[Note: this is likely to be a dynamic document, but any edits will be explained.]
The recent years have highlighted a number of incidents, mainly on US campuses, where social justice activists have rallied against the insidious power of microaggressions (MAs). Microaggression theory (or MAT, as we’ll call it here) finds it’s most comprehensive and cited form in the 2010 book “Microaggressions in Everyday Life” (MIEL) by Derald Wing Sue. The implicating of microaggressions in human discourse has reached a zenith. Modern “progressive” social activism often implicates MAs in its explanation for the genesis and maintenance of social inequality. Within these social justice movements, the notion that people in social minority groups are subject to “everyday verbal, nonverbal, and environmental slights, snubs, or insults, whether intentional or unintentional, that communicate hostile, derogatory, or negative messages to target persons based solely upon their marginalized group membership” (Sue, 2010) appears to be taken as axiomatic.
In extreme cases to deny the existence and pervasive nature of MAs is, in itself, microaggressive and evidence of its pervasiveness. Aside from a few “letters to the editor” (e.g., these from Kenneth R. Thomas and Rafael S. Harris Jr.) in the wake of a piece for the American Psychological Association, MAT has not received a great deal of critique. [Edit 12/6/18: I wrote this prior to the publication of Scott Lillenfield’s thorough and well-received critique in Jan 2017.] But what would such a critique of MAT find? The following sections will contend that such a critique would find MAT to be a poorly-constructed and poorly-articulated psychological theory.
So, what does a good psychological theory look like? Clinical forensic psychologist Tony Ward provided a systematic foundation for how psychological theories might be evaluated (well-summarised on page 6 of this paper by Caoilte Ó Ciardha and Tony Ward). To his mind, a good theory: (1) accounts adequately for all instances of the phenomenon under investigation; (2) avoids internal contradictions; (3) makes few special assumptions; (4) is parsimonious with relevant established theory; (5) can be falsified and generates predictions that can be empirically tested; and (6) explains the deep underlying causal mechanisms. Based on this framework the MAT found in MIEL does not come out so well.
Let’s start with the big picture problem. MAT is commonly established in the psychological discipline of social cognition. A combination, obviously, of cognitive and social psychology, social cognition is the study of the way in which we as humans perceive, interpret, resolve, and consequently act on and in the social environment. It will be shown later that the phenomenon of MAs is a social cognitive process; however, MAT as it is presented in MIEL doesn’t appear to be a social-cognitive theory.
That is not to say that: (1) prejudice (a cognitive concept) does not engender discrimination (the behavioural expression of prejudice); (2) expressions of prejudice can be both overt/explicit and/or subtle/implicit; (3) the effects of discrimination on the well-being of those targeted are negative and can be severe; (4) discrimination is not an issue for many individuals and groups in society; or (5) it is often the case that majority groups are afforded greater opportunity to discriminate against minority groups. It is simply that MAT provides an insufficient explanation for the psychological processes involved.
But what’s wrong with Sue’s MAT? Let’s start with whose social-cognition we’re talking about. Reading MIEL and its language and rhetoric you would be excused for believing that MAT is a theory of cognitive processes in the brain of, let’s call them, the interlocutor – the person who utters the microaggressive statement. Indeed, the term microaggression itself implies that it is an act directed toward another person. MIEL, to me, reads that way with its description of ‘perpetrators’ and their ‘unconscious biases’. In fact, that implication is misleading as the social-cognitive processing that drives the phenomenon of MAs actually occurs in the mind of, let’s call them, the recipient. The only cognitive process that is explained by MAT is that of the recipient inferring the intention(s) of the interlocutor and interpreting the meaning of the communicated item.
As a result, the only cognitive structures, processes, or products on which the MAT can or should draw conclusions are those located in the brain of the recipient. We’ll get to issues of measurement and data later, but the smoking gun here is that the empirical data used as evidence of MIEL is self-report by recipients. Furthermore, MAT, as explained in MIEL, cannot (and should not) draw objective conclusions about the intentions of the interlocutor or their cognitive processes – particularly any processes considered unconscious or implicit, since these are neither objective in nature or are attempts made to measure them. Keep these premises in mind as this evaluation continues.
Nonetheless, this establishes the fact that MAs are manufactured in a brain and is a function of social information processing. So it must have something to do with social cognition. But to consider MAT social cognition would represent something of a retrograde philosophical step for social cognition. Given its reliance on ‘power structures’ as an explanatory mechanism, one might even go so far as to say that MAT is an example of sociological rather than psychological theory.
Unlike MAT, most theories of social cognition do not categorise people into their broad demographic factors like race, gender, and sexuality. And for good reason: a brain is a brain is a brain. It seems a truism that the cognitive products (e.g., beliefs or attitudes) of different groups would differ, given the variety of social environments and personal experiences, but the evidence supporting the existence of broad demographic differences in cognitive structures and processes is still being debated. But however that debate pans out, in drawing broad demographic lines MAT suggests that there is something inherently different in the cognitive function of one broad demographic group compared to another – an ironic proposition for a theory that identifies exactly that kind of discriminatory action as problematic and subsequently claims to explain it.
The implication in MAT is that ‘majority’ demographic brains (in MIEL these are basically white, male, cisgender, heterosexual brains) and ‘minority’ brains (non-white, non-male, non-cisgender, non-heterosexual brains) process social information in different ways and with varying degrees of ‘success’. This notion seems to be both evidentially and ethically flawed. It creates false dichotomies of cognition: white vs. non-white, male vs. non-male, cisgender vs. non-cisgender, heterosexual vs. non-heterosexual. MIEL (and examples of other related writings by Sue) is clear in its conceptualisation of MAT as being specific to “well-intentioned White folk” and MIEL bemoans the pervasiveness of “white European psychology” in Western teaching of the subject (the latter, albeit, perhaps quite rightly).
MAT then infers unsupported (and implausible) representations of the qualities of these ‘majority’ and ‘minority’ brains. It implies that ‘majority group’ cognition is entirely at the mercy of fast/automatic processing – that those individuals lack the ability to inhibit behavioural responses, in this case an inability to stop from acting on one’s prejudices. It also implies that ‘minority group’ cognition is entirely endowed with slow/deliberative processing – that individuals in minority groups have the ability to inhibit their responses, conceptualised as a willingness to identify MAs and react in a measured way. Indeed, MIEL specifically states (again, with little evidential backing) that minority group members have a greater ability for inferring the mental states of majority group members than vice versa.
According to MAT, majority group members surrender to bias and minority group members transcend bias.
Contemporary social cognitive psychology acknowledges the fact that we are members of a variety of societal in-groups. These in-groups include those based on the broad demographic categories that are the focus of MAT (ethnicity, gender, and sexual orientation/identity). Individuals hold memberships in a combination of majority and minority groups. For example, consider a person who is demographically black, male, cisgender, and homosexual. Or a person who is demographically white, transgender, and heterosexual. As the ‘social’ element in social cognition would imply, context is everything. Across contexts these broad demographic in-groups change in their minority/majority status – being male, or white, or cisgender may make one a majority-group member in one context but a minority in another. One might even say that our in-group memberships are (gasp) intersectional?
It is also well-established that we hold cognitive biases that lead us to favour our in-groups over related out-groups. For example, Brewer (2007) conducted a theoretical review of in-group bias and ethnocentric attachment and concluded that (a) in-group attachment and positivity are primary and independent of out-groups, (b) security motives underlie universal in-group favouritism, and (c) attitudes toward out-groups vary as a function of intergroup relationships and associated threats to belonging and distinctiveness. Pretty tight (or at least based on empirical evidence).
MAT states that it is “nearly impossible” for individuals not to inherit racial, gender, or sexual-orientation biases through their ecological inheritance and MIEL asserts that “all citizens are exposed to a social conditioning process that imbues within them prejudices, stereotypes, and beliefs that lie outside their level of awareness”. That’s not a particularly controversial statement. But the MAT expounded in that book provides no inclusion of MAs being delivered by all citizens and therefore seems to contradict that well-established notion by suggesting that it is possible for minority racial, gender, and sexual-orientation groups to avoid that fate.
Given that: (1) each of us has a variety of dipole in-group and out-group memberships and consequent in-group preferences; and (2) since MAT, being founded in sociological concepts of power, inequality, and social justice, and proposing that as a consequence minorities groups cannot commit microaggressions, only majority groups can, it becomes almost impossible to establish exactly who can commit a microaggression against who.
The converse of this leads into the next problem – the inability to generate predictable outcomes. Not only is it impossible to work out who can and can’t microaggress, in MAT, any statement – no matter how benign – could represent a microaggression, because MAs are subjectively measured in the recipient. This inability to say, “this is a microaggression”, “these are the conditions under which we expect it to occur”, and “this is population in which we expect it to occur” makes it almost impossible to evaluate MAT as a psychological theory. This makes MAT unfalsifiable and as such impractical in explaining the social cognitive processes it is intended to.
So how is MAT evaluated; what evidence is presented to support MAT? Well, not a lot really. And the evidence that is presented as evidence of MAT – or at least that is presented in MIEL – is problematic in four key ways.
The first problem is that the psychological literature is uncritically evaluated (and, in my option, regularly cherry-picked) and plausible alternative explanations for the phenomena it seeks to explain are disregarded. Much of the research presented is inappropriately extrapolated and involves unrepresentative samples and, at least as it is covered in MAT-related reviews, studies in which the behavior of majority group members towards minority group members are studied (and not vice versa) are over-represented and over-emphasized. For example, MIEL describes research on helping behaviours as evidence of unconscious biases towards minority groups – however, much of the literature presented only tests that phenomenon in the majority-towards-minority demographic context.
Many substantive claims are made that are not grounded in sufficient evidence. Take, for example, the statement “according to women, gender microaggressions occur frequently.” That may be true, but provision of some source of evidence that establishes what women, when, where, and under what conditions this occurs is typically requisite in academia.
Other evidence presented discounts or ignores plausible alternatives for theory-consistent ones. This includes the first example given in MIEL, of US Senator John McCain’s response to an elderly white woman who stated, “I don’t trust Obama. He’s an Arab”. McCain’s response of a shaken head and “No ma’am. He’s a decent family man, a citizen that I just happen to have disagreements with” is presented as a microaggression exemplifying his unconscious belief that Arabs are bad people who can’t be trusted and can’t be decent, family people.
Such an interpretation, however, presumes a great number of things (not least of all, that McCain definitely heard the speaker’s second utterance) – which, whether intended or not, is an interpretation that is (a) simplistic (in that it includes the presumption that McCain’s response is primarily directed to address the second utterance about race) and (b) uncritically MAT-consistent. Another seemingly legitimate interpretation is that McCain’s response was intended to address the first part of the utterance and that being a decent, family man are characteristics of trustworthy individuals and Obama meeting those criteria at least garner McCain’s trust, if not his political support. I’m not saying that was his intention, simply that it is just as plausible as an explanation. But that explanation is not even considered, let alone considered and subsequently tested, found to be dubious, and disregarded.
The second problem is the nature of the data itself – primarily the reliance on self-report. I’ll very quickly skim over the fact that MIEL presents the author’s own personal anecdotes and even some fictional examples as if they were self-reported real-world empirical research data. Even so, empirical data is presented, so the benefit of the doubt is given to MIEL acknowledging that those anecdotes and hypotheticals were intended to augment the data.
Despite the confusion over whose brain we’re in, we do at least find that MAs are measured in the recipient given that it is where the processing occurs. Nonetheless, in most studies of MAT, the data is self-report taken from samples taken from what are considered to be marginalised populations – in particular, in MIEL, the Asian-American population. Self-report data is itself problematic in the sense that it is reliant on the recipient for verbatim and truthful recall – but in exploratory work and when examining a phenomenon for which no instruments yet exist, it is often all that is available and can generate unique and interesting data. Even if we accept, however, that self-report is acceptable for neophyte theories and exploratory studies, the assumption that there is no value in even attempting to verify that the phenomenon requires measuring in all demographic groups is partisan from the outset.
The third problem is the use of biased measurement methods – particularly the use of leading questions in both the quantitative survey-based and qualitative research methods. Examples of questions used to provide evidence for MAT include: (1) “What are some subtle ways that people treat you differently because of your race?”; (2) “Describe a situation in which you felt uncomfortable, insulted, or disrespected by a comment that had racial overtones”; (3) “Think of some of the stereotypes that exist about your racial group. How have others subtly expressed their stereotypical beliefs about you”. All of these questions presume, a priori, that the participant has already experienced the phenomenon under investigation and implicitly require endorsement of that presumption in order to answer the question. These are leading questions and render the data subsequently generated as compromised.
The fourth and last problem is that MAT speculates wildly, draws outlandish conclusions, and takes huge theoretical leaps of faith in interpretations of the already limited and compromised data. Sue’s first example of a ‘second-class citizen’ microaggression is to mistake a person of colour for a service worker. The “hidden message”, to use Sue’s terminology, that this is purported to convey is that people of colour are servants to whites and couldn’t possibly occupy high-status positions. This theoretical leap is extrapolation ad absurdum, disregards plausible alternative theory-contradictory interpretations, and is neither grounded in empirical data or traditional social cognitive theory.
Similarly, in the section on MAs in their popular textbook Social Cognition, Fiske and Taylor recite the example of an Asian-American woman who has been pulled-over on the highway by a white male policeman who asks her “Is this your car?” This is presented as an MA. This is surely extrapolation ad absurdum?! Leaving aside the fact that asking such a question is a plausible standard operating procedure for an officer of the law prior to asking to see one’s license and registration, exactly what prejudicial view is it supposed to be subtly communicating? I lived in the US for some time – although, I admit that I am not always au fait with the state-of-the-ire in identity politics – I was not aware there was a stereotypical view of young Asian-American women as persistent car-jackers?!
A final criticism is that MAT makes practical resistance futile. I remember, as a student at the University of Birmingham in England, looking upon the nearby offices for the Royal Society for the Prevention of Accidents and wondering how one prevents a chance occurrence that is, by definition, unexpected and unintentional. One cannot avoid MAs. Since you as the interlocutor have no internal control over the cognitive processes that generate an utterance and all and everything that you say could be considered a microaggression under external conditions not within your control, it cannot be wilfully or purposefully avoided. So what’s with all these training programs being developed? Isn’t that a theoretically-affirmed waste of resources?
So, in sum, I don’t think MAT is a good example of explanatory theory in social cognitive psychology. But then how do we explain what is by nature a social cognitive phenomenon? Well, I believe we can explain this phenomenon, using social cognition in a way that is commensurate with contemporary social cognition, explains the phenomenon in all its manifestations, and does not resort to discriminatory practice. It’s grounded in three social-cognitive concepts: (1) out-group homogeneity effects; (2) confirmation bias; and (3) attribution bias.
To explain this new theoretical explanation, let’s use Sue’s own favourite example in MIEL – his “Flight Attendant” anecdote. (I should stress that I believe using Sue personally within this example is justified, since (a) the anecdote is in the public domain and (b) Sue appeared to be comfortable with publicly shaming a flight attendant as a racist.) In this scenario, Sue recounts a flight in which he (himself a self-described Asian-American) and a colleague (a fellow passenger and described by Sue as also a person of colour) are asked by a white female flight attendant to move to seats at the back of the plane, her stated reason being that it was necessary in order to distribute the weight of the plane evenly. She makes this request, as both Sue and his colleague notices, shortly after three white male passengers had been told they could sit anywhere and were not asked to move.
Sue interprets this as a MA: that they have been chosen to move based on their racial characteristics and that the request is the consequence of prejudice and is therefore evidence of racially-discriminatory practice. As a result, Sue describes experiencing emotions of resentment, irritation, and anger, and physical reactions of increased heart rate, blood pressure, and blushing. He confronts the flight attendant by highlighting to her that she has moved two people of colour to the “back of the plane”. She replies with what Sue calls “a righteously indignant tone” that she had no racist intentions, that she doesn’t “see color”, and that she was doing her job.
Social cognition can indeed explain this scenario. We can acknowledge that the flight attendant’s behaviour and her statements are ambiguous and therefore would invite social cognitive interpretation. But let’s remember that the cognitive processing in this scenario – the only cognitive processing we have evidence for from Sue’s self-report is going on in Sue’s brain.
The flight attendant is ‘out-group’ to Sue in a number of key ways – but primarily in this situation, by race. As noted above, well-established social cognitive theory would predict that Sue holds a positive preference for his racial in-group (whether that be membership of a specific racial group or the ‘non-white’ group) over his racial out-groups. Furthermore, out-group homogeneity effects would suggest that Sue also holds a perception that members of that out-group are more similar to each-other than his related in-group – so he will implicitly align the behaviours of the flight attendant with his previous experience of the behaviour of others from that out-group.
Furthermore, Sue’s response might also be augmented by confirmation bias. Nickerson (1989) defines confirmation bias as a “one-sided case-building process” as a result of “unwitting selectivity in the acquisition and use of evidence”. We are biased towards interpreting behaviour in ways that confirm, rather than contradict, our beliefs about the world and how it works. Contradictory information is either ignored or re-formulated as to be belief-consistent. Sue is vigilant to the possibility that he may be the victim of discrimination (some might suggest uniquely vigilant, given his authorship of MIEL!) and thus an interpretation of the Flight Attendant’s ambiguous behaviour as racially-discriminatory confirms Sue’s beliefs regarding the behavior of whites towards non-whites. It is therefore emphasised over other plausible alternatives.
This is accompanied by the contribution of a third cognitive bias, an attribution error. According to Ross and colleagues (1977), attribution theory relates to the way in which humans understand the causes and implications of the events that they witness, to explain behaviour and draw inferences on others and their environments. The data upon which humans can draw in order to do this is often sourced from different places (first-hand, second-hand, mass media, etc.), is often of varying representativeness and randomness, requires sufficient processes for coding, storing, and retrieving, and later requires summarising, analysing, and interpreting (rules, formulas, schemata to extract meaning and form inferences). This is a social interpretative project made even more difficult by the fact that humans often fail to repeat the same behaviours consistently under the same or similar circumstances.
Central to this, is the fundamental attribution error: that we are biased to attributing the behaviour of others (particularly behaviours we perceive as negative) to their disposition, internalising them – “it’s because they have psychological flaws”. In concert, we are also biased to attributing our own behaviour (particularly behaviours that we perceive as negative) to situational factors, externalising them – “it’s a result of this or that thing happening at the time”. So Sue attributes the Flight Attendant’s ambiguous behaviour to her disposition (her holding racial prejudices and being willing to engage in discrimination) as opposed to situational factors (the need to balance the weight of the plane – or myriad other situational factors) and concludes that it was unintentionally discriminatory.
It’s important to note that Sue’s reported interpretation may well have been totally correct. He and his colleague may have been on the sharp end of the discriminatory actions of a person unable or unwilling to control their racial prejudices.
The point here is that Sue does not – and cannot – know this. Furthermore, Sue’s interpretation of the Flight Attendant’s actions as discriminatory and the resultant emotional and physical turmoil he experiences does not constitute evidence that his interpretation is correct.
I don’t know if I can emphasise this point enough. So I’ll just repeat it: Sue does not – and cannot – know this. Furthermore, Sue’s interpretation of the Flight Attendant’s actions as discriminatory and the resultant emotional and physical turmoil he experiences does not constitute evidence that his interpretation is correct.
The consequent experience of being the victim of a ‘microaggression’ is a result of his interpretation of this particular ambiguous social situation. The irony of Sue’s example of course, is that by interpreting the Flight Attendant’s behaviour as the result of her character and attributing that character flaw primarily on her membership of a racial group, according to the tenets of MAT it is Sue who delivers the only demonstrable microaggression when he utters a statement subtly communicating the “hidden message” that she is engaging in racial discrimination.
Et tu, Sue…?!
Furthermore, Sue perceives his own reaction not as being not the result of his own biases or his interpretation of the situation, but of the situation itself. Again, by describing her response as “righteous indignation” Sue demonstrates that he has: (1) interpreted the behaviour in a confirmatory fashion; and (b) attributed it as evidence of her righteousness. The scenario only serves to illustrate Sue’s social cognitive processes, not the Flight Attendant’s.
But what of the Flight Attendant’s experience? This is pure speculation, but consider a plausible (admittedly, idealized) situation in reverse. She asks two people to change seats and one of them points out (arguably with what could also be called “a righteously indignant tone”) the racial group-membership of those to whom she has directed that request. She interprets this as a comment on her moral integrity, that the interlocutor is communicating the “hidden message” that she is engaging in racial discrimination, and she experiences emotions of resentment, irritation, and anger. The flight attendant could also experience precisely the same reported symptoms described in MAT.
MAT in MIEL could not explain both of these explanations of the situation, the behaviours, and the resulting experiences both of the parties involved – despite them having the same aetiology, implicating the same cognitive processes, and resulting in similar biophysical and psychological outcomes. Sue’s description of the scenario gives primacy to his interpretation: his experience is legitimate and authentic. Her experience of the situation doesn’t even merit acknowledgment.
Social cognition does explain both experiences. It suggests that of the two groups she could ask to move she is likely to have been influenced by her preference for the in-group (if Sue had been female, like his colleague, this story may be different) and thus might less inclined to move individuals in her in-group. However, Sue’s response is also consistent with a social cognitive approach as his natural in-group bias and her out-group membership would lead him to interpret her behaviour as being in-group-preferential (and thus out-group-discriminatory) and process that ambiguous information in a biased and schema-consistent manner. Thus, well-established contemporary social cognition appears to be better able to explain the phenomenon that MAT seeks to explain.
Criticism of MAT is not intended to marginalise the experiences of many individuals and groups who have been the subject of discrimination and racism. Simply put, this analysis sought only to suggest that these concepts are already explained by standard social-cognitive theory (in-group/out-group preferences and the many “-isms” that are based on objective, measurable, and intended statements of prejudice) and not this ‘folk psychology‘ original sin of subjectively-perceived insult.
A social cognitive explanation also does not exclude a recipient from taking offence to comments that they interpret as marginalising and exclusory, and exemplary of prejudicial beliefs in the interlocutor: that experience and the physical reactions that can result are real. It is simply to say that the recipient cannot then claim that this subjective interpretation is an objective truth without further independent evidence of the intention(s) of the speaker.
Humans are notoriously bad at inferring the mental states of others and such inferences are typically the result of egocentrism (using one’s own beliefs as an anchoring point and adjusting from there) and existing stereotypes. One should not attempt to infer the mental states of others without at the very least realising that such an interpretation may be the product of one’s own cognitive biases. To borrow the social justice vernacular, to do so would be a failure to check one’s cognitive privilege.
As a species, we have long contended with the ethical conundrum of the fine but distinct line between prejudice and discrimination. We all hold prejudice at a cognitive level and we need to have pre-judgements of individuals and social groups. Such cognitive heuristics are essential to our being able to function in the social world without being overloaded with information and to process a great deal of information efficiently.
On the whole, most people are able to avoid allowing their prejudices to develop into literal acts of discrimination and many appear to be able to identify and address their own prejudices a priori. (Although, it is worth noting that our inferences about our own cognition is notoriously poor and subject to bias.) Indeed, the world pays its comedians to riff on our many prejudices for our collective entertainment.
Also, the likely real-world effect of MAT is further distance between social groups as members trip over language and semantics in everyday conversation – or even avoid conversations with out-group members – for fear of their words being considered microaggressive. As Pettigrew and Tropp’s 2006 meta-analysis of the effects of intergroup contact on prejudice suggests, what we need to be doing is encouraging contact between groups and thus reducing anxiety about intergroup contact.
Thus, pushing the faulty logic of MAT is likely to be counter-productive towards efforts to reduce discrimination and the various related –isms. Furthermore, requests that people be mindful of what they say – again, bear in mind that in MAT any statement can be considered a microaggression – is to demand a level of deliberative social information processing beyond the capacity of most human beings.
It would appear, to me, to be important that we continue to acknowledge that the simple fact that our cognitive systems produce in us prejudicial beliefs is not innately a harmful thing and that we should not shame people for simply having beliefs – beliefs that we are typically able to control. MAT seems to be a lesson in turning our innate prejudices – again, essential to our ability to process social information – into a form of cognitive “original sin” that we should bear.
Given that in MAT, the only explicit cognitive processing that we know is subject to bias is that of the recipient, to advocates of MAT I say, let he who is without sin cast the first stone as it were. It also has the dangerous and noxious effect of excusing racist beliefs. The determinism in MAT – the implication that those people in majority groups cannot help but express their prejudices against the minority – provides a useful excuse for those who wish to do so. “Well, it’s not my fault I keep uttering these derogatory statements about women: it’s inevitable that me, a majority group member, would do so.”
The suggestion that groups of humans are uniquely unable to abstain from racism, or to understand when a statement is or isn’t tainted with discrimination is not only implausible, but provides those who delight in doing it with an impost to go ahead.
Thanks for reading.
Ian A. Elliott, Ph.D.
June 8, 2016