What Does Recent Social Cognitive Neuroscience Research Contribute To Our Understandng Of Steereotyping?

Stereotypes can be considered as cognitive structures containing categorical knowledge of social groups (Fiske, 1998). Conventionally, social psychologists seize to understand the processes of stereotyping by utilising measures such as reaction time (RT) and recall (Bartholow, 2010). However, these methods limit the inferences that can be drawn about the underlying processes. For illustration, a behavioural response from person to stimulus results from perceptual, cognitive, and motor processes (Coles, Smid, Scheffers, Otten, 1995). A recent field, Social Cognitive Neuroscience (SCN), seeks to address these constraints. It aims to understand phenomena not merely at the social level, but also the cognitive and neural levels (Oschner Lieberman, 2001). The approach employs brain measurement techniques such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI), to reveal where, when, and how these processes reside in the brain. This paper is concerned with the contribution contemporary SCN has provided to our recent understanding of stereotyping. Our focus is on person perception (Cunningham et al., 2012), Stereotype activation (Hehman, Volpert, Simons, 2013), and application (Devine, 1989). We first consider prevailing social theories, and then imperative SCN findings. We propose that SCN has profoundly enhanced our knowledge of stereotyping to date. However, we note that caution must be applied when interpreting the validity of these often tentative claims. With advancements in brain measurement and data analysis, this field will inevitably provide greater utility in time.

It is becoming apparent that stereotyping is initiated as early as person perception (Quadflieg et al., 2011), the first processes occurring following exposure to individuals. Traditional research has tended to overlook these processes, favouring study of person construal at the levels of stereotype activation and application (Bodenhausen Macrae, 1998 Fiske Neuberg, 1990). This can be attributed to methodologies employed in social research, where typically participants are shown category labels (e.g. Black), and one then investigates the effects of these on ensuing responses in tasks (e.g. LDT) (Moskowitz and Li, 2011). This consequentially makes it difficult to investigate the perceptual processes involved in the classification and recognition of targets (e.g. Blair, Judd, Fallman, 2004 Gilbert Hixon, 1991), because of their early temporal and implicit nature. Yet, research in SCN suggests person perception is crucial to stereotyping, drawing conclusions beyond those attainable via conventional research.

Extensive cognitive literature documents a core person perception network comprising of the occipital face area (Gauthier et al., 2000), fusiform face area (Kanwisher, McDermott, Chun, 1997), extra-striate body area (Downing, Jiang, Shuman, Kanwisher, 2001), and fusiform body area (Peelen Downing, 2005). These regions consistently activate during perception of human faces/bodies, and appear to activate in response to stereotypes. In research by Quadflieg and colleagues (2011) participants reported either the sex (person categorisation) or dot colour (colour classification) from photographs of men and women depicted in stereotypic or counter-stereotypic occupations. The authors reported that activity in person perception brain regions increased in response to targets violating stereotypic beliefs (e.g. female airline pilot), along with regions associated with executive control (e.g. DLPFC). Such activations were found only under social processing conditions (i.e. not in the dot colour task) (Quadflieg et al., 2011). On this basis, they argued that implicit stereotypes influence the perception of individuals, and posit top-down influences from prefrontal regions to perceptual processing. Thus stereotypes likely influence person perception, a process that occurs almost instantaneously following exposure to a person. But can these findings be taken as wholly valid?

The above inferences are limited by the inadequate temporal resolution of fMRI. It is possible that the activity found in the person perception brain regions fed-forward information to the right DLPFC, rather than vice versa. Yet, this is unlikely considering that there were no obvious perceptual differences between unexpected rather than expected social targets that would lead to variations in person perception activity to be fed-forward (Quadflieg et al., 2011). Whatever the true nature of the above findings, compelling further research demonstrates how our implicit racial biases can influence early perceptual processes (Brosch, Bar-David, Phelps, 2013).

The Implicit Association Task (IAT) measures ones implicit stereotypes, particularly, the degree to which individuals associate traits with selected stereotyped groups. Faster RT to congruent than incongruent pairings implies a strong stereotype to that group (Brosch et al., 2013). Brosch and colleagues (2013) showed Black and White faces to subjects and utilised an analysis technique in fMRI called Multi Voxel Pattern Analysis (MVPA) to predict the race of faces participants were viewing. MVPA uses a machine learning algorithm to detect differential patterns of neural activity that derive from different stimulus sets. If patterns are sufficiently different, the algorithm can distinguish between the stimuli types based on these. Thus, when new data is fed to the algorithm, it can predict the stimulus a participant is viewing (e.g. Black/White faces). They found that the algorithm could better predict the race of face viewed by participants when they exhibited a greater implicit race bias (IRB) as determined by the IAT. Further, these patterns were expressed in the FFA (Brosch et al., 2013). Therefore, it can be argued that one s proponent stereotypes alter the degree to which race is encoded during face processing in the FFA, a region implicated in perceiving the identity of faces. This suggests that our implicit stereotypes may alter initial perception of persons, biasing how we perceive an individuals identity relative to non-stereotyped persons.

However, such results are not ubiquitous, with findings that activity in these person perception regions are greater influenced by salient in-group/out-group dynamics, rather than race indicators in various situations (Van Bavel, Packer, Cunningham, 2011). This posits that person perception may not always be biased by racial cues under some circumstances, casting doubt on the notion that these early perceptual processes are the start of our racial stereotypic responding.

Specifically, Van Bavel and others (2011) recruited White participants and had them randomly assigned to one of two novel, mixed race groups. They then responded to faces (Black/White) from their in-group or out-group. Comparing the effects of race and group membership, they reported greater activation in the FFA to the out-group, but found no effect of race (Van Bavel et al., 2011). Thus, perhaps the greater activity to racial out-groups in prior research could instead be accredited to the FFA responding differentially to in-group/out-group dynamics that are made salient by our goals in the present situation. This contests that the FFA may bias how much we individuate individuals identity on the basis of the salience of group dynamics in the current environments. However, such findings should not be taken at face value, as they could simply reflect the analysis of the fMRI data (Ratner, Kaul, Van Bavel, 2012).

As with the research by Brosch and colleagues, Ratner et al., (2012) utilised a MVPA approach, and used this to re-analyse the data from Van Bavel et al., (2011)`s study. They reported that unlike the conventional univariate approach, MVPA revealed that the FFA could decode the race of faces above chance. This is empirically imperative because the previous study suggested that when the in-group/out-group distinctions were made salient, FFA did not code for the race of individuals, as the in-group consisted of mixed races. Thus, implying that the salient in-group out-group distinctions of the current situation determine how activity in the person perception network discriminates between individuals. However, this re-analysis adds to a body of evidence suggesting that the core person perception network does code for race. It is likely that the FFA is influenced by our current processing objectives, so that when race is not the best way to categorise individuals in the current situation, FFA responds more greatly to other categorical markers, but may still process the individuals to a degree based on cues like race. So, it appears that person perception is influenced by our stereotypes, and how we choose to categorise individuals for cognitive efficiency in the current context (Ratner et al., 2012). Further research has demonstrated just exactly how efficient and dynamic this process is on a temporal scale (Cunningham et al., (2012).

Of pertinence are SCN findings that imply that person perception, a process that ERP research suggests occurs as early as 100ms post stimulus exposure is dynamic (Cunningham et al., 2012). The P100 (Bentin, Allison, Puce, Perez, McCarthy, 1996), and N170 (Herrmann, Ehlis, Meuhlberger, Fallgatter, 2005) are early ERP components known to sub serve face processing. A study by Cunningham and colleagues (2012) employed an EEG paradigm in which participants were briefly shown faces (Black/White). There were two main block types: in one, participants used a joystick to pull faces towards them (approach block), and in the other (avoid block), were requested to push them away. They found that during the avoidance blocks (seen to induce an avoidance motive), the P100 component exhibited an IRB, whereby this early perceptual component was stronger to own race faces. However, in the approach condition, this bias was attenuated (Cunningham et al., 2012). These findings suggest that firstly, an IRB is evident at as little as 100ms post face exposure, providing evidence of person perception processes being influenced by racial stereotypes. But further imply that individual motivations (i.e. to approach/avoid a social target) can modulate early perception processes to other races. These findings seem to advance traditional theorising that suggests early biases associated with stereotypes are automatic, and can only be controlled by later processing (Devine, 1989). In sum, SCN research clearly reveals top-down influences on early person perception processes, and demonstrates with high temporal resolution the speed at which these processes are initiated. Following person perception, one activates their implicit stereotypes. This stereotype activation process has been extensively researched in social literature.

The prevailing social psychological theory of stereotype activation is that stereotypes are embedded within an associative system (Devine, 1989). This is supported by findings that group primes (e.g. a category relevant word/image) consistently influence ensuing speed of processing of stereotype relevant information (Bargh, Chen, Burrows, 1996). For instance, participants primed with the stereotype of African Americans acted with more hostility following a vexatious request by an experimenter (Bargh et al., 1996). Arguably this was due to the initial prime activating in a spreading fashion, the trait concept of hostility, which is associated with the African American stereotype. These theories have profuse support in the literature, but could benefit from more direct evidence of their associative nature in the mind. lt;/p>

SCN research has exploited prior EEG/ERP findings of a signature ERP component, the N400, elicited by semantic incongruency (Kutas Hillyard, 1980) to test the notion that stereotypes are represented as associative semantic knowledge in the brain (e.g. White, Crites, Taylor, Corral, 2009). In one demonstration, Hehman and colleagues (2013) had participants primed with Black or White faces, which were followed by Black/White stereotypic traits (positive/negative). They reported a greater N400 elicited on trials where the trait was incongruent with the prime, relative to when traits matched the prime. Further, the N400 amplitude was modulated by participants` self reported racial prejudice (Hehman et al., 2013). Essentially, this demonstrates the associative nature of stereotypes. Traits incongruent with a prior racial face elicited an increase in the N400 amplitude, showing that this stereotype incongruency altered the ERP process indicative of semantic incongruency. These findings are in line with social theorising, and serve to add an extra dimension of evidence to these. However, there is much debate in the social literature concerning the extent to which stereotype activation, and its subsequent application, reflect distinct processes (Moskowitz Li, 2011).

Specifically, in some research stereotype activation is construed as implicit/unconscious, and stereotype application, controlled/explicit (Devine, 1989). This proposition has received some support in behavioural research (Kunda Spencer, 2003). However, some social enquiry demonstrates that effects including associative learning (Kawakami et al., 2000), and processing goals (Moskowitz Li, 2011) can disrupt even initial stereotype activation. The finding of goals impacting stereotype activation suggests that control may be exerted at an implicit level, and thus not necessarily be explicit. For instance, Moskowitz and Li (2011) in 4 experiments reported that the activation of stereotypes as measured via RT tasks (i.e. LDT and Stroop) can be inhibited in those possessing egalitarian goals following a manipulation getting participants to contemplate past failures at being egalitarian. This is found even in the absence of conscious attempts to inhibit stereotypes during responses, and no reported awareness of inhibition (Moskowitz and Li, 2011). These findings have since been replicated and extended (Moskowitz and Stone, 2012).

Further, in the social literature there has been a proliferation in research investigating the variability of egalitarian motivated individuals in their ability to control stereotypic responses, particularly based on proposed differences in their motivations at an internal (IMS personal) and external (EMS motivational) level (Plant Devine, 1998). It appears that individuals with high IMS but low EMS are most proficient at regulating their responses, compared to high IMS/high EMS and low IMS/EMS participants (Devine, Plant, Amodio, Harmon-Jones, Vance, 2002). Johns et al., (2008) propose that this may be due to the ability of internally motivated people to automatically activate egalitarian goals which inhibit stereotypes. They based their arguments on findings that even despite racial primes (faces) being presented subliminally, subsequent responses by internally motivated egalitarian individuals were more altruistic towards African Americans in an ensuing point allocation task (Johns et al., 2008). However, social research is limited in teasing apart the evidence to support one theory of stereotype processes over the other, and the underlying mechanisms that may enable individuals to inhibit stereotypes.

SCN is of use at ameliorating this debate. SCN ostensibly favours a notion that these two processes are widely distributed and highly dynamic, and may interact at all stages of cognitive processing (Devine Sharp, 2009 Forbes Grafman, 2013). In research by Amodio et al., (2008), they sought to investigate the mechanisms underlying low IMS/high EMS superior ability to regulate stereotypic responses (e.g. Devine et al., 2002). They employed the Weapon Identification Task and looked at activity in the Error Related Negativity (ERN), an ERP component sensitive to conflicts resulting in errors (Yeung, Botvinick, Cohen, 2004). In the WIT participants are shown Black/White faces, followed by tools/guns. It is argued that slower or inaccurate identification of a tool following a Black face is indicative of an IRB. They replicated Devine and colleagues findings that high IMS/low EMS individuals possess superior control of stereotypic responses, and found that this could be attributed to greater activity within the ERN component in such individuals (Amodio, Devine, Harmon-Jones, 2008). They posited that the goal of egalitarianism is internalised by high IMS/low EMS participants such that the goal has become automated and quickly conflicts with stereotypic tendencies prior to a response. This suggests that effective control may be initiated without awareness (Amodio et al., 2008). It could well be that these mechanisms underlie the behavioural results of participants in both of the above two social cognitive studies (e.g. Devine et al., 2002 Johns et al., 2008). Further, counter to the suggestion above my Moskowitz and Li (2011) that goal shielding may inhibit the initial activation of stereotypes, these could be explained by these SCN findings that perhaps their egalitarian goals had been sufficiently internalised that they led to instantaneous inhibition of stereotype tendencies by the conflict monitoring system. Thus SCN findings appear to favour the account for automatic control of stereotypes, and reveal the possible mechanisms that underlie this ability, beyond the interpretations that could be made in behavioural social research. However, noteworthy SCN also provides some support for the dual process account too (Jia et al., 2012).

It may well be that in certain low prejudice individuals that egalitarian motives may be sufficiently internalised to automate a stereotype inhibiting process, but this may not hold for all low prejudice individuals, of whom may still initiate differential neural stereotype activation and application processes. In a study by Jia et al., (2012) they utilised an affective valence, and subsequent content identification task. In their sample of Chinese participants, they found that differences in the ERP components of the P2, N2 and Late Posterior Positivity indicated a stereotype activation process. Whereas, the P2b and P3 appeared to activate to a greater degree in the ensuing content identification task, and these components had different source locations to those components related to stereotype activation (Jia et al. 2012). Though this does suggest distinct engagement of neural processes during each stage of stereotyping, subsequent findings are not fully supportive of this contention.

More contemporary research has investigated the early P2 component in relation to the stereotyping process, but in stark contrast to the above, has argued that it may be malleable on the basis of motivation/goals (Amodio, 2010). Amodio again employed the WIT. He reported that in individuals motivated to respond in an accurate manner appeared to modulate P2 activity via top down mechanisms (evidenced by alpha wave activity prior to the prime) to tune perceptual processes to race. Amodio found that this resulted in better action control in these egalitarian motivated individuals (Amodio, 2010). Perhaps then, motivation to respond without stereotypes leads to top down modulation of perception prior to presentation of the stereotyped group member, which leads to selective attention to racial cues, signalling a need to respond with caution (i.e. without the influence of stereotypes). Thus suggesting that control may not be initiated post stereotype activation, but rather that our goals may impact how we respond to targets, perceive them, and subsequently the likelihood of stereotyping processes to be initiated. These findings seem to support the notion in some social literature (e.g. Moskowitz et al., 2011) that the processes of stereotyping are not necessarily distinct.

In sum, it is clear that social research on stereotyping posits that a myriad of processes may be recruited during stereotyping. It is suggested that stereotypes are stored as semantic associative information, which can be activated by primes and these stereotypic associations are evident in ensuing behavioural responses (Bargh et al., 1996). However, the nature of the methodologies utilised in social paradigms restricts the ability of social research to fully investigate processes with early time courses, prior to stereotype activation. SCN has demonstrated that these are imperative to the stereotyping process, allocating differential perceptual resources on the basis of categorical cues (Ratner et al., (2012), and demonstrating a role for top down influences on these early processes (Quadflieg et al., 2011). Social scientists have conducted profuse research on the distinctions between stereotype activation, and ensuing control of stereotypes. From this, it appears that under some circumstances, individuals are able to control the effects of these stereotypes (Moskowitz et al., 2011). But, there is debate concerning the extent to which these reflect dual processes that is, stereotype activation as implicit/automatic, and control as explicit/controlled (Devine, 1989). SCN has advanced our understanding of these processes, and revealed some counterintuitive findings. For instance, it appears that control may actually precede stereotype activation in some instances (Amodio, 2010), that control can become an internalised automatic response to stereotypes in certain low prejudice individuals (Amodio et al., 2008), and that stereotype activation and application are probably better construed as existing on a temporal continuum (Forbes and Grafman, 2013).

In conclusion, one cannot negate the imperative role for social research on understanding stereotyping. However, with the availability of brain measurement techniques, with their complementary benefits in resolution and ability to determine causality (e.g. TMS), it would be criminal not to utilise these to provide consolidation to prevailing social theories. They further help ameliorate ongoing debates in the literature by providing direct evidence for one line of theory over the other. Particularly, SCN has enabled us better construe the dynamic, flexible, and efficient nature of stereotypes.