The Structure of Magic is the two-volume book by Richard Bandler and John Grinder in which Neuro-Linguistic Programming first appeared in print. Volume I was published in 1975 with the subtitle A Book About Language and Therapy, and Volume II followed in 1976 as A Book About Communication and Change, both from Science and Behavior Books in Palo Alto. Volume I sets out the Meta Model, a set of questions built by watching how Fritz Perls and Virginia Satir used language with clients. Volume II widens the frame to representational systems, incongruence and family sessions. Volume I opens with a foreword by Virginia Satir and a preface by the anthropologist Gregory Bateson, which tells you a great deal about the company the book was keeping.
| The short version | Volume I (1975) | Volume II (1976) |
|---|---|---|
| Subtitle | A Book About Language and Therapy | A Book About Communication and Change |
| Main contribution | The Meta Model | Representational systems, incongruence, family work |
| Modelled from | Fritz Perls and Virginia Satir | Mainly Virginia Satir’s family sessions |
| Read it if | You want the Meta Model at its source | You want the roots of VAK and congruence |
What is The Structure of Magic actually about?
It is about a single claim: that the results some therapists get, which look to observers like personal charisma or intuition, have a describable structure that other people can learn. Bandler and Grinder were not trying to write a therapy manual. They were trying to prove that skill which appears to be innate can be written down.
Their method was to watch practitioners who were unusually effective, write down what those practitioners did rather than what they said about it, and test whether a stranger could reproduce the effect from the description. That process is what NLP calls modelling, and this book is its first published output.
The word magic in the title is provocation, not a claim about the supernatural. What looks like magic from outside is a sequence of moves from inside, and the authors argue that the sequence is teachable. Read as a promise of wizardry, the book disappoints. Read as an argument about method, it holds up.
What did Volume I introduce?
Volume I introduced the Meta Model, which remains the most durable thing NLP has produced. The argument runs like this. Language is not experience. Between what a person lives through and what they say about it, three things happen: information gets deleted, it gets distorted, and it gets generalised. The Meta Model is a set of questions that reverse those three processes in the specific places where they show up in a sentence.
The theoretical scaffolding came from Noam Chomsky’s transformational grammar, particularly the distinction between deep structure (the full meaning a speaker holds) and surface structure (the compressed sentence that reaches the listener). Grinder was a linguist by training, and Volume I reads like it. There are tree diagrams. There are exercises with answer keys. This is why so few people finish it.
The patterns the book teaches you to hear include:
- Deletions. “I am upset.” Upset about what, and by whom?
- Unspecified verbs. “She rejected me.” Rejected you how, exactly?
- Nominalisations. “There is no communication in this team.” A frozen process turned into a noun you cannot touch. Who is not communicating what to whom?
- Universal quantifiers. “Nobody ever listens.” Nobody? Not once?
- Modal operators. “I can’t say no.” What stops you, and what would happen if you did?
- Mind reading. “He thinks I am incompetent.” How do you know?
- Lost performatives. “It is wrong to want that.” According to whom?
- Cause and effect. “Her tone makes me anxious.” How does her tone cause that, step by step?
Written out like that it can look like a party trick. In practice it is slow, careful work, and the difficulty is not knowing the questions but choosing which one to ask and when to stop. Ask all of them and you sound like an interrogation. That judgment is the part no book teaches, which is why the Meta Model is taught with a live partner across the seven days of the Practitioner certification we run in Lisbon rather than as a handout.
How is Volume II different from Volume I?
Volume II moves away from syntax and towards the person in the room. Where Volume I asks what a sentence leaves out, Volume II asks which sensory channel a person is thinking in, and what happens when their words and their body say different things.
Two ideas from it went on to become standard NLP vocabulary. The first is representational systems, the notion that people process experience mainly in pictures, sounds or feelings, and that their word choice leaks which one they are using. The second is incongruence, the mismatch between the message a person speaks and the message their posture and voice deliver at the same moment. Volume II treats incongruence as information to work with rather than dishonesty to catch out, and that framing is still the humane part of NLP.
It is also the volume that has aged less well as evidence. The strong claim, that you can identify a person’s preferred sensory channel and get better results by matching it, has not held up in controlled research. The weaker and more useful version, that listening to how someone describes their experience tells you something worth knowing, survives.
Is The Structure of Magic still worth reading?
Yes for Volume I, with conditions. No, or at least not urgently, for Volume II unless you are interested in the history.
Volume I is worth it because it shows you the reasoning behind each Meta Model question. Most modern trainings hand you the list of thirteen patterns and move on. The book explains why those particular patterns and not others, and that explanation changes how you use them.
The honest warnings are these. The linguistics is a product of the mid 1970s and parts of the theoretical apparatus were already contested when the book appeared. The therapy transcripts reflect the norms of that decade and can read as blunt today. And the book makes claims with a confidence that the following fifty years of research has not fully supported. Read it as a founding document that shaped a field, not as a settled account of how minds work.
How do you read it without giving up?
Most people who abandon the book do so in the first forty pages, in the linguistics. There is a way around that.
- Read the foreword and preface first. Satir tells you what she thought the authors had captured, and Bateson places the work in a wider intellectual setting. Five pages, high value.
- Skip ahead to the Meta Model chapters and read them before the theory. The examples make the theory readable in a way the theory does not make the examples readable.
- Do the exercises out loud, with a real sentence from your own week. Silent reading of Meta Model material produces almost nothing.
- Go back to the grammar chapters only if you want to. Plenty of good practitioners never have.
- Pair it with a modern introduction. The book explains why, a current guide explains how it is taught now.
If you want a sense of where it sits among everything else on the shelf, our ranked list of the NLP books actually worth reading puts it in context alongside eight other titles.
What the book changed in how I teach
I came to The Structure of Magic after I was already certified, which is the wrong order and also the usual one. What it changed for me was not technique but pacing. Reading the reasoning behind each question made me slow down. I stopped running the Meta Model as a sequence to get through and started using one question at a time, then waiting.
That is now how I teach it. I would rather a group leave a Practitioner week with three Meta Model questions they use well than thirteen they can recite, and the book is the reason I hold that line. For a fuller account of where the model came from, our piece on the history of how NLP developed covers the years around its publication.
Carolin Mallmann, Licensed Trainer of The Society of NLP, Lisbon
One note on scope. Both volumes are books about therapy, written for therapists in the 1970s. NLP complements professional psychological or medical care and does not replace it. If you are reading them for help with a diagnosed condition, treat them as background and speak with a licensed clinician about treatment.
Frequently asked questions about The Structure of Magic
Who wrote The Structure of Magic?
Richard Bandler and John Grinder wrote both volumes. Bandler was a student at the University of California, Santa Cruz at the time, and Grinder was an associate professor of linguistics there. Volume I also carries a foreword by the family therapist Virginia Satir and a preface by the anthropologist Gregory Bateson.
When was The Structure of Magic published?
Volume I appeared in 1975 and Volume II in 1976, both from Science and Behavior Books in Palo Alto. In the same year as Volume I, the authors also published Patterns of the Hypnotic Techniques of Milton H. Erickson, M.D., the source of the Milton Model.
Is The Structure of Magic the first NLP book?
Yes, it is the first book of the field, although the term Neuro-Linguistic Programming does not yet appear in Volume I. The name came slightly later. What the book contains is the method and the first model, which is why it is usually treated as the starting point of NLP.
Do I need to know linguistics to read it?
No, but you will need patience with the chapters that assume it. Grinder builds the argument on Chomsky’s transformational grammar, including deep and surface structure, and does not simplify. Readers who skip straight to the Meta Model chapters and return to the theory afterwards report a much better experience.
What is the difference between the Meta Model and the Milton Model?
The Meta Model, from The Structure of Magic, asks precise questions to recover the detail a person left out. The Milton Model, from the Erickson books, does the reverse: it uses deliberately vague language so the listener fills in their own meaning. One narrows, the other opens, and practitioners use both depending on the moment.
Should I read Volume I or Volume II first?
Volume I, without question. It contains the model that everything else builds on, and Volume II assumes you have read it. If you only ever read one, read Volume I and treat Volume II as optional history rather than required material.
Keep exploring
- The history of NLP, for what was happening in Santa Cruz in the years either side of 1975.
- Nine NLP books worth reading, if you want to know what to pick up once this one is behind you.
Carolin Mallmann is a Licensed Trainer of The Society of NLP, trained by Richard Bandler. She has trained more than 300 people and runs English-language NLP trainings in Lisbon, in groups of no more than ten.
Read how the Meta Model is taught across a Practitioner week in Lisbon.
