The Liking Gap: Did That Conversation Go Better Than You Think?

In Boothby and colleagues' 2018 paper, the first study recruited 36 people from around Yale. They talked in same-sex pairs for about five minutes, with suggested icebreaker questions, then answered questions separately. One pair was excluded because they were already close friends, leaving 34 participants.

Participants rated their liking for the other person and how much they believed that person liked them. Each measure combined four questions on seven-point scales.

The model-estimated averages were 5.82 for liking the partner and 5.17 for estimated liking from the partner. These are scale scores, not probabilities of friendship. Across the paired sample, the average liking given is also the average liking received. The lower estimates therefore indicated underestimation overall.

This small, structured study cannot establish every individual's accuracy or represent every culture.

A 2025 paper by Tissera and colleagues distinguishes two comparisons:

• Actual liking gap: the difference between someone's report of liking you and your estimate of that report.
• Perceived liking gap: the difference between how much you like someone and how much you think they like you.

The published abstract reports three first-impression samples totalling 2,753 people. It describes the gaps as psychologically distinct and reports that neither was reliably associated with interaction quality. Our account of this newer paper is limited to its abstract; we have not checked its full methods here.

The distinction matters because feeling that affection is unequal is not, by itself, evidence that your estimate is wrong. You need the other person's answer to check that.

Imagine Maya and Noor have just met at a class. For illustration, they privately use a simple scale from one to seven:

• Maya rates how much she likes Noor as 6.
• Maya estimates Noor's liking for her as 4.

Maya feels a two-point imbalance. But these two answers come entirely from Maya. They cannot establish Noor's view.

Now consider two possible versions of the same scene:

Version A: Noor independently rates her liking for Maya as 6. Maya's estimate of 4 was too low. There is both a perceived imbalance and underestimation of Noor's report.

Version B: Noor rates her liking for Maya as 4. Maya's estimate was accurate, even though she liked Noor more than Noor liked her.

Maya's own answers are identical in both versions. The missing piece changes the interpretation. That is why a headline about people feeling less liked should prompt the question: did the study collect both people's reports, or only one person's beliefs?

These invented numbers are an explanation of measurement, not data, a diagnostic score or a recommendation to rate your friends. In ordinary life, we rarely have a private research survey from the other person. The honest answer may simply be that we do not yet know.

Mastroianni and colleagues' 2021 paper extended the question beyond pairs. In one study, 159 participants talked in groups of three, choosing a conversation length between one and 45 minutes. They then separately rated their liking for each partner and estimated each partner's liking for them. Average estimates again fell below reported liking.

A different study in that paper analysed survey responses from 312 working adults after exclusions. Feeling more liked was associated with greater reported comfort asking colleagues for help and with other workplace outcomes.

That workplace survey did not independently obtain each named colleague's liking report. It therefore cannot establish the accuracy of each worker's estimate. Its associations also do not prove that changing a belief would improve teamwork or job satisfaction. A supportive workplace could affect both how liked people feel and how comfortable they are asking for help.

The group-conversation result and the workplace survey address related questions with different evidence. Treating them as interchangeable would hide an important limitation.

A study average cannot decide whether to trust a particular person. It cannot turn an explicit refusal into hidden interest, or make exclusion, harassment or disrespect harmless. Other people's boundaries still mean what they say.

There is also a difference between liking, compatibility and willingness to continue a relationship. Someone can enjoy a conversation and still be unavailable for friendship. A pleasant encounter does not create an obligation on either side.

An ordinary invitation, when welcome and appropriate, leaves room for an ordinary answer: “I enjoyed talking. Would you like to join us next time?” That is an editorial example of clear communication, not a tested liking-gap intervention. Respect a refusal or lack of reciprocity; the research is not a reason to keep pursuing someone.

For broader questions about finding welcoming settings and making contact easier, see our guide to building social connection. For understanding why company and belonging are different, read loneliness and being alone.

The most useful question after reading a liking-gap headline is not “How much more do people like me?” It is “Which person's judgment was measured, and what is still unknown?” That keeps an intriguing finding from becoming a promise it cannot support.

Published by PsychPod. Evidence checked 21 October 2026. Prepared with AI-assisted writing and editorial source checking. The 2018 and 2021 papers were read in full-text form; the 2025 discussion uses its published abstract. This is education about social psychology, not an assessment or treatment guide. Maya and Noor are fictional and do not represent participants or testimonials.