Erasing and Replacing Race

Are humans doomed to attend to racial categories?

Please note: This is a repost of an essay I wrote for Aporia, very slightly edited from the original. The editors were kind enough to countenance cross-posting this essay here at Fossils. For subscribers who came here from Aporia, apologies for the redundancy.

When I was a graduate student at the University of California Santa Barbara (UCSB) in the psychology department, most of the people studying human social behavior were, unsurprisingly, in the social psychology area. Those of us studying human social behavior from an evolutionary perspective were cordoned off in the Development and Evolution area, known as Devo. In my first year of graduate study, I took a required seminar in social psychology and discovered a new world of counterintuitive “facts” about the human mind, most of which struck me as unlikely to be true, a suspicion that has been plentifully vindicated in the fullness of time.

One claim I was repeatedly subjected to was the idea that there were three dimensions along which people automatically categorized others: age, sex, and race. The idea was that when you encountered a “social target”—a person, for those of us not wedded to obfuscating terminology—you could not help but notice those three things about them.

I have regrets about my time in graduate school. No, I’m not talking about the time we loaded the department photocopier with paper that had images of animals in, eh, compromising positions printed on one side. I mean, yes, I regret that as well. (Loy, now you know. I suspect you suspected.) But what I really regret is that my research was motivated in no small part by an interest in showing that a bunch of Really Smart People were wrong.¹

Now, look, science proceeds through falsification, so my regrets are modest. But still. Most of my early work was creatively destructive in that sense, with a bit more emphasis on destruction than creation, if I’m honest. Sometimes I wish I had focused a bit more on rewards of building new ideas and less on the satisfaction of undermining old ones.

In any case, from an evolutionary perspective, I reasoned, there was good reason to think that age and sex could be special when it comes to the social world. If you want to make good predictions about others’ preferences and likely behavior, age and sex would have been pretty useful. So maybe evolution selected for systems that parsed people this way automatically, outside of conscious awareness. That hypothesis was at least plausible.

On the other hand, race doesn’t occupy a similar position. In the past, people didn’t travel nearly as far as they do now. Most people would only encounter people who were similar to themselves, at least as far as what we think of as “race,” today. So it seemed unlikely that humans would automatically categorize people’s race.

At the time, social psychologists in general and the researchers at UCSB in particular were talking quite a lot about race. At a social psychology conference some time later, a count of the words in the program revealed that racial appeared 179 times, race appeared 146 times, and stereotype 160 times. Meanwhile, “evolution-related terms (evolutionary, evolved, evolution) came in at 30, 15, and 11, respectively the same amounts as Purdue, embodiment, blue, and… seldom.” (I used to be snarky, you see.)

If you’re wondering how things have changed, in the recent 2024 Society for Personality and Social Psychology conference program, race appears 88 times, racial appears 126 times, and some form of stereotype appears 111 times. Following terms such as pavilion and poster, the word diversity is the 35th most common word. Following that are identity, gender, race, bias, inclusion, and equity.

In contrast, evolutionary appears 11 times. Evolution appears once. Darwin did not make it at all. He was selected out, I suppose.

In any case, I spent my 2^nd and 3^rd years of graduate school—roughly 1994 to 1996—gathering data to see if I could find evidence that race was not, in fact, encoded automatically.

This work took a long time to be published for, well, reasons, with the title, too clever by half, Can Race be Erased? (Credit, or blame, for this title, lies squarely on the shoulders of John Tooby, god rest his soul.)

I bring all this up again now because of a new paper forthcoming in the journal Evolution and Human Behavior, extending this work, entitled Can Race be Replaced?

The balance of this post reviews the paper and attempts to draw some larger lessons from the work.

First, the theory.

When we were explaining our reasoning in the Erased paper, we wrote: “we propose that no part of the human cognitive architecture is designed specifically to encode race. We hypothesize that the (apparently) automatic and mandatory encoding of race is instead a byproduct of adaptations that evolved for an alternative function that was a regular part of the lives of our foraging ancestors: detecting coalitions and alliances” (p. 15387).

In plain English, we were saying, look, humans don’t have a special ability or propensity to notice race. We do, however, work hard to identify what coalition people belong to: which groups do they belong to and cooperate with? So, today, racial groups sort of hijack this function, so it only looks like we automatically notice and remember race.

Similarly, in the new paper, Sng, Boyd-Frenkel and Williams—I’ll refer to this as SBW, for short—write: “Evolution may not have selected for a psychology of race. However, evolution has selected for a psychology of coalitions—the groups with which people typically cooperate and/or compete (Pietraszewski et al., 2014),” citing my friend and former lab mate David Pietraszewski for the idea in a piece that appeared more than a decade later than mine, for reasons best known to SBW. I hastily note that Pietraszewski improved on my methods in his subsequent work, replicating and extending my findings using cleaner techniques. So, credit where credit is due.

So how does one answer questions such as this one?

I, and they, used a method known as the “who-said-what” paradigm. It goes like this.

Experimental participants are shown a series of pictures of people. The stimuli are constructed with cross-cutting categories. So, in my work, half of the people seen by participants were on the red team and half were on the blue team. Within each team, half of the players were black and half were white. Participants see a dialog: the images of people paired with the statements they made.

Who said what?

A little while after they watch this little dialog, participants are presented with a surprise recall task. They are re-presented each sentence and their job is to identify which person made the statement. Participants cannot remember all the people and statements accurately, so they always make errors, confusing which person said which statement. To measure how the participant categorized the people making the statements, we look at the pattern of errors that they make. More specifically, we are interested in how often participants confuse one person with another according to a feature like race, sex, or age.

There are two kinds of mistakes that a participant can make: 1) Within-category errors (e.g., attributing a black person's statement to another black person) indicating categorization and 2) Between-category errors (e.g., attributing a black person's statement to a white person) indicating lack of categorization. By tallying up these measures and doing some statistics, you can make inferences about how people are misremembering according to category features that they automatically encoded.

In my work, I used “teams” (the red/blue team indicated by shirt color) to interfere with the category of race. When participants misremember who said a statement, would they confuse a speaker of the same race or a speaker on the same team? If team is a more powerful indicator of coalition than race, and people are typically “automatically” encoding race as a proxy for team, then shirt color could overpower skin color in how participants categorize the speakers.

SBW were interested in a different cross-cutting category, what they term ecological harshness or unpredictability. So in my case, people were wearing different colored uniforms, a marker of their team. In SBW, the background of the image was changed, with some backgrounds being “hopeful” and some being “harsh,” as evaluated by raters. (You can see the stimuli here. To my eye it’s basically West Philadelphia versus a swanky suburb, but that might just be me.)

Remember, my friends in social psychology told me that race was automatic, so you shouldn’t be able to reduce racial categorization, in general, and certainly not with a five-minute slide show. In both my work and the new work by SBW, however—not to mention work by others—it did.

Ok, so what?

First, I think this work is interesting in itself. I’m not sure that I would have predicted that crossing race with nice versus less-nice houses would have had this effect. These studies seem to be well powered, with hundreds of participants. Generally, this reinforces the suspicions I began with decades ago: racial categorization is, at least to some extent, fragile, and not automatic. Maybe you can’t easily erase it, per our perhaps overly aggressive title, but you can at least suppress it.

Second, and related, some readers might be familiar with the so-called replication crisis in psychology. In some respects, the work by SBW is a conceptual replication of the initial work done back in the 90s. It’s not the same, but it tests—attempts to falsify—the same hypothesis. This idea has now been shown to be at least somewhat robust. Again, I commend to readers the careful work by David Pietraszewski, whose work on this topic is, candidly, better than mine and better powered.

I note this in particular in the context of a less happy outcome for a body of work that also came out recently. When I was at UCSB, a very popular line of research looked at “stereotype threat.” The idea was that if you remind someone of their social category before a test and there’s a negative stereotype about people of that category—e.g., women are bad at math—then their performance would suffer if the test-taker was a woman and the test was on math. This research has been under fire for some time, but a recent failed attempt to replicate the effect I think might finally put the nail in the coffin of this work.

Other lines of work to do with stereotypes and discrimination have similarly fared poorly. In some cases, there just wasn’t any there there. In other cases, people just made stuff up.

Obviously, one successful conceptual replication in no way settles such a major point of debate, but I think about the continual toppling of core research findings in social psychology in the context of my experience at UCSB. There was a certain rivalry between the people who used an evolutionary meta-theory to derive hypotheses—the Devo people—and the social psychologists, who mostly relied on folk psychology, their intuitions about human nature. (These intuitions were mostly that people were lazy and racist. I’ll leave it to the clinicians to suss out why they thought that.) The social psychology crowd claimed that we Devo people were just “telling stories” rather than doing science. In some sense, they were correct insofar as what they meant by “science” was not what we meant by the term. They mostly thought science was a way to dress up political activism with impenetrable prose and a p value to make it sound more authoritative. You can see this in the topics they studied and how well—really, how poorly—they held up. I highly recommend work by Lee Jussim on this topic.

Now, look, I’m not saying that all the work from an evolutionary perspective has been rock solid. The work on ovulatory cycle effects has largely (though not completely) fallen apart  and researchers should probably have known that their studies were underpowered. So-called “Error Management Theory” has now been recognized as a rebrand of signal detection theory. And, more generally, there is room for improvement in the field to be sure. All fields have people with more and less talent. Such is life.

Still, I’m pleased to see that an idea derived from an evolutionary theoretical perspective continues to be vindicated by the data and that the overall idea is generating new research. This is more than can be said of many ideas in science, especially the social sciences. To my mind, the successes of hypotheses generated from the evolutionary perspective illustrates the power of starting with a firm theoretical foundation, particularly compared with starting from intuition.

There is, in addition, I think, a larger lesson about human nature and the current political infatuation with race. The work discussed here illustrates that we don’t have to see race. These relatively gentle manipulations in the laboratory reduce the extent to which people see race—or, at least, encode/recall race—in a matter of minutes. What could be done over years?

Race is not the same sort of fundamental way to parse the world that sex and age are.

Humans don’t have to see race, focus on race, obsess over race.

Doing so is a choice.

1. On the other hand, I have no regrets regarding the alleged political motivation of my work. After the research I discuss below attracted some attention, I was accused of political pandering with regard to race blindness which, at the time, was the politically correct position. This played no role in my motivation. Anyway, how times have changed…