“‘Me and her’ meets ‘he and I’:
Case, person, and linear ordering in English coordinated pronouns”,
by Thomas Grano (2006) seems relevant. Grano says
The puzzling distribution of case in English coordinated pronouns has sparked a number of attempted
Some explanations take case variation to be syntax-internal, and among these, some model it
as resulting from optionality (Schwartz 1985, Johannessen 1998) or underspecification (Parker et al. 1988) in the syntax, and others model it as constraint reranking (Sadock 2005) or reweighting (Quinn 2005). Still, other explanations take case variation to be the result of a natural variation-free grammar pitted against explicit prescriptive instruction, through which speakers adopt either
ad hoc transformational rules (Emonds 1986) or grammatical viruses (Sobin 1997) in an attempt to conform to the prescriptive standard.
Maybe one of the options is not "grammatical", only "acceptable"
One solution is to say that the "naive" opinion of native English speakers is wrong, and one of the two options is ungrammatical after all.
When you ask a speaker to judge a sentence, you're really getting an acceptability judgement, not a grammaticality judgment in the technical sense of "grammaticality". And there are reasons to suspect a sentence may be judged acceptable even if it is not "grammatical" from a theoretical perspective. Norbert Hornstein has made various posts about this on the generative-grammar blog "Faculty of Language"; e.g Acceptability and Grammaticality, Degrees of grammaticality?, Judgments and grammars
This might seem like a reversion to the pre-scientific viewpoint of the prescriptivists, but there are various reasons to suppose that speakers sometimes systematically produce and accept constructions that are not "grammatical" in a generative-grammar sense. E.g. do we really expect a synchronic grammar of English to account for the remnant uses of the subjunctive in sentences like "God save the queen"?
Preminger (2017) seems to take this viewpoint about the case of coordinated noun phrases in English: he says sentences like "My mother and I went to the market" aren't actually grammatical.
(18) a. Hec1 is here on time.
b. Herc2 and himc2 are here on time.
➻ I’m assuming, with Sobin (1997), that the other forms are just prescriptive (hyper)correction
- that they exist doesn’t mean we should shove them in the grammar
- any more than the existence of “Numeral NP do/does not a NP make” means we should make the grammar of English verb-final
Preminger, Omer. Case in 2017: some thoughts. UMD Department of Linguistics & Maryland Language Science Center (Workshop in Honor of David Pesetsky’s 60th Birthday)
Sobin explains the acceptability of constructions like "My mother and I" as the result of a post-grammatical "virus" that changes "and me" to "and I". This is supposed to account for the over- and underextension of the rule.
This position with respect to examples like these is discussed in more detail in my and Ron Maimon's answers to When do I use “I” instead of “me?” on ELU.
Maybe there are multiple valid sources of case assignment
Maybe there is an optional process that results in an alternative case assignment. While I'm less certain about this position, I think I encountered it in a paper by Joan Maling and Rex A. Sprouse, "Strutural Case, Specifier-Head Relations, and the Case of Predicate NPs" (1995).
Maling & Sprouse provide the following example of variation in case assignment from Icelandic:
a. Jón bað mig, [CP að PRO vera fljótur/fljótan] Pred AP
John asked me-ACC to be quick-NOM/quick-ACC
b. Jón bað mig, [CP að PRO vera dyravörður/dyravörð] Pred NP
John asked me-ACC to be doorkeeper-NOM/doorkeeper-ACC
They explain it by saying
Let us consider this to be an instance of the phenomena traditionally known as Case Attraction. Under our assumptions it is not obvious why this alternation is possible. However, let us assume that the case of the controller may optionally be copied onto PRO (as long as PRO is not assigned lexical case by the embedded predicate). Then the paradigm in (11) above and in (12)-(13) below will follow straightforwardly. In (11), PRO may either receive structural nominative from I or it may receive structural accusative via Case Attraction. In either case, I will serve as the source for case features for the predicate NP...