Implicit Intergroup Bias 1 Implicit Intergroup Bias: Cognitive, Affective, and Motivational Underpinnings

ed inductively through experimentation, and thus it does not necessarily reflect the way that information is actually represented or how the mind actually operates. The advantages of representational models are balanced by some important limitations. These include a general disconnect with the non-cognitive systems (e.g., emotion, attention), inconsistencies with functional neuroanatomy, and a lack of connection to actual behavior. For example, several influential dual-process models posit that implicit associations are learned through a slow, associative process in memory (Smith & Decoster, 2000). However, affective associations learned through a classical conditioning occurs rapidly, often after a single exposure to an association (LeDoux, 2000). Therefore, traditional representational models may provide a good account for how semantic associations with social groups (i.e., stereotypes or evaluative associations) are learned and stored, but they do not provide an adequate account of affective forms of bias. Another critical limitation of representational models is that few, if any, specify a connection between mental processes and behavior, and thus they are silent regarding the mechanism through which implicit racial bias is expressed in behavior. That is, representational models do not address how basic emotional processes, such as autonomically-aroused states like anxiety, fear, anger, or compassion, influence the activation and expression of implicit biases. Some theorists have attempted to address this issue by proposing cognitive representations of affect, which are then assumed to interact in a network with cognitive representations of bias (Gawronski & Bodenhausen, 2006; Storbeck & Clore, 2008). This approach typically focuses on how affect shapes cognitive representations. However, the approach of treating emotions as cognitive structures may not fully capture the nature of a true emotional state or the process through which it influences behavioral expressions of racial bias. Similarly, Strack and Deutch (2004) proposed a model through which cognitive and motivational systems influence Implicit Intergroup Bias 31 “behavioral schemata” (i.e., a representation of behavior), but the mechanisms through which schemata translate into actual behavioral responses remain unclear. For these reasons, representational models are limited in their ability to account for emotional aspects of implicit intergroup processes and their behavioral expression. We should note, however, that these limitations refer to broad and long-standing questions about the cognition-affect interface with which the field has grappled. Although these are general issues, we see them as critical to the understanding of implicit racial bias effects. Memory systems model of implicit bias Although representational models have dominated research on implicit social cognition, alternative approaches have recently emerged from research integrating models of learning and memory from the human and non-human neuroscience literatures. Amodio’s Memory Systems Model (MSM) of implicit bias applies an integrative social neuroscience approach to address questions of how implicit racial biases are learned, stored, and expressed in behavior (Amodio, 2008; Amodio & Devine, 2006; Amodio et al., 2003; see also Carlston’s Associated Systems Theory, 1994). Past theory and research has demonstrated multiple forms of implicit learning and memory, each associated with distinct neural substrates (Figure 2; Squire & Zola, 1996; Poldrack & Packard, 2003). This model departs from traditional representational models of implicit processing derived from dual-process accounts of automaticity and control, which assume that implicit processes reflect a single system of associations characterized by a uniform processing mode. The MSM posits that different implicit systems learn according to different parameters, and that they influence emotions, perceptions, cognition, and behavior via different neural and neurochemical circuits. A large body of behavioral, neuroimaging, and brain-lesion research now supports the MSM view (Poldrack & Foerde, 2008). Implicit Intergroup Bias 32 In an effort to better understand the mechanisms of implicit bias and their expression in behavior, Amodio and colleagues have applied the MSM approach to the study of intergroup bias (Amodio, 2008; Amodio et al., 2003; Amodio & Devine, 2006). They noted that affective forms of implicit bias correspond to affective forms of learning and memory, such as classical fear conditioning, which are supported by the amygdala and its associated subcortical circuitry. By contrast, implicit stereotyping reflects semantic associations, which involve conceptual forms of learning and memory, linked to regions of the neocortex such as the left prefrontal cortex (e.g., Brodman areas 45/47) and the medial temporal lobe (Thompson-Schill, D’Esposito, Aguirre, & Farah, 1997). Although most research to date has focused on comparing affective and semantic systems underlying implicit evaluative bias and stereotyping, other aspects of implicit bias likely involve additional systems, such as those associated with habit learning and reward. The MSM is useful because it generates hypotheses for how different forms of implicit bias should influence judgments and behavior. For example, if implicit affective bias reflects a system that governs the activation of autonomic arousal and triggers avoidance behaviors in response to threat, then measures of implicit evaluations should predict basic inhibition and avoidance behavior. If, by contrast, implicit stereotyping reflects the operation of semantic memory systems, which have stronger connections to neural regions involved in judgment formation and goal representation, then implicit stereotypes should influence impressions of outgroup members and goal-driven behaviors. This distinction has been supported by studies of behavior (Amodio & Devine, 2006; Amodio & Hamilton, 2009) and neural activity (Potanina, Pfeifer, & Amodio, 2009). It is notable that, according to the MSM, an implicit evaluation may reflect a combination of affective and semantic associations. In line with classic models of attitudes, an evaluation may be driven by a combination of affective and cognitive (i.e., semantic) processes (Eagly & Chaiken, 1998). Behavior-based measures of implicit bias, such as Implicit Intergroup Bias 33 the IAT, are unable to parse the specific contributions of affect and cognition. Nevertheless, Amodio and Devine’s (2006) findings suggest that, barring abnormal brain function (Phelps, Cannistraci, & Cunningham, 2003), measures of implicit evaluative bias may reflect affective processes. The MSM also generates specific hypotheses for how affective and semantic forms of implicit bias may be learned and unlearned. For example, classically-conditioned associations are learned rapidly, often after a single experience. Once learned, they are tenacious and may never be fully extinguished (Bouton, 1994). By contrast, semantic associations are learned slowly, after repeated and highly-probably pairings between two stimuli (Reber & Squire, 1994). Semantic associations are presumably unlearned in a similarly slow fashion, after repeated nonpairings. It is notable that these predictions are different than those suggested by representational models, which assume that implicit associations are learned and unlearned slowly (Smith & Decoster, 2000; Rydell, McConnell, Strain, Claypool, & Hugenberg, 2007). Amodio (2008) has suggested that past social cognitive models correspond well to the implicit semantic memory system, but do not account for affective forms of implicit bias. Thus, the MSM is not inconsistent with representational models per se, but suggests that representational models pertain to a subset of the range of implicit processes relevant to race bias. A major advantage of the MSM approach is that it posits a model of implicit bias that is integrated with perceptual, emotional, motivational, and cognitive systems, and it delineates pathways from different memory systems to behavior. This model will become increasingly useful as researchers turn more attention to understanding how implicit biases are expressed in social behavior. Although our discussion has focused on the implications of the MSM approach for issues of racial bias, the MSM describes general, basic-level processes that apply to all attitude objects, social and non-social alike. It will be interesting for future work to consider the MSM’s Implicit Intergroup Bias 34 predictions for behavior toward groups that are perceived with varying degrees of affect (cf. the Stereotype Content Model; Fiske, Cuddy, Glick, & Xu, 2002). More broadly, we expect that integrative approaches such as the MSM will become more common as the field of psychology becomes increasingly interdisciplinary. Remaining questions and controversies Although an enormous amount of research has been conducted on implicit bias, many important questions remain. In this section, we touch on two such issues. The first concerns the meaning of responses on an implicit measure – how should responses on implicit tasks be interpreted? The second concerns the broader controversy of whether bias on an implicit task should be considered a mark of true prejudice. Issues in implicit measurement Measurements of implicit bias have a mystique about them. How do they work? How can they measure our hidden thoughts? This mystique has cultivated a view that implicit tasks provide a direct probe into the unconscious mind, such that responses on an implicit bias task provide a pure representation of our mental processes and contents. To be clear, “implicit” tasks measure behavior, or in some cases, physiological responses. The logic is that if a concept is cued (e.g., by a racial prime), then its effect on a subsequent behavior may be observed. For example, Bargh, Chen and Burrows (1996) primed subjects with subliminal pictures of Black faces and measured the extent to which it led to more hostile behavior toward an experimenter. Similarly, Devine (1989) primed African American concepts and measured the degree of stereotyping applied in later judgments of a story character. In both cases, the object of interest is cued (e.g., Black people), and its effect on behavior is measured. In this same way, an “implicit task” primes the object of interest and then measures its effect on a behavioral response (e.g., speed to respond to a target). The main difference is that, in an implicit task, the behavioral Implicit Intergroup Bias 35 outcome is contained within the task, and the measurement is repeated across several trials. In this sense, an implicit task may be thought of as a “behavioral assay,” or a circumscribed index of how the actual behavioral effect would occur in a social situation. We suggest that a useful distinction between explicit and implicit tasks is that an explicit measure assesses the reporting of a belief, or proposition, whereas an implicit measure assesses a behavioral or physiological response. Considered this way, the critical difference between implicit and explicit measures is the channel of expression through which the response is made, rather than a hypothetical process. As noted previously, an implicit task does not provide a pure measure of implicit or automatic processes (Amodio et al., 2008; Payne, 2001, 2005), but rather a combination of processes that are expressed through behavioral channels. Similarly, explicit measures may also assess a combination of underlying processes, though they may be particularly sensitive to explicit beliefs. For this reason, it is useful (and practical) to interpret implicit task responses as behavioral expressions rather than as pure implicit processes. Is implicit bias really prejudice? To be clear, prejudice and discrimination remain strong and pervasive in American society. Controversy and debate surrounding the meaning of implicit race bias measures does not question the existence of prejudice in America. Indeed, the finding that most Americans show more favoritism toward Whites than Blacks on measures such as the IAT cannot be dismissed or explained away – it truly reflects that at some level of processing, people in America tend to have racist tendencies, and these tendencies are often expressed in behavior (Jost et al., in press). This is not controversial. What is controversial concerns a more subtle issue about how implicit racial bias relates to conscious beliefs and overt behavior. Setting aside the issue of whether research on implicit bias reveals a real form of prejudice in American society (it does), this section addresses some of the finer points in evaluating the meaning of implicit bias. Implicit Intergroup Bias 36 In his seminal paper on the measurement of implicit racial evaluations, Fazio et al. (1995) dubbed their sequential evaluative priming task the “Bona Fide Pipeline.” This name was a reference to Jones and Sigall’s (1971) “Bogus Pipeline” – a fake physiological contraption that purported to assess subjects’ true racial attitudes. When connected to the bogus pipeline, Jones and Sigall’s (1971) subjects reported higher levels of prejudice than control subjects, with the belief that any attempt to conceal their true attitudes would expose them as liars. Fazio et al.’s (1995) sequential priming method purported to be a direct conduit to one’s true attitude, obviating the need for “bogus” procedures used in the past. Similarly, when the IAT was introduced, it was heralded as a measure of one’s “true” attitude (Banaji, 2001). Given that the vast majority of Americans, including non-Whites and egalitarians, showed an Anti-Black bias on the IAT, this view was quite threatening to many people (e.g., Arkes & Tetlock, 2004). In essence, it pointed a finger at most people and accused them of bigotry. Several researchers voiced the concern that laypeople completing the IAT online on web sites would be misled into believing that they were unconscious bigots (e.g., Blanton & Jaccard, 2006). The “true attitude” view contrasted with Devine’s (1989) theory that automatic tendencies reflected passive learning in a historically racist culture, but that one’s true belief could only be expressed with the aid of controlled processing (see also Amodio et al., 2003, 2008; Devine et al., 2002). Indeed, several researchers have made a specific point to avoid using the term “prejudice” to describe implicit processes because prejudice is a complex construct that is associated with a wide range of attitudes, beliefs, and behaviors, particularly as the term is used colloquially (see Payne & Cameron, this volume). We ascribe to this principle of usage; the reader may have noticed the absence of the term “implicit prejudice” in the present chapter. A compromise position was proposed by Wilson et al. (2000), who argued that implicit and explicit measures assess different attitudes that exist in different modes of psychological Implicit Intergroup Bias 37 processing. According to the dual-attitudes approach, an individual may simultaneously possess negative implicit attitudes and positive explicit attitudes toward an outgroup. This approach acknowledges ownership of associations that exist within one’s mind, even if they were formed without one’s intention and contradict one’s explicit beliefs. Importantly, both Devine (1989) and Wilson et al. (2000) argue that implicit attitudes and stereotypes can be overridden with controlled processing, and thus the responsibility for the expression of implicit race bias ultimately resides with the individual. In the end, the question “is implicit bias prejudice?” is too complex for a simple yes or no answer. The discussion of whether implicit bias constitutes prejudice corresponds to legal distinctions concerning punishment based on intent vs. harm (Heider, 1958). If a person is held accountable based on their intent, then implicit bias is not prejudice. If their intent is irrelevant, but rather harm (i.e., the expression of implicit bias as discrimination) is the key issue, then implicit bias may constitute prejudice. We will leave this debate to the legal scholars (e.g., Lane, Kang, & Banaji, 2007). We hasten to add, however, that from a social psychological point of view, the question of “true prejudice” is not the critical question. That is, the goal of research on implicit bias is not to identify whether a person is prejudiced, but to understand the mechanisms of the social mind as they relate to intergroup processes and social behavior.