not just but also

How to spot it

Start with the grammar, because the grammar is innocent. Not only X but also Y keeps the first thing and adds a second, the way a parcel can arrive late and, on top of that, opened. Both halves stand, and the heavier one usually goes last. The reflex keeps that frame and bends it. A writer names a small version of something, treats it as beneath mention, and hands you a bigger version before the sentence has ended: "not just a broken heater, but a shrug." The heater is still granted, but only so the shrug can look larger beside it.

Put your thumb over everything up to the "but" and read what is left as though it had arrived alone. Does it give you a mechanism, a cost, or something a stranger could go and verify? Then the turn is holding an argument up, and you can leave it alone. Does it repeat the first half in a bigger coat? Then the sentence holds nothing but the escalation, and the not just but also turn has been handed the job that evidence was supposed to do.

The first half is probably the faster of the two tells, because its cheapness shows. "Not just a version bump" waves off a claim nobody made; the modest reading was invented so the sentence would have something to climb over. After that, ask whether you could argue with the second half at all. "A rethink of how we store customer records" can be wrong, and you would know how to say so. "Something deeper" cannot be wrong, which is a polite way of saying it is not yet a claim.

Why models overdo it

Likelihood picks the next words, not proof. After "not just" the likeliest continuation is an upgrade, so an upgrade is what turns up. And the first half costs nothing to produce, because calling a thing not merely X asks for no evidence when nobody said X was all there was. By the time the second half lands, the reader has already been told to expect something larger, and the larger thing collects credit it did nothing to earn.

Now the concession, which matters here. The shape works. A good essayist uses it to walk a reader off the obvious reading and onto the one she wants argued, and it survives in speeches people still quote a century on. Any model trained on that writing picks the move up along with everything else. What does not come with it is the judgement about when the second half has earned its billing, since nothing in the process keeps a ledger of what was promised against what was paid.

The result reads well formed and slightly swollen. Every point sounds like the important one. Read a few hundred words like that and the effect turns on itself: once every claim gets upgraded, the upgrade stops carrying any information, including on the claims that were fine as they stood.

When it's fine

Most of the time it is fine, and a reader who starts flagging every instance will be wrong more often than right. Obama's Nobel remarks say that nations will at times find the use of force "not only necessary but morally justified." Both halves are claims. The first is the hard-headed one; the second is the moral one, the claim a listener can actually dispute, and it goes last because it weighs more. Delete the first half and the sentence loses the concession it stands on.

The King James wording of Philippians does the same thing at a slower pace: it is given "not only to believe on him, but also to suffer for his sake." The second half names a cost, plainly, that a reader could not have guessed from the first.

Douglass goes a step further and turns the first half down outright: it is not light that is needed, but fire. That is close to correction in the old sense, a weaker word taken back and a stronger one set in its place, and the refusal is his argument. He is standing in front of an audience that came for a polite Fourth of July address, telling them gentleness will not do this job, and what he puts in the place of light is a demand they can hold him to.

So keep your own turn when you can say what the second half adds and who might want to fight you over it. When you cannot, put the two facts down plainly and let the heavier one go last. The sentence loses a few words, the claim gets easier to believe, and that is usually the trade you were after.

Where it comes from

The additive frame is ordinary grammar, and Latin had it long before English did, as non solum ... sed etiam: not only this, but also that. Both halves stand, and nobody takes anything back. The figure the rhetoricians named works the other way round. They call it epanorthosis, from the Greek for straightening out again, and the Latin manuals carry it as correctio: a speaker takes back a word he has just used and puts a stronger one in its place, the way someone might say he is angry and then correct himself to furious. Roman handbooks listed it among their figures, and Renaissance schoolbooks carried it into the training of generations of English preachers and lawyers. Preachers valued it because a congregation cannot turn back a page, so the weaker word has to be heard before the stronger one can replace it. The outright denial, not this but that, sits close to it. Matthew's Gospel has Christ say he came not to send peace, but a sword, and Douglass turned light down for fire in front of an audience whose polite expectation was exactly the thing standing in his way. The reflex this lesson is about borrows that move and softens it. "Not just a broken heater" never takes the heater back; it keeps the smaller thing in name only, as the weak word waiting to be corrected, so the larger one arrives with the authority of a correction and none of its proof. In the hands that did it well, none of this was decoration. A listener who has just heard one word replaced by a stronger one expects the stronger one to be true, and the orators who lasted are the ones who made sure it was.