What a wonderfully thought provoking essay and impressive body of work! I remember well my early conversations with Mahzarin about my interest in examining the effects of unconsciously priming stereotypic concepts of Black Americans. My friend did not give away her surprise or concern that I might be exploring something quite controversial let alone illicit or dangerous. I certainly had no sense of the potential perils that provoked her trepidation. My goal was to understand why those who renounced prejudice – self-professed egalitarians – showed evidence of bias that belied their conscious intent. It seemed to me that the main challenge was to understand the dissociation between what people consciously believed about their attitudes and what their behavior showed (Crosby et al., 1980). Other theorists championed the theme of the inevitability of prejudice, a theme that seemed tremendously pessimistic and antithetical to the idea that people can change. Could it be possible that there was something below the surface that was driving the inconsistency between conscious and unconscious responses? To explore processes below the surface required asking questions about and using methods to detect these unconscious processes.

What I remember most about my discussions with Mahzarin was that she was interested in what I was doing and was supportive. (And, she clearly understood my need to do research on impression formation!) She asked deep and probing questions and quietly encouraged me to explore this as yet unchartered territory. Her support was reassuring because neither my mentor nor other faculty members at OSU at the time seemed the least bit interested in what I was doing. What I didn’t know was that these early conversations and my early work would encourage my dear friend to join the effort to explore and understand what she refers to as the “dark side of the mind.”

Here we are all these years later, both still interested in “trying to figure it all out.” The Implicit Association Test (IAT) has been a tremendously useful tool in directly assessing the presence of implicit biases. Prior to its development, much of the reasoning about the impact of implicit biases was based on indirect evidence that was consistent with the idea that unconscious processes affected perception and behavior. Being able to directly measure these biases has enabled researchers to establish the prevalence of these biases and brought into focus puzzling patterns of data that Mahzarin so eloquently reviewed and challenged us to think about. Having limited time to respond, let me comment on just two issues that have puzzled Mahzarin as she has explored the nature and consequences of implicit biases – the developmental stability of implicit race preferences revealed by the IAT and the question of such biases being malleable. The issues are, I think, very much interrelated.

Mahzarin and others have been struck by the stability of implicit race preferences among Whites across development. That is, whereas explicit responses fall more in line with socially prescribed norms over time, with the majority of people expressing increasingly egalitarian views as they move from childhood to adulthood, implicit biased favoring Whites over Blacks remains high and constant over development. I don’t find this finding surprising or puzzling. We have known for a long time that Whites show bias favoring their group from a very early age. The IAT and other methods have simply revealed to us that the biases that were originally assessed with self-report measures can also be revealed with implicit indicators. Whatever learning processes lead to the development of implicit preferences for Whites over Blacks, the primitive attitude to which Mahzarin refers, produce strong effects on the responses to which we have conscious access and those to which we do not. Social leaning about race tolerance and egalitarianism leads people to examine their beliefs and to consciously wrestle with whether the racial attitudes align with more abstract principles of equality. For those who sincerely embrace egalitarian values, their conscious beliefs change. The problem is that they are not likely to be aware of their unconscious processes or the need to made adjustments that would prevent them from showing bias in unintended ways (i.e., they may not understand the need for work to undo early learning about what the social world signals about the status or value of social groups). Consequently, a great many people are unwittingly complicit in the perpetuation of discriminatory outcomes. People simply won’t (can’t) change something they are not aware of!

This perspective, however, raises the question of whether people would be motivated to make efforts to rid themselves of unconscious biases if they were made aware of them. And, if motivated, what should they do. For many years, research in my lab showed that self-professed egalitarians struggled with knowledge that they don’t always live up to their egalitarian standards. When confronted with evidence of their bias, they felt guilty. The guilt motivates them to expend effort to understand why they show bias and to be interested in information about reducing bias. This evidence is compelling but indirect. They seemed sincere but would they work at overcoming bias? Yes, they will. When their bias is revealed to them using the IAT, those who are personally motivated to overcome prejudice will spend time on a task they are told will help to eliminate unconscious prejudice. Again, the evidence, though compelling, is indirect – we provided no evidence that implicit prejudice was reduced. More recently, we developed and tested the effectiveness of an intervention designed to (1) increase awareness of implicit bias (2) provide education about the nature of implicit bias (likening implicit to habitual responses) and (3) teach strategies that if practiced would help people reduce the prejudice habit. The goal was to help participants become sensitive to their own implicit biases and then equip them tools to combat the biases.

To raise awareness of their implicit bias, our participants completed the IAT and received feedback about their performance. As is true in the literature, the vast majority of our participants showed implicit bias favoring Whites over Blacks. Control participants were then dismissed. Those in the intervention group watched a 45-minute narrated slide show that included the educational (i.e., explanation of the IAT, summary of evidence linking the IAT to discriminatory outcomes in a wide range of contexts) and training (i.e., stereotype replacement, counterstereotype imagery) components. The training strategies had previously shown to produce incidental reductions in implicit bias (i.e., the strategies were practiced at the behest of the experimenter without any intention to reduce bias or awareness that the strategies may produce lead to reductions in implicit bias). Could these techniques be used proactively to bring one’s implicit responses in line with their conscious intentions? We brought participants back to the lab on two additional occasions, 4 and 8 weeks after the initial IAT assessment, during which we administered the IAT (without feedback).

The intervention was successful! At the outset, the level of implicit racial bias was equivalent for our intervention and control groups. The level of bias remained unchanged for the control group participants – despite having been made aware of the implicit bias. However, by week 4 and extending through week 8, those in the training group showed a dramatic drop in their IAT scores. This study provides the first evidence that people can use the power of their conscious minds to reduce implicit biases. It seems clear, however, that awareness of one’s implicit bias is not sufficient to break the prejudice habit, though it may be necessary. It appears that people also need guidance on how to break the prejudice habit. Once equipped with strategies, those concerned about discrimination and their role in perpetuating it, however unwittingly, will do the work to reduce their biases.

So, is the consistency in the magnitude of implicit bias unusual? I would say only to the extent that one assumes that changes in beliefs translate immediately to unconscious processes. And there is good reason to believe this assumption is not valid. Are implicit biases malleable? Previous work has shown that they are but had left unknown whether (1) the reductions in bias were enduring or (2) whether people could use strategies intentionally with the goal of reducing implicit bias. For those who are personally concerned about ridding themselves of bias, the news is good. A little hard work directed intentionally toward the goal of reducing implicit bias can bring unconscious responses in line with conscious intentions and shine light into the darkness.

(posted for Patricia Devine by G. Comstock)

What advice would you give parents and educators concerned with raising more tolerant beings?

What about the first 3-6 years of life primarily contributes to the formation of these attitudes?

Do you think that implicit attitudes toward non-social/human objects (eg. food, animals, etc) would be similarly stable across development? Or, is the stubbornness to change something particular to how our mind treats other people?

Anthony Greenwald has a downloadable pdf of studies showing connections between IAT scores and real-world behaviour. You can get it here: http://faculty.washington.edu/agg/.

Banaji’s article, in many respects, is exactly what we would expect from an extremely prominent social scientist writing about something she really knows about—namely, a methodology (the IAT) which she helped discover, and which has generated a voluminous data base to which she has been a continuing contributor. The article is superbly written, personally engaging, compelling and forceful in its insights and contextual perspectives, and simply loaded with interesting information—major findings, conceptualizations, and several provocative unanswered questions waiting further inquiry. Indeed, this inquiry will doubtless occur, as I find the essay so stimulating that I can easily see it having a research-stimulating effect—perhaps the highest praise one can bestow upon a review of research. As for her substantive observations, there are many to consider. For myself, the stability of the IAT for race (but not gender) across a vast age range—what Banaji terms the “stark stability of the preference across development– is as interesting a set of data as one is likely to encounter, and the group differences between white and black populations beg for an explanation, as do the individual differences in IAT scores about which Banaji ponders at some length. Individual differences are particularly important not to ignore in a paper that is devoted primarily to strong group effects and, it would seem, generalizations about the mind itself.

Allow me, however, to throw in here a couple of tiny caveats. First, I think Banaji could have left out the dig at President Obama, i.e., her phrase in the last paragraph, “But until the president develops a spine…” I’m hardly defending the president here, nor necessarily refuting Banaji’s observations regarding his orthopedic shortcomings. Nevertheless, I think the phrase does not add to the force of her paper which is both very readable and sophisticated, and it is also a bit distracting, as I found myself musing about her motives or reference point, i.e., just what is it about the president or his policies that Banaji finds so onerous. And once one is prompted to play this guessing game and think about such things, one is not thinking instead about the many other things that Banaji is so uniquely qualified to tell us about. Perhaps I am over-reacting here—the phrase occupies 7 words.

A second issue is perhaps less of a criticism than an elaboration of a point well made. Banaji frames her paper, at least in part, by commenting upon the courage it took for Patricia Devine, as well as Banaji and her colleagues, to engage in the probing of unconscious racism in white people, what she terms “the hidden biases of good people.” She does this very early in the paper, as well as at the end, where, after quoting from Martin Luther King Jr., she concludes, “We have told it like it is about the research discoveries, whether palatable or not.” Having taught many undergraduate courses on the subjects of stereotyping, stigma, and prejudice, I agree wholeheartedly with Banaji’s point. It is not easy to tell “good” people that they are not quite as good as they could be on matters pertaining to race (and related social/ethnic/religious groups). Many people do not like this message and they will not like the messenger either. However, I think it is worth noting here that many social psychologists have told it “like it is,” have discovered unpalatable truths, and continue to do so. Historically, the names Stanley Milgram and Philip Zimbardo come immediately to mind, but many others could be added to this large list. Banaji is clearly not claiming a courageous or heroic status solely for the research at issue in her paper, but I do think it should be recognized that many social psychologists have told it like it is, and often paid a severe price for doing so in terms of criticisms leveled at them not only from resistant lay persons but professional colleagues as well. (The same would be true undoubtedly for many research projects in the natural sciences). To a degree, of course, this is precisely as it should be. Some might say the angst instigated by many scientific discoveries is simply part of the scientific enterprise, a natural (if at times vexing) reaction to unexpected, emotion-laden findings.

Finally, I would like to comment on the title of Banaji’s paper. Titles are interesting phenomena in their own right. All authors want their work to be read, and titles are undoubtedly chosen in an effort to attract readers. Banaji’s title, in my view, is particularly interesting. In fact, when I first read the title, I thought it read “the dark side of the mind.” It took a second reading of the paper before I caught my mistake—I had missed the second “dark.” The title is a kind of double entendre. In one sense, the title’s meaning is clear and straightforward. Referring to her colleague Patricia Devine’s classic study on automatic and controlled processes in racial prejudice, Banaji describes it as a “breakthrough that would open a brand new gate to theorizing and experimentation about the double dark side of the mind.” One of these “darks” refers to the unconscious, and the second to Blacks (“humans with high amounts of melanin in their epidermis”). Thus, taken literally, Banaji’s title is simply a somewhat imaginative description of the unconscious racial bias that is an essential focus of the paper.

However, I find another aspect to the title, beyond the literal one noted above. Banaji seems to be addressing human nature itself. That is, she is indicating that we all have a dark (even dark dark) side to our minds. Most of us are, of course, motivated to think rather highly of ourselves, and social psychologists seem to have a penchant for describing most people as good. Consider, for example, Banaji’s observations: “Psychologists and neuroscientists have found it profitable to think about the mind’s fractures by relying less on the malign motives of bad social actors as the explanation, and more on elucidating the evolutionary presses that created the minds we have and the sociocultural and situational presses that exert influence on us more proximally. Over the past 25 years, we have studied the hidden biases of good people, i.e. self-professed egalitarians.”

After considering the research in this paper (and the innumerable other instances of good people behaving badly in social psychology experiments for the past 7 decades or so), I am left to wonder a bit about this idea of “good” people.” Who exactly are these people? I guess it depends on the meaning of “good.” I think we can agree that Banaji’s focus is not on “the malign motives of bad social actors,” presuming some agreement on her precise meaning of “malign” or ”bad.” To the degree that research participants are not aware of their actions in the IAT laboratories around the world, there would certainly seem a correspondingly low degree of intentionality. But, that aside, there clearly is plenty of bad behavior to be seen here. Banaji suggests that her participants are largely good social actors without malign motives. Are we to ice the cake further and let them off the responsibility hook as well? I think a quick answer here is “yes,” a somewhat unsettling note upon which to end this commentary.

Arthur G. Miller

I test neurtal on the IAT for race, and yet I think I am very race conscious. Race is an annonying salient feature in my experience, and I make some effort to keep this somewhat under wraps.

Here’s what I think produced the neutral reading. I grew up in a white community in a very racist part of the country. In approximately 1986, sitting in our car at a stop light on Nassau Street in Princeton, my adolescent son asked me if I have noticed how much better looking blacks are than whites. In general, of course. I saw immediately what he meant and that it could easily be true. And that the same idea would never have occurred to me. I felt so ashamed that my perception was so skewed and racist. So I started working on it. I tried to uncover the narratives in my mind that would go on when I’d see/meet/read about a black person. I worked quite hard to understand some of the factors behind features of some blacks that might draw criticism and substitute explanations that did not denigrate them. And I got active in trying to promote diversity in my university (not Princeton).

That worked. I was surprised at the IAT result, especially since I clanged on women and science, about which I care a great deal, but not VERY surprised. But then I haven’t done the work on women and science. E.g., I’ve sat on hiring committees and simply noted the prejudices about women and science I have without really trying to alter them. (Not voting on them is hard enough!)

First, if it turns out that the IAT is tracking conceptual associations, rather than revealing propositional attitudes, that might explain some of the puzzling features about the IAT findings that you mention, like the stability of IA’s through development, the matching of results between children and adults, and the relative ease with which the associations can be altered. Those things might be expected if the IAT is revealing what the subject knows about the content of a stereotype, rather than beliefs or values the subject holds.

Second, insofar as the issue above is unsettled, I’d like to urge much more caution in characterizing IAT results in terms of “implicit bias,” especially in cases where those results appear to contradict the subject’s avowed values and attitudes. It cannot be just assumed that the IAT results constitute evidence against the subject’s own reports — in advance of independent evidence, the mismatch could equally well be taken as evidence that the IAT is not tapping into stable or deeply held attitudes. I say this not because I think that we human beings are infallible about our own mental states — demonstrably, there are cases in which our own motivations are opaque to us. But we also know that associations can outlive changes in beliefs and preferences. (Having been an atheist now for nearly forty years, I find I still think “full of grace” to myself whenever I’m introduced to someone named “Mary.”) In cases where a person has devoted a great deal of thought to an issue, or when a person has made a great effort to change her behavior, there is reason to think she knows her own mind, and we should not be casual about contradicting her. The charge of “bias” is incendiary, even if it’s explained that the bias is presumed to be unconscious, and hence not something for which the subject is necessarily responsible.

So like Susanna and Eric, I’d be very interested to hear more about the way you are thinking about what the IAT measures, and I’d welcome more information about studies that examine the relations among IAT scores and behavior that would independently support a charge of bias. Eric reports finding Payne’s results surprising, but they seem to me to be rather what one would expect if the IAT is not really tapping into the cognitive and affective states that control and explain much of our behavior.

Thanks for the article, and the opportunity to comment!

1) Have you (or any of your colleagues) tried correlating individual differences in the salience of race (as a perceiver) with IAT scores? You note that anecdotally it appears that people who somehow avoid having an implicit bias also seem oblivious to race in certain ways. This seems like a testable hypothesis. For example what if you gave a description task where you introduce subjects to characters of different races, professions, and backgrounds, and ask the subjects to describe those characters after some waiting period. If those who mention race later on in their description (or don’t mention race at all) correlate highly with a neutral IAT preference, that seems like it may be a start to showing the anecdotal hypothesis holds. Do you know of any studies that have attempted running something like that?

2) What do you think the relation is between implicit bias and weapon bias that Payne finds? Payne cites that the correlation between those who show a high implicit racial bias and those who fall prey to the weapon bias is low, which I found quite surprising. Do you think that the two effects are tapping into different cognitive systems or attitudes? Does the weapon bias show a different developmental curve (or lack thereof) than the one your report for the IAT?

Thanks again for the great read.

Q1. Is there evidence for whether underlying state is:

(a) an association of concepts, on par with “hot-cold”, “salt-pepper”, “abbot-costello”, “smoke-fire” — movements of the mind from one concept to the next that lack the predicative structure that would make them the kind of state that could be true or false?

…or whether it the underlying state includes:

(b) an attitude that attributes properties to the groups in the test? On this option, someone who shows strong association of white+good and black+bad does so because they have an implicit attitude attributing negative properties to blacks and positive ones to whites.

What kinds of experiments do you think would speak to this question?

Q2. Another dimension of the underlying attitude is negative affect. What do you think the main theoretical options are for where negative affect fits in the implicit attitude?

It seems to me that it could be either downstream of a truth-evaluable attitude (eg, you have negative affect toward a group, because of what you covertly believe about them), or it could be upstream of it (you covertly believe they have negative features, because of your negative affect).

Perhaps something like these same downstream/upstream options apply in the case of association: either implicit bias starts life as negative affect, and grows into an association from there; or the associations come first, and negative affect comes after.

Is there developmental evidence that speaks to whether negative affect toward a group is developmentally prior to full-blown implicit attitudes?

Thanks again
–Susanna

For what its worth, I am a 61-yr old white male. I grew up in a Lilly-white community — Palo Alto and Los Altos, CA. I couldn’t have been any older than how old I was in the second grade but might have been much younger when I noticed my first black person (a man). He and I were both standing in the aisle of a grocery store. Apart from the fact that he was much taller than I was, what struck me most what how white his teeth were. I was envious.