Meta Model questions are the precise questions that recover what a speaker left out. When someone says “nobody listens to me” or “I can’t do that”, their sentence has deleted, distorted or generalised the actual experience behind it. A Meta Model question asks for the missing piece: Who specifically? What stops you? Compared to what? This article gives you the questions themselves, grouped by the pattern they answer, plus practice drills you can run this week.
If you want the theory first, the glossary entry on the Meta Model in NLP explains where the model came from and how it fits the rest of the toolkit. This page is the practice side: examples, scripts and what goes wrong.
The short version
| What the speaker does | Example sentence | Question that recovers it |
|---|---|---|
| Unspecified noun | “They don’t respect me.” | Who specifically doesn’t respect you? |
| Unspecified verb | “She hurt me.” | How specifically did she hurt you? |
| Comparison with no reference | “This is worse.” | Worse than what? |
| Universal quantifier | “I never get it right.” | Never? Not once? |
| Modal of necessity | “I have to take the job.” | What would happen if you didn’t? |
| Modal of possibility | “I can’t say no to her.” | What stops you? |
| Mind reading | “He thinks I’m useless.” | How do you know what he thinks? |
| Cause and effect | “His tone makes me anxious.” | How does his tone cause the anxiety? |
| Nominalisation | “There’s no communication here.” | Who is not communicating what, to whom? |
| Lost performative | “It’s wrong to ask for more.” | According to whom? |
What is a Meta Model question actually for?
A Meta Model question moves a person from the words they used back to the experience those words came from. That is the whole job. The model was built by Richard Bandler and John Grinder in The Structure of Magic (1975), drawn largely from how Virginia Satir questioned families, and it borrows its vocabulary from transformational grammar.
The practical point is smaller than the theory. People describe their lives in summaries. The summary is where the problem lives, because a summary cannot be tested. “My boss undermines me” cannot be examined. “My boss rewrote my slide deck on Tuesday without telling me” can. Once a statement is specific enough to be checked, a person usually starts solving it on their own.
Which Meta Model questions handle deletions?
Deletions are the pieces of information simply left out of the sentence. You recover them by asking for the missing noun, verb, or standard of comparison.
- Simple deletion. “I’m upset.” Ask: “Upset about what?”
- Unspecified noun. “People are judging me.” Ask: “Which people, specifically?”
- Unspecified verb. “He let me down.” Ask: “How specifically did he let you down? What did he do?”
- Comparative deletion. “I’m doing badly.” Ask: “Badly compared to what, or to whom?”
- Lost performative. “Good coaches don’t take notes.” Ask: “Who says so? Where did that rule come from?”
Comparative deletion is the one most people skip, and it is often the most useful. Almost every discouraging self-assessment hides a comparison to an unnamed standard. Name the standard and the judgement usually loses most of its weight.
Which questions handle generalisations?
Generalisations turn one or two experiences into a rule about all of them. The two main forms are universal quantifiers (always, never, everyone, nothing) and modal operators (have to, must, can’t, should).
- “I always mess up presentations.” Ask: “Always? Has there ever been one that went fine?”
- “Everyone in that team dislikes me.” Ask: “Everyone? All nine of them?”
- “I have to answer emails at night.” Ask: “What would happen if you didn’t?”
- “I should have known better.” Ask: “According to what standard? Known it how?”
- “I can’t ask for a raise.” Ask: “What stops you?”
Two questions carry most of the load here. For necessity (“have to”, “must”, “should”), ask what would happen otherwise. For possibility (“can’t”, “impossible”), ask what stops them. Those two open more stuck situations than any clever variation.
Which questions handle distortions?
Distortions rearrange experience into a shape it did not have. Mind reading claims knowledge of another person’s inner state. Cause and effect claims one thing produces a feeling automatically. Nominalisation freezes an ongoing process into a thing.
- Mind reading. “She thinks I’m not committed.” Ask: “How do you know that? What did she do that told you?”
- Cause and effect. “Open-plan offices make me unproductive.” Ask: “How does the office cause that? Are you unproductive there every single day?”
- Complex equivalence. “He didn’t reply, so he’s angry.” Ask: “Has anyone ever not replied to you without being angry?”
- Nominalisation. “Our relationship has no trust.” Ask: “Who is not trusting whom, about what?”
- Presupposition. “Why do I always ruin things?” Ask: “What tells you that you ruin them?” The question smuggles in the assumption; see the entry on presuppositions in NLP.
Turning a nominalisation back into a verb is the single most reliable move in the whole model. “Communication”, “commitment”, “trust”, “support”, “respect” are all frozen verbs. Unfreeze them and you get an action someone can actually take on Monday morning.
How do you practise this without sounding like an interrogator?
Badly used, the Meta Model is unpleasant. A string of precision questions fired at someone who is upset feels like cross-examination, and rapport collapses long before the information arrives. This is the most common failure we see in Practitioner trainings, and it has a straightforward fix: pace before you question. Match the person’s tone and register, reflect what they said, then ask one question. The material on pacing and leading covers the sequence.
- The single-question rule. For one week, ask exactly one Meta Model question per conversation. One. Notice what it does before you add a second.
- Soften the frame. “I’m curious, who specifically?” lands differently from “Who specifically?” The softener buys you the question.
- Transcribe and mark up. Take five minutes of a recorded conversation or a podcast, write it out, and underline every deletion, generalisation and distortion. You will find more than you expect.
- Run it on your own writing. Your journal and your work emails are full of your own patterns, and nobody’s feelings are at stake.
- Pair drill. One person talks for two minutes about a mildly annoying situation. The other asks only questions, no advice, no reassurance. Swap. Then compare what each of you found.
One honest caveat about scope. The Meta Model recovers information. It does not by itself resolve trauma, and precision questioning can push someone into distress faster than they expected. Used in a coaching context it complements, and does not replace, professional medical or psychological care. If the material turns clinical, refer.
In seven-day Practitioner groups of no more than ten people, the Meta Model is usually the pattern participants find hardest to keep gentle. Everyone learns the questions in an hour. Learning to ask one and then stay quiet takes the rest of the week. It is also the skill people report back on most often a year later, because it changes ordinary conversations, not only sessions. Carolin Mallmann, Licensed Trainer of The Society of NLP, Lisbon.
Where does the Meta Model sit among the other tools?
It is one half of a pair. The Meta Model makes language more specific. The Milton Model, built from Milton Erickson’s work, makes it deliberately vaguer so the listener can fill in their own meaning. Practitioners learn both and switch depending on whether the moment calls for clarity or for space. You can see the rest of the toolkit in the overview of what NLP is, and if you want to work through these questions live rather than from a page, they are taught hands-on in the seven-day NLP Practitioner training in Lisbon.
FAQ
How many Meta Model questions are there?
The original set in The Structure of Magic lists twelve or thirteen language patterns, depending on how the categories are counted, each with its own challenge question. Most trainings teach between nine and twelve. In practice, four or five questions cover the large majority of real conversations: who specifically, how specifically, what stops you, and what would happen if you did.
What is the difference between the Meta Model and the Milton Model?
The Meta Model asks for precision and recovers deleted information. The Milton Model does the opposite: it uses deliberately vague language so the listener supplies their own content. They use the same linguistic categories in opposite directions. Practitioners typically use Milton Model language to build rapport and a receptive state, then Meta Model questions to get at specifics.
Can I use Meta Model questions in everyday conversation?
Yes, and that is where most of the value is, but sparingly. One well-placed question in a meeting or a difficult conversation is useful. A sequence of five sounds like an interrogation and damages rapport. The reliable habit is to notice the pattern silently and ask only when the answer would genuinely change what happens next.
Is the Meta Model scientifically supported?
The Meta Model’s linguistic roots in transformational grammar are well documented, but NLP as a whole has a thin and contested evidence base, and the Meta Model has not been tested as a standalone intervention in large controlled trials. What can be said honestly is that asking for specifics is standard practice in several established counselling approaches. Treat it as a useful questioning discipline rather than a proven clinical technique.
Which Meta Model question should I learn first?
“What stops you?” It responds to the “I can’t” statements that block most decisions, it is short, and it rarely sounds aggressive. The answer is usually either a practical obstacle, which can be solved, or a belief, which can be examined. Once that question feels natural, add “what would happen if you did?”
Do I need a certification to use these questions?
No. Nothing stops anyone from asking better questions, and the patterns are published in books. A training matters for a different reason: you get live feedback on how your questions land, which no book can give you. Reading the model takes an evening; being able to use it under pressure without damaging the relationship takes practice with other people.
Keep exploring
- NLP coaching questions, for how these questions sit inside a full coaching conversation rather than a single exchange.
- Personal development retreats in Portugal: an honest guide, if you would rather practise this in a week away than in your own kitchen.
Carolin Mallmann is a Licensed Trainer of The Society of NLP, trained by Richard Bandler. She has trained 300+ people and runs English-language NLP trainings in Lisbon (groups of max. 10).
