The Sandwich Board

In the mid-1970s, a Stanford undergraduate showed up for what she thought was a study on communication. A researcher met her, explained that the study was about “unusual methods of communication,” and made a request: would she walk around campus for thirty minutes wearing a large sandwich board that read “EAT AT JOE’S”? She would have to wear it through the busiest parts of campus, in plain view of everyone she passed. Then she would report back on what she had observed.

The researcher made clear that the choice was hers. She could decline and a different task would be assigned instead. About half the students who got this request said yes. About half said no.

But the sandwich board was not the point. The point came next. After each student decided, the researcher asked a second question: of your fellow students, what percentage do you think would also agree to wear the board?

The students who had agreed to wear it estimated that roughly 62 percent of their peers would also agree. The students who had refused estimated that only about 33 percent would agree — meaning two-thirds of their peers, like them, would refuse.

Read that again. The two groups were drawn from the same student body, asked the same question, in the same setting. Yet each group estimated that its own choice was the majority choice. The yes-sayers saw a campus full of yes-sayers. The refusers saw a campus full of refusers. Both could not be right, and the gap between them — nearly thirty percentage points — was not noise. It was a systematic, predictable distortion in how people read the social world.

That distortion is the false consensus effect: the tendency to overestimate how much other people share our beliefs, preferences, judgments, and behaviors. Lee Ross, David Greene, and Pamela House named it and demonstrated it in 1977, and unlike a great many findings from that era of social psychology, it did not crumble when the field’s replication crisis arrived. It is, by any reasonable standard, one of the most robust effects the discipline has ever produced.

It is also one of the most expensive to ignore — because every founder, marketer, pollster, and product manager who has ever said “everyone wants this” was, in that moment, a participant in the sandwich-board study.

What Ross, Greene, and House Actually Did

The 1977 paper in the Journal of Experimental Social Psychology was not one study. It was four, deliberately stacked to rule out easy objections.

The first two were paper-and-pencil studies. Ross and his colleagues handed Stanford undergraduates short vignettes describing everyday dilemmas — a “Supermarket Story” in which a shopper is asked to sign a release allowing a recorded comment to be used in a television commercial, a “Term Paper Story,” and others. For each vignette, participants did three things: they said what they personally would do, they estimated what percentage of their peers would make each choice, and they rated the personality traits of the hypothetical people who chose each way.

The pattern was clean. Whatever a participant chose, they estimated that choice to be more common among their peers than did the participants who chose the other way. If you said you’d sign the release, you thought most people would sign. If you said you’d refuse, you thought most people would refuse. The self served as a silent anchor for the estimate of everyone else.

The third and fourth studies took the effect out of the questionnaire and into the world. This is where the sandwich board lived. Real students faced a real, slightly mortifying choice, made it, and then estimated how their peers would behave. The field version produced the famous numbers — the roughly 62 percent versus 33 percent split — and it mattered because it killed the most obvious counter-explanation. You could argue that paper vignettes are abstract and people guess lazily. You cannot argue that about a student who just had to decide whether to actually wear an embarrassing sign in front of everyone they know.

There was a second finding folded into the same studies, and it is the one strategists most often miss. Ross and colleagues also measured how participants explained other people’s choices. When someone made the same choice as you, you tended to see that choice as the natural, situationally-determined response — “of course, anyone would do that.” When someone made the opposite choice, you tended to attribute it to their personality — they’re the kind of person who would do that. In the language of attribution theory, your own behavior reads as a response to the situation; deviations from your behavior read as evidence of someone’s character.

That second finding is why the paper’s subtitle is “an egocentric bias in social perception and attribution processes.” The false consensus effect is not just a counting error. It is a two-part machine: you overestimate how many people are like you, and then you treat the people who aren’t like you as somehow defective, idiosyncratic, or revealing of a flawed disposition. The salesperson who can’t understand why a prospect didn’t buy (“they just don’t get it”), the founder baffled that users churned (“those weren’t our real users”), the partisan who concludes the other side must be stupid or malicious — all of them are running the attribution half of the 1977 finding.

The Meta-Analysis That Made It Stick

A single clever study, even a four-study package, is not proof of a durable effect. The history of social psychology is a graveyard of clever studies that did not replicate. So the relevant question is: did the false consensus effect hold up when it was tested again and again, by other labs, in other domains?

It did. In 1985, Brian Mullen and colleagues published a meta-analysis of 115 hypothesis tests of the false consensus effect in the same journal. Pooling across all those tests, they found an average effect corresponding to a correlation of roughly r ≈ 0.30 — a robust effect by the standards of social psychology, and one that appeared with remarkable consistency.

The crucial word is consistency. The false consensus effect was not a fragile lab curiosity that only worked on Stanford sophomores under one specific procedure. Mullen and colleagues found it across student and non-student samples, across attitudes, behaviors, and personal traits, across studies that asked people once and studies that asked them repeatedly. It was the kind of finding that shows up no matter how you slice it — which is exactly the profile of a real cognitive regularity rather than a procedural artifact.

This is what separates the false consensus effect from the ego depletion and power posing findings that the replication crisis dismantled. Those effects depended on specific paradigms, showed shrinking effect sizes under scrutiny, and in some cases vanished entirely in large pre-registered replications. The false consensus effect, by contrast, sits in the same robust company as confirmation bias, hindsight bias, and the spotlight effect — biases that have been measured so many times, in so many ways, that doubting their existence is no longer a defensible scientific position. The honest critique of these effects is not “do they exist” but “how large are they, when, and what causes them.” That is the mark of a mature, well-replicated finding.

Why It Happens: The Four Mechanisms

In 1987, Gary Marks and Norman Miller published an empirical and theoretical review in Psychological Bulletin summarizing the first decade of false-consensus research. Their most useful contribution was to organize the competing explanations into four mechanisms — and the important thing for a practitioner is that these are not rival theories where one wins. They operate together, and each one maps onto a different failure mode you can actually design around.

1. Selective exposure. We do not sample the population at random. We surround ourselves with people who resemble us — same neighborhoods, same professions, same income brackets, same political tribes, same group chats. Your actual social environment really is more like you than the general population is. So when you generalize from “the people I know” to “people,” you are extrapolating from a biased sample, and the bias points exactly toward your own position. This is the mechanism most relevant to founders and product teams, who are typically surrounded by other early adopters who look nothing like the mainstream market.

2. Salience and focus of attention. When you take a position, that position becomes vivid and front-of-mind. The reasons for your choice are concrete and available; the reasons someone might choose differently are abstract and faint. Because your own view dominates your attention, it dominates your estimate of what’s typical. This is the same cognitive-availability machinery behind the availability heuristic: what comes easily to mind feels more frequent than it is.

3. Logical information processing (rational construal). Some of the effect is, ironically, defensible reasoning gone slightly wrong. If you genuinely believe a situation calls for response X, it is not crazy to assume that other reasonable people, seeing the same situation, will also respond with X. The problem is that you and they are not seeing the same situation — you are each construing it through your own values, experiences, and priorities, and you systematically underestimate how differently other people construe the same facts. You think you’re reasoning about the world; you’re actually reasoning about your version of it.

4. Motivation. Believing that others agree with us is reassuring. It validates our choices, protects self-esteem, and reduces the discomfort of standing alone. Consensus, even imagined consensus, feels like being right. The motivational account explains why the effect tends to be stronger for choices we care about and for positions tied to our identity — the places where being in the majority matters most to us emotionally.

Marks and Miller’s review is also where the honest scientist earns their keep: they noted that the effect varies in size by domain and method, that some of the early estimates were inflated by weak measurement, and that “false consensus” is partly a misnomer because some projection from self to others is statistically rational. None of that overturns the finding. It refines it. A bias can be real, robust, and still smaller and more conditional than its first headline suggested — and saying so is how you tell a survivor of the replication crisis from a casualty.

Where It Costs Real Money

The false consensus effect is not an academic curiosity. It is a load-bearing error in how organizations make decisions, and it shows up in at least four places I see repeatedly.

Market sizing built on intuition. The most common version of this mistake begins with a sentence that sounds like insight: “Everyone has this problem.” The founder has the problem. Their friends have the problem. Their early users — recruited from their own network — have the problem. So the addressable market feels enormous. But “everyone I’ve talked to” is the selective-exposure sample, and it is the single least reliable input to a market size you can possibly use. The general population is not your network. Total addressable market estimates anchored on founder intuition are false consensus with a spreadsheet attached. The discipline of base-rate thinking — what fraction of the actual population behaves this way, not the fraction of people who already orbit you — is the corrective.

Product teams who assume users think like them. Product designers are not normal users. They are technically sophisticated, deeply familiar with the product, and emotionally invested in their own design decisions. When a designer says “users will obviously understand this flow,” they are reporting their own salience-distorted intuition, not a fact about users. The entire discipline of usability testing exists to break the false consensus of the builder. Every time you watch a real user fumble through something your team thought was self-evident, you are watching the 1977 effect in real time — and the attribution half kicks in immediately (“that user just wasn’t paying attention”) unless you discipline yourself against it.

Political and polling misreads. The classic symptom is the post-election sentence: “I can’t believe X won — I don’t know a single person who voted for them.” That is not evidence about the electorate. It is a confession about the speaker’s social bubble. Pollsters and forecasters fight false consensus structurally, by sampling representatively and weighting demographics, precisely because individual intuition about “what people think” is so reliably wrong. When pundits are blindsided by a result, the failure is rarely the data — it’s that they substituted their own milieu for a representative sample.

Hiring and culture. Interviewers overestimate how widely shared their own values, working styles, and standards are, then read deviation as a character flaw rather than a different-but-valid approach — the attribution machine again. “Culture fit” assessments are especially vulnerable: they often measure similarity to the interviewer rather than fit to the role, and the false consensus effect makes that substitution feel like objective judgment.

What unites all four is the same structure as the sandwich board: a person generalizes from a non-representative sample (themselves and their immediate surroundings) to a population, and then treats the people who don’t fit as anomalies. The error feels like knowledge. That is what makes it dangerous.

The Antidotes

You cannot introspect your way out of the false consensus effect, because the bias operates on introspection — the more vividly you consult your own view, the more the distortion grows. The corrections that work are structural. They replace your intuition about “what people think” with something external to your head.

Use real base rates, not felt ones. Before betting on “everyone wants this,” find the actual number. How large is the population that exhibits this behavior, according to data collected from people who are not in your network? If you can’t find that number, treat your market estimate as a hypothesis, not a fact. The discipline here is identical to the one that defeats base-rate neglect: start from the population frequency, then adjust, rather than starting from your vivid case.

Sample diversely and deliberately. Selective exposure is a sampling problem, so fix the sampling. Talk to users who are not early adopters, not in your industry, not in your demographic. Recruit interview subjects who you expect to disagree. The goal is not balance for its own sake — it’s to puncture the illusion that your environment is the world.

Run premortems. Before committing to a decision built on “people will obviously respond well to this,” run a premortem: imagine it’s a year from now and the bet failed. Why? Forcing yourself to generate reasons for the opposite outcome counteracts the salience mechanism, because it manufactures availability for the views you’d otherwise underweight. This is the same logic behind the outside view that makes forecasters more accurate than pundits.

Separate the choice from the attribution. When someone disagrees with you — a user who churned, a prospect who said no, a colleague who pushed back — resist the reflex to explain it with their character. Ask instead: what about their situation, construed through their values, makes their choice reasonable? That single question disarms the attribution half of the false consensus effect, and it is the difference between learning from disagreement and dismissing it.

Aggregate independent judgments. When you need a forecast, don’t poll your own intuition or the consensus of a like-minded room. Collect independent estimates from diverse, non-colluding sources. This is precisely the mechanism behind the wisdom of crowds: independence and diversity cancel out individual biases, including this one. A room full of people who all share your bubble does not produce a crowd’s wisdom — it produces an echo, amplified.

The Strategist’s Takeaway

The false consensus effect is one of the cleanest demonstrations in psychology that the inside of your own head is a terrible instrument for measuring other people. Your sense of what’s normal, popular, obvious, or widely-shared is not a window onto the population. It is a mirror, lightly fogged, reflecting yourself and the people you happen to stand near.

For a strategist, the practical content is brutal and simple: whenever a decision rests on an unexamined “everyone wants / thinks / does this,” you are almost certainly looking in the mirror and calling it a window. The bigger the decision and the more confident the intuition, the more urgent it is to go get the external number — the base rate, the representative sample, the independent forecast, the user who isn’t like you.

And here is the part worth sitting with, because it separates this finding from the discarded ones: the false consensus effect is robust. It is not power posing or ego depletion, fragile artifacts that dissolved under replication. It survived the crisis with its effect size intact across 115 tests and four decades. So you don’t get to wave it away as “just another psychology study that didn’t hold up.” It held up. It is operating on you right now, on the very question you are most confident about — and the only reliable defense is to stop trusting the mirror and go count the world directly.

Sources

  • Ross, L., Greene, D., & House, P. (1977). The “false consensus effect”: An egocentric bias in social perception and attribution processes. Journal of Experimental Social Psychology, 13(3), 279–301. https://doi.org/10.1016/0022-1031(77)90049-X
  • Mullen, B., Atkins, J. L., Champion, D. S., Edwards, C., Hardy, D., Story, J. E., & Vanderklok, M. (1985). The false consensus effect: A meta-analysis of 115 hypothesis tests. Journal of Experimental Social Psychology, 21(3), 262–283. https://doi.org/10.1016/0022-1031(85)90020-4
  • Marks, G., & Miller, N. (1987). Ten years of research on the false-consensus effect: An empirical and theoretical review. Psychological Bulletin, 102(1), 72–90. https://doi.org/10.1037/0033-2909.102.1.72

Frequently Asked Questions

What is the false consensus effect in one sentence? It is the well-documented tendency to overestimate how many other people share your own beliefs, preferences, judgments, and behaviors — to assume your choice is more common than it actually is.

What was the “Eat at Joe’s” sandwich-board study? In Ross, Greene & House (1977), Stanford students were asked to walk around campus wearing a sandwich board reading “EAT AT JOE’S.” Those who agreed estimated about 62 percent of peers would also agree; those who refused estimated only about 33 percent would agree. Each group assumed its own choice was the majority — the core demonstration of the effect.

Did the false consensus effect survive the replication crisis? Yes. Mullen and colleagues’ 1985 meta-analysis of 115 hypothesis tests found a robust effect (about r ≈ 0.30) that held across populations, domains, and methods. It belongs to the well-replicated category of biases, not the fragile findings that collapsed under scrutiny.

Why does the false consensus effect happen? Marks and Miller (1987) identified four contributing mechanisms: selective exposure (we surround ourselves with similar people), salience (our own view dominates our attention), rational construal (we assume others see situations the way we do), and motivation (imagined agreement validates our choices). They operate together rather than competing.

How does it hurt business and strategy decisions? It inflates market-size estimates built on “everyone wants this,” misleads product teams into assuming users think like them, blindsides political and polling forecasts anchored to a social bubble, and biases hiring toward people who resemble the interviewer. In each case, an unrepresentative sample (yourself and your surroundings) gets mistaken for the whole population.

What’s the most reliable antidote? Replace intuition about “what people think” with something external to your head: real base rates from representative data, deliberately diverse sampling, premortems that force you to imagine the opposite outcome, and aggregated independent judgments. You cannot introspect your way out of the bias, because it operates on introspection itself.

replication-crisis false-consensus-effect social-perception cognitive-bias evidence-evaluation
Share this article
LinkedIn (opens in new tab) X / Twitter (opens in new tab)
Atticus Li

Experimentation and growth leader. CXL-certified CRO practitioner, Mindworx-certified behavioral economist (1 of ~1,000 worldwide). 200+ A/B tests across energy, SaaS, fintech, e-commerce, and marketplace verticals.

About LinkedIn Newsletter