Comments on Mark Liberman's Left Dislocation posting drifted for a while into the vexed question of how to express possession when two (or more, though I'll restrict myself to two here) conjoined possessor NPs are involved. For the coordination of a 1sg (pronominal) possessor NP with a 3sg non-pronominal possessor NP, commenters came up with five possibilities (to which I can add many more from my files and from web searches).
Some people weren't comfortable with any of the alternatives, but some had a very clear preference for one of them, and different people's preferences were different. This is not an uncommon sort of variation, occurring what principle for connecting semantics to morphosyntactic form should apply. (What verb form to use with the subject either you or I: am, are, or is?) Different people opt for different solutions, and some people "opt out", rejecting all the solutions whenever possible, choosing instead some quite different formulation of the intended meaning.
I'll sidle up to the particular case the commenters were looking at by first considering some (apparently) much simpler cases.
[For the record, the versions the commenters considered were the ones labeled (3), (5), (6), (8), and (13) below. These labels will be bold-faced when their variants are introduced in the discussion below.]
Non-pronominal non-coordinate NPs have two phonologically distinct variants (to simplify things even further, I'll stick here to singular NPs), which I'll refer to as I and II. I is bare, unaffixed: Kim, the family dog, a friend from Chicago. II has a suffix (basic /z/, with two other phonological variants in the singular; the suffix is usually spelled 's in the singular), attached to the last word of the NP: Kim's, the family dog's, a friend from Chicago's. I'll call this suffix PS.
What happens with coordinate NPs X and Y of this sort, for example Kim and Sandy, would seem, at first glance to be a simple matter of "logic":
If the referents of X and Y are understood as forming a unit semantically (e.g. Kim and Sandy in concert), then the syntax should reflect that, with PS attached to the last word of the coordination: Kim and Sandy's cars 'cars associated with Kim and Sandy jointly'.
But if the referents are understood as separately associated, semantically, with the referent of the modified head, then the syntax should reflect that, with PS "distributed over" the conjuncts, attached to each of the conjuncts: Kim's and Sandy's cars 'car(s) associated with Kim and car(s) associated with Sandy'.
In this view, "holistic" syntax is associated with holistic semantics, and "distributed" syntax with distributed semantics. A pretty picture — but it's not quite the world we live in.
As far as I can tell, people's preferences are to match holistic with holistic and distributed with distributed, but it's easy to find the crossed matchings (we'll see some below). The thing is, the holistic syntax is somewhat simpler (Brevity!), while the distributed syntax is somewhat more parallel (Symmetry!), so people are sometimes willing to sacrifice a clean syntax-semantics matching in favor of getting a satisying syntactic structure. That's the way of the world: you can't always get everything you want.
Another important bit of background: for non-coordinate non-pronominal NPs, I and II each have families of uses. I gets used for subjects, as in Kim exulted, and for objects (of verbs, as in We noticed Sandy, and of prepositions, as in We looked at Terry), and also in some other circumstances (It was Kim at the door). Call the subject uses In (suggesting nominative) and the object uses Ia (suggesting accusative).
II gets used where it is uncomplicatedly a determiner, as in Kim's joy was evident, but also when it constitutes a NP on its own, representing a fusion of a determiner with a head, as for Kim's in Sandy's joy was muted, but Kim's was evident. Call these uses IId (suggesting determiner or dependent) and IIi (suggesting independent), respectively.
That bit of background is important because when we turn to personal pronouns, each of I and II splits in two in their forms (though there are some homophonies in the system). For 1sg, the system is
The distribution of these four variants is remarkably complex; we've posted a number of times on pieces of the puzzle (and on variation in the system). When you throw in coordination, things get even more complex, even if you look only at coordinations of a 1sg pronoun and a 3sg non-pronoun.
For coordinate subjects and coordinate objects, standard English uses In and Ia pronouns, respectively: Kim and I did it, They saw Kim and me. But there are two very common non-standard patterns: Ia in subjects (Me and Kim did it), In in objects (They saw Kim and I).
The order of the conjuncts makes a difference. There are two countervailing forces at work here. First, there is a much discussed general principle that, ceteris paribus, prefers to put the most salient or significant item first (this principle sometimes competes with another, which prefers shorter before longer); since people generally takes themselves to be the most salient or significant entities in their environments, this principle would prefer 1sg first (as in me and Kim). But against this is a potent (at least in our culture) "politeness" principle that says that 1sg should come second in coordinations, as in Kim and I, because we should defer to other people.
The major resut of the politeness principle is a strong dispreference for the In pronoun I as first conjunct (I and Kim did it, They saw I and Kim), though these orderings do occur. For objects, the order effect is much less strong for Ia pronouns (They saw me and Kim), though it's not absent. I'm summing up a huge amount of literature here, much of it accessible through Thomas Grano's Stanford honors thesis.)
I know, you're reeling from all of this, and we haven't gotten to the main act yet. Here it comes: when we propose to concoct a possessive version of a coordination (in particular, of 1sg pronoun and 3sg non-pronoun, in either order), there are a lot of options, none of which is obviously the best.
At first, I'll restrict myself to the case that exercised the Language Log commenters: coordinate possessor phrases that are attributive (that is, prenominal), in a context where IId variants are called for (my car(s)). Later, I'll look at one context were IIi variants are called for (The car is mine, The cars are mine); these turn out to be more straightforward than the attributive cases, though not without their own wrinkles.
So in what follows I'll emphasize that we're looking specifically at attributive uses by adding Nom (for Nominal, a N-headed expression lacking a determiner — ((very) happy) cat in a/the/that/my/Kim's ((very) happy) cat) after the formula for conjoined possessors.
Step 1. Coordinate the two NPs and affix PS to the whole business, for the holistic interpretation. The coordination is either I and Kim or Kim and I (with the second much preferred on politeness grounds). That gives the posssessives in (1) and (2):
(1) I and Kim's Nom
I and my wife's wedding photos. (link)
(2) Kim and I's Nom
If any room in this house deserves to be called a shithole, it is without doubt Justin and I's room. (e-mail exchange forwarded to me by Edwin Rodriguez)
(These possessives will also be available for the distributed interpretation, of course.)
Most people are uncomfortable with both (1) and (2), no doubt because of the strong association of I with subjects, though here we're dealing with possessive determiners rather than subjects. Type (1) is also troubling on politeness grounds, and (2) is troubling because of the word I's, which looks like a possessive version of I — but the possessive version of I is my, right?
So one way to "fix" (2) is to use my instead of I's. That gives:
(13) Kim and my Nom
(This is numbered (13) only because it could arise by a slightly different route, which I'll get to later.) Things like (13) seem to be fairly popular. Here's an example from correspondence (with a 2sg pronoun):
I wanted you to know how much I appreciated Betty and your presence at the services for Pat.
(letter from Richard Nixon to Gerald Ford, reprinted in Harper’s March 2007, p. 23)
Step 2. Another way to "fix" (1) and (2) would be to opt for the non-subject forms of the pronoun, using Ia forms rather than In forms:
(3) me and Kim's Nom
Its me and my wife's first visit to the OBGYN this afternoon, what can I expect? (link)
(4) Kim and me's Nom
a big thanks goes out for duane for being my wife and me's guide. (link)
Type (3) seems to be especially frequent (though it's extremely hard to judge these frequencies through searches). Indeed, on one of the many English usage sites where people have posted queries about coordinate possessives, one respondent specifically recommended this variant:
While due to prescriptive influence you may hear things like "my wife and I's", the more natural word order is "me and my wife's"… One important note is that in everyday speech, it is normal to actually use oblique case rather than nominative case for pronouns in such coordinate constructions
(Well, that's one person's opinion.) Going against both (3) and (4) is the prescriptivist ranting against Ia pronouns in functions other than object. And (4) has the possibly problematic word me's, which looks, once again, like a possessive variant of the 1sg pronoun — which "should", of course, be my, bringing us back to (13).
Step 3. Distributed syntax is available for holistic semantics (see above). That would allow (5) and (6), with parallel IId forms:
(5) my and Kim's Nom
Well here are my and my wife's cats. (link)
(6) Kim's and my Nom
The closest I came was getting my third daughter to wrestle and pillow-fight on my wife's and my bed. (link)
Parallelism good, semantics complex. You get plenty of hits for these two types, but most of them are for distributed interpretations, which are not problematic.
Step 4. For pronouns, the IId and IIi variants are distinguished by more than syntax: the IId variants are usually subordinated in accent to the head they're in construction with, while the IIi variants have (like noun heads) phrasal accent. At the same time, coordinations normally have accent on each conjunct. Putting these two considerations together: there is some pressure to have IIi variants in coordinations, as in:
(7) Kim's and mine Nom
38 On Central: My wife's and mine new place to eat. (link)
(8) mine and Kim's Nom
The photo board for LASC 2000-01 is done. Currently it is being displayed at mine and Tami's desk in the cave, but will be more widely available in a few days. (e-mail exchange found by Chris Potts)
Step 5. The affix PS is a clear and consistent marker of possession on NPs, so it's natural to use it for this purpose, whatever the internal composition of an NP, even a coordinate one. If the second conjunct itself has PS, as in my/mine and Kim's, then an "external" PS can't be detected, since instances of PS+PS are regularly haplologized: only one PS survives, as in a friend of my brother's bear 'the bear of a friend of my brother's'. But if the second conjunct is a pronoun, it can pick up a PS that "belongs to" the whole coordination:
(9) Kim's and my's Nom
He even wants to be in on my husband and my's romantic date night! (link)
(10) Kim's and mine's Nom
This is my wife's and mine's first place together, just something to hold us off for 6 months, … (link)
(Variant (10) might be encouraged by a friend of mine's bear, where the IIi pronoun mine happens to be the last word of the NP that's possessivized with PS.)
And then, with the mark of possession not distributed to the first conjunction, we get:
(11) Kim and my's Nom
This is to be my husband and my's special day to celebrate our commitment to each other. (link)
(12) Kim and mine's Nom
My wife and mine's show hits NYC this summer … (link)
Next, some speakers who might otherwise opt for type (11) might balk at the double marking of possession on my's and simplify it, giving
(13) Kim and my Nom
(See above for other routes to this variant. Nothing says that a single account works for everyone.)
The fat lady isn't quite ready to sing.
Step 6. A great many English speakers can use reflexive pronouns, especially 1sg myself, as emphatic, accented variants of Ia forms, as in They gave it to Kim and myself. There's a large complaint literature about such usages — and some more temperate literature noting that these variants can serve useful functions on many occasions — but there's no denying that plenty of people, including some "serious" writers, use untriggered myself on occasion.
The point is that the variant is out there and deployable in the troublesome coordinate cases, where it's available to "rescue" me (or I), as in
(14) = (2') = (9') Kim's and myself's Nom
(15) = (4') = (11') Kim and myself's Nom
(16) = (1') = (3') myself and Kim's Nom
All of these types are attested.
There ARE logically possible coordinations that seem to be vanishingly rare, in particular, those with I pronoun + PS as the first conjunct:
I's and Kim's Nom
me's and Kim's Nom
myself's and Kim's Nom
(though I'd guess that the last of these is likely to turn up eventually).
The story so far: when you try to coordinate possessives and use the result attributively, you're in some trouble, at least if you're going for the holistic interpretation. None of the possibilities — and at least 16 are attested — is entirely satisfactory. There is, in a very real sense, a gap in your (internal) grammar, which you have to cope with my patching, as best as you can. Some people choose one variant and run with it; many people dither (and, often, comment on their perplexity). Meanwhile, you have to speak, or write, SOMETHING.
The gap phenomenon has been noted explicitly for some time; my own recognition of it came from Chuck Fillmore, at least thirty years ago (I am not attributing priority to him, only saying that this is where I got the idea.) In any case, gaps provide yet another source of variability in language.
A crucial point here is that though each variety of a language does exhibit great systematicity in many respects, it's also a kluge (nod to Gary Marcus's recent book Kluge here), of regularities from here and there, pasted together. Sometimes these conflict with each other, sometimes there are cases not covered by any of the regularities.
This HAS to be the case. There is no Great Programmer of Language L who writes the code for L, debugs it, and checks to make sure it covers all the eventualities. Even if there were, how could people access the code? Instead, what people have to do is discern, as best they can, regularities in the language around them from what they hear and read. This is piecemeal process, and it works differently for different people (because they have different experiences and because no one has, or could have, a panoptic view of the whole business).
Now, a few final remarks on IIi (non-attributive) coordinate possessives. This is the easy case: in standard English, both conjuncts are IIi:
This house is Kim's and mine / mine and Kim's
The two coats on the bed are Kim's and mine / mine and Kim's.
Note that semantics is irrelevant: whether the coordinate NP is most likely to be understood holistically (as in the first example above) or distributively (as in the second), the form is the same. (I suspect that there are some people who sometimes go for the variant my and Kim's in the holistic case, but this is hard to search for.)
The grammar of the language is, in this IIi case, crisp for most people. But IId coordinations are a minefield, and people cope as best they can. Please don't write me for advice; all I can say to you is that there are several reasonable (and reasonably popular) choices. You can avoid making a choice, but that's a minefield of its own: the alternatives are all longer than the coordinate possessives (the presence of you and Betty), and it's often not easy to jiggle them appropriately (note the awkward the house that belongs to me and Kim). Speakers and writers would like something compact, but if they try for it, they can fall into perplexity.
[A note of thanks: over the years, many people have contributed coordinate possessive examples to me. Some of them are credited above, but special thanks go to John Singler, for examples and discussion of them.]