Reflection
First of all I would like to say that I'm not completely sure what to say here, mainly because there was little to no information given about it. So I will do my best and write something about the process I used to combine my own insights with the comments from the supervisor. As a main guideline I always mailed my supervisor when I came across an issue I needed an immediate answer to in order to continue my writing. Besides that there were also co-students who could answers less important questions, for example about details that had been forgotten regarding the precise setup we used. This communication was always good (both between the group and the supervisor) and I think in the end we all had benefit from it (the group members at least). Furthermore I generally 'fixed' every point made in the evaluations of the semi papers, unless I
specifically disagreed or thought otherwise. This was very rare to be fair (just 2 of all the points that were made), but the points are stated later on for reference. All in all I think I did a good job in integrating external critique and revising my own work by re-reading it a couple of times.
Regarding the process a whole, I feel like there were no real hurdles along the way. The group worked very nicely together and the communication with our
supervisor was always fast and accurate. On some single occasions though I feel like we could have shown more anticipation on getting things done early. Some things were really done on the last moment (like sending a letter to the Commissie Ethiek) and I think we could have been much better off if those things got dealt with earlier. That would have made sure our research would have been online earlier and we could have had more participants enroll early on. Besides that I would have liked to program a bit more, but due to a tight schedule and a lot of other things to do for other aspects of the study, it did not happen. Nevertheless I did fix a few minor bugs while the experiment was live and that gave me insight in the code and code language, a thing that might be valuable some day.
Regarding the strengths and weaknesses of my paper, as a whole I am quite happy with the results. It was the first time I wrote a serious paper in English, so naturally that was a big step up. All in all I am happy I did it though, because not only could I stay more true to the terminology and syntax used in real articles, sometimes it felt like the point came across better than when I translated it to myself in Dutch. On an additional note I think the paper looks more 'professional' being in English and with the current trend of the world progressively making English the main language for almost everything, I am somewhat surprised the University doesn't really promote it that much. In the beginning the language used in this paper might have been a bit popular , that was mainly because I had to find my way in writing scientifically in English because we never practiced it before in class.
Looking at the content discussed in this paper, I think I did a good job in
managing different (3) hypotheses and integrating the IAT (which was a subresearch on its own). I would have liked to get deeper into the IAT results, but considering my paper was already far above the actual word limit I decided not to. There was no way for me to significantly reduce the amount of words used in this paper without really letting
important information go, so I hope I at least tell a convincing story. Especially since I examined three different hypotheses, integrated the IAT, and did my best to fully analyze the data to its maximum potential, the word limit was just too low for me.
Even though I used 27 different citations in my paper, I feel like on some points I could have cited a bit more to integrate the paper as a whole. I did my best however to
not let too many sentences without citation, although I have a habit of thinking and reasoning for myself without proper look-up first.
The ethical aspects of the research did not really entail too many problems and I think we dealt with them correctly. The main issue we had of course was that we were dealing with a topic of racism and because of it we had to elaborate the specific purpose of this research well in a proper debriefing. That did not lead to any problems though, so all in all I'd say we dealt with ethics pretty good.
As a final word I would like to thank my group and supervisor (Marte Otten) for the entire process and Marte in particular for her fast responses and general
helpfulness. Below are a few (but not even close to all) of the most important changes that were made as a result of the earlier evaluations.
● As a result of early revisions I completely overhauled my hypotheses. This was a procedure I learned a lot from specifically.
● Changed the order in which certain topics were discussed in the Introduction. Mainly the IAT and literature about implicit activation of concepts was dragged to the beginning of the
Introduction and was elaborated extensively upon compared to the first draw.
● Completely removed literature about the somatic marker hypothesis as in retrospect it did not at all contribute to the paper.
● Added a large section on the IAT in the methods section, split between stimuli and procedures. ● 'Faced' has been removed from the paper, as it was corrected in the abstract. I assumed it was incorrect in the whole paper. I used the word 'faced' to emphasize the primes are faces, but I acknowledge that it sounds kind of odd. Therefore I removed it entirely from the paper. ● Data reduction section was merged with the participants section.
● Did not change the Δ to further specify that this meant Δ reaction time, because the sentence it is used in states that the data used to analyze is in fact reaction times.
"Contrasts showed that the main effect of primes originated from a slower mean
response time for black faces when compared to Moroccan faces (Δ = 7.42 , p = .021), white
faces (Δ = 12.43 , p = .012) and angry white faces (Δ = 8.95 , p = .004)."
● Did specify the trend I found as being a trend instead of just describing it as such.
● Added a part summarizing the results and speculating about future research in the discussion. ● Made the last discussion point more 'open' instead of stated, as if it could be mentioned by someone else. I kept it in because it is an example of something others could mention about the research as a flaw, but in my discussion I counter it by immediately stating why it is not. ● Elaborated on the use of angry white faces in the discussion. That is, while besides some possible confounding effects it was still the best choice to test the hypothesis.
The Effect of Different Prime Types on Gun/Tool Decisions
By Jordy Muller
Abstract
The current research examines the effect of four different prime types (black, white, Moroccan & angry white faces) on identifying a briefly shown object as either a gun or a tool. Participants (N=44) were given 512 trials consisting out of a prime, followed by a target (gun or tool), followed by a mask. Both reaction times and errors were recorded and analyzed. Analysis of reaction times showed that for black primes, a significant difference between gun and tool reaction times was evident. Further analysis showed that this effect originated from a relatively slower response to tool stimuli compared to gun stimuli. No significant difference in reaction times between gun and tool stimuli was found for either white, Moroccan or angry white primes. Analysis of error rates showed a significant interaction effect between primes and targets. Post hoc analyses showed a higher rate of error for tool responses on black prime trials compared to the error rates found on all of the other prime type trials. Concluded is that of all of the examined prime types, only black faces produced significant effects on gun/tool decisions.
Jordy Muller 10334955
Introduction
In December 2014, an unarmed black man named Rumain Brisbon got shot by a white police officer in Phoenix, who used lethal force upon perceiving an object in the victim’s hands which he allegedly believed to have been a weapon. The item in the victim’s hands, however, turned out to be nothing more than an ordinary pill bottle. The incident caused a statewide uproar and spawned many protests across the United States. It seems every now and then a story comes out in which a white cop shoots and kills a black man, who in hindsight turns out to have been completely unarmed. Usually these cases end up in media-driven trials, carefully followed on the sidelines by activists of anti-racism groups.
From a psychological standpoint however, these cases are interesting not because of their moral or legal implications, but because of their dramatic display of psychological processes that guide our judgments. How is it possible for so many cops to misidentify harmless objects like pill bottles as firearms, and make such drastic
decisions based on those misperceptions? One proposed explanation of this phenomenon arises from theory regarding dissociation between automatic and
controlled processes (Payne, 2001). Normally, automatic processes take place in a fast and rigid manner to alleviate the workload that is being placed on conscious control. These automatic processes however are not flawless and make extensive use of heuristics to guide their reasoning. Therefore, automatic processes are especially vulnerable to biases.
Racial prejudice, or racial bias, seems to be subject to automatic attitude activation (Fazio, Sanbonmatsu, Powell, & Kardes, 1986). This finding was very
important for scientists interested in research of prejudice and stereotypes because it gave them the opportunity to unobtrusively measure racial attitudes using paradigms based on facilitation of attitude congruent versus attitude incongruent primes and stimuli. The unconscious aspect of stereotypes and prejudice has been shied by many scientists for a long time (Banaji & Greenwald, 1994), even though the methodological limitations of introspection and explicit measures had been well documented (Nisbett & Wilson, 1977). When measuring social attitudes, the preferred method should always be an indirect measure, as first stated by Campbell (1950). By using automatically
generated attitudes without explicit knowledge of this activation scientist can measure racial prejudice or bias in an indirect way without having to deal with added drawbacks of introspection and participants knowingly giving socially acceptable answers
(Greenwald & Banaji, 1995).
Probably the best known procedure to measure implicit attitudes using a paradigm based on automatic attitude activation is the implicit associations test (IAT; Greenwald, McGhee, & Schwartz, 1998). In order to unveil implicit attitudes, the IAT uses four categories of words (or images), for example black, white, pleasant and unpleasant. The IAT builds on the assumption that if two concepts have a strong association with each other, that pair is processed with greater ease than when those concepts are more loosely associated with each other (e.g., Bargh, Chaiken, Govender, & Pratto, 1992; Fazio et al., 1986; Perdue, Dovidio, Gurtman, & Tyler, 1990). Using the categories stated before, two possible pairs may be formed, for example black - pleasant and white - pleasant. If one has a stronger association with black people and pleasant things than with white people and pleasant things, matching black names or faces with pleasant words is performed with greater ease (i.e., faster and typically with less errors) than matching white names or faces with pleasant words. If the reverse association is
present, meaning one has a stronger association with white people and pleasant things than black people with pleasant things, the reverse results will be found (i.e., processing white - pleasant pairs is performed with greater ease than black - pleasant pairs). By calculating the difference in performance between attitude congruent and attitude incongruent trials, a measure of difficulty difference provides an indication of one's implicit attitudinal difference between the two target categories.
Implicit racial bias might be a key factor in situations wherein people quickly have to decide whether an item in someone’s hand is a weapon or something harmless. Negative stereotyping of black people as being violent leads participants to identify an object faster as a gun when they are being primed with black faces than when they are being primed with white faces (Payne, 2001). In subsequent studies, other researchers have found that participants are more likely to virtually shoot a person (by pressing a button) that is holding a harmless object when the face of that person is black than when the face of that person is white (Correll, Park, Judd, & Wittenbrink, 2002). Inspired by those studies, Greenwald, Oakes and Hoffman (2003) showed in an experiment using virtual reality that in addition to just a racial bias, there might even be an effect of race on perceptual sensitivity. The participants in this study did not only experience a higher difficulty in distinguishing guns from harmless objects in the hands of black criminals, but also in the hands of black co-worker police officers. Furthermore, a study conducted by Eberhardt, Goff, Purdie and Davies (2004) showed that activating stereotypic
associations caused participants to detect crime relevant stimuli at a lower perceptual threshold than crime irrelevant stimuli. The study used black and white faces as primes and degraded crime relevant (weapons) and crime irrelevant (common objects) stimuli as targets. The results of the study showed that when participants were primed with black faces, the threshold of identifying crime relevant study was much lower than when they were primed with white faces. All of the studies discussed above suggest that there is at least a marginal effect present of race on implicitly activating race-specific negative (often crime-related) associated concepts.
The current research mainly follows the gun-tool decision paradigm as
introduced by Payne (2001). The gun/tool decision task he utilized in his experiment used black and white faces as primes and a set of handguns and tools as targets. In each trial of the task a face would be shown, followed by either a gun or a tool. The
participants then have to classify the target as quickly as possible as being either a gun or a tool. Both reaction time and accuracy of the responses are being measured.
In addition to using black and white faces as primes, together with guns and tools as targets, we added another class of primes to our research. Since this experiment is being conducted in a Dutch population, we decided to add Moroccan faces to our set of primes. Moroccan people in the Netherlands are generally subject to negative
stereotyping (Gordijn, Koomen, & Stapel, 2001), and they also perceive themselves as being negatively stigmatized (Hermans, 2006). Assuming misidentifying tools as guns is a result of implicit racial bias, we expect Moroccan faces to exhibit comparable
misidentification rates as black faces in the Dutch population.
Using IAT inspired reasoning, one possible explanation for the effects of black faces on gun/tool decisions could be that participants have many negative race-specific associations with black people (e.g., black – violence, black – criminal). Another possible explanation, using the same line of reasoning, would be that the effects caused by black faces on gun/tool decisions are not necessarily caused by negative race-specific
associations, but are more due to non race-based associations regarding general
paves the way for a different way of thinking about the results found in gun/tool decision experiments. What if the effects are not caused by race-specific associations, but by negative associations in general? One example of a universally negative
association is the emotion anger. Even though a lot of research has been conducted on the topic of cross-cultural differences in perceiving emotion (Elfenbein & Ambady, 2002), anger is always regarded as a negative one. If the effects on the gun/tool decision task are not caused by race-specific associations, but by negative associations in general, we could hypothesize that other negative associations, namely angry faces, should produce effects on the gun/tool decisions comparable to those of black faces.
Previous research on emotional faces have suggested that humans are
predisposed to detect emotional expressions on faces . Studies using positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) showed that neural activity in the human amygdalae differed when people were exposed to different emotional facial stimuli (Morris, Öhman, & Dolan, 1998; Whalen et al., 1998). Upon perceiving an emotional face, people spontaneously react with distinct facial
electromyographic (EMG) reactions in the emotion-relevant muscles of the face, even if the emotional face is shown briefly or is being blocked from consciousness by
backwards masking (Dimber, Thunberg, & Elmehed, 2000). Other research has shown that a threatening expression is detected faster than a neutral expression on faces (Öhman & Mineka, 2001). Because of the predisposition humans have of detecting and processing emotional faces, angry faces seem to be a valid choice as a prime to represent a racially non-specific generally threatening concept.
The main hypothesis examined in this research is whether black faces (implicitly) activate the concept of guns. This is the same main hypothesis as stated in the original study by Payne (2001) and it is examined to see if the results found in his experiment are replicable in a Dutch population. We expected to see significantly faster reaction times for gun stimuli compared to tool stimuli on black prime trials but not on white prime trials, in line with what was found in Payne's (2001) study. We also expected more error rates on tool trials than on gun trials because we expected participants to misidentify tools as guns rather than guns as tools in the critical black prime trials. We ought it to be more likely that someone wrongly perceives a tool as being a gun than a gun as being a tool, because the latter would be a strictly more harmful mistake in a real life setting.
The next hypothesis examined in this research is that the effects found in
gun/tool decision tasks exist due to a plain bias against the race used as prime. Because the Dutch have race-specific negative associations with Moroccan people (Gordijn et al., 2001; Hermans, 2006) in roughly the same way as Americans have with black people, we expected to see significantly faster reaction times for gun stimuli than for tool stimuli on Moroccan trials but not on white trials, comparable to the results found for black faces. Just like on black trials, we also expected to find more errors on tool trials than on gun trials for Moroccan primes.
Lastly we examined the hypothesis that the effects found on gun/tool decision experiments do not originate from strong race-specific negative associations (e.g., black - violent, black - criminal) but are more due to associations of the concepts with non race-based general negativity (e.g., gun - bad, anger - bad). So far science has suggested that the effects of primes on gun/tool decisions are due to race-specific associations (e.g., Payne, 2001; Correll et al., 2002; Greenwald et al., 2003; Eberhardt et al., 2004) , therefore we expected to see no significant effect of angry white primes on gun/tool decisions in contrast to the race-based primes. More specifically, for angry white faces
we expected to see no significant difference in reaction times for gun or tool stimuli, nor did we expect to see more or less errors on either gun or tool trials.
Methods
ParticipantsForty-four participants (13 male, 31 female) of ages ranging from 15 to 58 years (M = 24.1 , SD = 8.6) have been recruited by using a participation network, combined with direct and open invitations. Where possible, undergraduate students received a participation point for their effort. Originally 46 people had been recruited. However, one participant's data was lost due to equipment failure (a computer crashed during the experiment) and one participant noticed he could keep pressing the space bar all the way till the end of the experiment, effectively making all his data worthless due to a very large amount of null responses. Therefore the number of participants was reduced from 46 to 44. The remaining sample consisted out of 39 participants born in the Netherlands and 5 participants born in foreign countries, however due to a lack of pre-experimental disclosure on how to handle foreign-born participants in our experiment we decided not to remove them from our data pool.
Stimuli
Gun/Tool decision task
Photographs of emotional neutral black, white and Moroccan faces together with a set of angry white faces were used as primes. White and Moroccan faces were selected from the Radboud Faces Database (RFaD; Langner, Dotsch, Bijlstra, Wigboldus, Hawk, & van Knippenberg, 2010). Black faces were selected from the Chicago Face Database (CFD; Ma, Correll, & Wittenbrink, in press). For each type of face we selected 16 different variants, making the total number of unique faces used 64. Pictures were cropped to size 5,3 x 4 cm in order to match the size of our gun and tool stimuli. The gun and tool stimuli were the same set as used by Payne (2001). This set includes six different handguns, two kinds of pliers, a socket wrench, a nutcracker, a screwdriver and an electric drill. This equals to six different handguns and six different tools. The mask consisted of an irregularly black and white colored rectangular pattern in the same size as the rest of the stimuli.
Implicit associations test
Two separate IATs were used in this experiment, one to measure implicit associations with black people (IAT-B) and one to measure implicit associations with Moroccan people (IAT-M). The IAT-M used the Dutch words 'Allochtoon' and 'Autochtoon' as categories, which roughly translates into 'Foreigner' and 'Native'. The foreigner category included the Dutch words 'Marokkaan, Allah, Buitenlander, Islam, Marokko, Couscous, Moskee, Hoofddoek', which translated means 'Moroccan, Allah, Foreigner, Islam, Morocco, Couscous, Mosque, Kerchief'. The native category included the Dutch words 'Nederlander, Nederland, Hollander, Molen, Christendom, Katholiek, Protestants, Kerk', which translated means 'Dutchmen, Netherlands, Hollander, Mill, Christianity, Catholic, Protestant, Church'. The IAT-B used the Dutch words 'Zwart' and 'Blank' as categories, which translated means 'Black' and 'White'. Instead of using words in these categories, eight white and eight black faces were used as items to sort.
The second set of categories was the same for both IATs. The categories used here were the Dutch words 'Goed' and 'Slecht', which translated means 'Good' and 'Bad'. The Good category included the Dutch words 'Plezierig, Stabiel, Gezellig, Betrouwbaar, Mooi, Prettig, Fantastisch, Leuk', which roughly translates into 'Pleasurable, Stable, Cozy, Reliable, Pretty, Nice, Fantastic, Fun'. The Bad category included the Dutch words 'Naar, Gemeen, Onprettig, Agressief, Vreselijk, Pijnlijk, Verschrikkelijk, Vies', which translates roughly into 'Nasty, Mean, Unpleasantly, Aggressive, Terrible, Painful, Horrible, Dirty'. Procedure
Gun/Tool decision task
Participants were told that they would participate in two unrelated experiments. The experimenter told the participants they were about to perform a sorting task that measures both speed and accuracy. The experimenter also explained that in the task, each time a pair images would be shown. Participants were told to ignore the first image, which would always be a face, and to only react to the second image, which would always be either a gun or a tool. Participants were told that all they had to do was to classify the object on the second image as either a gun or a tool, as fast as possible. They were instructed to press a button on the keyboard (A or L) that corresponded with respectively either gun or tool.
The entire task consisted out of 512 trials, each containing a face as prime, a gun or a tool as target and a mask. The prime was shown for 200 ms, after which it got replaced by the target. After 50 ms, the target got replaced by a mask, which remained on the screen until the participant pressed the key corresponding to either gun or tool. After another 500 ms had passed since the last button press, another trial began. After every 32 trials, a short break of one minute would be possible. A randomized list of trials was created for each individual participant, in such a way that every participant
completed the same amount of trials for each category (e.g., black face + tool, white face + gun, etc.). The specific prime or target used in each trial was randomized as well. If for example a trial was coded to be a black face together with a gun stimulus, a random black face and a random gun stimulus were assigned for that particular trial.
Implicit associations test
After the gun/tool task had been completed participants had to make two separate IATs, the IAT-B for implicit associations with black people and the IAT-M for implicit
associations with Moroccan people. The order in which they made these IATs was randomized across all of the participants, in such a way that half of the participants made the IAT-B first and half of the participants made the IAT-M first. The IATs
consisted out of 5 blocks each. Blocks 1, 2 and 4 can be considered introductory blocks and consisted out of 20 trials per block. Blocks 3 and 5 are considered the critical blocks and consisted out of 60 trials each. In each trial participants were told to sort words or faces that appeared in the center of the screen to categories that were anchored to the top-left and top-right corners of the screen. They could sort these items by pressing the keys "E" for left and "I" for right. In the first block, the initial target concept
discrimination was introduced. For the IAT-M that meant participants had to sort words to their respective category (e.g., Catholic-Native, Allah-Foreign). In the IAT-B this phase consisted out of sorting faces to their respective category, black faces to 'Black' and white faces to 'White'. The second block introduced the evaluative attribute
discrimination. For both IATs this meant participants had to sort pleasant and unpleasant words to the categories 'Good' and 'Bad'. The third block was the first combined task. Both the 'Good' and 'Bad' categories were on screen together with the categories 'Foreign' and 'Native' for IAT-M and 'Black' and 'White' for IAT-B. Participants had to sort pleasant and unpleasant words to their respective category (Good, Bad), together with words to the categories 'Foreign' and 'Native' for the IAT-M and black or white faces to the categories 'Black' and 'White' for the IAT-B. The forth block was essentially the same as the first block, with the only exception that the categories were reversed. Categories that were anchored to the top-left were now on the top-right and vice versa. The fifth block then was essentially the same as the third one, with the only exception that the categories were reversed as well, just like what was practiced in block 4.
Other measures
A few extra manipulations were included in the experiment, unrelated to this particular research. The entire study was conducted by five different students, each with slightly different research questions. As a result of this, eye-gaze was manipulated on some of the trials. For that research, half of the faces were gazing to the right and half of the faces were gazing to the left. In addition to the manipulation of eye-gaze, a questionnaire was presented at the end of the study to measure the participant’s familiarity with foreign races. Results of this extra manipulation and questionnaire will not be discussed further in this paper, and instead will have their own individual paper. Possible hindering effects of these extra manipulations with regards to this research will be taken into account in the discussion of the results.
Data analysis
For the gun/tool task, both reaction time and error rates will be analyzed. For the analysis of reaction time, a repeated measures analysis of variance (ANOVA) will be used to look at possible differences in reaction times between gun and tool stimuli for each of the four prime types. To analyze error rates, another repeated measures analysis of variance (ANOVA) will be used to see if there are differences in error rates between gun and tool stimuli for each prime type. IAT scores will be analyzed by looking at the differences between the scores in the association congruent and association incongruent trials.
Results
Implicit associations
To ensure the presence of an implicit racial bias in our participants we analyzed IAT scores by looking at the difference in mean latency between the compatible and non-compatible conditions. For IAT-B, the group showed a moderately strong bias against black people (M = .47 , σ = .41). The IAT-M showed a strong bias against Moroccan people (M = .60 , σ = .36). Both IATs suggest that there is in fact a racial bias present against both of the races used as experimental primes in the current research.
Reaction times
First we examined the effects of different prime types on the reaction times of
identifying objects as either guns or tools. Figure 1 shows the means and errors of the reaction times on gun and tool responses for each of the different prime types.
Figure 1. Mean Reaction Time Scores (in Milliseconds, ms) for Gun/Tool Responses for Each Prime Type. Error
Bars Denote the 95% Confidence Interval.
Data were analyzed using a repeated measures ANOVA with primes (white, black, Moroccan, angry white faces) and targets (guns, tools) as variables. Mauchly's Test of Sphericity indicated that the assumption of sphericity had been violated for both the main effect of primes, χ2(5) = .603 , p = .001, and the interaction effect of primes and targets χ2(5) = .616 , p = .001. Therefore degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity for primes (ε = .74) and the interaction-effect of primes and targets (ε = .79). A significant main interaction-effect of primes was found F(2.21 , 95.11) = 3.873, p = .021, ηp2 = .083, indicating an effect of the prime stimuli on reaction times. Contrasts showed that the main effect of primes originated from a slower mean response time for black faces when compared to Moroccan faces (Δ = 7.42 , p = .021), white faces (Δ = 12.43 , p = .012) and angry white faces (Δ = 8.95 , p = .004). No
390 400 410 420 430 440 450 460 470 480 490
White Black Moroccan Angry White
Re ac tio n Time (ms ) Guns Tools
significant main effect of targets was found F(1, 43) = 2.978, p = .092, ηp2 = .065, showing no direct effect of the target stimuli on reaction times. The interaction effect between primes and targets however turned out to be significant F(2.36 , 101.39) = 7.244, p = .001, ηp2 = .144, indicating that the response time differences between the two types of targets (gun/tool) is differentially influenced by the preceding face.
Since we are interested in finding which prime or primes specifically influence gun/tool reaction differences, post hoc paired-samples t-tests were performed to see which primes significantly produce a difference in response times for gun and tool stimuli. There was no significant difference found in the reaction times for white faces being followed by guns (M = 443,16, SD = 85,93) or tools (M = 438,21, SD = 112,29); t(43) = .664, p = .510. There was however a significant difference found in the reaction times for black faces when they were followed by guns (M = 437,27, SD = 85,60) or tools (M = 468,97, SD = 95,57); t(43) = -4.857, p < .001. Results indicate that participants responded faster to guns than to tools when they were being primed with black faces. This result is in line with the original hypothesis as stated by Payne (2001) that black faces influence gun/tool decisions in such a way that guns are being identified faster than tools. There was no significant difference found in the reaction time scores for Moroccan faces being followed by guns (M = 439,55, SD = 93,67) or tools (M = 451,84, SD = 93,18); t(43) = -1.227, p = .226. In contradiction to our predictions, Moroccan faces did not produce the same effects as black faces did in our experiment. In addition to that, there was no significant difference found in the reaction time scores for angry white faces being followed by guns (M = 443,07, SD = 83,97) or tools (M = 445,28, SD = 101,51); t(43) = -.315, p = .755. The latter also defies our hypothesis that angry white faces produces effects in the same way as black faces do. So far, the data seems to show a unique effect of black faces on gun/tool decisions, unmatched by those of Moroccan faces, white or angry white faces.
Interestingly, instead of responding relatively faster to guns as a function of different prime types, the data as shown in Figure 1 suggests participants instead responded relatively slower to tools as a function of the different prime types. This finding is fundamentally different from the findings gathered by Payne (2001), in whose experiment participants responded faster to guns as a function of the different prime types. To quantify this observational finding, two additional repeated measures ANOVAs using Greenhouse-Geisser adjustments were run. The first one contained only the means of the reaction times to gun responses and the second one only contained the means of the reaction times to tool responses. No significant main effect of guns was found F(2.40 , 103.02) = 1.028, p = .372 , ηp2 = .023, indicating that there was no significant difference in reaction times for gun responses as a function of different prime types. The second ANOVA, containing tool responses, did show a significant main effect of tools F(1.86 , 79.94) = 7.336, p = .002 , ηp2 = .146, indicating that there was in fact a significant change in reaction times for tool responses as a function of the different prime types. Combining these two ANOVAs the analyses suggests the change in reaction times found in the experiment originated from a decrease in reaction time for responses to tool stimuli, rather than a faster or slower response to gun stimuli.
Error rates
Error rates were expressed by calculating a percentage of correct responses to gun and tool stimuli for each prime type condition. Figure 2 shows the calculated error rates (in percentage correct) for gun and tool responses per prime type.
Figure 2. Error Rates Expressed in Percentage of Correct Responses to Gun and Tool Stimuli for Each Prime
Type. Error Bars Denote the 95% Confidence Interval.
The data were analyzed by using another repeated measures ANOVA with primes (white, black, Moroccan, angry white faces) and targets (guns, tools) as variables. Mauchly's Test of Sphericity indicated that the assumption of sphericity had been violated for both the main effect of primes, χ2(5) = .665 , p = .004, and the interaction effect of primes and targets χ2(5) = .710 , p = .014. Therefore degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity for primes (ε = .77) and the interaction-effect of primes and targets (ε = .81). No significant main effect was found for either primes F(2.32 , 99.61) = 1.601, p = .203, ηp2 = .036, or targets F(1, 43) = .191 , p = .665, ηp2 = .004, indicating no direct effect of either the primes or the targets alone to error rates. There was however a significant interaction effect found between the
primes and the targets F(2.44 , 104.71) = 4.962, p = .005, ηp2 = .103, indicating a possible effect of different prime types on the error rates of gun/tool responses.
To further examine this interaction effect, post hoc paired-samples t tests were performed. No significant difference was found in error rate scores for white faces followed by guns (M = .896, SD = .078) or tools (M = .911 , SD = .173); t(43) = -.518, p = .607, black faces followed by guns(M = .913, SD = .085) or tools(M = .864, SD = .166); t(43) = 1.856, p = .070, Moroccan faces followed by guns (M = .914 , SD = .072) or tools(M = .895 , SD = .165); t(43) = .746, p = .460, or angry white faces followed by guns(M = .894 , SD = .089) or tools(M = .904 , SD = .154); t(43) = -.385, p = .702. It has to be noted however that even though the difference in error rates for guns and tools was not statistically significant for black primes in the paired-samples t test (p = .070), a distinct trend towards significance is shown. There does seem to be a relatively large difference in gun and tool error rates for black primes compared to the differences found in other prime type conditions. This trend is made visible in Figure 2 where the 95% confidence intervals of gun and tool error rates for black primes do not overlap,
78 80 82 84 86 88 90 92 94 96
White Black Moroccan Angry White
Per cen ta ge Co rr ec t (% ) Guns Tools
possibly showing a hidden effect of black faces on error rates in either gun or tool decisions. Interestingly enough the data for error rates show a larger change in tool error rates than gun error rates as a function of the different prime types, in the same trend as shown in the analyses of reaction times. That is, the critical difference in performance does not seem to lie in gun responses, but rather in a worsened
performance of tool responses. To further quantify this effect for error rates another repeated measures ANOVA using Greenhouse-Geisser corrections was performed using only the data for tool responses. A significant main effect of primes was found F(2.046 , 87.99) = 4.855, p = .010, ηp2 = .101, indicating an effect of different prime types on tool error rates. Contrasts showed that this effect was mainly caused by the difference in error rates as shown by black primes when they were compared to white primes (Δ = -.048 , p = .009), Moroccan primes (Δ = -.030 , p = .012) and angry white primes (Δ = -.040 , p < .001). The analysis showed that for black faces, error rates for tools were
significantly higher than for white, Moroccan or angry white faces. Discussion
Results of this research partly support the hypothesis that black faces implicitly activate the concept of guns and facilitate responses to gun stimuli. From all of the examined prime types, black faces were the only prime type that showed a significant difference in reaction times to gun and tool stimuli. However, in contrast to what was found in the original study by Payne (2001), the difference in reaction times between gun and tool stimuli did not seem to arise from a faster response to gun stimuli, but rather from a relatively slower response to tool stimuli compared to all other prime types. Instead of facilitating gun responses, the data suggests an effect of black faces on inhibiting tool responses, a finding surprisingly inconsistent with previously conducted research. The effect can be explained however by assuming that because the concept of guns was activated during black prime trials, participants experienced a higher difficulty in
responding to tool stimuli because it contested their initial implicitly activated response. This possible explanation is supported by the finding that there were also significantly more errors on tool trials than on gun trials with black faces as primes. This shows that participants generally had more difficulty in accurately responding to tool stimuli during black prime trials, likely because the black primes implicitly activated the concept of guns during these trials.
No real support was found for the hypothesis that the effects found on gun/tool decision tasks originate from a plain bias against the race used as prime. As the IATs showed, participants exhibited a larger bias against Moroccan people than against black people. Nevertheless, contrarily to what we expected, Moroccan faces did not nearly produce effects comparable to those of black faces on either reaction times or error rates of gun or tool responses. Therefore the effects found on gun/tool decision tasks did not seem to arise from a plain bias against a specific race alone. It could be however that even though participants exhibited a larger bias against Moroccan people in total, the content of this bias might have been different from the bias against black people. The explicit content of racial bias differs from race to race (e.g., Katz & Braly, 1933; Karlins, Coffman, & Walters, 1969; Dovidio, Evans, & Tyler, 1986) and also on an implicit level the content of one's bias differs from race to race, as shown in several lexical decision tasks (e.g., Gaertner & McLaughlin, 1983; Wittenbrink, Judd, & Park, 1997). It is possible that the bias against black people is very gun oriented, while the bias against Moroccan people could be more knife oriented or is grounded in other crimes and stigma's. A
difference in activated concepts related to the races could explain why Moroccan faces did not produce the same effects as black faces did on gun/tool decisions. Further research will be needed to test this hypothesis.
Regarding the hypothesis that the effects found on gun/tool decision experiments do not originate from strong race-specific negative associations (black - violent, black - criminal) but are more due to associations of the concepts with general negativity (gun - bad, black - bad, Moroccan - bad, anger - bad), no real support was found either. In line with what we expected, angry white faces did not produce significant effects for either reaction times or error rates. Of all of the prime types examined in this research, angry white primes and regular white primes exhibited the largest resemblance in scores for both reaction time and error rates. The larger effects found for both of the experimental race primes (black, Moroccan) compared to the angry white primes suggest that the effects found on gun/tool decisions tasks are in fact due to race-specific associated concepts, and not to associations of the primes with general negativity.
Rather surprisingly, the data suggests that black faces are unique in their influence on gun/tool decisions. Regarding reaction times, black faces were the only prime type that exhibited a significant difference between gun and tool responses and regarding error rates black faces were the only prime type that showed a significantly larger error rate for tool stimuli. Could it be then that black faces really are unique in this regard? In order to further analyze this, two possible facets have to be examined. First of all, it has to be examined if the effects found on the gun/tool task are due to the content of the bias against black people specifically, and if so, if this content differs from that of biases against other races. The content of a bias differs from race to race (e.g., Dovidio et al., 1986; Gaertner & McLaughlin, 1983; Wittenbrink et al., 1997) and a gun-specific bias against black people could very well explain why black faces seem to exhibit unique results on gun/tool decision task. Secondly, there is a possibility that the effects found are not due to a specific bias centered on guns, but are due to a more innate mechanism. To examine this possibility, the gun/tool task has to be conducted worldwide across different cultures while controlling for the effects of the content of race-specific biases. Considering the fact that the second possibility builds on the first one, upcoming research should first focus on examining the role of the content of the bias against black people on gun/tool decisions.
It has to be noted that if we had decided to remove the five foreign-born participants from the analyses, the outcome of the analyses would have been
substantially different. Most notably, if we had removed the foreign-born participants from the data pool, a significant effect of Moroccan faces on reaction times would have been found. This effect of reaction time had the same profile as the effect found with black primes, with the effect not originating from a faster response to gun stimuli but rather from a relatively slower response to tool stimuli. Apparently the five foreign-born participants shifted the effects found on Moroccan prime trials, partly concealing the effects found on Dutch-born participants. This adverse effect could be due to the fact that foreign-born participants regard themselves as immigrants as well and are therefore more likely to have fewer negative thoughts about other immigrant groups. While immigrants are an important part of the Dutch population, it will be interesting to see in future research if there is a profound difference in the reaction of Dutch and foreign-born people to Moroccan faces on gun/tool decisions.
The choice of angry faces as a prime type to test the third hypothesis is also debatable. While the innate preposition of processing angry faces can be seen as a benefit, it can also form a confound. Previous researches found unique neurological
activity in the amygdala when participants looked at faces with angry expressions (e.g., Morris et al., 1998; Whalen et al., 1998). On the other hand, other research has also shown that amygdala activity was larger when participants were looking at African-American faces than when they were looking at Caucasian-African-American faces (Lieberman et al., 2005), suggesting a similar neurological route of processing emotional and racial information of the faces. It remains unknown however if race and expression produce similar effects in the amygdala, a question open for future research. It could be that the effects triggered by emotion and race facilitate each other, or possibly cancel each other out. Both of which can skew the data found on this experiment. That being said, for the hypothesis examined in this research, the benefit of humans universally processing an angry face as something negative (Elfenbein & Ambady, 2002) provided us with the least race-dependant experimental face possible. For the purpose of testing the hypothesis that the effects found on gun/tool decisions tasks are due to associations with general negativity instead of race, we had to come up with a prime type that was as least race-dependant as possible while still being a face. If we consider white faces as the baseline condition of our experiment with black and Moroccan faces as experimental conditions to test race-based effects, using white faces with something negative (i.e., angry expression) gave us the best possibility of testing non race-based effects (i.e., something generally negative).
One last thing to mention about the procedure we used in this research is that we did not properly introduce the gun and tool stimuli to our participants. In the original study by Payne (2001) a series of practice trials was implanted to acquaint the
participants with the target stimuli and how to classify them quickly. In our experiment we did not include a few practice trials, and therefore we heard complaints from the participants that they did not know exactly what the target stimuli were and that they missed the first few trials because the experiment started to abruptly. While this could be considered a procedural flaw, because all of the participants had to cope with the same lack of knowledge about the target stimuli there is no reason to suspect the outcome of the experiment would have differed significantly if we had implanted a few practice trials, especially since the experiment consisted out of a large number of trials.
All things considered we can conclude that the effects found by Payne (2001) are replicable in a Dutch population, albeit in a slightly different way. The main difference found in the current research is the emphasis on a higher difficulty in responding to tool trials, instead of a facilitated response to gun trials. Both effects (facilitated reaction to guns, inhibited reaction to tools) however arise arguably from the same implicit activation of the concept guns as an automatic reaction to seeing a black face.
Furthermore we can conclude that the different effects found for each of the prime types are most likely due to race-specific associations with guns. Future research however will have to indicate if the content of one's race-specific associations is directly related to the extend in which their gun/tool responses are being biased.
Literature
Banaji, M. R., & Greenwald, A. G. (1994). Implicit stereotyping and prejudice. The psychology of prejudice: The Ontario symposium, 7, 55-76.
Bargh, J. A., Chaiken, S., Govender, R., & Pratto, F. (1992). The generality of the automatic attitude activation effect. Journal of Personality and Social Psychology, 62, 893– 912.
Campbell, D. T. (1950). The indirect assessment of social attitudes. Psychological Bulletin, 47(1), 15.
Correll, J., Park, B., Judd, C. M., & Wittenbrink, B. (2002). The police officer's dilemma: Using ethnicity to disambiguate potentially threatening individuals. Journal of Personality and Social Psychology, 83, 1314–1329.
Dimberg, U., Thunberg, M., & Elmehed, K. (2000). Unconscious facial reactions to emotional facial expressions. Psychological science, 11(1), 86-89.
Dovidio, J. F., Evans, N., & Tyler, R. B. (1986). Racial stereotypes: The contents of their cognitive representations. Journal of Experimental Social Psychology, 22(1), 22-37. Eberhardt, J. L., Goff, P. A., Purdie, V. J., & Davies, P. G. (2004). Seeing black: race, crime,
and visual processing. Journal of personality and social psychology, 87(6), 876-893.
Elfenbein, H. A., & Ambady, N. (2002). On the universality and cultural specificity of emotion recognition: A meta-analysis. Psychological Bulletin, 128, 203–235. Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On the automatic
activation of attitudes. Journal of Personality and Social Psychology, 50, 229–238. Gaertner, S. L., & McLaughlin, J. P. (1983). Racial stereotypes: Associations and
ascriptions of positive and negative characteristics. Social Psychology Quarterly, 46(1), 23-30.
Gordijn, E. H., Koomen, W., & Stapel, D. A. (2001). Level of prejudice in relation to knowledge of cultural stereotypes. Journal of Experimental Social Psychology, 37, 150–157.
Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: attitudes, self-esteem, and stereotypes. Psychological review, 102(1), 4-27.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464–1480.
Greenwald, A. G., Oakes, M. A., & Hoffman, H. G. (2003). Targets of discrimination: Effects of race on responses to weapons holders. Journal of Experimental and Social Psychology, 39, 399–405.
Hermans, P. (2006). Counter narratives of Moroccan parents in Belgium and the Netherlands: answering back to discrimination in education and society. Ethnography and Education, 1, 87–101.
Karlins, M., Coffman, T. L., & Walters, G. (1969). On the fading of social stereotypes: studies in three generations of college students. Journal of personality and social psychology, 13(1), 1.
Katz, D., & Braly, K. (1933). Racial stereotypes of one hundred college students. The Journal of Abnormal and Social Psychology, 28(3), 280-290.
Langner, O., Dotsch, R., Bijlstra, G., Wigboldus, D. H. J., Hawk, S. T., & van Knippenberg, A. (2010). Presentation and validation of the radboud faces database. Cognition & Emotion, 24(8), 1377-1388.
Lieberman, M. D., Hariri, A., Jarcho, J. M., Eisenberger, N. I., & Bookheimer, S. Y. (2005). An fMRI investigation of race-related amygdala activity in African-American and Caucasian-American individuals. Nature neuroscience, 8(6), 720-722.
Ma, Correll, & Wittenbrink (in press). The Chicago Face Database: A Free Stimulus Set of Faces and Norming Data. Behavior Research Methods.
Morris, J. S., Öhman, A., & Dolan, R. J. (1998). Conscious and unconscious emotional learning in the human amygdala. Nature, 393(6684), 467-470.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological review, 84(3), 231-259.
Öhman, A., & Mineka, S. (2001). Fears, phobias, and preparedness: Toward an evolved module of fear and learning. Psychological Review, 108, 483–522.
Payne, B. K. (2001). Prejudice and perception: The role of automatic and controlled processes in misperceiving a weapon. Journal of Personality and Social Psychology, 81, 181-192.
Perdue, C. W., Dovidio, J. F., Gurtman, M. B., & Tyler, R. B. (1990). Us and them: Social categorization and the process of intergroup bias. Journal of Personality and Social Psychology, 59, 475–486.
Wittenbrink, B., Judd, C. M., & Park, B. (1997). Evidence for racial prejudice at the implicit level and its relationship with questionnaire measures. Journal of personality and social psychology, 72(2), 262.
Whalen, P. J., Rauch, S. L., Etcoff, N. L., McInerney, S. C., Lee, M. B., & Jenike, M. A. (1998). Masked presentations of emotional facial expressions modulate amygdala activity without explicit knowledge. The Journal of neuroscience, 18(1), 411-418.
