People Like You More Than You Think. Five Studies Found the Gap Lasts for Months, Even With Your Roommate

The Conversation After the Conversation

You walk out of a conversation and the replay starts: the joke that landed flat, the pause that stretched too long, the story that sounded like bragging, and most of us leave quietly certain the other person noticed every stumble and liked us a little less. The evidence says the replay is fiction, because in 2018, a Yale team led by Erica Boothby ran five studies on what people believe others think of them after a conversation, and the pattern never wavered: people systematically underestimated how much their partners liked them and enjoyed their company. They called it the liking gap. It showed up in the lab, in a workshop of strangers, and in a dormitory over a full school year.

Five Minutes, Five Studies, One Persistent Illusion

The cleanest demonstration was also the simplest: thirty-six strangers from the Yale community were paired into eighteen dyads, talked for five minutes, then rated how much they liked their partner and guessed the reverse. Average actual liking was 5.82 out of 7 against an average guess of 5.17, and that 0.65-point shortfall (95% CI 0.42 to 0.87, p < .001) was roughly medium-sized, pointing the same direction in nearly every pair.

The team then stress-tested the finding from every angle. Participants who listed their thoughts during the conversation revealed the mechanism plainly: the harsher the self-critical replay, the bigger the gap grew. Independent observers watching the recordings could see the liking in the footage, plain as daylight, while the participants, stuck inside their own heads, could not. The gap survived conversations of different lengths, and survived the lab entirely at a personal-development workshop with members of the British public. Then came the study that turned a curiosity into something stranger: 102 first-year students tracked from September to May, rating their suite-mates and guessing the reverse, present in the fall, still present in February, only beginning to close around May. A bias that survives months of daily evidence is not a misread; it is a design feature of the social brain.

The Arithmetic Nobody Did

That persistence is where the paper gets unsettling, and where nobody has run the numbers, so consider what "the gap survived until May" means in units of evidence. Conservative read: a suite-mate you chat with three times a week, over thirty weeks, hands you ninety data points about whether they enjoy your company. Ninety verdicts. The brain barely moved.

The authors' proposed mechanism is that self-critical thoughts hog the spotlight: you remember your awkwardness in high definition and their warmth in a blur, so your model of their opinion is built mostly from your own noise. The persistence data let us quantify that claim. If dozens or hundreds of warm interactions cannot close a 0.65-point miscalibration, then the brain is weighting its inner monologue tens or hundreds of times more heavily than the other person's actual behavior, which means the liking gap is not a lack of evidence but evidence being systematically outvoted. That ratio has never been computed in the literature. It should worry anyone who trusts their social instincts.

The Strongest Counterargument

The fiercest challenge to this story comes from the people who replicated it. Jiamin Li, Zhenchao Zhong, and Lei Mo ran an independent replication in 2020 and found the gap again, larger this time at d = 0.83, but they also found it in participants who reported no negative self-thoughts at all, which cuts directly against the paper's headline mechanism. They argue the original paradigm is conceptually unclear, and the gap may reflect a general psychological defense, a low-expectations posture in first contact, rather than self-criticism. A replication that confirms your effect while rejecting your explanation is the most uncomfortable kind of confirmation, and the follow-up came from inside the liking-gap research family itself: in 2025, Hasagani Tissera and colleagues, including Boothby et al. co-author Gus Cooney, ran a preregistered study of 2,753 people that confirmed the gap while arguing the classic measure conflates two phenomena: the actual gap (how much others like you versus how much you think they do) and the perceived gap (how much you like them versus how much you think they like you). Only the perceived gap tracked insecurity: lower self-esteem, higher social anxiety, more neuroticism. The descriptive finding stands. The neat single story about why it happens does not. The original also carries the caveats of its era: no preregistration, a flagship study that stopped collecting data when the term ended, and a first experiment run on thirty-six people, which is why the replications, not the original alone, are the reason this article exists, even as the mechanism is still being negotiated seven years on.

What We Didn’t Prove

This paper does not prove everyone likes you. It proves a directional bias in one specific judgment: after getting-acquainted conversations, estimates of the other person's liking run roughly half to two-thirds of a scale point below reality. It does not cover deep friendships, marriages, or job interviews; the lab chats lasted five minutes, the dorm sample was first-years at one university, and the ratings were self-reports. The 2025 refinement adds a wrinkle: some of the "liking gap" may be people projecting their own warmth, a different bias with a different fix. None of this says social anxiety is irrational: in related work, the more socially anxious were liked slightly less by strangers, a real cost the gap literature does not erase.

What You Can Do

Treat the post-conversation replay as noise, not data. The observers in the third study could see the liking that participants could not, which means your inner critic works with worse information than a stranger watching the tape. When the replay starts, label it: this is the audit and the audit is rigged, so if you walk away thinking a new acquaintance rated you a 5 out of 7, the evidence says they probably rated you closer to a 6, so calibrate the memory, not the behavior.

Say the thing you are withholding, because the same research family finds people systematically underestimate how warmly compliments land, and the compliment you fear sounds awkward is exactly the one being undercounted on the other side. The cost is a few seconds; the miscalibration is months deep, so give new relationships a full season: your read on a new person is not trustworthy until the evidence outvotes the noise. Schedule the second hangout even when the first felt flat, and remember, for the socially anxious, that the 2025 study ties the perceived gap to insecurity: the people most convinced they are disliked are, on average, the most wrong. The feeling is real. The accuracy is the part you can borrow from the science.