A simple example of ellipsis is:
Peter likes to eat apples, and Mary oranges.
(Peter likes to eat apples, and Mary [likes to eat] oranges.)
Recently, I've been engaged in a debate about a sentence in an academic paper that I wrote (that is not on language or linguistics, as my field of study is in the sciences). The sentence in question has the same grammatical structure as:
?Circles correspond to apples, and squares oranges.
(Circles correspond to apples, and squares [correspond to] oranges.)
I was told that the sentence above is ungrammatical, and the correct sentence should be:
Circles correspond to apples, and squares to oranges.
The argument I was given was the following example:
*Peter talks to Sally, and Mary Henry.
Peter talks to Sally, and Mary to Henry.
While I agree with this particular example, I am still of the opinion that my sentence (on circles and squares) is grammatically correct; or at least, its grammatical correctness is open to interpretation. To me, what my friend is saying is that gapping cannot apply to a phrase ending with a preposition. (She did not phrase it this way.) However, this argument seems very prescriptivist to me, and I can immediately offer a rebuttal, using the famous phrasal verb "to put up with":
Peter puts up with Sally, and Mary puts up with Henry.
*Peter puts up with Sally, and Mary up with Henry.
?Peter puts up with Sally, and Mary with Henry.
?Peter puts up with Sally, and Mary Henry.
(The question marks indicate that I have been unsure about the grammatical correctness of the sentence since I entered into this debate.) So I think we can safely discard the rule that gapping cannot apply to a phrase ending with a preposition. Instead, what matters seems to me to be the "proximity" between the words near the boundary of the ellipsis: For example, the words "puts" and "up" in the example above are absolutely inseparable, which is what makes "Mary up with Henry" absurd. In comparison, the bond between "puts up" and "with" is less tight, and this is why I think "Mary with Henry" sounds less wrong. The boundary between the phrasal verb "puts up with" and the object "Henry" is very clear, which is why I think "Mary Henry" sounds the most correct to me.
Looking back at the example "Peter talks to Sally", I can now explain that "Mary to Henry" sounds better than "Mary Henry" in this case because "talks / to Sally" is a better structural analysis than "talks to / Sally".
If my analysis is correct (which may not be the case), then this means that the group of words to be elided must contain the entirety of a phrasal verb and not just part of it, since the semantic meaning of the phrasal verb depends heavily on the togetherness of the verb and the preposition. (I am aware that some phrasal verbs can be cut up; for example, "to talk somebody into something". In this case, I am arguing that the words "talk" and "into something" should appear or disappear together:)
Peter talked Sally into joining the club, and Mary Henry.
*Peter talked Sally into joining the club, and Mary Henry into joining the club.
Unfortunately, I must admit that "corresponds to" is not necessarily a phrasal verb (depending on your point of view). However, I think that the words "corresponds" and "to" are sufficiently bound that "corresponds to / oranges" is at least as good a structural analysis as "corresponds / to oranges". My thinking here is supported by the fact that "corresponds to", which means "is associated to", differs in meaning from "corresponds with", which means either "corroborates" or "communicates with".
So the questions is:
What is the rule governing gapping when the verb in question has or needs (a) preposition(s)?
I have given a lot of thought to this (probably more than I should have), and I have read the Wikipedia articles on ellipsis and gapping. I have also tried to search online for relevant examples of ellipsis, but most books and articles I have found only give examples where the elided verb takes direct objects without a preposition.
Finally, I have read the answers to a possibly irrelevant question on this forum, as well as those to several others that seem even less relevant. (In the linked question, I tend to agree with the answer given by John Lawler, if it is important at all.)