Maybe I can state my concern over AEs and propositions to a general concern over Types and Tokens.
I seems that Mr. Realist would want to hold that:
red rot rouge
all are instances of the same word-type. The 'redness' type. I'm not totally sure how they would go about expressing this, probably just with saying: these words all express redness. Likewise:
△ ▽ ▷ ► ▼ ◬ ▿ ◿ ▲
all are instances of ▲ity, or triangularity.
Now it seems that Mr. Realist is going to have some problems. Maybe they are just small and bred by my personal confusion. But 'redness' seems to be an English word. If there is an English and a German red-type, this seems to be really the talk of a English and German red-token-class. (Same for the triangle case, with the variety of shapes and designs.) Thus, the Sellarsean move to introduce honest-to-God types: 'red' in English, 'rot' in German, 'rouge' in French all play the same role. Each are •red•s.
Mr. Realist cannot be happy with this, I don't think. Triangularity is supposed to be an abstract entity, not some sort of functional class! Likewise, the sign-designs *△*, *▽*, *▷*, etc. are all of the triangular-kind. Mr. Nominalist wants to say that each shape is called triangular, where Mr. Realist wants to say that each shape is an instance of a three-sided closed-plain figure, i.e. instances of triangularity.
To claim that each shape plays the triangular role, that each is a •triangle• cannot make Mr. Realist very happy at all. We might say that the German 'dreieck' and English 'triangle' each play the same linguistic role: each are •triangle•s. Likewise,
△ ▽ ▷ ► ▼ ◬ ▿ ◿ ▲
are each distinct token-classes of the triangle type. Each shape stands for triangularity, in that: "This is a triangle" is true of each shape.
Mr. Realist cannot seem to be happy about this at all.
It seems that Mr. Realist wants the type to be an entity in the full-blooded sense, distinct from its tokens. The type should be able to be real (or on some accounts, exist) even if its tokens do not exist (or are not real, on those same accounts). To claim that types are only functions seems be incompatible with Mr. Realist's general philosophy. Mr. Realist wants us to have to compare things in the world against things in Platonic heaven to see what they really are. We have to compare instances and exemplifications with the damn Universal! Not with a function!
Certainly we can claim that types or numbers are real, even if they are just functions or structures, but I don't think we are being full-blooded, honest-to-God realists any more. What need have we of Universals when we outsource their explanatory role to functions?
Also, I see a problem in that presumably "the universal blah" and "an instance of the universal blah" seem to be rigid designators, where "the structure that plays the role blah" does not seem to be. Consider:
The killer of Jones is Smith.
It seems that there could be possible worlds where Smith didn't kill Jones. So the killer of Jones isn't a single entity which we can necessarily identify with Smith. Other people could play the killer role. Or no one could. Doesn't the same problem crop up if we want Universals and their instances? It seems that if we appeal to a structuralist account, then in any world, we can literally identify a different entity with the structure! This cannot make Mr. Realist happy at all. That
is an instance of ◣ity seems to be necessary, not just a matter of the shape fitting a role! But this seems to be false, doesn't it.