Gaslighting vs disagreement: the psychology of "that never happened"

Gaslighting is a pattern in which one person repeatedly undermines another's confidence in their own perception, memory or judgement, so that the second person comes to depend on the first for a sense of what is real. It is not a disputed memory, a defensive denial, or an unkind "you're overreacting" — though all three can be part of it when they recur, in one direction, about things that can be verified.

The word comes from Patrick Hamilton's 1938 play Gas Light and its film versions, in which a husband dims the gas lamps, then insists to his wife that the light is unchanged and that her perception is failing. The point of the lamps is not the lamps. It is to make her unable to trust herself — and therefore dependent on him.

Philosopher Kate Abramson's analysis captures this: gaslighting aims to destroy the target's standing as a knower. The gaslighter does not merely want to win the argument about what was said at dinner; they want the other person to stop believing that their view of what was said at dinner could ever be reliable. That is why gaslighting so often shifts from the event to the person: "you're too sensitive", "you always do this", "everyone thinks you overreact".

Sociologist Paige Sweet argues that this cannot work on willpower alone. It works because of power — economic, social, emotional — and because of stereotypes that make the denial plausible to the target and to others: the "hysterical" woman, the "paranoid" partner, the "difficult" employee. Gaslighting borrows credibility from the surrounding culture.

The most common reason two people disagree about what happened is not manipulation. It is memory.

Elizabeth Loftus's decades of research on the misinformation effect show that memory is reconstructive: each time we recall an event, we rebuild it from fragments, and the rebuilding is shaped by what we have heard, felt and believed since. People confidently and sincerely remember details that were never there. They adopt a partner's account without noticing. They remember what they meant rather than what they said.

This has two consequences for anyone trying to work out whether they are being gaslit.

First, a single disagreement proves nothing. Two honest people remembering a conversation differently is the default state of human relationships, not a warning sign.

Second, you are not immune either. Your memory is also reconstructive. Being certain is not the same as being right — for either of you. This is why records matter more than conviction.

The table is a lens, not a verdict. Real relationships sit somewhere along each row, and a defensive but fundamentally honest person may show two or three features on a bad week. The pattern — most rows, most of the time — is what matters.

Until recently, most writing on gaslighting came from clinical experience and popular psychology — Robin Stern's The Gaslight Effect (2007) is the best known. In 2023 Willis Klein, Sherry Li and Suzanne Wood published one of the first systematic studies: a qualitative survey of 65 people who identified as having experienced gaslighting in a romantic relationship.

Their findings are sobering and useful:

• Gaslighting typically occurred inside a cycle that also contained affectionate behaviour. Participants did not describe relentless cruelty; they described warmth and undermining alternating — which is part of why it was hard to see and hard to leave.
• The consequences they reported included a diminished sense of self and increased mistrust of others, persisting after the relationship ended.
• Some participants also described post-traumatic growth — a stronger sense of their own values and limits afterwards.
• Recovery pathways centred on physical separation from the perpetrator, prioritising healthier future relationships, and engaging in meaningful, re-embodying activities that rebuilt trust in their own perception.

This is one study, with a self-selected sample. It does not tell us how common gaslighting is or who does it. It does give the experience a shape that matches what clinicians have described for decades — and it points to the same thing the problem page recommends: protect your access to independent reality.

Resist the temptation to reach straight for a diagnosis. "That never happened" is produced by at least four very different engines.

Honest memory. Covered above, and the most common.

Shame and defence. Admitting a hurtful remark means feeling, briefly, like a person who says hurtful things. Some people cannot tolerate that and deny by reflex. The denial is self-protective rather than controlling — but if it is never repaired, its effect on you can be similar.

Escape. Some people deny or minimise to end a conversation they find overwhelming — a cousin of the stonewalling that John Gottman's observational research links to relationship breakdown. The motive is relief, not domination.

Control. And some people learn, consciously or not, that if their partner can be made to doubt their own perception, arguments are won before they begin and leaving becomes almost unthinkable. Where this sits alongside isolation, monitoring, financial control or intimidation, it belongs to the pattern of coercive control described by Evan Stark — a criminal offence in England and Wales since 2015.

The aim is not to catch anyone out. It is to give honest memory every chance to explain things — and to notice if it cannot.

1. Record, briefly and close to the time. Date, what was said, by whom. Facts, not feelings.
2. Keep written exchanges. Do not delete threads in anger.
3. Propose a shared rule. "When we disagree about what was said, we check before we argue." Note the response. Good-faith partners accept this with relief; it costs them nothing.
4. When you find evidence, watch what happens next. "You're right, I did say that — sorry" is repair. "That was about something else", "you're twisting it", or a sudden change of subject to your flaws is the thing to track.
5. Track direction over a month. Symmetrical errors: memory. Asymmetrical errors: something else.
6. Keep outside reference points. Friends, family, a therapist, a diary. A single person's reality can only become the only reality if nobody else's is available to you.

"We remember this differently and I'm not going to argue about who's right. I'd rather look at the messages together."

"Maybe I am sensitive about this — that's a separate conversation. Right now I want to stay on what was said."

"I hear that you don't remember it that way. I'm going to trust my notes on this one."

"This keeps happening and it's wearing on both of us. Can we agree to check rather than argue?"

None of these lines accuses anyone of anything. Each declines the invitation to debate your sanity and returns to the event. The response you get is the data.

Seek support if you have begun to doubt your perception in general, if denials extend repeatedly to things you have in writing, if the people who might offer an outside view have become "the problem", or if the pattern sits alongside control, threats or fear.

And if, having run the test, what you find is a partner who remembers badly in all directions and apologises when shown the message — then what you have is a human being with a human memory. Keep the shared rule, and let the word go.