The Meta Model: NLP Questions That Recover the Missing Detail
The Meta Model is a set of questions from NLP that recover the detail people leave out when they turn experience into words. When someone says “Nobody respects me here”, the Meta Model gives you precise questions — “Nobody at all?”, “Respects you how, specifically?” — that take the sentence back towards the events it came from.
It was the first model Richard Bandler and John Grinder published, and it is still the most practical piece of language patterns work for anyone who coaches, manages or interviews. This note covers where it came from, the three processes it listens for, every core pattern with an example and a question, a worked dialogue, and how to use it without sounding like a cross-examination.
What is the Meta Model?
Richard Bandler and John Grinder set out the Meta Model in The Structure of Magic, Volume I (1975). They had studied the language of two unusually effective therapists, Virginia Satir and Fritz Perls, and noticed that both kept asking a particular kind of question: one that made the client’s vague description concrete again. Bandler and Grinder borrowed ideas from transformational grammar, in particular the distinction between the surface structure of a sentence (what is said) and the deep structure behind it (the fuller meaning), and catalogued the ways the first loses information from the second.
The working assumption is simple. The words a person uses are a map of their experience, not the experience itself. A map that is too thin limits the choices the person can see. Asking the right question fills in the map, and often the choices reappear. The model sits at the base of the NLP communication model, which describes the same filtering from the perception side.
Deletion, distortion and generalisation
Every Meta Model pattern is an instance of one of three processes. None of them is a mistake. They are how a mind compresses a great deal of experience into a sentence someone else can follow. They become a problem only when the compressed version is the one the person is living by.
- Deletion — part of the experience is left out. “I’m annoyed” omits about what, and at whom.
- Distortion — the experience is reshaped: a process turned into a thing, a guess presented as a fact, one event said to cause another. “His tone made me angry.”
- Generalisation — one or a few experiences stand in for all of them. “Meetings here are always a waste of time.”
Training schools sort some patterns differently — nominalisation is filed under deletion in some manuals and distortion in others — so treat the categories as a listening aid rather than a fixed taxonomy.
Meta Model questions with examples
Each pattern below is sometimes called a “Meta Model violation”, a term from the original book meaning a place where the surface sentence has lost information. The question in the third column is the standard challenge. In a real conversation you would soften it, as the later sections show.
| Pattern | Example sentence | Meta Model question | What it recovers |
|---|---|---|---|
| Simple deletion (D) | “I’m frustrated.” | “Frustrated about what?” | The object or subject that was left out |
| Comparative deletion (D) | “That went badly.” / “She’s a better fit.” | “Badly compared with what?” / “Better than whom?” | The standard being used for comparison |
| Lack of referential index (D) | “They don’t value my work.” | “Who, specifically, doesn’t?” | The actual person or group |
| Unspecified verb (D) | “My manager undermines me.” | “How, specifically, does she undermine you?” | The observable behaviour behind the verb |
| Nominalisation (D/Di) | “The communication in this team is poor.” | “Who isn’t communicating what to whom, and how?” | The ongoing process frozen into a noun |
| Mind reading (Di) | “He thinks I’m not ready.” | “How do you know he thinks that?” | The evidence behind the assumption |
| Lost performative (Di) | “It’s wrong to ask for help.” | “Wrong according to whom?” | The source of the rule or judgement |
| Cause–effect (Di) | “Her emails make me anxious.” | “How does reading an email cause you to feel anxious?” | The steps between stimulus and response, and the person’s own part in them |
| Complex equivalence (Di) | “He didn’t reply, so he’s angry with me.” | “How does not replying mean he’s angry?” | The rule equating one thing with another |
| Presupposition (Di) | “When will they realise I’m leaving?” | “What leads you to believe you’re leaving?” | The assumption taken for granted inside the sentence |
| Universal quantifier (G) | “I always freeze in interviews.” | “Always? Has there been a time you didn’t?” | The exceptions that the generalisation hides |
| Modal operator of necessity (G) | “I have to say yes to every request.” | “What would happen if you didn’t?” | The feared consequence behind the rule |
| Modal operator of possibility (G) | “I can’t delegate this.” | “What stops you?” / “What would happen if you did?” | The perceived barrier and what lies beyond it |
D = deletion, Di = distortion, G = generalisation. If you want to check your recognition of these patterns, the NLP Practitioner knowledge quiz includes Meta Model items with an explanation for each answer.
A worked example: the Meta Model in a coaching conversation
This is an illustrative exchange, not a transcript. A team lead has asked for help because she feels she is failing in a new role. Notice that the coach asks one question at a time and lets the answer decide the next one.
Client: I’m just not cut out for leadership. Everyone can see it.
Coach: Everyone? Who in particular have you noticed seeing it? (universal quantifier, referential index)
Client: Well — my director, mainly. She thinks I’m out of my depth.
Coach: How do you know she thinks that? (mind reading)
Client: She took over the client call on Tuesday.
Coach: And her taking over the call means you’re out of your depth? (complex equivalence)
Client: …I suppose she might have just known the client. She’s worked with them for years.
Coach: So what would you want to find out from her? (moves to action)
Client: Whether she stepped in because of me, or because of the relationship.
The coach did not tell the client she was wrong or offer reassurance. Three questions moved the conversation from a global verdict (“not cut out for leadership”) to a single, checkable event (one call on Tuesday) and a question the client can actually go and ask. That shift from verdict to event is what the Meta Model is for.
How to use Meta Model questions without sounding like an interrogation
Asked in a row and without warmth, Meta Model questions feel like being cross-examined. Early NLP trainees are known for exactly this. The skill is less about knowing the questions than about choosing when, and which one, to ask.
Build rapport first
A precise question from someone who has been listening feels like interest. The same question from someone who has not feels like a challenge. Reflect back what you have heard before you probe it, and watch the person’s response as you ask — the shifts in breathing, posture and tone covered in sensory acuity tell you whether a question opened something or closed it.
Soften the question
Instead of “Who, specifically?” — “I’m curious who you had in mind when you said that.”
Instead of “How do you know?” — “What have you noticed that tells you that?”
Instead of “Always?” — “Has there ever been an interview that went a bit differently?”
Choose the one pattern that matters
A single sentence can contain five patterns. Challenging all of them is pointless. Ask yourself which deletion, distortion or generalisation is holding the problem in place, and question that one. Usually it is a modal operator (“I can’t”, “I have to”), a cause–effect claim, or a universal quantifier, because those are where a person’s choices disappear. For a structured way of tracking which questions you used in a session, see the coaching session checklist.
Meta Model vs Milton Model
The two models are mirror images. The Meta Model takes vague language and makes it specific. The Milton Model, drawn from the hypnotherapist Milton H. Erickson, uses the very same patterns deliberately — mind reads, nominalisations, presuppositions, universal quantifiers — so that the listener fills in the meaning from their own experience.
| Meta Model | Milton Model | |
|---|---|---|
| Direction | General to specific | Specific to general |
| Aim | Recover detail and expand choice | Leave room for the listener’s own meaning |
| Typical use | Coaching, interviewing, problem definition | Teaching, presenting, therapeutic suggestion |
| Original source | Satir and Perls, via The Structure of Magic (1975) | Erickson, via Patterns of the Hypnotic Techniques of Milton H. Erickson, M.D. (Vol. I 1975; Vol. II 1977, with Judith DeLozier) |
A skilled practitioner uses both: precise questions to find out what is going on, and looser language to help someone imagine something new.
When not to use the Meta Model
- When someone needs to be heard, not questioned. A person describing grief or shock is not presenting a thin map to be repaired.
- When precision is not the point. Small talk, humour and social warmth run on deletion. Asking “Better than what?” at a dinner party is not insight.
- When there is a power imbalance. A manager meta-modelling a junior employee’s complaint can feel like being caught out. Use it to understand, never to win.
- When the material is clinical. Trauma, anxiety or depression need a qualified clinician. The Meta Model is a conversational tool, not a treatment.
Free instruments for this note
NLP Practitioner Knowledge Quiz
Test your grasp of the core NLP Practitioner toolkit: rapport, sensory language, anchoring, precision questions, outcomes and reframes.
Open itQuestions people ask
What is the Meta Model in NLP?
The Meta Model is a set of language patterns and questions published by Richard Bandler and John Grinder in The Structure of Magic (1975). It identifies where a sentence has deleted, distorted or generalised experience, and offers specific questions — such as “Who, specifically?” or “How do you know?” — that recover the missing detail and widen the choices a person can see.
What are the three processes in the Meta Model?
Deletion, distortion and generalisation. Deletion leaves part of the experience out, distortion reshapes it (for example, treating a guess as fact or one event as the cause of another), and generalisation lets one or a few experiences stand for all of them. All three are normal features of language. They matter when they hide choices from the person speaking.
What is a Meta Model violation?
It is the original term for a place in a sentence where information has been lost: an unspecified noun, a vague verb, a mind read, a “must” or “can’t”, and so on. “Violation” sounds harsh, but it refers to the grammar, not the speaker. Many trainers now simply say “Meta Model pattern”.
Who did Bandler and Grinder model for the Meta Model?
They studied the language of Virginia Satir, the family therapist, and Fritz Perls, the founder of Gestalt therapy. Both habitually asked questions that turned clients’ vague statements back into specific experience. Bandler and Grinder combined these observations with ideas from transformational grammar to produce the model.
Is there scientific evidence for the Meta Model?
There is little controlled research on the Meta Model itself, and NLP as a whole has limited and contested scientific support. The questions are best understood as a disciplined form of clarifying questions, which are a standard part of good interviewing and coaching. Judge them by whether they produce more specific, useful information in your own conversations.
Related
One useful observation,
no motivational sludge.
One useful observation about human behaviour, language, or decision-making — sent when there’s something worth saying. Unsubscribe any time.