Here’s a pattern most people know from their own experience. Two people are trying to talk about something they disagree on. Neither starts out angry. Within five minutes, both are angry. Within fifteen, the conversation has moved from the original topic to grievances that go back years. By half an hour in, both parties are saying things they’ll wish they hadn’t said, and the original question — the one the conversation was supposed to be about — hasn’t been discussed for some time.
This is, for many people, roughly what disagreement looks like. It’s also, according to research and clinical practice, not how disagreement has to go. There’s a different pattern — calmer, more honest, less expensive — that some people seem to fall into naturally and others have to learn deliberately. What most of the research suggests is that this calmer pattern can be decomposed into specific moves, and that with some practice almost anyone can learn to use them when conversations start to slide in a damaging direction.
The most widely-taught framework for this comes from a psychologist named Marshall Rosenberg, who developed what he called Nonviolent Communication, usually shortened to NVC. Rosenberg spent decades teaching the framework to mediators, teachers, therapists, activists, and prisoners — contexts where the ability to have difficult conversations without destroying the relationship had genuinely high stakes. The framework has limitations, and we’ll get to them. But it offers something many people haven’t been given: a specific script for what to do when an important conversation is about to go wrong.
The four moves
Rosenberg’s framework breaks a difficult message into four components, each of which does specific psychological work.
Observation without evaluation. Describe what’s actually happening in terms a camera could capture, without the judgements or interpretations that usually come attached. Not you’re always late. That’s an evaluation. This is the third Monday this month you’ve arrived after 9am. That’s an observation. The difference sounds pedantic. It isn’t. Evaluations provoke defence; observations invite discussion. Most conversations go wrong at this first step, because what feels like a simple statement of fact is usually loaded with judgement, and the judgement triggers the defensiveness that derails everything that follows.
Feelings. Say what you’re feeling, in specific emotional language, rather than what you think or what the other person has done. Not I feel like you don’t respect me. (That’s not a feeling; it’s an interpretation disguised as one.) I feel frustrated, and a little hurt. The distinction matters because genuine feelings can be acknowledged without agreement about the underlying issue. I feel hurt is something the other person can hear without having to defend anything. You don’t respect me is something they have to push back on, because it’s an accusation dressed as an emotional report.
Needs. Name the underlying need that’s producing the feeling, in a form the other person might share. Not I need you to be on time. I need to feel that this time matters to you too. The distinction is between a need expressed as a demand and a need expressed as a genuine human requirement that the other person might also, if they thought about it, share. Universal human needs — to be respected, to be heard, to feel safe, to feel useful, to matter — can usually be named in ways that don’t sound like accusations.
Requests. Make a specific, actionable request that the other person could reasonably act on. Not I want you to care more. That’s not actionable. Could we agree that if you’re running late, you’ll send me a text by nine? That’s a request. Requests should be specific enough that both parties can tell whether they’ve been met, and they should be genuinely optional — the other person can say no, and the conversation continues.
The framework, put together, produces sentences that feel slightly stilted at first. When you arrived after nine on the last three Mondays, I felt frustrated because I need to feel that our time together matters. Would you be willing to send a text if you’re going to be more than fifteen minutes late?
This is, admittedly, not how people ordinarily talk. But the stiffness is partly the point. The structure prevents the conversation from skipping over the components that actually matter — the specific observation, the honest feeling, the underlying need, the actionable request. Most natural disagreements skip several of these, producing exchanges where nobody is quite sure what’s actually being said or asked.
The underlying psychology
Rosenberg’s framework didn’t emerge from nowhere. Its core ideas draw directly on the person-centred therapy tradition of Carl Rogers, which we looked at in the listening article. Rogers’s central insight — that people become less defensive when they feel heard and accepted — is the psychological foundation NVC rests on.
The specific value of the framework is that it operationalises Rogers’s principles. Rogers taught that genuine empathy and unconditional positive regard produce therapeutic change. This is true but difficult to put into practice. NVC gives a specific script for applying Rogers’s principles outside the therapy room, in the ordinary difficult conversations of family, work and community life. It’s a prosthetic until the underlying capacities develop. For people who haven’t learned them from their families of origin — which is most people — the script provides something to lean on while the deeper capacities form.
The framework has been extended in many directions. Marshall Rosenberg trained mediators in conflict zones — Palestine and Israel, Rwanda after the genocide, Northern Ireland. He trained teachers, therapists, managers, couples, and many others. The Center for Nonviolent Communication he founded continues to run training programmes worldwide. The framework has become embedded in fields including therapy, education, restorative justice, and activist movements.
The research evidence
The empirical research on NVC is smaller than the framework’s influence suggests. A 2020 review by the Spanish researcher Oriol Juncadella and colleagues found that the available studies on NVC interventions showed generally positive but modest effects on communication quality, conflict resolution and relationship satisfaction. The research base isn’t as large or as methodologically rigorous as we might want — much of NVC’s credibility rests on practitioner experience rather than on randomised trials — but the findings that do exist are encouraging.
This is worth being honest about. NVC has a strong evidence base for its theoretical foundations (the Rogers tradition) but a less developed one for the specific framework as taught. It’s a genuinely useful practical framework whose specific empirical support is thinner than you might expect from how widely it’s taught. The same is true for many practical frameworks — they often outpace their formal research base because they’ve been developed by practitioners rather than academics.
The counter-thread worth hearing
Before endorsing NVC wholesale, an honest critique that some thoughtful practitioners have raised.
The structure of NVC, done rigidly, can feel manipulative or performative. When every difficult sentence is packaged into the observation-feeling-need-request format, the other person can feel they’re being communicated at in a technique rather than being genuinely engaged with. The framework’s very specificity — the careful phrasing, the universal-needs language — can become its own signal of distance, particularly when used by someone who has recently learned it.
A related concern, articulated by some critics including the writer Suzanne Kelman, is that NVC’s pacifist framing can undervalue the legitimate uses of confrontation. Not every difficult conversation should aim at mutual understanding and compromise. Some situations call for firm opposition, clear moral positions, or the willingness to end a relationship rather than preserve it. NVC, rigorously applied, can in certain cases obscure genuine power imbalances and teach people to be softer than their situations warrant.
And there’s a subtler concern: NVC is heavily practised in particular subcultures — therapeutic, progressive, education-adjacent — and using it in contexts where it’s unfamiliar can read as preachy or superior. The person who insists on structuring conflict according to their particular framework is, in some sense, imposing a communication style on a partner who never agreed to it. This isn’t a fatal objection — any communication style imposes something — but it’s worth holding.
So the honest framing is that NVC is a genuinely useful tool, with real theoretical grounding, that has helped many people have better conversations than they otherwise would have. It’s also a tool with limitations, better suited to some situations than others, and more valuable as a source of specific moves to draw on than as a rigid script to follow in every conversation.
What to actually practise
For most people, the useful takeaway isn’t to learn NVC as a full system. It’s to practise the specific moves in ordinary difficult conversations.
Notice when you’re about to evaluate rather than observe. You’re not paying attention to what I’m saying is an evaluation. I’ve said this three times now is an observation. Catching yourself in the second or two before the evaluation leaves your mouth, and swapping in an observation, is one of the single most effective conversational moves available.
Distinguish what you’re feeling from what you’re accusing the other person of. I’m frustrated is honest. You’re being frustrating is something else — it’s attribution wearing emotional clothing. The first invites conversation. The second provokes defence.
Name the underlying need rather than the surface demand. The surface demand — be on time — is about behaviour. The underlying need — to feel that my time matters to you — is about something humans actually share. Getting to the underlying need usually produces better conversations than staying at the surface demand, because it makes what’s actually at stake visible.
Make requests that can be acted on. If the other person, in good faith, asks what would you like me to do differently?, you should have an answer that’s specific enough to act on. Be more supportive isn’t that. Ask me how my week went on Sunday evenings is.
And critically — the part that the rigid framework sometimes obscures — stay in the actual conversation rather than the scripted one. If the other person responds in ways that don’t fit the NVC script, don’t force them back into it. The framework is meant to help you have the conversation, not to replace it.
The question that remains
The deepest thing the NVC framework, at its best, teaches is something subtle. Most of what destroys difficult conversations isn’t the content being discussed. It’s the way the content is framed. The same underlying issue, raised as an accusation, produces a fight; raised as an observation plus a feeling plus a need plus a request, often produces a conversation. The issue didn’t change. The way it entered the room did.
This is both encouraging and humbling. Encouraging because it means that many conversations that currently end badly could end better, with a few small adjustments in how they begin. Humbling because the adjustments are skills that have to be practised, and most of us have been practising the opposite patterns our whole lives.
The question worth carrying, especially after a difficult conversation that went worse than it needed to:
If you’d opened the conversation with an observation instead of an accusation, and named what you were actually feeling instead of what you thought they were doing — would the conversation have been able to go somewhere useful rather than where it actually went?
Key research referenced: Marshall Rosenberg, Nonviolent Communication (2003 and earlier editions); Carl Rogers’s person-centred therapy as the theoretical foundation; Oriol Juncadella and colleagues’ 2020 review of NVC research; Suzanne Kelman and other critical perspectives on NVC.