faces, and 6 of which were pictures of Black faces. The stimuli were equally proportioned in terms of male and female faces in order to prevent gender confounding implicit associations. Each test condition differed in terms of stimuli presentation according to the block (Refer to appendix 1). The facial stimuli were then to be categorised with one of 8 either positively valanced or 8 negatively valanced central words per trial. These written stimuli were necessary in order to fully assess the extent to which the hypothesised implicit associations were prevalent amongst participants (Refer to Appendix 2.)
The experiment was conducted and data was recorded using a software package called Gorilla.sc (www.gorilla.sc/about) on desktop computers in the lab at UCL. Gorilla was developed specifically for the use of behavioural research.
Procedure
In an attempt to account for any potential biases within the sample, the Gorilla software was used to randomly allocate the participants into one of the two conditions; ‘congruent first’ or ‘incongruent first’. Counterbalancing and a true experiment were thus achieved, reducing the influence of order effects on subsequent test performance.
During the experimental brief, participants were instructed to categorise a central stimulus (either a face or a word) by pressing either the E or I key corresponding to options in the top right or left corners of the screen. These options were either standardised words or Black and White faces (Refer to appendix 3). They were told that their task was to classify a set of words or images as quickly as possible with the intent of making as few mistakes as possible. Participants were given 5 sets of sorting exercises to complete. After the correct classification was made, their reaction time for the trial was recorded and the next stimulus was presented. The type of central stimuli, and the options given varied according to the block presented (Refer to appendix 1). After completion of the first condition trials, participants were made aware that the positions of the two concepts would be switched.
Participants were first presented with a 12-trial image practice block irrespective of their designated condition; followed by a word practice block consisting of 16 trials. Each set of either congruent or incongruent trial blocks consisted of 28 trials, given to participants in a random order determined by the software. Participants in the ‘congruent first’ condition were given a congruent mapping word practice block and congruent trial block first; participants in the ‘incongruent first’ were presented with the opposite (Refer to appendix 1). Mean response times for each trial were recorded.
In between each trial, participants saw a fixation cross for 500ms. Between each block, a screen alerting them of the upcoming set of stimuli and instructions as to how to complete the next task were presented until the spacebar was pressed, allowing them to proceed.
Results
In order to eliminate potential outlying results, reaction times longer than 5 seconds were discarded and assumed to have been a result of participant inattention or computer faults.
Descriptive Statistics –
For each participant, their average reaction times for both the congruent and incongruent trials were calculated; giving us a ‘meanRT’ value. The variance was calculated by computing the SD and the SE Mean for congruent and incongruent trials (See Table 1). An IAT score for each participant was also calculated by subtracting each participants Congruent condition score from their Incongruent condition score.
Table 1.
Congruent vs Incongruent trials
Variable No. Mean reaction time (ms)
Standard Dev. SE Mean
Congruent
171 791.2503 172.1652 13.172
Incongruent
171 887.3253 197.9523 15.138
Inferential statistics –
when μd = μcongruent – μincongruent
Null Hypothesis H0: μd = 0
Alternative Hypothesis HA: μd ≠ 0
Mean reaction times for congruent and incongruent trials were calculated and analysed using a paired-samples t-test. This test allowed us to measure the differences between participant scores on both trials; and was thus necessary to inferentially analyse our repeated measures design experiment. The results indicated that participants mean reaction times were significantly faster statistically when categorising the congruent, in comparison to the incongruent stimuli t(170) = -7.4897, p < 0.001 (2 tailed). The probability value shown was within the experimental rejection region indicating that the results were significant as p < 0.05.
These findings were similarly reflected in the histogram representing the data (See Figure 1). Despite the distribution of scores for both conditions being positively skewed, there was both significantly smaller variance in mean reaction times with a 95% confidence interval for true mean difference (-121.5, -70.8) (See Table 1) and a smaller distribution of scores for the congruent trials. 125 of 171 participants (73%) were faster at categorising on the congruent trials. The graphical depiction of the data reinforces the notion that participants were consistently much faster at sorting congruent than they were at sorting incongruent stimuli.
DISCUSSION
The first task performed interferes with performance on the second. Likewise, people who respond more quickly on average tend to show smaller IAT effects than people who respond more slowly. Identification of extraneous influences such as these provides an opportunity to reduce or remove their influence. The “order effect” and “average response time” influences have been reduced with procedural and analytic innovations, respectively. The coming years of research will continue to refine the methodological features of the IAT for association measurement. (COKIN TUCKER AND NOSEK)
one argument is that the significant difference between the trials is evidence of implicit prejudice. Or in other words, what makes it ‘incongruent’ to pair good words and Black faces is the negative associations that participants already have towards Black faces. How this relates to discriminatory behaviour, explicit attitudes that people say they hold, and whether we can say it is evidence of personal ‘prejudice’: that is all a matter of debate and further evidence.
Drawbacks of a Within-Subject Design
This type of experimental design can be advantageous in some cases, but there are some potential drawbacks to consider. A major drawback of using a within-subject design is that the sheer act of having participants take part in one condition can impact the performance or behaviour on all other conditions, a problem known as a carryover effect.
Finally, performance on subsequent tests can also be affected by practice effects. Taking part in different levels of the treatment or taking the measurement tests several times might help the participants become more skilled, meaning, they may be able to figure out how to game the results in order to do better on the experiment. This can skew the results and make it difficult to determine if any effect is due to the different levels of the treatment or simply a result of practice.
(read online so only paraphrase)
Appendix:
Appendix 1 – Order of stimuli presented in blocks to participants during each experimental test condition.
A. Congruent Condition
(1)
Image Practice Block
12 trials
(2) Word Practice Block
(Congruent mapping)
.16 trials (3) Congruent Trial block
28 trials (4) Word Practice Block II (Incongruent mapping)
16 trials (5) Incongruent Trial block
28 trials
B. Incongruent Condition
(1)
Image Practice Block
12 trials
(2) Word Practice Block
(Incongruent mapping)
16 trials (3) Incongruent Trial block
28 trials (4) Word Practice Block II (Congruent mapping)
16 trials (5) Congruent Trial block
28 trials
1) Image Practice Block
– Black face on Left, White Face on right
– Image to be categorised in the center.
2) Word Practice Block – congruent mapping
– “Bad” Text category on left, “Good” Category on right
– Word to be categorised in the center
3) Congruent Trial Block
– “Bad” and Black face on left, “Good” and white face on right.
4) Word Practice Block II – Incongruent mapping
– “Good” Category on left, “Bad” Category on right.
– Word to be categorised in the center
5) Incongruent Trial block
– “Good” and Black face on left, “Bad” and white face on right
Appendix 2 – Words used (written stimuli):
Positively valanced – (Joy, Love, Peace, Wonderful, Pleasure, Glorious, Laughter, Happy) Negatively valanced – (Agony, Terrible, Horrible, Nasty, Evil, Awful, Failure, Hurt)
Appendix 3 – An example of facial stimuli displayed to participants during the image practice block.