What is this linguistic phenomenon?

I’m sorry, but it’s not “very well attested” that window-curtain was a common phrase. According to Google ngrams (which you also cited, so presumably trust), those two words used together has never been particularly common, any more than you’d expect from straightforward collocation.

The word curtain has come to - except in the theatre - nearly always mean the kind you use on a window. The meaning has become narrower. Agreed. And it interesting the way words change meaning over time, from general to narrow, or from one specific type of narrow meaning to a different but related one.

But it’s not a case of a word dropping a modifier that was used as part of the word, like motor-car dropped the word motor. “Window-curtain” with or without the hyphen just was never very common as a phrase.

Funnily enough I was actually looking recently for some door curtains, and it was annoying because everywhere comes up with regular curtains (they have slightly different features), so I do understand what you mean about the meaning of the word having narrowed in general use. What I dispute is the claim that “window curtains” was ever common. And that’s pretty much indisputable, or should be. And it means it’s not an example of a compound word that dropped its modifier.

So I’m not sure if what you’re thinking of is a form of semantic narrowing, which would apply to the way the word curtains has changed, or something more specific about dropping a modifier.

Well, I suppose that dispute is fundamentally unresolvable without some objective quantitative criterion for the term “common”.

True, but you can do that when comparing two words or compound words. I’d think that the Google ngram shows that “window curtains” was always extremely uncommon when compared to the word curtains alone. Even accounting for the fact that every instance of the word curtains will include the ones that said window curtains, it’s a really marked difference.

Anyway. Is semantic narrowing anything like what you were thinking of, or is it something a little different?

I think what I’m talking about doesn’t have the “narrowing” part. Some semantic narrowing did happen in the shift in meaning of the more general term to default to the more specific form of it (“heels” means by default “heels that are high”, “curtains” means by default “curtains on a window”, etc.).

What I’m talking about is specifically the loss of the modifier because the more general term now defaults to the more specific meaning.

Would “electronic calculator” and “electronic computer” be examples? A “calculator” and a “computer” were terms talking about people in the early 20th century and before, so when devices were created to do the same job, the modifier was needed - but now the modifier is dropped, and it’s taken for granted that a “calculator” or “computer” refers to a electronic machine

Yes, I think those are illustrating the same sort of thing.

Elided modifier.

Just to toss another word onto the bonfire - consensus.

It boils my piss (yes I know it’s unreasonable) when people say or write “consensus of opinion”.

What’s the consensus here?