Somewhere between the news and the dinner table, most of us have asked it quietly.
Can this actually change?
Or is it just how people are?
The question is not abstract. If you believe racism is permanent, the weight of it settles in your chest as hopelessness. If you believe it disappears on its own, you stop looking at what is still there. Neither belief helps you carry your life well.
So this article does something different. It looks at what researchers have actually measured, says plainly what they found, and says just as plainly where the evidence runs out.
The short version: change is real, it is uneven, and it does not happen by accident. Some things have moved a great deal. Some have barely moved at all. And the research points to specific things that make a difference.
Is It Just Human Nature?
Start with the biggest claim: that people are simply built to divide into groups and distrust one another, and race is where it lands.
Part of that is true. Human minds sort people into categories fast. We notice who is like us and who is not.
But noticing is not the same as hating, and the research on this is more hopeful than most people expect.
In 2001, researchers Robert Kurzban, John Tooby, and Leda Cosmides ran experiments in which people watched a conversation and later tried to remember who said what. The mistakes people made showed which categories their minds were using. When other cues about who was on whose team became useful, the habit of sorting people by race dropped sharply. In some conditions it became statistically undetectable.
What that shows is that racial sorting in the mind responds to context. It is not fixed.
What it does not show is just as important. It was a short laboratory task. It measured how people categorize, not whether hostility, unfair treatment, or unequal institutions disappear. The researchers’ hope that racism might be erased went beyond what the experiment itself measured.
Studies of children point the same way. Infants as young as three months look longer at faces like those they see most often. That is familiarity, not hatred. Researchers Katherine Kinzler and Elizabeth Spelke found that 10-month-olds accepted toys equally from adults of different races, and that 2.5-year-olds shared toys equally. By age five, the White children in their study showed a preference for their own race. A large review of 113 research reports by Tobias Raabe and Andreas Beelmann found that prejudice tended to rise in early childhood, peak around ages five to seven, and ease somewhat after that. The amount varied with how much contact children had and with the status of the other group.
These are findings from particular tasks and particular groups of children, not a universal timetable. But they make one thing clear. Neither “children are born prejudiced” nor “children only learn it from their parents” is established. What children take in depends heavily on the world around them. That is where the opening is. For what happens when a child absorbs the wrong lessons, see The Enemy Within.
What Has Changed, and What Hasn’t
Here is where the picture gets honest, because two things are true at once.
Stated attitudes have changed dramatically. In 1958, Gallup found that 4 percent of Americans approved of marriage between Black and White people. In 2021, the number was 94 percent.
Measured bias has moved too, but more modestly. Tessa Charlesworth and Mahzarin Banaji analyzed about 4.4 million Implicit Association Tests taken by volunteers between 2007 and 2016. Over that decade, the average White-over-Black preference moved about 17 percent toward neutral, and the preference for lighter over darker skin moved about 15 percent. Those percentages describe change in an average score, not the share of people who stopped holding bias. Volunteers are also not a perfect picture of the country.
Behavior has moved the least. In 2017, Lincoln Quillian, Devah Pager, Ole Hexel, and Arnfinn Midtbøen combined 24 field experiments, 54,318 job applications in all, in which employers received matched applications that differed only in racial signals. White applicants received about 36 percent more callbacks than equally qualified Black applicants. The authors found no decline in that gap since 1989.
Put those side by side. Approval of interracial marriage went from 4 percent to 94 percent. The callback gap for equally qualified applicants did not shrink over the period studied.
What that suggests: attitudes and behavior did not move together on their own. Changing what people say they believe is one task. Changing what happens in a hiring decision, a classroom, or a hospital is another. Both can change. They just have to be worked on separately.
One more caution from the researchers. Even among the IAT studies, scores predict behavior only weakly in general. A large 2019 review of 217 reports, led by Benedek Kurdi, found a small unique relationship between implicit scores and behavior (β = .14). The strongest results, correlations of about .37, came from just 13 studies designed to match the test closely to the behavior. The weakest, about .02, came from six studies that did not. So an implicit score is not a verdict on anyone’s character, and it can’t be used to say what a person will do.
What Moves People
If attitudes and behavior can change, what does the research say moves them?
Real contact, under good conditions
(For the human side of this idea, see One Love.)
The best-known body of evidence is on contact between groups. In 2006, Thomas Pettigrew and Linda Tropp combined 515 studies, with 713 samples and about 250,000 people in 38 countries. A little over half of the samples involved racial or ethnic groups. They found that more contact went with less prejudice, with an average correlation of about −.21. That is a pattern, not a claim that contact cuts racism by 21 percent. The most rigorous studies in the set showed larger effects.
Contact tended to work better when people met as equals, worked toward shared goals, cooperated, and had the backing of an institution. But the review did not find those conditions to be required. Contact helped even when they were missing; the conditions made it stronger.
Now the honest critique. Elizabeth Levy Paluck, Seth Green, and Donald Green pointed out in 2019 that about 95 percent of the studies in that collection did not randomly assign who had contact. People who are already less prejudiced may simply seek out more contact. When the researchers gathered only the randomized studies, they found 27, and only 11 of those looked at race or ethnicity. None of those 11 studied anyone older than 25. Effects on racial and ethnic prejudice were weaker than effects on other kinds of prejudice.
The takeaway is not that contact doesn’t work. It is that the evidence for lasting change in adult racial attitudes is thinner than the overall trend suggests.
Arrangements, not just intentions
The strongest causal evidence tends to come from situations where an institution set up the contact. In a Norwegian field experiment by Henning Finseraas and Andreas Kotsadam, military recruits were randomly assigned to share rooms. Sharing a room with minority recruits improved views of immigrants’ work ethic. It did not change views on welfare policy or immigration overall. One stereotype shifted. The rest stayed.
In workplaces, a 2006 study in the American Sociological Review by Alexandra Kalev, Frank Dobbin, and Erin Kelly followed 708 private-sector workplaces from 1971 to 2002 to see what changed who ended up in management. Efforts to establish responsibility for diversity led to the broadest increases in managerial diversity. Diversity training and diversity evaluations were the least effective at increasing the share of women and Black men in management. That study was observational, not randomized, so it shows a pattern rather than proving a cause.
What has not lasted
Quick fixes have a weak record. Paluck and her colleagues reviewed 418 experiments in 2021 and found that 76 percent tested “light-touch” interventions. Publication bias was a concern, and few studies measured whether anything lasted or changed behavior. That review did not prove that everything fades. It found that for much of this work, nobody had checked.
Where researchers did check, the results were sobering. Calvin Lai and colleagues tested nine ways of reducing implicit racial preference in two studies with 6,321 people. All nine worked right away. None worked after a delay of a few hours to a few days. A 2019 review of 492 studies by Patrick Forscher and colleagues found that implicit scores can change but that behavior mostly did not follow. A randomized study of online diversity training among 3,016 employees found some attitude improvements and mixed behavior results, mostly measured on gender.
What this says: a one-time workshop is a weak tool. That is not a reason to give up. It is a reason to build things that last, such as repeated contact, shared goals, and structures that hold people accountable for results.
What Allies Can Do
Many people ask what they can do when they hear something wrong. The research here is real but small.
In 2006, Alexander Czopp, Margo Monteith, and Aimee Mark ran three experiments in which people made a stereotyped remark about Black people and were then challenged. Those who were confronted gave fewer stereotyped responses afterward. This held even when the confrontation was sharp, and even when the person confronting was Black. The cost was social: people disliked whoever challenged them, especially when the challenge was hostile.
So being disliked in the moment does not mean the correction failed. And hostility is not required for it to work. The study did not test which approach is best, and the effects were measured shortly after, in a lab, not across years in a workplace. A reasonable reading is to name the specific remark or decision, explain the concern, and leave the person room to do better. That is a sensible inference, not a tested script.
Two more points. First, a 2024 scoping review by Sally Grapin and Peter Goldie found only 16 empirical studies on performative allyship, which tells you how young this field is. The research cannot yet rank ally behaviors or predict long-term effects. What it does point toward is follow-through on specific, consequential things, not symbolism alone.
Second, nothing in this research says the person on the receiving end of prejudice owes anyone a confrontation. If you are the one who was targeted, protecting yourself is a valid choice. The work of correction belongs to those with the standing and the safety to do it.
“Reverse Racism”
This phrase carries a lot of heat. It helps to slow down and ask what people mean by it, because it is used in at least three different ways.
Meaning one: individual prejudice. Here, racism means treating someone badly, or believing they’re inferior, because of their race, and it can come from anyone toward anyone. The strongest case for this view is that a person’s dignity doesn’t depend on which group they belong to. Being mistreated for your race is wrong whoever you are.
Meaning two: prejudice plus institutional power. Here, racism refers to patterns that are built into institutions and carried across generations, as distinct from individual attitudes. The strongest case for this view is that a hurtful remark and a lasting pattern of exclusion are different things, and need different answers. Matthew Clair and Jeffrey Denis’s 2015 sociology review describes both approaches and notes that this framework is influential but is not the only definition scholars use.
Meaning three: the claim that the direction has flipped. This is the empirical claim, that discrimination against White people now outweighs discrimination against others. This is a question that evidence can speak to.
What does the evidence say? On the legal side, the Supreme Court held in McDonald v. Santa Fe Trail Transportation Co. (1976) that civil-rights law protects White employees on the same terms as everyone else. In that case, White employees alleged they were fired for misconduct while a Black employee involved in the same incident kept the job. The Court said their claim could go forward. So an individual White person who is discriminated against has legal protection, and being protected does not depend on whether discrimination is common.
On the research side, the matched-application studies described earlier measured a specific, ongoing gap that favors White applicants over Black applicants, with no decline since 1989. Those studies measured callbacks for particular jobs, not every kind of prejudice anywhere.
There is also research on perception. In 2011, Michael Norton and Samuel Sommers asked Black and White Americans to rate how much discrimination each group faced in every decade from the 1950s to the 2000s. White respondents on average saw less bias against Black people as going hand in hand with more bias against White people, and rated anti-White bias as larger by the 2000s. Black respondents did not show that pattern. This measured how people remembered and judged the past, not what actually happened in each decade. And it does not describe every White person.
Put together: individual prejudice against anyone is real and wrong, and the law protects against it. A documented, lasting hiring gap against Black applicants is also real. Neither fact cancels the other. Keeping the two apart is what lets you talk about both honestly. If you want to be precise, say what you mean: individual prejudice, discrimination in a particular decision, or a pattern in an institution. Those are different claims, and each deserves its own evidence.
Schools, Briefly
One place where these questions land hard is school discipline, and the wider picture is covered in The Weight That Was Never an Accident and Understanding the Mental Health Access Gap. Two points belong here.
The federal government’s Civil Rights Data Collection for the 2021–22 school year reported that Black children were 18 percent of public preschool enrollment, but 38 percent of the preschool children suspended and 33 percent of those expelled. The report itself says Black preschool children were suspended at a rate over twice their enrollment. The counts are small: nearly 2,700 preschool children suspended and about 370 expelled, out of 1.4 million enrolled. And the data shows who was disciplined, not why.
For the why, look to experiments. In 2015, Jason Okonofua and Jennifer Eberhardt gave teachers records of a student’s minor misbehavior and randomly changed whether the student’s name suggested a Black or White child. The differences showed up most clearly after a second incident, when teachers were more likely to see the Black-named student as a troublemaker with a pattern, and to favor harsher discipline. Those were hypothetical records, not real suspensions, and the study cannot say how much of the national gap this explains. But it shows a way that small judgments can build on each other.
Separately, Seth Gershenson and colleagues, using national data in which two teachers rated the same student, found that non-Black teachers expected less educational attainment from Black students than Black teachers did. Teachers were not randomly assigned, so this is a pattern and not proof of a cause.
What We Want You to Know
Can racism change? The honest answer is yes, in specific ways, and no, not automatically.
Public attitudes have changed enormously. Some measures of bias have moved. Some behavior, like who gets a callback, has barely moved at all. Quick fixes fade. Structured contact, shared goals, and accountability seem to matter more, though the strongest evidence is still being gathered.
You don’t have to choose between “nothing ever changes” and “it’s all fixed.” Both are easier than the truth, and neither helps you.
Here is what the evidence supports: how people sort each other is flexible. Institutions can be built to bring people together under fair conditions. Individual people can be asked to do better and sometimes do. And none of it happens on its own. It happens when people build it.
That is a different kind of hope than comfort. It is the kind that asks something of us. It is also the kind that holds up when you look closely.
You do not have to carry this by yourself, and you do not have to wait for the world to finish changing to take care of your own mind. Understanding what is known, and what isn’t, is one way of standing on solid ground.
If You Need Support Right Now
- 988 Suicide & Crisis Lifeline: Call or text 988 — 24/7
- Crisis Text Line: Text HOME to 741741
- 211: Dial 2-1-1 for local mental health and community resources
- SAMHSA National Helpline: 1-800-662-4357 — free and confidential
- The Steve Fund (young people of color): Text STEVE to 741741
LEGH.org — Love Enabled Growth & Hope. For the people who deserve reliable resources — and have always deserved better. No appointment. No insurance. No gatekeeping. Just reach out.