The NLP meta model: precision questioning guide

What is the meta model?

The meta model is a structured set of questions designed to bring vague language into specificity. The premise: most human distress runs on imprecise descriptions of reality - deletions, distortions, and generalizations that the client has accepted as fact. The meta model challenges each type with a specific question that recovers the missing material.

Developed by Bandler and Grinder in the early 1970s, the meta model draws from transformational grammar and the questioning patterns of Virginia Satir and Fritz Perls. It is published in their first book, The Structure of Magic (1975).

The three categories

1. Deletion (information missing)

• Simple deletion: "I am upset." -> "About what specifically?"
• Comparative deletion: "It is better this way." -> "Better than what?"
• Lack of referential index: "They don't respect me." -> "Who specifically?"
• Unspecified verbs: "She rejected me." -> "Rejected you how specifically?"

2. Distortion (facts warped)

• Mind reading: "He thinks I am incompetent." -> "How do you know?"
• Cause-effect: "Her tone made me anxious." -> "How specifically does her tone cause your anxiety?"
• Complex equivalence: "If she loved me, she would call." -> "How does calling mean love specifically?"
• Lost performative: "It is wrong to ask for help." -> "Who says it is wrong?"
• Nominalizations: "We have a communication problem." -> "Who is failing to communicate what to whom?"

3. Generalization (one becomes all)

• Universal quantifiers: "I always mess up." -> "Always? Every single time?"
• Modal operators of necessity: "I have to please everyone." -> "What would happen if you didn't?"
• Modal operators of possibility: "I can't speak in public." -> "What stops you?"

How meta-model questions work in practice

The questions do three things at once:

• Recover information the client deleted from their own awareness.
• Loosen frozen generalizations by surfacing counter-examples.
• Make the problem workable by moving it from abstraction to specifics.

Often a single well-aimed meta-model question dissolves what looked like a complex problem. "I have to please everyone" sounds insoluble. "What would happen if you didn't?" usually surfaces a specific fear that is much smaller than the generalization implied.

How to use the meta model without breaking rapport

• Pace before you challenge. Mirror the client's predicates and feeling first. Cold meta-modeling reads as interrogation.
• Soften the form. "I want to understand more about that. When you say 'always', is there ever a time it isn't true?" is gentler than "Always? Every time?"
• Be selective. Most sentences contain three or four violations. Ask the one whose answer will move the work, not all four.
• Listen to the answer. The point is the information that surfaces, not your demonstration of skill.

When the meta model is the wrong tool

• Mid-emotion. A client in acute distress needs to be heard, not questioned for precision.
• Identity-level statements without preparation. "I am unlovable" is a generalization the meta model can address, but only after rapport is firmly established.
• When the client already has the information. If they know what is specific, asking is annoying. Move on.