Psychological Priming is Real. Some of the Studies are Bogus.

Correction, August 2026: This piece originally ran in October 2024 under the title "The Power of Psychological Priming." It cited a quote attributed to Daniel Kahneman that I have since been unable to source anywhere, along with several studies that have not survived independent replication. It has been rewritten rather than patched. The original argument, that environment shapes behavior below awareness, still holds.

Thirty-four college students.

That is the entire foundation of one of the most repeated ideas in marketing psychology. Not thirty-four per group. Thirty-four total, split three ways, about eleven people per condition.

In 1996, John Bargh and his colleagues at NYU had students unscramble sentences. One group got words associated with rudeness. Another got words associated with politeness. A third got neutral words. Then each student walked down the hall to find the experimenter deep in conversation with someone else, and the researchers timed how long it took them to interrupt.

Sixty-seven percent of the rude-primed students broke in. Only sixteen percent of the polite-primed ones did.

I have told that story in print. So has almost everyone who writes about consumer behavior. It is a wonderful finding, and it launched a thousand keynote slides about the hidden power of suggestion.

Two things about it are worth knowing.

The first is that the number in the middle usually gets left out. The neutral group interrupted thirty-eight percent of the time, which sits much closer to the rude group than to the polite one. Read the full set and the story shifts. Rude words did not make people rude. Polite words made people patient.

The second is that in thirty years, no independent lab has ever published a replication of that experiment.

Not a failed one.

Not any.

Then the field checked its own work

The rudeness study was not a one-off. It belongs to a research area called social priming, and social priming has had a rough decade. In 2022, a group of researchers gathered up seventy attempts to repeat forty-nine social priming experiments and looked at how they went. When one of the original authors helped run the do-over, it worked about two thirds of the time. When an independent lab ran it alone, there were fifty-two attempts.

Zero worked.

Not one.

Bargh's other famous experiment got the same treatment. That is the one where students who unscrambled sentences about aging then walked more slowly down the hallway. In 2012, a Belgian team ran it again with 120 people, using infrared timing gates instead of a stopwatch. Primed group: 6.27 seconds. Control group: 6.39 seconds. Nothing.

Then they did something clever. They ran it a third time. Same setup, with one change. They told half their research assistants to expect slow walking and the other half to expect fast walking. The students were told nothing different at all. The famous result came back. But only for the assistants who had been told to expect it.

The words were never slowing anyone down. The people holding the stopwatch were.

There is one more piece, and it is the one that should have been caught in 1996. Do the math on the original and it was almost built to mislead. A study with thirty people has roughly a one in five chance of catching an effect that size, even when the effect is completely real. Four times out of five, it comes back empty. Bargh's team reported a hit twice in a row. Those are about one in twenty-five odds.

Psychological Priming is Real. The Landmark Studies Aren't

When I first wrote about priming, I quoted Daniel Kahneman saying that an effective prime must be "strong enough to impact behaviour, but not so strong that it enters conscious thought."

I cannot find that quote anywhere. Not in Thinking, Fast and Slow, not in any interview, not in any collection of his work. Searching the exact phrase returns a marketing blog, and then my own article. Somewhere along the line a sentence acquired a Nobel laureate's name and started traveling.

What makes this worse is that Kahneman did say something about priming, publicly, and it is the opposite. In 2017, responding to a statistical critique of his own book, he wrote:

"What the blog gets absolutely right is that I placed too much faith in underpowered studies. Clearly, the experimental evidence for the ideas I presented in that chapter was significantly weaker than I believed when I wrote it. This was simply an error. I have changed my views about the size of behavioral priming effects. They cannot be as large and as robust as my chapter suggested."

The man whose name I attached to a sentence endorsing priming had already published a retraction of the priming chapter of his own book. That is a real quote, it is checkable, and it is far more interesting than the one I used.

How a Number Becomes a Fact

Here is one worth watching in real time. You have probably read that playing classical music in transit stations cuts crime. The figures usually quoted are a 75% drop in calls for emergency service, a 50% drop in vandalism, and a nearly 20% drop in crime. I have quoted them myself.

Trace them back and they end at one LA Metro spokesperson, talking to local news in March 2023, about one station. Westlake/MacArthur Park. There is no study behind them. No report, no board document, no stated baseline period, nothing that says what the numbers were measured against.

There are three problems with them, in ascending order of seriousness. The claim always says stations. It was one station. The music was not classical, at least not at first. The pilot opened with a four minute dark ambient string loop by the composer Adrián Berenguer, played on repeat. Nobody asked him. "I have not given any permission for my music to be used for these purposes," he said publicly. Metro switched to a royalty-free Bach and Vivaldi playlist after complaints.

The music was not the only thing that changed. Metro also installed floodlights at both ends of the platform, closed a secondary entrance, added cameras, brought in transit ambassadors, increased police presence, and added cleaning staff. Lighting is the one design factor with definitive evidence in the transit crime literature. Crediting the music means ignoring five other interventions that ran at the same time.

There is a fourth problem, and it has nothing to do with Los Angeles. Calls for emergency service are not crime. PolitiFact ran into a nearly identical claim in 2011 about Portland's light rail, where a reported 41% crime drop turned out to be a drop in calls for service, which police noted could mean anything from disorderly conduct to a fare check. The definitive review of transit crime research, covering fifty years of studies, does not list music as an intervention at all.

None of this means Metro was lying. The station probably did get quieter. It just was never a measurement of music, it was never a study, and somewhere between a spokesperson's quote and a few hundred articles it became one.

So Does Environment Shape Behavior or Not?

Yes. Emphatically yes. The idea was never the problem. The 1990s laboratory demonstrations of it were the problem, and the good evidence now looks completely different.

Here is what a defensible version looks like.

In 2024, a team led by Paul Lohmann ran a preregistered experiment with 4,008 adults on a simulated food delivery app, with real money on the line. They tested three ways to shift people toward lower-carbon meals. A ten percent tax on meat dishes. Carbon labels on every item. And simply reordering the list so lower-carbon options appeared first.

The tax did nothing. The labels did nothing.

Reordering cut basket emissions by twelve percent, dropped calories by fifty-five, cost essentially nothing to implement, and left diners more satisfied with their choice than the control group.

Two of those three interventions asked people to think. One of them just changed what they saw first. Only one worked.

That pattern holds across the field. Romain Cadario and Pierre Chandon pooled 299 effect sizes from ninety-six real-world field experiments and sorted every intervention into three families. Behavioral nudges, meaning portion size, placement, and convenience, landed at 0.39. Affective nudges, meaning appeals to taste and pleasure, landed at 0.24. Cognitive nudges, meaning information and labels, landed at 0.12.

Change what people do. Then change how they feel. Changing what they know comes last, and it comes a distant last.

What this Means if You Run a Restaurant

The environment is doing enormous work on your guests. Just not through the mechanisms the 1990s promised.

Position beats persuasion. Items at the first or last spot in a menu category take fifty-five to fifty-six percent of orders, against forty-five percent for everything in the middle. That has held in a controlled lab study and in a real Tel Aviv cafe. A randomized trial across 7,968 lunch purchases moved actual behavior by 5.6 percentage points using nothing but menu order. Single digits, for free.

Defaults beat encouragement. A sixteen-week experiment across 4,099 items sold made vegetarian dishes the visible default, with meat still available on request. Vegetarian sales went from 641 to 1,325. Meat fell from 604 to 168. The effect was still there at a ten-week follow-up, and nobody got annoyed.

Tempo is real, and it lands where you might not expect. Ronald Milliman's 1986 study of 1,392 dining parties found slow music lengthened meals and raised bar spend. Not food sales. Bar spend.

And a hungry guest is not reading what you think. A 2025 study using eye tracking and computational modeling found that hunger did not just make people care less about nutritional information. It made them discount that information even while looking directly at it. Your health messaging is being read by a brain actively down-weighting it.

The One Lab Study That Survived, and Why

Not everything from that era fell apart, and one survivor teaches a better lesson than the studies it replaced. In 2017, Yann Cornil, Pierre Chandon and Aradhna Krishna gave 154 young men in Paris an identical cocktail. Same vodka, same Red Bull, same fruit nectar, same alcohol content. The only thing that changed was what the drink was called.

Men handed a "vodka-Red Bull cocktail" reported feeling meaningfully more intoxicated than men handed the identical drink called an "exotic fruits cocktail." They took modestly more risk on a lab gambling task and rated themselves more likely to approach women in photographs.

But here is the finding that makes the study worth keeping. The effect appeared almost entirely among men who already believed energy drinks amplify alcohol. Among men who did not hold that belief, there was no effect whatsoever.

The label did not create the expectation. It activated one that was already there.

That is the honest version of priming, and it is more useful than the magic version. You are not installing beliefs in your guests. You are triggering the ones they walked in with. Which means the work is knowing what they already believe, not finding clever words to slip past their defenses.

Oh, and one more thing from that study, which nobody quotes... the men who got the Red Bull label also said they would wait about fourteen minutes longer before driving. Feeling drunker made them more careful, not less.

Primed for Power

A Boston College study put players behind the wheel of identical race cars in a video game. The only thing that changed was the livery. One car carried Red Bull's logo and colors. The others carried Guinness, Tropicana, and Coca-Cola.

Players driving the Red Bull car drove the way the brand sells itself: fast, aggressive, willing to take a risk. What makes the finding worth repeating is where that led. Researcher S. Adam Brasel described it this way. They "either do very, very well or they push themselves too far and crash." Very little middle ground. And none of them connected any of it to the logo on the hood.

*One caveat worth stating. This is a single 2011 experiment, and no independent lab has replicated it.

The Red Bull effect shows advertising and marketing can push beyond the sale. They can influence behavior in a way that consumers aren't even aware of.

Primed for Honesty

Because priming works through associations you already hold, almost any stimulus can trigger it. Images, words, scent, light, sound, temperature, touch. There is a wealth of research here. It is also uneven, so it is worth separating what survives scrutiny from what does not.

Logos. In a 2008 study, students briefly exposed to Apple logos generated more uses for a brick, and more creative ones, than students exposed to IBM logos. The detail that usually goes missing is that the experiment had no control group, which the researchers pointed out themselves. When they ran a follow-up that added one, Apple beat IBM clearly but only barely edged out a no-logo control. So the finding is less "Apple makes you creative" and more "Apple and IBM pull in opposite directions from roughly the same starting point." Interesting, and narrower than it usually sounds.

Scent. Of all the senses, smell has the most direct anatomical line to the brain's emotion centers. Three subregions of the amygdala receive single-synapse input from the olfactory bulb, which makes the amygdala part of primary olfactory cortex. No other sense is wired that way, and olfactory cortex has stronger connectivity to the hippocampus than vision, hearing, or touch do.

When a smell does trigger a memory, that memory tends to be older, often from before the age of ten, and to feel more transporting than one recalled through words or pictures. What smell does not do is give you more memories, or more accurate ones. Real estate agents worked out the emotional half long before anyone measured it. The apple pie in the oven is folklore, but it is folklore with a real mechanism underneath it.

The evidence that scent changes behavior is thinner than its reputation. In a 2005 study, people eating a crumbly biscuit in a room with a hidden citrus cleaning scent tidied up after themselves more than people in an unscented room. That experiment had twenty-two participants. A separate field study in 2012 found less litter in citrus-scented train cars.

Both are worth knowing, and both come with the same caveat: when the original research group went back and added proper odor controls, including scents that were pleasant but had nothing to do with cleaning, the effect tracked how pleasant the smell was rather than what it meant. Their conclusion was blunt. Odors do not make for good semantic primes.

Malcolm Gladwell, author of the bestselling book Blink: The Power of Thinking Without Thinking, describes priming this way:

What we think of as freewill is largely an illusion: much of the time, we are simply operating on automatic pilot, and the way we think and act—and how well we think and act on the spur of the moment—are a lot more susceptible to outside influences than we realize.

---