Chapter 7 of Syntactic Structures puts the machinery introduced in earlier chapters — the transformation — to work on English negatives and questions. It’s the most concrete, and most striking, stretch of the book.

Building a Negative

The starting point is the analysis of the auxiliary: Aux as “C(M)(have+en)(be+ing),” where C carries tense and agreement. Chomsky treats every sentence as falling into one of three segments:

Negation (T-not) simply inserts “not” (or “n’t”) between the second and third segments. Applied to “they - 0+can - come,” it gives “they - 0+can+n’t - come” — eventually, “they can’t come.”

Something interesting happens with “John - S - come” (the form underlying “John comes”). Apply T-not and you get “John - S+n’t - come.” But at this point the “S” (third-person-singular affix) hasn’t yet attached to the verb — the rule that attaches an affix assumes the sequence “Af+v” with nothing wedged in between, and “not” has broken that assumption. What’s left is a stranded affix, floating with nothing to attach to.

“Do” as the Bearer of a Stranded Affix

To deal with this stranded affix, Chomsky adds one more rule: insert “do” right before it, as the word that will carry it. Apply that to “John - S+n’t - come” and the result is, finally, “John doesn’t come.”

The very same rule works for questions. The question transformation (T-q) simply swaps the first and second segments of the same three-part analysis. Applied to “they - 0 - arrive,” it gives “0 - they - arrive,” and “do” steps in to carry the stranded “0,” yielding “do they arrive.” Applied to “they - 0+can - arrive,” “can” needs no bearer — it comes out directly as “can they arrive.”

NP - C - V… (kernel skeleton) T-not: NP - C+not - V… insert "not" between segments 2 & 3 T-q: C - NP - V… swap segments 1 & 2 John doesn't come do they arrive
Negation and questions are the same skeleton with two different operations applied. Either way, a stranded affix gets the same "do"-insertion rule. Diagram: ToL

Chomsky himself underlines the economy of this:

The crucial fact about the question transformation Tq is that almost nothing must be added to the grammar in order to describe it.

The “three-segment” analysis and the “do”-insertion machinery, both built for negation, carry straight over to questions. The only genuinely new ingredient is the tiny rule that swaps the first two segments.

Why There’s No Inversion in “Who Ate an Apple”

Even more interesting are wh-questions. “What did John eat” and “who ate an apple” are both questions, yet one has the “did”-inversion and the other doesn’t. Why?

In Chomsky’s analysis, both start from the same underlying kernel string — “John - C - eat + an + apple.” T-q first swaps the first two segments. Then a further transformation (T-w) picks one noun phrase and replaces it at the front of the sentence with “who” or “what.”

The key is which noun phrase T-w picks. Choosing the object, “an apple,” yields a form like “an apple - past - John - eat,” which becomes “what did John eat” — the inversion T-q already performed survives as “did.” But choosing the subject, “John,” makes T-w operate in a way that exactly undoes the swap T-q just made. T-w cancels T-q’s effect. The upshot: “who ate an apple” shows no visible inversion at all.

From one mechanism — the three-way segmentation, T-q, and do-insertion — four kinds of sentence (negatives, yes/no questions, and wh-questions of two different shapes) fall out with almost no extra machinery. This is the concrete substance of what Chomsky means by “simplicity.”