Compare these two sentences.

“Colorless green ideas sleep furiously.” “Furiously sleep ideas green colorless.”

Neither one means anything. But any speaker of English can feel that the first one is shaped like a real sentence, while the second is just words scattered in the wrong order.

That feeling is the starting point of a small 1957 book by Noam Chomsky, Syntactic Structures. Its first question: what actually decides whether a sentence is “grammatical”?

Not “Does It Mean Something,” and Not “Is It Common”

The two obvious answers are: because it means something, or because it’s a familiar way of speaking. Neither holds up. “Colorless green ideas sleep furiously” means nothing, and almost certainly nobody had ever said it before Chomsky wrote it down — and the same is true of the scrambled version. Yet only one of the two feels like a sentence.

So “grammatical” is something other than meaning, and other than familiarity. The rest of the book is Chomsky’s attempt to say what that something is.

A Machine That Just Guesses the Next Word Isn’t Enough

One natural idea: a machine that guesses each word from the one before it. After “I,” the odds favor “am” or “have”; keep chaining guesses and, with enough branching paths, you could in principle build sentences of any length.

But there’s a wall this approach can never get past. Consider:

“If it rains tomorrow, then we’ll stay home.”

Here, “if” demands to be answered by “then,” no matter how much gets packed in between — and what gets packed in between can be an entire sentence of the very same “if… then” shape. Like Russian nesting dolls, this can go many layers deep.

A machine that only ever looks at the word just before it cannot handle this kind of nesting. To pick the right “then” far down the line, it would have to remember the “if” from much earlier — and nothing in its design lets it remember anything at all. Chomsky doesn’t just say this is hard to build; he shows it is impossible in principle. So “guess the next word from the last one” cannot be the right idea about language.

Thinking of a Sentence as Boxes Inside Boxes

The next idea is to break a sentence into boxes. “The man hit the ball” splits into “the man” and “hit the ball”; the second box splits again into “hit” and “the ball” — something like diagramming a sentence in grammar class.

This does much better than the guessing-machine. But it still runs out of road in a couple of places.

One is swapping. Turning “the dog chased the cat” into “the cat was chased by the dog” — the passive — means swapping subject and object outright. Boxes alone have no way to express “swap.”

The other is pieces that come apart. A form like “has been reading” is really one unit — have…en plus be…ing — with other material wedged in between its two halves. Plain boxes struggle with pieces like that too.

Rewriting a Whole Sentence: The Idea of a “Transformation”

So Chomsky reaches for a different move: rewrite one whole kind of sentence into another whole kind of sentence. He calls this a “transformation.”

Start from the plainest possible sentences — the “kernel” — and build up every complicated sentence by layering these rewrite operations on top. That’s the book’s central invention.

What’s striking is how few moving parts this takes. The machinery behind negation and the machinery behind questions turn out to share most of their parts. A rule built for one purpose quietly solves an unrelated puzzle as a side effect — the book has several moments like this. Why “have” behaves oddly in English questions (“have you a book…?”) gets explained, almost as an afterthought, by this same machinery.

The Same Words, Two Different Readings

One more thing the book shows: “the shooting of the hunters” can mean either that the hunters did the shooting, or that someone shot the hunters.

Looked at on the page, both readings have exactly the same shape. What tells them apart is only which simpler sentence each one grew out of — “the hunters shoot” on one hand, “[someone] shoots the hunters” on the other.

In other words, an “ambiguous” sentence is one where a single surface hides two different histories. This suggests something bigger: understanding a sentence might mean retracing how it was built, not just reading the words on the page.

What the Book Did

In short, the book set out to answer one question:

Can we write down — as simply as possible — a system that produces every correct sentence and no incorrect ones?

It got there by refusing to lean on meaning or frequency, replacing “just line words up” with “boxes,” and then patching what boxes still couldn’t do with rewriting whole sentences into others. The technical details get intricate, but the question at the root is simple, and it still sits near the center of linguistics today.

To walk through each idea in more depth, one at a time, see Five Ideas from Syntactic Structures, One at a Time.