You Reread Her Two-Word Reply Looking for the Tone
The Mind · The Reading Room, Vol. 17 · 9 min read · What the liking gap studies found, when your reading of a room is the one to trust, and the five-second correction that works better than reassurance
She replied “Fine, thanks.” Two words, no full stop after thanks, sent at 4:52 — and you have now read it three times looking for something that was never in it.
You read it once, carried on, and then went back to it twice more before six. You checked whether she used your name. She usually does. She did not this time.
What follows is a reading of published social psychology on that habit. I am not a clinician, the findings belong to the researchers named below, and only the plain-English version is mine.
The short answer: if you have ever wondered why do I assume people are annoyed with me on evidence this thin, the answer is that you are running an estimate that never gets corrected — the uncorrected read. A 2018 study in Psychological Science found that after conversations, people systematically underestimate how much their partner liked them and enjoyed their company, an effect the researchers named the liking gap.¹ It showed up between strangers in a lab, between adults at a workshop, and between first-year students who had shared a room for most of an academic year. The estimate reads low, and nothing in ordinary life ever comes back to tell you so.
The skim version
- People consistently think they are less liked than they actually are after talking to someone.¹
- The effect held across short, medium and long conversations, and across three different settings.¹
- Among students sharing a room it persisted for most of the academic year before closing.¹
- One cause is that you judge your own conversational performance far more harshly than anyone else does.¹
- Nobody sends the message saying they enjoyed talking to you, so the estimate is never corrected — that is the whole mechanism.
- A change sustained over weeks is different data and should be believed. That case, and four others, are in the differential below.
The source for this piece
Prof. Erica J. Boothby, PhD
A social psychologist with a doctorate from Yale, now Clinical Professor of Business Administration in the organisational behaviour group at the Tuck School of Business at Dartmouth, previously at Wharton. Her research is on the persistent gap between how we think other people see us and how they actually do — in first impressions, in workplaces and in negotiations. She is the lead author who named the liking gap, which is the finding underneath almost everything written on this subject since.
What we read: Boothby, Cooney, Sandstrom & Clark, Psychological Science (2018), alongside a 2024 replication of the effect in online interaction.¹²
Where to follow her work: LinkedIn · Tuck School of Business · Google Scholar
Why do I assume people are annoyed with me when nothing was said?
Because you are working from the only evidence you have access to, which is your own performance, and you grade it harder than anyone else in the room ever will.
Erica Boothby and colleagues put people into conversations and then asked both sides two things: how much did you like your partner, and how much do you think your partner liked you. The second number came in consistently below the first. People walked away from conversations that had gone well believing they had gone worse than they had.¹
They tested it in three separate settings. Strangers meeting in a lab. Adults getting to know each other at a personal development workshop. And first-year students who shared a room for an academic year, where you might expect the illusion to correct itself within a fortnight. It did not. The gap persisted for most of the year.¹
Part of the reason is structural. Conversations run on politeness, so nobody tells you they enjoyed talking to you, and nobody tells you they did not either. You are left interpreting a silence that means nothing in particular.
And part of it is that while your partner was listening to what you said, you were listening to how you said it.
Why does two words feel like enough to go on?
Because the ambiguity is doing the work, and your mind will not leave a gap empty.
“Fine, thanks” carries no tone. It has no volume, no face, no pause before it. That absence is not neutral to you, because a mind already running a low estimate of how it lands will fill an empty space with the reading it already holds. You are not detecting coldness. You are supplying it, and then reading it back.
Definition: the liking gap is the reliable difference between how much a conversation partner actually liked you and how much you estimate they did — measured by asking both people, separately, straight after the conversation.¹
What makes it so persuasive is that the correcting evidence never arrives. A colleague who thought you were good company does not send a message saying so. The absence of reassurance gets filed as mild confirmation, and the estimate stays where it was, quietly, for years.
This is the same machinery that makes one critical line outlive nine good ones. Ambiguous social information gets resolved in the unfavourable direction and then stored as though it were fact.
| What happened | Your reading | What the research would predict |
|---|---|---|
| Short reply, no name | She is annoyed with me | She was replying between two meetings |
| Nobody responded to your point | It landed badly | People are poor at signalling approval in groups |
| A conversation you enjoyed | They probably tolerated it | They likely enjoyed it more than you think¹ |
| No follow-up message | Confirmation of the above | Almost nobody sends unprompted reassurance |
| Your own performance | Awkward, too long, too keen | Judged far more harshly by you than by them¹ |
If this has got heavier with age rather than lighter, that is worth ten quiet minutes on its own. The Quiet Audit is a private pass through what you are monitoring and what it is costing you. Nobody sees your answers. Start here.
Why has this got worse rather than better with age?
A few things stack, and none of them are about you becoming worse company.
The channel changed. A great deal of your social contact now arrives as text, where tone has been stripped out entirely and there is nothing to read except word count and punctuation. Twenty years ago the same exchange happened in a doorway, with a face attached to it.
The stakes changed. When you are senior, other people manage their expressions around you, so the honest signals thin out exactly when you have more reason to want them.
And there is a quieter one. By fifty-two you have three decades of accumulated readings, almost all of them made by the same biased instrument, and they feel like a body of evidence rather than a repeated measurement error.
The wider version of this is worth reading on its own, because it applies well beyond a two-word reply: a thought arriving with certainty is not a thought arriving with proof.
Between us — what the research found
The 2018 paper is a set of studies rather than one, which is why the effect stuck: laboratory conversations between strangers, adults at a personal development workshop, and roommates followed across an academic year, with both sides rating each other so the two numbers could be compared directly.¹ A 2024 study found the same direction of error in online interaction.²
The limitation, stated plainly: the samples were mostly young — students and workshop attendees — and largely Western. Nobody has run this study on women in their fifties reading work messages from a colleague of fifteen years, which is the exact situation most readers of this piece are in. The effect is also about how much someone liked a conversation, not about whether they are annoyed with you over something specific that happened. Those are related but not identical, and the research speaks to the first one directly and the second one only by implication.
When your reading is probably right
There is a version of this that deserves to be trusted rather than corrected, and glossing over it would make everything above less useful.
The liking gap is about a single interaction — the estimate you form after one conversation, from one set of signals, with no follow-up information. That estimate reads low, reliably.
A pattern is different. If someone who used to loop you into decisions has stopped, if invitations that were automatic now need chasing, if three separate exchanges over two months have been shorter than they used to be, you are not reading tone off a two-word message. You are observing a change in behaviour over time, which is real data and worth taking seriously.
So the test is about the unit of evidence. One message, one meeting, one exchange — apply the correction. A change sustained across weeks — believe it, and act on it.
Most women who run that test honestly find they have been treating a single ambiguous reply as though it were a trend. Occasionally they find a genuine pattern they had been talking themselves out of seeing, which is its own useful discovery and usually more actionable than the anxiety was.
Either way you get information instead of a mood.
Five Things That Look Like This and Are Not This
Everything above assumes an ordinary low estimate in an ordinary week. Sometimes something else is producing it, and a five-second correction is the wrong instruction to give a woman in any of these five situations.
- A real change in behaviour over weeks. The correction applies to single interactions. A sustained shift across a month is data, and treating it as a bias is how women talk themselves out of seeing something they needed to see.
- Social anxiety as a condition rather than a habit. If the fear of being judged is shaping what you agree to do — which meetings, which invitations, which phone calls — that is a recognised and treatable pattern, and it responds to specific treatment rather than to a better reading of one message.
- Depression’s negative interpretation bias. Depression pulls ambiguous information downwards across the board, not just socially. The distinguishing question is what happens on a genuinely clear weekend: depletion lifts a little, anhedonia does not.
- The perimenopausal pattern. Sensitivity to perceived rejection that arrives out of proportion to the trigger and tracks the cycle rather than the events has a different mechanism — and it is reported by most women in the transition.
- Hearing that has quietly narrowed. In a noisy room, missing three words in a sentence leaves a gap, and a gap gets filled with tone. Mid-range hearing loss is common from the mid-forties, arrives without an obvious moment, and is worth a test before it gets read as coldness for another five years.
None of these means the liking gap is not also operating. They mean the correction below is the cheap first move, not the whole answer.
Who Else Has Measured This
One paper names an effect; it takes other work to show it holds. Three lines converge here.
The online replication. A 2024 study in Computers in Human Behavior Reports found the same direction of error in online interaction: people are liked more than they think there too.² That matters for a two-word reply, because text strips out the signals that might have corrected you.
The stranger experiments. Nicholas Epley and Juliana Schroeder, writing in the Journal of Experimental Psychology: General in 2014, had commuters predict how a conversation with a stranger would go before having one. The predictions were reliably worse than the experience. Different question, same shape of error: people forecast social interactions more pessimistically than the interactions turn out.
The group version. Follow-up work by Adam Mastroianni, Gus Cooney, Erica Boothby and Andrew Reece, published in Organizational Behavior and Human Decision Processes in 2021, found the gap in workplace groups and teams as well as in pairs — which is the setting most relevant to a woman rereading a colleague’s message at 5:40.
Gillian Sandstrom, a co-author on the original paper and now at the University of Sussex, has spent the years since studying what happens when people talk to strangers anyway. The consistent finding is that the anticipated version is worse than the real one.
You are not detecting coldness. You are supplying it, and then reading it back.
What This Actually Changes
Not her reply. It still says “Fine, thanks” and it still has no full stop.
What it changes is the standing of your reading of it. You have been treating the conclusion as an observation — something you noticed rather than something you produced. It is an estimate, made by an instrument with a documented direction of error, in conditions that guarantee no correcting information ever comes back.
It also changes what to do about it. Reassurance does not work, because reassurance asks you to believe something about her, and you have no evidence about her. Correction works, because it only asks you to accept something about your own instrument, which is a much smaller thing to ask.
The rule: correct in the known direction
You cannot talk yourself out of this in the moment, because the feeling arrives with the confidence of a conclusion rather than a guess.
So do not argue with the reading. Adjust it. When you catch yourself deciding how something landed, apply one sentence: “My estimate is low, so the real number is higher than this.” Not “she probably liked it.” That is reassurance, and it will not hold.
It takes about five seconds and you do it without resolving anything.
There is a second half, and it is the part most people skip. Once a week, tell one person something specific you appreciated about them. Not a compliment about how they look. Something they did. That does not fix your estimate of yourself. It does something more useful: it puts real information into a system that runs on almost none, which is the actual reason the gap survives.
If the loop is mostly about work conversations replaying afterwards rather than reading messages, the mechanism is adjacent but the fix differs, and it is worth reading about rehearsing a meeting that already finished.
When this stops being a reading problem
Speak to a doctor rather than practising the correction if the checking has started to take real time out of your evenings; if you are avoiding meetings, calls or invitations to avoid the aftermath; if the low reading now covers most of your relationships rather than one or two; if the flatness has lasted more than a few weeks and does not lift on a genuinely clear one; or if you are waking two hours early with the day already dreaded. Social anxiety and depression both have treatments with good evidence behind them, and neither is a character verdict. If you are having thoughts of harming yourself, speak to someone today rather than reading further.
What to Say, and What You Will Probably Hear Back
The instinct is to ask her directly whether she is annoyed. That is the one move that reliably backfires, because it asks her to manage your discomfort and hands her the job of reassuring you — and reassurance, as above, does not hold.
Short enough to read off a phone screen
To yourself, in the five seconds after the reading arrives: “My estimate is low, so the real number is higher than this.”
If something genuine needs raising, raise the thing rather than the mood: “Can we go back to Thursday’s decision — I want to check I have it right.” Never “are you annoyed with me?”
Once a week, to one person: “The way you handled that call on Tuesday made the whole thing easier. Thank you.” Specific, about something they did, no reply required.
If the checking has been running since long before this job, the Quiet Audit is ten minutes of questions about where the monitoring came from and what it was originally keeping safe. No advice, no programme. Start here.
Questions women actually ask about this
What if I am right and she actually is annoyed?
Sometimes you will be. The research does not say your reading is always wrong; it says it is biased low, consistently, across settings. The correction is worth applying every time even though it will occasionally overshoot. The alternative is trusting an instrument you know reads under.
Should I just ask her?
Rarely, and not in the form you are imagining. “Are you annoyed with me?” puts her in the position of reassuring you. If something genuine needs raising, raise the thing itself rather than your anxiety about it.
Is this social anxiety?
Not necessarily. The liking gap turned up in ordinary participants rather than in an anxious sample.¹ It is a general feature of how people read conversations. It can be more intense in social anxiety, which is a separate matter and worth discussing with a clinician if it is shaping what you do.
Does it apply to messages, or only to conversations?
The original work studied conversations. Later research found the same effect in online interaction.² Text strips out the signals that would have corrected you, so if anything the conditions are worse.
Why do I do this most with people I like?
Because the estimate matters more, so you check it more often. The people whose opinion you are indifferent to never get reread at 5:40. Worth knowing when you catch yourself doing it — the checking is at least pointed at someone whose regard you actually want.
The one move
The next time you catch yourself deciding how something landed, do not argue with the reading. Apply the correction out loud or under your breath: my estimate is low, so the real number is higher than this. Then leave it unresolved and carry on with the evening.
A correction, not a reassurance. You are not deciding what she thinks. You are adjusting a measurement you already know reads under.
Where to Go Next
In order, and each for a reason.
1. Why a thought arriving with certainty is not a thought arriving with proof — the general case, and the one that generalises beyond messages.
2. Then rehearsing a meeting that already finished, if the replaying is louder than the rereading.
3. Then the refusal you write on someone else’s behalf, which is the same instrument pointed at a different question.
You checked whether she used your name. She did not, this time. It remains the thinnest possible evidence, and you now know which direction your reading of it errs in.
References
1. Boothby, E. J., Cooney, G., Sandstrom, G. M., & Clark, M. S. (2018). The Liking Gap in Conversations: Do People Like Us More Than We Think? Psychological Science, 29(11), 1742–1756. Read the study
2. The liking gap online: People like us more than we think. Computers in Human Behavior Reports (2024). Read the study
3. British Psychological Society. The “liking gap”. Research Digest. Read the summary
On the researchers. Erica Boothby is a social psychologist and business school faculty member; Gus Cooney, Gillian Sandstrom and Margaret Clark are academic psychologists. None of them is a clinician, none of them is your clinician, and nothing in their published work is advice about a specific relationship or workplace. None of them has any commercial relationship with this publication. The two studies cited with findings are linked in full above, and the settings and sample types are described in the body rather than buried.
Sourcing, disclosure and disclaimers
This is not medical or psychological advice. Blue Leaf Journal publishes general information for a general readership. Nothing here is a diagnosis, a treatment recommendation, or guidance about a specific relationship, colleague or workplace.
We are not clinicians. Blue Leaf Journal is an independent publication. We read published research and translate it. The findings belong to the researchers and institutions named and linked above; the plain-English rendering is ours, and so is any error in it.
No affiliation and no endorsement. Blue Leaf Journal is not affiliated with Erica Boothby, the Tuck School of Business, Dartmouth College, the University of Sussex, Yale University or the British Psychological Society. None of them has reviewed, approved or endorsed this article, and none of them is responsible for it.
No commercial relationship. Nobody named above paid for or was paid for this coverage. There are no affiliate links, no sponsored placements and no gifted products in this article.
How this was checked. The two studies carrying the findings are linked in the references; the Epley and Schroeder and Mastroianni papers named in the corroboration section are cited by journal and year and are not open access. Where the research does not cover women in midlife reading work messages — which it does not — we have said so rather than implied it does. Sources were checked on 28 August 2026. If you find something we have got wrong, write to norawhitfield@blueleafjournal.com and we will correct it and say that we did.
Nora Whitfield writes the Mind desk at Blue Leaf Journal: attention, rumination, and the readings we make of rooms we have already left. She has reread a message with no punctuation in it four times in one evening, decided it was fine, and then reread it once more at eleven to be certain. The correction works. It does not work on the first attempt.
Written by Nora Whitfield for Blue Leaf Journal. Updated: 28 August 2026.
