Specific language tells people what to think. Artfully vague language lets people find out what they think. That second skill is rarer, older, and more misunderstood, and it is the subject here: how to be vague on purpose, with enough control that the vagueness lands where you aim it. Erickson could spend ten minutes composing one sentence because he knew exactly which vagueness did which work. The rest of us can learn the pattern set.
One man built two models of language, and they point in opposite directions. In the mid-1970s, John Grinder, working with Richard Bandler, studied how language shapes experience and produced the Meta Model, a set of questions for pinning language down. Then he studied Milton Erickson and produced the Milton Model, a set of patterns for opening language up. Where Erickson's method came from is its own story.
Most NLP training treats them as rival systems. They are not rivals. They are two directions of travel on the same road, and the skill is knowing which direction a conversation needs.
The Meta Model: language that recovers information
The Meta Model came first, documented in The Structure of Magic (1975). Grinder and Bandler borrowed a framing from linguistics: the idea that what people say (the surface structure) is a compressed version of what they actually experienced (the deep structure). In compressing it, people delete, distort and generalise. The Meta Model is a set of precise questions that reverse that compression.
The patterns are simple to show:
- Deletion. "I'm unhappy." What, specifically, is unhappy about what, exactly?
- Unspecified nouns. "Everyone hates me." Everyone? Who specifically?
- Unspecified verbs. "She ignored me." Ignored you how, exactly?
- Mind reading. "I know he thinks I'm stupid." How do you know?
- Cause and effect. "You make me angry." How, specifically, does my behaviour cause your feeling?
- Universal quantifiers. "It's always like this." Always? Every single time?
The purpose is clarity. The Meta Model cuts through vague complaint and gets to testable specifics, which is why it is the workhorse of therapy and coaching. A client says "I can't do this" and the practitioner asks "Can't, or haven't yet? What stops you?" That single question is the Meta Model in action. The full question set, with worked examples, has its own piece.
The risk is equally clear: fired without rapport, Meta Model questions sound like an interrogation. Precision is a tool, not a personality.
The Milton Model: language that opens experience
The Milton Model came second, documented in Patterns of the Hypnotic Techniques of Milton H. Erickson (1975). Erickson was a psychiatrist and hypnotherapist who used indirect, artfully vague language to work around conscious resistance and let people's own minds fill in the gaps.
Where the Meta Model asks for specifics, the Milton Model deliberately withholds them:
- Mind reading. "You're wondering why I asked that." (No, they weren't, but now they are.)
- Lost performatives. "It's important to relax." Who says so? The pattern hides the source.
- Cause and effect. "As you breathe out, you can feel yourself becoming calmer."
- Presuppositions. "And when you find this useful, which you will, notice what changes first." The sentence assumes the outcome.
- Ambiguity. Words and phrases that can be taken two ways, so the listener's own experience supplies the meaning.
A classic Milton sentence sounds like this: "And as you sit here, you might begin to notice something about the way your breathing has already started to change." It is vague on purpose. The listener does the work of making it true, which is exactly the point. The Milton Model is the backbone of Ericksonian hypnotherapy and of storytelling that opens rather than closes.
Vagueness as a craft, not a fog
The title of this piece deserves an honest defence, because "be vague" sounds like advice to be slippery. It is the opposite. Artful vagueness is a craft with tight tolerances. The pattern selection is precise: presupposition rather than command, nominalization rather than accusation, story rather than instruction. The calibration is precise: you watch the listener's body answer while your words stay open. The delivery is precise: pace, tempo, where the emphasis falls. Erickson's vagueness worked because it was engineered; a fog is just a fog.
The everyday use does not require a trance room. Telling a stressed colleague "somewhere in this mess there is usually one piece that, once it moves, lets the rest follow" hands them a search instruction without handing them a conclusion. Telling a client "people often find the change arrives before they have finished arguing for the old way" gives permission without pressure. Vague, yes. Precisely vague, which is what makes it work.
When to use which
The two models are opposites, and both are necessary:
- Use the Meta Model when you need precision: problem solving, accountability, therapy, cutting through vagueness, challenging a limiting belief.
- Use the Milton Model when you need openness: building rapport, trance work, creativity, and any moment where the client's own answer will beat yours.
A skilled conversation cycles between them. You pace with Milton, then challenge with Meta. You gather specifics with Meta, then open the frame with Milton so the person can reorganise what they just discovered. The direction of travel depends on where the person is stuck: too vague, go Meta. Too rigid and over-analytical, go Milton.
The ethical line
Both models can influence people. The Milton Model, because it works below conscious attention, is the one that raises the question, and it deserves a straight answer.
Vague language used to steer someone toward a decision they did not ask for, without their awareness, is manipulation, no matter how it is dressed up. The line we hold: transparency about intent, influence only toward what the person has said they want, and no use of these patterns to trap, pressure or deceive. Neither model is mind control. People accept or reject suggestions all the time, and anyone who claims otherwise is selling fantasy.
How IONI teaches them
The teaching order matters. The programme starts with the Meta Model in the first months, because precision and calibration come first: you cannot challenge someone's language usefully until you can hear its structure and hold rapport while you do it. The Milton Model comes later, once participants have earned the relational skill to use vagueness without weaponising it. Both are drilled until they are fluent, not just recognised.
For more on the wider toolkit these patterns belong to, see our guide to NLP techniques. To see how the two models are trained in sequence over twenty-four weeks, see the Spirit of NLP programme, or start with the free webinar.