You have been trying to explain this for months, to a friend or to yourself, and every attempt collapses into an example that sounds like nothing.
So you conclude you are being oversensitive, and you go back, and the feeling comes back too.
Why looking at their behaviour does not work
Their behaviour is contestable, reinterpretable, and in the moment usually defensible. Every instance has an innocent reading available, and you will find it, because you would rather be wrong about this than right.
Paige Sweet's sociological analysis of gaslighting (2019) makes a related point that is worth sitting with: this kind of dynamic works through the erosion of someone's standing to make claims, including to themselves. The question "am I overreacting" is not incidental to the pattern. It is often the thing the pattern produces.
A standing count is what a referee gives a fighter who has not gone down. The count happens anyway, because the referee has learned not to take the fighter's word for it in the one situation where the fighter is least able to assess themselves.
The move: the Standing Count
Method at a glance
The Standing Count
- Count the signal in yourself, not the behaviour in them. Yours is not contestable.
- Include what your body is recording, and apply the asymmetry test to it.
- Tell one person. Isolation is the mechanism, and a private log does nothing about it.
What to actually count
Not what they did. What you do.
Rehearsing before ordinary interactions. Leaving a conversation less certain of something you were certain of going in. Apologies you cannot reconstruct the cause of. Checking your own memory with a third party. Relief on the good days.
You are not building a case against anyone. You are taking a measurement of yourself, which is the only instrument you have that they cannot adjust.
The signal most people discount
The body keeps its own record and it is often ahead of the reasoning. Sleep that goes wrong on Sunday nights and comes back on Friday. A stomach that reacts to a name appearing on a screen. The particular exhaustion of a day where nothing happened.
And the one worth watching most closely, because it is easiest to misread as happiness: the good days landing as relief rather than as normal. That is what a nervous system does when it has been braced for a while.
The asymmetry test
None of that tells you anything about the other person, and it is not a verdict on you either. Plenty of things produce it. What makes it usable is one question:
Does it track this one relationship, and does it lift when you are away from it?
That is the test. Log the signals beside each other and let the count decide, rather than trying to adjudicate any single evening.
Tell one person
This is not optional, and it is not the same as venting.
Isolation is the mechanism that makes the whole thing work, and a private log does nothing about it. One person who knows the outline, who you can check a memory against, changes the situation in a way that no amount of private record-keeping does.
How common this is
More common than the silence around it suggests. Meta-analytic work on workplace bullying prevalence (Nielsen, Matthiesen, & Einarsen, 2010) found rates varying substantially with how it is measured, but consistently indicating that a meaningful proportion of people experience it.
The reason the number matters is not statistical. It is that the experience of being the only one this is happening to is itself part of what keeps people from naming it.
Two honest limits
This is a recognition signal in a communication book, not medical guidance. If any of the physical signals are persistent or getting worse, that belongs with a doctor rather than in a log.
Going the other way: if you are someone whose sleep and stomach react to everything, that class of signal carries less information for you, and the behavioural entries should carry the weight instead.
When this is the wrong tool
If there are threats, if you are afraid, or if leaving feels unsafe, this is not a communication problem and no method in a book is the right resource. That belongs with a domestic abuse service, a lawyer, or the police.
Naming a pattern is also not diagnosing a person. Clinical terms describe conditions assessed by clinicians, not labels to apply to someone from across a kitchen table, and reaching for them tends to weaken your position rather than clarify it.
Where to start
Keep the count for four weeks. Only your own signals, dated, with a note about whether that day involved them.
Then look at whether the pattern tracks the relationship and lifts when you are away from it. That question is answerable, which is more than can be said for the one you have been asking.
Common questions
How do I know if I am overreacting?
You probably cannot answer that by examining their behaviour, because individual incidents almost always have an innocent reading available. A more answerable question is whether your own signals track this one relationship and lift when you are away from it.
What should I actually be tracking?
Your own behaviour rather than theirs: rehearsing before ordinary interactions, leaving conversations less certain than you arrived, apologies you cannot reconstruct the cause of, checking your memory with someone else, and relief rather than ordinary contentment on good days.
Is a physical reaction meaningful evidence?
It can be, with a test attached. Sleep that goes wrong on Sunday nights and recovers by Friday is worth logging, but only the asymmetry makes it usable: does it track this relationship, and does it ease when you are away from it. If it is persistent or worsening, that belongs with a doctor.
Should I tell someone?
Yes, and it is the part people most often skip. Isolation is the mechanism that keeps this kind of dynamic working, and a private log does nothing about it. One person who knows the outline and can check a memory with you changes the situation.
Sources
Sweet, P. L. (2019). The sociology of gaslighting. American Sociological Review, 84(5), 851–875. · Nielsen, M. B., Matthiesen, S. B., & Einarsen, S. (2010). The impact of methodological moderators on prevalence rates of workplace bullying: A meta-analysis. Journal of Occupational and Organizational Psychology. · Stinson, F. S., Dawson, D. A., Goldstein, R. B., et al. (2008). Prevalence, correlates, disability, and comorbidity of DSM-IV narcissistic personality disorder. Journal of Clinical Psychiatry. · Wink, P. (1991). Two faces of narcissism. Journal of Personality and Social Psychology, 61(4), 590–597. · Kjærvik, S. L., & Bushman, B. J. (2021). The link between narcissism and aggression: A meta-analytic review. Psychological Bulletin, 147(5), 477–503.