DARVO is the shape a conversation takes when raising something becomes the offence. The term comes from the psychologist Jennifer J. Freyd, who uses it for a reaction that perpetrators of wrongdoing — particularly sexual offenders — may show when they are held accountable: deny, attack, and reverse victim and offender. [1]
What it looks like in an online argument
It is easier to recognise as a sequence than as a single message:
- Deny — it did not happen, or not like that, or it was a joke
- Attack — you are obsessive, dishonest, unstable, or running a vendetta
- Reverse victim and offender — you are the one harassing them, and they are being persecuted for something trivial
By step three the original conduct has left the conversation entirely, and the question in front of any audience is whether you have behaved acceptably.
Why it is effective
It exploits the fact that onlookers usually arrive late. Someone joining at step three sees a person being publicly accused, defending themselves, and describing distress. That reads as sympathetic, and reconstructing what actually started it takes effort nobody watching is obliged to make.
It also creates a genuine cost to continuing. Pressing the point now looks like piling on, so the cheapest option for you is to drop it — which is the outcome the pattern produces.
What the research actually found
DARVO began as a concept rather than a measurement, and the empirical work came later. That work includes a 2017 peer-reviewed open-access paper, Perpetrator Responses to Victim Confrontation: DARVO and Victim Self-Blame, by Harsey, Zurbriggen and Freyd. [2] Quoted on Freyd's own page, it reported that DARVO was commonly used by the individuals who were confronted in the situations studied. [3]
The finding most worth knowing is the one about self-blame, because it describes something you might currently be feeling. Higher levels of exposure to DARVO during a confrontation were associated with increased perceptions of self-blame among the people doing the confronting. [4] The authors put the significance plainly: studying DARVO helps explain how perpetrators are able to enforce victims' silence through the mechanism of self-blame. [5]
Two things need saying about how far that goes. It is an association rather than a demonstrated cause, and it is one study — it is not established that DARVO produces self-blame, only that more of the one went with more of the other. And the framing assumes an actual perpetrator, which is the question a bystander cannot answer.
What it is legitimately good for is this: if you raised something and came away convinced you were the problem, that reaction is documented, and the authors' own reading is that it is the mechanism by which silence gets enforced. That is worth knowing even though it settles nothing about who was right.
When it is an institution doing it
Freyd also names an institutional form: DARVO committed by an institution, or with institutional complicity — her example is police charging rape victims with lying. [6]
That matters here because the online version rarely stays purely online. A report is dismissed, a complaint is turned back on the person who made it, an employer treats raising something as the misconduct. The shape is the same and the power behind it is not, which is why the advice below — return to the specific, once — is about preserving a record rather than winning an exchange.
A caution worth taking seriously
This is one of the easiest concepts on this site to misuse.
The three components — denial, counter-accusation, claiming to be wronged — are also exactly what an innocent person does when accused of something they did not do. There is no way to tell the two apart from the structure alone, and anyone who tells you otherwise is selling certainty that does not exist.
So it is a lens for understanding why a conversation has become unwinnable, not a test for who is lying. Announcing "that's DARVO" mid-argument is both unfalsifiable and, to a bystander, indistinguishable from the pattern itself.
What to do
Return to the specific, once: what was said, when, and where the record is. Specifics resist reframing in a way that characterisations do not.
Then accept that the conversation is unlikely to resolve. The audience — not the person — is who the specifics were for, and repeating them past that point achieves nothing except making the exchange longer.