What the data actually shows
The clearest evidence comes from a 2018 study by Erica Boothby, Gus Cooney, Gillian Sandstrom and Margaret Clark, published in Psychological Science, which named the effect the 'liking gap.' Across several experiments — including strangers paired for short conversations and people tracked over longer acquaintance — participants reliably underestimated how much their conversation partner liked them and how much the partner had enjoyed their company.
The gap was not a one-off first-meeting jitter. In the studies that followed people over time, including university students living together over an academic year, the liking gap was still detectable months in. People who genuinely came to like each other continued, to a degree, to under-read how positively the other felt.
The pattern fits a broader body of work showing people hold overly negative views of how they come across socially. We tend to assume our nervousness is visible, that our small fumbles registered, and that others are quietly less impressed than they actually are. The general direction of the error is consistent: we are harsher judges of our own social performance than the people in the room with us.
Why this feels different from how it actually is
You have full access to your own self-criticism and none to anyone else's actual opinion. During a conversation you can hear the awkward pause, the joke that did not land, the thing you wish you had said better — and you cannot hear that the other person barely noticed any of it. Your inner critic fills the silence with a verdict the other person never delivered.
That internal commentary then leaks into your estimate of what they think. Research on this suggests people partly use their own self-assessment as a stand-in for the other person's view, so a low opinion of your own performance quietly becomes a low guess about how much you were liked. The other person, meanwhile, was mostly attending to the content of the talk, not auditing your delivery.
There is also an information asymmetry baked into politeness. People rarely tell you directly that they enjoyed talking to you, so the warm signals are muted and easy to miss, while your own doubts are loud and constant. The result is a reading skewed toward 'that went worse than it did.'
What the research says to do about it
The most useful move is simply knowing the gap exists. Studies on debiasing social perception suggest that being told about a systematic error like this can soften it — when you catch yourself concluding 'they probably didn't like me,' you can treat that conclusion as a likely artifact of the liking gap rather than as information.
Assume the warmer interpretation as the default, because the evidence says it is usually the more accurate one. After a conversation that felt merely okay, the calibrated guess is that the other person enjoyed it somewhat more than you did and thought somewhat better of you than you assume.
Acting on that calibration tends to compound. Because people underestimate how much they are liked, they often hold back from following up — not reaching out again, not extending the next invitation. Since the other side frequently felt more positive than you realised, taking the small social risk of following up is, on the evidence, less risky than it feels.
What the research says does not help
Replaying the conversation to find what went wrong does not help, because it feeds exactly the self-critical loop that produces the gap. The more weight you give your own internal review, the further your estimate drifts from how the other person actually felt.
Generic 'just be confident' advice misses the mechanism. The issue is not usually a lack of confidence in the moment but a biased reading of the aftermath — a wrong guess about someone else's mind. Confidence affirmations do little for an error that is about misjudging another person rather than yourself.
Seeking constant reassurance also tends to backfire over time. It can erode the very relationships it is meant to protect, and it treats the symptom rather than the cause. The more durable correction is recalibrating your default assumption, not collecting fresh confirmations after every interaction.
Real numbers in context
The liking gap is a directional bias, not a precise figure — but its key feature is that it is consistent across studies and conversation types. In the original 2018 research, strangers paired for conversations reliably rated how much their partner liked them below how much the partner reported actually liking them, and the effect reappeared across follow-up samples and longer time frames.
It is worth holding alongside its cousin, the spotlight effect: people also overestimate how much others notice and scrutinise them. Taken together, the research points the same way — others are paying less critical attention to you than you fear, and liking you somewhat more than you assume. The honest reading of most ordinary conversations is more generous than the one your own mind defaults to.