Each point from the overview article, pulled out here one at a time with a short example. Every section stands on its own.

1. What Counts as a “Sentence”

Chomsky first pins down the object of study: a language is “a set of sentences built from a finite set of parts.” Not a vague sea of words, but the boundary line between the set of correct sentences and the set of incorrect ones. A “grammar” is simply the device that draws that line.

2. “Grammatical” Isn’t “Meaningful”

“Colorless green ideas sleep furiously” is nonsense but grammatical; reversed, “furiously sleep ideas green colorless” is neither grammatical nor sensible. Both are equally meaningless, yet only one feels like a sentence. From this, Chomsky draws the conclusion that the standard for correctness isn’t found in meaning at all.

3. “Grammatical” Isn’t “Familiar” Either

Neither of those two sentences had likely ever been said before. Pile up statistics on “how often has this been said” and both come out equally rare. Yet only one feels natural — so the standard for correctness isn’t found in frequency either.

4. A Word-by-Word Guessing Machine Isn’t Enough

A machine that picks each word from the one before it — like rolling dice down a chain of words — looks simple and powerful. But it cannot, in principle, handle sentences where two distant words are locked together (“if… then”), especially when the whole structure can be nested inside itself. The machine has no way to “remember” the earlier word at all.

5. Boxes Inside Boxes: Phrase Structure

Split a sentence into a “subject box” and a “predicate box,” then split each of those further — something like diagramming sentences in school. This handles far more than the guessing-machine could. Chomsky formalized this as a system of rewrite rules.

6. But Boxes Alone Aren’t Enough Either

Explaining when two pieces can and can’t be joined by “and” turns out to require knowing not just the words’ current arrangement but how they were built — their history. A rule that only looks at the present shape of a string can’t see that. The passive sentence, which swaps subject and object wholesale, hits the same wall.

7. Rewriting a Whole Sentence: Transformation

Enter the “transformation”: a rule that rewrites one whole shape of sentence into another. Start from the plainest possible sentence — the kernel — and build up complex sentences by layering on transformations for negation, questions, the passive, and more. What’s surprising is how few parts this needs. The machinery for negation and the machinery for questions largely overlap.

8. One Rule, an Unexpected Bonus

Why English “have” behaves oddly in questions (“have you a book…?”) turns out to have nothing to do with meaning — it’s a side effect of how the transformational machinery happens to work. A rule built for one purpose ends up explaining a seemingly unrelated puzzle — this is the strongest evidence for the idea that simpler theories are better theories.

9. Same Surface, Different History: Ambiguity

“The shooting of the hunters” can mean either that the hunters did the shooting or that they were shot. One surface, two different histories — which simple sentence it grew from, and by which transformation. The reverse also happens: sentences made of entirely different words (“John played tennis,” “my friend likes music”) turn out to share the same underlying shape. Understanding a sentence, Chomsky suggests, may just mean retracing how it was built.

For more depth, see the detail articles: the limits of the finite-state model, the limits of phrase structure, the analysis of negation, questions, and do, and structural ambiguity and the autonomy of syntax.