Humans Are Overconfident About Almost Everything. Five Studies Found the One Exception: After Conversations, People Systematically Underestimate How Much Others Liked Them.

The Paradox of Social Confidence

Psychology has a name for our tendency to think we're special. It's called the above-average effect, and it is pervasive. Most drivers believe they're safer than the median. Most students believe they're smarter. Most married couples believe their relationship is less likely than average to end in divorce. Decades of research, from Svenson's 1981 driving study to Dunning and Kruger's 1999 experiments, have confirmed that people possess a stubborn conviction that they're better than they are.

So when Erica Boothby, a postdoctoral researcher at Cornell, started studying what happens after two strangers talk, she expected to find the same thing. She found the opposite.

Five Minutes, One Mistake

In 2018, Boothby and her collaborators—Gus Cooney at Harvard, Gillian Sandstrom at Essex, and Margaret Clark at Yale—recruited 36 adults and paired them for five-minute conversations using standard icebreaker questions. Afterward, each person rated how much they liked their partner and how much they believed their partner liked them.

The results were unambiguous. Participants liked their conversation partners significantly more than they thought their partners liked them (actual: M = 5.82 vs. perceived: M = 5.17 on a 7-point scale; b = −0.65, t(34) = −5.83, p < .001). Since both people in a pair can't logically like the other more than the other likes them, the discrepancy means someone is making an error. The data showed who: virtually everyone was underestimating how well they'd been received.

Boothby named it the "liking gap."

The Signals You're Not Seeing

The natural explanation is that people conceal their enjoyment during conversation, leaving their partner with nothing to read. To test this, the researchers had trained coders watch video recordings of the conversations. The coders could accurately predict how much each participant actually liked their partner (b = 0.71, p < .001). But their ratings failed to predict how much participants thought they were liked (b = 0.38, p = .51).

The warmth was right there. Smiles, nods, genuine laughter. Outside observers could read it easily. The participants themselves could not, because they were listening to a different signal: their own inner critic.

Your Inner Critic Is Louder Than Their Laughter

A second study revealed the mechanism. Eighty-four Yale students had unstructured, open-ended five-minute conversations. Afterward, each person listed the three most salient moments they believed shaped their partner's impression of them and rated each as positive or negative. They did the same for the moments shaping their own impression of their partner.

The pattern was stark. People believed the defining moments of how their partner judged them were more negative than the moments defining their own judgment of their partner. This self-critical bias directly mediated the liking gap (indirect effect b = −0.14, 95% CI [−0.22, −0.07]). The harsher the inner monologue, the wider the gulf between how liked people actually were and how liked they believed they were.

"Did I talk too much?" "Was that story boring?" "Why did I bring up my ex?" These are the questions ricocheting through your head after a conversation. Your partner isn't asking any of them. Your partner is thinking about what they said wrong.

Neither Time Nor Familiarity Fixes It

Longer conversations didn't help. In Study 3, 102 people had conversations ranging from 2 to 45 minutes. The liking gap persisted at every duration (b = −0.38, p < .001), and an "enjoyment gap" appeared alongside it: people also underestimated how much their partner enjoyed the conversation (b = −0.52, p < .001).

The most striking finding came from tracking 102 college freshmen and their randomly assigned suite mates over an entire academic year. At five time points from September through May, students reported how much they liked their suite mates and how much they believed their suite mates liked them. The liking gap persisted at every assessment for nine months (b = −0.36, t(1131) = −6.16, p < .001), finally disappearing only at the last measurement in May, when students were actively deciding whether to live together the following year. Offering a $100 bonus for accurate estimates didn't close the gap. People weren't performing modesty. They genuinely believed they were less liked than they were.

Beyond the Original Lab

Independent teams have since extended the finding. Mastroianni, Cooney, Boothby, and Reece (2021) showed the liking gap operates in small groups and engineering teams, not just pairs, and that it predicts worse collaboration outcomes. Wolf, Carlson, and colleagues (2025) analyzed 2,753 first impressions across three large samples and found that 59% of individuals show a negative liking bias. Their work also revealed that the perceived gap correlates with low self-esteem, high social anxiety, and neuroticism, while the actual gap does not. This suggests something important: feeling unliked is a trait-level vulnerability, not a response to evidence.

Here's a calculation the original authors didn't run: if 59% of people underestimate how much a new acquaintance likes them, and this error discourages follow-up contact, then in a room of 50 people at a networking event, roughly 30 are walking out convinced they failed to connect. Multiply that by the approximately 11 conversations the average American adult has with non-household members each day (Bureau of Labor Statistics, 2023), and the liking gap generates billions of phantom social rejections per year across the population.

The Strongest Counterargument

The most serious challenge is that the liking gap may reflect rational self-protection rather than a cognitive error. Social rejection carries real costs. Defaulting to pessimism about how others view you means you never overcommit to someone who doesn't reciprocate, avoiding potential embarrassment or exploitation. Assuming someone dislikes you costs a missed friendship; assuming someone likes you and being wrong risks humiliation. From an evolutionary standpoint, a negativity bias in social inference could be adaptive, and participants may be applying a sensible discount to the noisy signal of a stranger's behavior. If this is calibrated risk management, calling it a "gap" overstates what's actually a functional feature of social cognition. Boothby and colleagues counter that the gap persists for nine months among roommates who have abundant daily data on each other's feelings, well past the point where uncertainty-based discounting should correct itself.

What We Didn't Prove

All five original studies drew from elite university populations or self-selected workshop attendees in the UK. Participants were predominantly young, educated, and English-speaking. Later replications have extended the finding to workplace teams, non-English speakers, and cross-cultural samples, but the original evidence base is narrow. The liking gap was measured with self-report scales, not behavioral outcomes. We know people think they're less liked, but the studies don't directly measure whether this misperception prevents friendships from forming. None of the five studies were pre-registered, and the researchers started Study 1 with a convenience sample, collecting as many participants as possible before the semester ended. The gap disappearing at month nine in the dorm study raises the question of whether it truly persists or simply takes longer to resolve than the earlier time points suggested.

What You Can Do

Next time you leave a conversation assuming the other person thought you were dull, consider this: the data says you're probably wrong, and so does the other person about you. The practical steps are simple. Reach out. The liking gap means the email you're hesitating to send, the text you're debating, the coffee you're not sure you should suggest, would almost certainly be welcomed more warmly than you expect. Notice the signals. Outside observers can see the warmth in your conversations even when you can't. The problem isn't that people hide their enjoyment; it's that you're not watching for it because you're too focused on your own performance. And recalibrate your inner critic. The gap between how you think you came across and how you actually came across is consistently wrong in the same direction. You're not the exception. Nobody is.