Social expectation and visual
perception
Name: Jihane Chaara, 10003589 Supervisor: Marte Otten
The aim of this thesis is to explore how social expectation could influence perception. Detailed knowledge about visual stimuli is often activated and mobilized before the stimuli is fully analyzed in detail. Bar (2003; 2006; 2007; 2009) has proposed a model that provides a potential mechanism for the way knowledge is retrieved from memory in situations where split-second decisions have to be made. This mechanism was explored by performing a gun/tool-task, primed with white or Black faces. These faces were presented in low, high and broad spatial frequencies. The low spatial frequency pictures were expected to travel through a ‘quick and dirty’ system, causing quicker responses to racial biases. The expectations were only partly met. This is to say that white faces were facilitative for tools, but Black faces were not facilitative for guns. The expected three-way interaction only presented itself in low spatial frequencies.
The socially skewed nature of our society is undeniable. Racism, sexism, homophobia, and every other type of judgement-based injustice are abundant, despite our endeavours to nip them in the bud with legislation and making them frowned upon. Research on these topics is extensive, particularly on racial biases. Several phenomena have been broadly explored and are currently accepted as unfortunate truths, such as the tendency for police officers to ‘stop and frisk’ coloured people far more often than White people (Gelman, Fagan, & Kiss, 2007); the disproportional incarceration of coloured people as opposed to White people in the United States (Blumstein, 1982) and Australia (Snowball & Weatherburn, 2007); the incredible underrepresentation of people of colour in mainstream movies (Smith, Choueiti, & Piper, 2013); the slight favouring of school teachers towards White children as opposed to coloured children (Tenenbaum & Ruck, 2007) and the subsequent tendency to penalize coloured students more severely upon transgressions at school (Skiba, Michael, Nardo, & Peterson, 2002).
Gender biases, too, are found in nearly every facet of our societal existence. Even at schools, where our futures are being moulded, persistent predispositions exist that favour male students over female students. Fewer women take up positions in school administrations; females continue to experience harassment from both male students and adults, and females are dramatically lagging behind in scientific fields (Zittleman & Sadker, 2002). Blickenstaff (2005) wrote an insightful paper on the underrepresentation of women in fields such as mathematics, engineering, technology and science. He elaborately discusses numerous possible explanations for this gap, such as the biased manner in which schoolteachers tend to treat girls in science-classes, the lack of female role models in scientific fields, and a cultural pressure to conform to traditional gender roles. He does explore the possibility that there might be inherent biological differences that would prevent women from excelling in scientific fields, but quickly dismisses them as being unhelpful and over-exaggerated. The fact of the matter is, however, that many people, including educators, do believe that there are inherent cognitive differences between boys and girls. This view can be incredibly harmful, given
educators tend to give girls less attention than boys in class, ask boys more complex and abstract questions and compliment boys on the conciseness and factuality of their work, whereas they more often compliment girls on the neatness of their work (Owens, Smothers & Love, 2003). This pattern continues well into universities. A study performed by Moss-Racusin, Dovidio, Brescoll, Graham, and Handelsman (2012) revealed that university science faculty professors who were given identical papers but with differing name-labels (either female or male), rated the ‘male students’ as significantly more competent, more hireable and they were more willing to mentor the ‘male student’ as opposed to the ‘female student’.
How is it that we frown upon racism and sexism and experience it so pervasively nonetheless? There’s a stark contrast between, for example, explicit racism (i.e.: openly expressing racist ideas) and implicit racism (i.e.: unintentional biases at the expense of coloured people). Devine (1989) performed a series of experiments to explore the processes involved in prejudice. In these experiments, Devine established that both high and low-prejudiced individuals have equal knowledge of cultural stereotypes that are being perpetuated. The difference between these two groups isn’t that either group is more or less susceptible to prejudice, but simply that the low-prejudice individuals put in a conscious effort not to be bigoted. When individuals can’t consciously monitor stereotypical thought, as was the case in Devine’s second experiment where priming was used, the automatic stereotypes show themselves in equal measure.
It is of endless significance to realize where these prejudices come from. The conclusion that both high and low-prejudiced individuals experience the same automatic biases when unable to monitor themselves, doesn’t in any way imply that humans are, for example, inherently racist. Instead, this conclusion demonstrates that our societal system is a flawed one; that whereas we strive for and make enormous leaps towards equality, we have a long way to go. Mainstream media portrayal plays a large role in shaping our attitudes regarding a variety of topics and issues. As Briggs and Cobley (2002) wrote in their book on media portrayal: “It matters profoundly what and who gets represented, who and what regularly and routinely gets left out; and how things, people,
events, relationships are represented. What we know of society depends on how things are represented to us and that knowledge in turn informs what we do and what policies we are prepared to accept.”
These findings are neither reassuring nor are they surprising. The relevance of researching these biases is of endless importance, as they form more than sheer inconveniences to people’s lives. These injustices can lead to injuries or even death, as is shown in a study by Keith Payne in 2001. In this study, participants were asked to identify guns whilst being primed with either Black or White faces. The result was not at all surprising: when primed with a Black face, participants were quicker to identify a gun. More interestingly is Payne’s second experiment, in which he set out to find out whether errors in judgement would occur if the participants were under time pressure. It turns out that when presented with either a gun or a tool, and whilst being primed with either a Black or White face, participants were more inclined to make mistakes and misjudge a tool for a gun when primed with a Black face. This outcome has significance in the real world, especially in light of the recent happenings in Ferguson in the United States where an unarmed Black man was killed by a police officer, which then incited a series of grand protests. This, in turn, started a very necessary conversation about the biases experienced by police officers.
This tendency to overwhelmingly perceive people of colour in a negative light is somewhat of an unfortunate trend. Shapiro, Ackerman, Neuberg, Maner, Becker and Kenrick (2009) did an interesting study on the perception of anger on White and Black faces. Participants evaluated neutral-looking White male faces as less threatening when they were followed by relatively angry White faces. This effect wasn’t found when Black neutral faces were followed by angry Black faces – this is to say, even the neutral Black faces were perceived as threatening.
All this raises important questions on how perception works in relation to social expectation. As mentioned before, people may mistake a tool for a gun due to racial biases, and Black faces are perceived as more threatening even when they are neutral. These kinds of alterations in perception
the current model for visual perception doesn’t adequately explain these phenomena. There are several frameworks to clarify how perception functions and it’s been established that perception works both in ascending and descending pathways. This is to say that visual information is both processed in lower-functioning regions upwards and in higher-functioning regions downwards (Bar, 2003).
Despite the knowledge that perception is not strictly hierarchical in its functioning, the general widespread notion for object recognition remains that information is first processed in the lower-level cortical areas, such as V1, V2 (the cells in V1 and V2 react to basic features such as orientation and retinal location (Hubel & Wiesel, 1968)), and V4 (these cells respond best to features of intermediate complexity, such as simple geometrical shapes (Pasupathy & Connor, 1999)). The visual information is subsequently projected to higher-level cortical areas, such as the inferior temporal cortex (IT). The cells in IT are particularly sensitive to complex patterns, viewpoint-invariant properties and faces (Chelazzi, Miller, Duncan, & Desimone, 1993). Within this bottom-up framework, the idea is that recognition occurs when the input image is associated with a representation stored in memory. In a model like this, with a strong emphasis on the hierarchical connections, an object is only recognized after the last cortical area in the visual recognition pathway has received and sufficiently analysed the input. This framework greatly underestimates the top-down connections in visual perception (Bar, 2003).
The bottom-up framework is extensively studied, to the point where it is generally understood that visual processing starts in V1 and continues to IT and beyond. There have been recent studies considering the importance of top-down connections, though there is not yet a consensus on how such a process is initiated. Bar (Bar, 2003; Bar, 2007; Bar, 2009) has proposed a model which not only incorporates top-down processes, but also serves to illuminate how social expectations and stereotypes could influence perception in a split-second timeframe. This potential mechanism consists out of three parts: first, the idea is that the low spatial frequencies in an image are quickly projected directly to the prefrontal cortex (PFC) via anatomical shortcuts. The second
part consists out of these low frequencies triggering expectations in the PFC. The final part involves the back-projecting of these expectations (or, rather: multiple hypotheses) to IT, where they set off the corresponding object-representations, which can then be integrated with the bottom-up processes (Bar, 2003).
The differing spatial frequencies (high- and low spatial frequencies) convey different information about what the stimulus looks like. High spatial frequencies represent abrupt spatial changes in the image such as edges, and they say something about the fine details. Low spatial frequencies, on the contrary, represent global information about the stimulus, such as the general orientation and proportions (Bar, 2003). Indeed, two studies (Tamura & Tanaka, 2001; Sugase, Ueno, & Kawano, 1999) have shown that area IT is quite predisposed to recognizing the low spatial frequencies, much quicker than the high spatial frequencies. This is to say that IT receives low spatial frequencies before it receives the high spatial frequencies. In other words, the rugged and unrefined properties of an image are processed before the fine details. This is in line with the premise that these unrefined properties are quickly sent to the PFC, in order to form initial ‘quick and dirty’ guesses and expectations.
In order to test whether social expectations and stereotypes influence perception via the mechanism proposed by Bar, we will perform a gun/tool task with faces of different ethnicities (Black and white). These faces will be presented in low spatial frequency, high spatial frequency and broad spatial frequency, followed by either a gun or a tool.
We have two hypotheses: first of, we believe that the differing ethnicities should incite varying speeding rates on the gun-response and the tool-response. Indeed, Black faces should elicit negative associations and therefore be facilitative for the gun response, whereas white faces should not show this negative bias. The second hypothesis is that we believe that low spatial frequency images should vary from one another – that is, low spatial frequency faces should elicit different responses compared to high spatial frequency faces, for example.
Indeed, our predictions are that low spatial frequency and broad spatial frequency images of Black faces will be facilitative for the ‘gun response’, whereas white faces won’t have this effect. In fact, white faces may even be facilitative for the ‘tool response’. High spatial frequency Black faces should not have any particular effect on the gun or tool response, as we predict that the high spatial frequency faces would have to travel through the relatively slow bottom-up system, rather than being able to use the ‘quick and dirty’ system Bar proposed.
Methods
Participants
Participants were all university students. Most of them were university students collecting research credits. A total of 62 participants took part in the study. Due to the distortions that outliers can have on valid measurements, the cut-off score for reaction time was set at 1000ms. Trials that violated this cut-off score were excluded (Payne, 2001). Furthermore, three participants were
eliminated from analysis due to a system error, thus leaving the measurements of N = 59 to be examined.
Materials
The materials used in this study were a computer, a computer screen, a mouse and a keyboard. The pictures used to represent white faces were retrieved from the Radboud Faces
Database, which has proven to be valid and useful for its purposes (Langner et al., 2010). The Black faces were retrieved from the Chicago Face Database. The images were selected on the following criteria: they were all of adult males facing forward with a neutral facial expression. The faces were put in grey scale. There were a total of 40 faces; 20 of which were white and 20 were Black.
Initially, a third condition was added: images of houses. These houses were supposed to form a baseline as opposed to the faces. However, after reconsideration, the houses were deleted from further exploration.
The faces and the houses were all offered in low spatial frequency, high spatial frequency and broad spatial frequency. These spatial frequencies were created in Adobe Photoshop 7. The high spatial frequencies were created by using the ‘Emboss’ function; the low spatial frequencies were created by using the ‘Gaussian Blur’ function. The images were created to mimic the stimuli used in a study by Vuillemier, Armony, Driver and Dolan (2003).
The images of tools were retrieved from the Bank of Standardized Stimuli (BOSS) (Brodeur et al., 2012; Brodeur, Dionne-Dostie, Montreuil, & Lepage, 2010; O’Sullivan, Lepage, Bouras, Montreuil, & Brodeur, 2012). The images of the guns were selected from the International
Affective Pictures System (LAPS) (Fox, Griggs, & Mouchlianitis, 2007). These images were all set to grey scale to prevent colour having any type of effect on perception, and to ensure that the pictures of the guns and the tools were as alike in properties as possible.
Design and procedure
Participants were instructed to sit behind a computer. Each trial started with an image of a face or a house for 40ms, followed by an image of either a gun or a tool presented for 50ms, followed by a pattern mask which remained on screen until the participants responded. Participants were asked to press “A” (left) for gun and “L” (right) for tool. There were three types of primes (house, Black face, white face), three types of prime presentation modes (low spatial frequency, high spatial frequency and broad spatial frequency) and two types of targets. For this 3*3*2 design, we show 30 trials in each condition, leading to a total of 540 trials. Trials were shown randomly intermixed, in blocks of 54 trials with 1-minute breaks between blocks.
The method of statistical analysis will be a within subjects Factorial Repeated Measures ANOVA. Both reaction time and the percentage of mistakes will be analyzed separately as dependent variables. The independent variables or factors are prime type (the different ethnicities (Black or white) and the houses), in the three different spatial frequencies of the prime (low spatial frequency, high spatial frequency, broad spatial frequency) and the two different types of target (gun or tool).
To determine whether the conducted experiment supports the theory proposed by Bar, I first examined whether participants identified guns quicker relative to tools depending on racial primes. After, the three way interaction effects came into play: did the spatial frequency levels weigh into the equation? Next, the same statistical comparisons were made using the percentage of correct responses when identifying a target (gun or tool).
The data successfully met the assumption of normality. The assumption of sphericity was violated, which was corrected by using the Greenhouse-Geisser correction.
A repeated measures ANOVA with a Greenhouse-Geisser correction showed that the mean reaction time between primes (Black or white) differed significantly from one another (F(1, 57) = 7,63, p = 0,008). This means that the reaction time was significantly less when a white face was introduced as opposed to a Black face, independent of target-type. Additionally, the mean reaction time between targets (gun and tool) differed significantly (F(1, 57) = 29,78, p < 0,001). The reaction time was significantly less when a gun was presented.
The hypotheses formulated in previous sections, however, rely entirely on the interaction effects. These hypotheses expect the frequency levels to impact how quickly people react to the targets (gun or tool). It was expected that Black faces would facilitate the reaction time to guns, especially in broad spatial frequency and low spatial frequency. Indeed, it was theorized that the white faces should not facilitate either a gun or a tool response, regardless of the frequency level of the presented prime.
The two-way interaction effect between primes and targets showed a trend towards
significance (F(1, 57) = 3,11, p = 0,083). Further analysis using a Paired Samples T-test indicated that participants were significantly quicker to identify a tool if it was prefaced by a white face (M = 441,959, SD = 60,755) as opposed to identifying a tool when prefaced by a Black face (M = 452,00, SD = 64,13), t(58) = -2,84, p = 0,006. When the same analysis was conducted to compare the differences in reaction time to a gun when the prime was a white face (M = 424,27, SD = 59,35)
was versus when the prime was a Black face (M = 43,455, SD = 58,07), the result was not significant, t(57) = -1,22, p = 0,228.
An essential part of this study is the three-way interaction effect. This effect between the primes, targets and frequency showed a trend towards significance (F(1,87, 106,58) = 2,606, p = 0,082). Indeed, the frequency levels (broad spacial frequency, low spatial frequency, high spatial frequency) did not sufficiently impact the reaction time when a gun or tool was identified prefaced by either a black or a white face.
In order to further analyse this trend, the effects of the individual frequencies on the interaction between the primes and the targets were analyzed using a repeated measures ANOVA. The interaction between primes and targets was not significant when influenced by the broad spatial frequency level (F(1,57) = 0,104, p = 0,748). When this two-way interaction was influenced by the high spatial frequency, it was not significant either (F(1, 57) = 0,027, p = 0,871. Lastly, the
interaction between primes and targets in combination with the low spatial frequency, turned out to be significant (F(1, 57) = 9,86, p = 0,003.
This significant interaction explains the trend found in the three-way interaction. In order to investigate this significant interaction (primes X targets X low spatial frequency), the differences within this interaction were thoroughly examined using Paired Samples T-tests. The difference in reaction time between guns (M = 425, SD = 68,28) and tools (M = 435,88, SD = 73,11) was not statistically different when it was prefaced by a low spatial frequency white face (t(57) = 1,496, p = 0,140). By contrast, the difference in reaction time between guns (M = 425,60, SD = 59,72) and tools (M = 458,72, SD = 71,88) when primed by a low spatial frequency Black face, turned out to be significant indeed (t(57) = 5,56, p < 0,001).
These findings are visualized in diagrams 1, 2, 3 and 4, in which diagram 1 contains the broad spatial frequency level, diagram 2 the low spatial frequency level, diagram 3 the high spatial frequency level and diagram 4 all the frequencies combined in order to give a global perspective, as
14 400, 417,5 435, 452,5 470, White Black RT Primes
Diagram 1: Broad Spatial Frequency
Tool Gun 435, 452,5 470, 487,5 RT
Diagram 2: Low Spatial Frequency
400, 417,5 435, 452,5 470, White Black RT Primes
Diagram 3: High Spatial Frequency
Tool Gun 400, 417,5 435, 452,5 470, White Black RT Primes Diagram 4: all frequencies
The analyses above have been carried out using reaction time as the dependent variable. However, the right and wrong responses were monitored as well. These responses have been scored into percentages of correctly identified guns or tools. A repeated measures ANOVA with a
Greenhouse-Geisser correction (due to the fact that the assumption of sphericity was violated) indicated that there is no statistically significant difference in the percentage of correct responses between the different primes (F(1, 58) = 0,51, p = 0,822), or between the different targets (F(1, 58) = 2,95, p = 0,091).
This was carried forth in the interaction effects. There was no interaction between the primes and targets (F(1, 58) = 1,96, p = 0,167), which indicates that the percentage of right responses was not influenced by the face prefacing the target.
The three-way interaction effect between primes, targets and frequencies not significant either, though it did show a clear and strong trend (F(1,71, 99,13) = 2,53, p = 0,093). To further examine this trend, the effects of the individual frequencies on the interaction between the targets and the primes were investigated. There was no effect in either the prime X target X high spatial frequency interaction, or the prime X target X low spatial frequency interaction (all p-values were > 0,2). However, the prime X target X broad spatial frequency interaction was significant (F(1, 58) = 4,05, p = 0,049.
In an attempt to resolve why the low spatial frequency level did influence the interaction between the primes and the targets, a Paired Samples T-test was performed to compare the percentage of mistakes made when identifying a tool whilst being primed by a white face (M = 0,93, SD = 0,08) , as opposed to identifying a gun whilst being primed by a white face (M = 0,87, SD = 0,15). This difference showed a trend towards significance (t(58) = 1,92, p = 0,059). The same analysis was performed to examine if there was a difference between the percentage of mistakes made when identifying a tool whilst being primed by a Black face (M = 0,90, SD = 0,10), versus identifying a gun when primed by a Black face (M = 0,90, SD = 0,12). This difference was not significant (t(58) = 0,319, p = 0,751.
17
The following graphs (graphs 5 through 8) demonstrate these results visually in the same manner as the previous graphs: they are sorted according to the spatial frequency, meaning that graph 5 contains the broad spatial frequency, graph 6 the low spatial frequency and graph 7 the high spatial frequency. Graph 8 offers a visual view on the primes X targets interaction.
80% 95% White Black C or rec t ans w er s % Primes
Diagram 5: Broad Spatial Frequency
Tool Gun 95% ec t ans w er s %
Diagram 6: Low Spatial Frequency
18 80% 95% C or rec t ans w er s %
Diagram 7: High Spatial Frequency
80% 95% White Black C or rec t ans w er s % Primes Diagram 8: all frequencies
Conclusion & Discussion
This study was aimed at comprehending the underlying mechanisms behind the influence of social expectations on perception. The framework used to grasp these mechanisms is the theory conjured by Moshe Bar. Bar theorized that the low spatial frequency components of an image travel via a ‘quick and dirty’ pathway to the prefrontal cortex, before the image has been fully analyzed and recognized via the bottom-up processes involved in perception. The expectation formed based on these low spatial frequency components is transmitted via a top-down process. This information — both the information that traveled via bottom-up processes and the information that travelled via
top-down processes — is integrated in the inferior temporal cortex (Bar, 2003). Thus, the formulated hypotheses of this study endeavored to closely examine how information on the
previously mentioned ‘quick and dirty’ pathway operates. The hypothesis were bipartite. Firstly, it was theorized that, when confronted with a gun-or-tool-task, people would be inclined give
reactions in line with social expectations. That is, white faces would facilitate the tool-response, whereas Black faces would facilitate the gun-response (Payne, 2001) regardless of spatial frequency. The second hypothesis focussed more intently on the theory proposed by Bar. This hypothesis stated that the low spatial frequency images would differ from the high spatial frequency and broad spatial frequency images. This is to say that the before mentioned two-way interaction between racial primes and targets (gun or tool) would be influenced by the type of frequency of the primed image. In particular, it was expected that the low spatial frequency components of the primed images would elicit a magnified interaction between the primes and targets.
The findings showed several interesting results. The results only partly agree with the hypothesis that expected a two-way interaction between the primes and targets. Indeed, it turned out that participants were quicker to recognize a tool when primed with a white face, as opposed to when they were primed by a Black face. In other words, white faces facilitated the tool-response. The finding that the gun response was not facilitated by the Black primes, conflicts with both the hypothesis and research that has been performed in the past.
The three-way interactions showed that there was no effect of the high spatial frequency or the broad spatial frequency components on the interaction between the primes and the targets. However, the low spatial frequency components did have an effect. Indeed, this result corroborates with the hypothesis. After all, an interaction between primes and targets was only seen when the low spatial frequency images were presented. This suggests that the low spatial frequency components of the images actively differed from the high spatial frequency and the broad spatial frequency images (which were theorized to be transported via slower anatomical routes). This result
frequency components of a scene, their split-second reaction is influenced by their social expectations.
This study not only examined the reaction time when identifying either a gun or a tool, but also kept track of the mistakes made in the process. There was no clear three-way interaction between the frequency levels, primes and targets. However, there was a trend towards an effect. Upon closer inspection, it turned out that only the broad spatial frequency images showed an
interaction between the primes and targets. In fact, participants made more mistakes when having to identify a tool after seeing a white face, versus when they had to identify a gun after seeing a white face. Once again, the tool-response seems to be linked to the white primes.
When combining these results, a pattern can be detected. After all, the expected interaction between primes and targets only exist when combined with low spatial frequencies, which suggests that there is, indeed, a difference in perception when only absorbing the low spatial frequency components of a scene. In this interaction, the most interesting finding was that participants were quicker to identify a gun when primed with a Black face, as opposed to identifying a tool when prefaced by a Black face. Indeed, a Black prime in low spatial frequency seems to trigger the readiness to identify a gun. On top of this, the findings repeatedly point towards the notion that white faces facilitate the tool-response. When seeing a low spatial frequency Black face, the notion of a gun seems more available. When seeing a white face, the tool is expected.
Let’s first discuss the ways in which these findings challenge the bottom-up framework that is currently the most mainstream framework when studying perception (Hubel & Wiesel, 1968; Pasipathy & Connor, 1999). This framework suggests that object recognition gets processed in a hierarchical manner. The information reaches the lower-level areas (V1, V2, V4) and consequently gets projected to higher-level cortical areas. The image, then, gets recognized once the processed information is associated with the representation stocked in memory. This hierarchy leaves little room for a possible feedback-loop or a top-down addition. The findings of this study argue that the theory proposed by Bar presents a far more likely framework; a framework that integrates the
notion that when split-second decisions have to be made, information might be projected onto pathways that offer the necessary recognition at once. Not only are there top-down processes at work, but these processes are influenced heavily by the societal expectations that surround us from an early age onwards.
Indeed, the implications of this study are ample. Problematic systems such a racism and sexism get interwoven in almost every aspect of society — even when people have the best intentions. Implicit racism, for example, is deeply ingrained in the way we perceive and react to stimuli around us (Devine, 1989). Well-intentioned human beings have been standing up to this marginalizing system our society is drenched in for centuries (Kandal, 1990). Today, still, movements exist with the purpose of battling issues of social injustice, such as the third wave of feminism all over the globe (Snyder, 2008), the Black Lives Matter movement in the United States (Jee-Lyn Garcia & Sharif, 2015) and the Zwarte Piet Niet movement in the Netherlands. These movements strive to battle inequality, which has to be prefaced by a deep understanding of the problem itself. Indeed, this study sheds light on how pervasive social expectations (and thus, toxic and implicit mindsets) get ingrained in the very core of our being: our brain.
It is of urgent importance that these topics continue to be thoroughly researched. For
example, does this framework work for other social expectations as well? Pictures of women might be presented to participants as primes (in differing spatial frequenties) in combination with a word association test. Such a study could be combined with a test that captures implicit sexism, so that the scores could be combined: do these findings correlate? The possibilities are truly vast.
We need to understand the profound mechanisms that are nurtured by a problematic society. A society creates human beings, who collectively hold up said society. Understanding the human beings, effectively, becomes the way we understand the world we live in.
References
Bar, M. (2003). A cortical mechanism for triggering top-down facilitation in visual object recognition. Journal of cognitive neuroscience, 15(4), 600-609.
Bar, M., Neta, M., & Linz, H. (2006). Very first impressions. Emotion, 6(2), 269. Bar, M. (2007). The proactive brain: using analogies and associations to generate predictions. Trends in cognitive sciences, 11(7), 280-289.
Bar, M. (2009). The proactive brain: memory for predictions. Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1521), 1235-1243.
Blumstein, A. (1982). On the racial disproportionality of United States' prison populations. J. Crim. l. & Criminology, 73, 1259.
Briggs, A., & Cobley, P. (Eds.). (2002). The media: an introduction. Pearson Education. Brodeur, M. B., Dionne-Dostie, E., Montreuil, T., & Lepage, M. (2010). The bank of
standardized stimuli (BOSS), a new set of 480 normative photos of objects to be used as visual stimuli in cognitive research. PloS ONE, 5(5), e10773
Brodeur, M. B., Kehayia, E., Dion-Lessard, G., Chauret, M., Montreuil, T., Dionne-Dostie, E., & Lepage, M. (2012). The bank of standardized stimuli (BOSS): comparison between French and English norms. Behavior Research Methods.
Chelazzi, L., Miller, E. K., Duncan, J., & Desimone, R. (1993). A neural basis for visual search in inferior temporal cortex. Nature, (363), 345-7.
Clark Blickenstaff, J. (2005). Women and science careers: leaky pipeline or gender filter?. Gender and education, 17(4), 369-386.
Devine, P. G. (1989). Stereotypes and prejudice: their automatic and controlled components. Journal of personality and social psychology, 56(1), 5.
Fox, E., Griggs, L., & Mouchlianitis, E. (2007). The detection of fear-relevant stimuli: Are guns noticed as quickly as snakes?. Emotion, 7(4), 691.
Gelman, A., Fagan, J., & Kiss, A. (2007). An analysis of the New York City police department's “stop-and-frisk” policy in the context of claims of racial bias. Journal of the American Statistical Association, 102(479).
Hubel, D. H., & Wiesel, T. N. (1968). Receptive fields and functional architecture of monkey striate cortex. The Journal of physiology, 195(1), 215-243.
Kveraga, K., Ghuman, A. S., & Bar, M. (2007). Top-down predictions in the cognitive brain. Brain and cognition, 65(2), 145-168.
Langner, O., Dotsch, R., Bijlstra, G., Wigboldus, D. H., Hawk, S. T., & van Knippenberg, A. (2010). Presentation and validation of the Radboud Faces Database. Cognition and Emotion, 24(8), 1377-1388.
Maass, A., Castelli, L., & Arcuri, L. (2000). MEASURING PRE]UDICE: Il\/IPLICIT VERSUS EXPLICIT TECHNIQUES.
Moss-Racusin, C. A., Dovidio, J. F., Brescoll, V. L., Graham, M. J., & Handelsman, J. (2012). Science faculty’s subtle gender biases favor male students. Proceedings of the National
Academy of Sciences, 109(41), 16474-16479.
O’Sullivan, M., Lepage, M., Bouras, M., Montreuil, T., Brodeur, M. B. (2012). North-American norms for name disagreement: Pictorial stimuli naming discrepancies. PLoS ONE, 7(10), e47802.
Owens, S. L., Smothers, B. C., & Love, F. E. (2003). Are girls victims of gender bias in our nation's schools?. Journal of Instructional Psychology, 30(2), 131.
Pasupathy, A., & Connor, C. E. (1999). Responses to contour features in macaque area V4. Journal of Neurophysiology, 82(5), 2490-2502.
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(2), 181.
Shapiro, J. R., Ackerman, J. M., Neuberg, S. L., Maner, J. K., Becker, D. V., & Kenrick, D. T. (2009). Following in the wake of anger: When not discriminating is discriminating.
Personality and Social Psychology Bulletin, 35(10), 1356-1367.
Skiba, R. J., Michael, R. S., Nardo, A. C., & Peterson, R. L. (2002). The color of discipline: Sources of racial and gender disproportionality in school punishment.The urban review, 34(4), 317-342.
Smith, S. L., Choueiti, M., & Piper, K. (2013). Race/Ethnicity in 500 Popular Films: Is the Key to Diversifying Cinematic Content Held in the Hand of the Black Director. Np: Annenberg School for Communication & Journalism.
Snowball, L., & Weatherburn, D. (2007). Does racial bias in sentencing contribute to Indigenous overrepresentation in prison?. Australian & New Zealand Journal of Criminology, 40(3), 272-290.
Sugase, Y., Yamane, S., Ueno, S., & Kawano, K. (1999). Global and fine information coded by single neurons in the temporal visual cortex. Nature,400(6747), 869-873.
Tamura, H., & Tanaka, K. (2001). Visual response properties of cells in the ventral and dorsal parts of the macaque inferotemporal cortex. Cerebral Cortex,11(5), 384-399.
Tenenbaum, H. R., & Ruck, M. D. (2007). Are teachers' expectations different for racial minority than for European American students? A meta-analysis.Journal of educational psychology, 99(2), 253.
Zittleman, K., & Sadker, D. (2002). Gender bias in teacher education texts new (and old) lessons. Journal of Teacher Education, 53(2), 168-180.
