It's orthogonal to the point of Mark's post, but in thinking about the expression, 'be friends with', a few questions sprang into my head, and having (I think) introspectively answered them to my satisfaction, I decided to inflict the whole train of thought on the rest of the world.
First, note that the expression 'be friends with' is one of them there symmetric predicates, related to the 'essentially plural' predicate 'be friends', which requires a plural subject:
(1) X and Y are friends
(2) a. X is friends with Y
b.Y is friends with X
It's like the way 'be married to' is related to the essentially plural predicate 'be married':
(3) X and Y are married
(4) a. X is married to Y
b. Y is married to X
"(Be) married" in this case is an adjectival passive, and "(be) friends" here looks like a predicate nominal. In their singular-subject forms, both involve idiosyncratically selecting a preposition to introduce the non-subject argument (to or with, respectively).
However, with respect to 'be friends', I think it's likely that an interesting process of idiomatization involving category change has occurred. In (5),
(5) X and Y are friends
the plural nominal predicate 'friends' is appropriate because the subject is plural. It's behaving like the straightforward nominal predicate in (6) below: plural subject, plural predicate. Singular predicates are not possible with plural subjects, and vice versa; nominal predicates must agree in number with their subjects:
(6) a. X and Y are doctors
b. *X and Y are a doctor.
c. *X is doctors.
d. X is a doctor.
Originally, X and Y are friends would have been symmetrical with the following singular and truly nominal predicate constructions, entailment-wise:
(7) a. X is a friend to Y
b. Y is a friend to X
These are behaving appropriately: nominal predicates with singular subjects take singular predicates, as in (6). Note that a plural predicate here sounds bizarre:
(8) *X is friends to Y
So why is "X is friends with Y" grammatical? If 'be friends with' is a real nominal predicate, it shouldn't be able to take a singular subject.
I think "X and Y are friends" must have undergone reanalysis, to become an idiomatic predicate adjective construction where the plural -s isn't really doing its plural job. That is, in the minds of speakers, [[friend]N -s]Pl changed to [friends]A .
A predicate adjective isn't marked for number with respect to its subject in English, the way nominal predicates are, so that the faux-plural form 'friends', now a predicate adjective, became acceptable with a singular subject. Hence, X is friends with Y is grammatical.
The proof that 'friends' is not the same nominal as 'friend' here is that as a predicate, it takes a different preposition than 'friend', namely 'with', rather than 'to' (though the change in preposition in itself doesn't confirm that friends is now an adjective in this expression -- but it does confirm that it's a different friend than the one in be a friend to).
If you're interested in the subtler semantic issues with symmetric predicates, you might check out the following, which I found with a quick Google: an abstract from Martin Hackl about these and other similar but not-the-same plural-subject predicates (from which I got the term 'essentially plural').
So a linguist's Sunday morning goes.
PS. What does it mean that 'be friends with' can be modified by 'best' without invoking a uniqueness presupposition?
John and Mary are best friends/Mary is best friends with John.
You can't do that with normal nominal predicates: John and Mary are *(the) best doctors.
And, of course, 'best' can't modify adjectives, being the superlative form of one itself. That's kind of problematic for my friends-as-adjective hypothesis.
"Be best friends (with)" is related, semantically, to the following real predicate nominal expressions, which do require the the:
John is the best friend of Mary.
Mary is the best friend of John.
Maybe it's just that 'be best friends' is also an idiom.
Posted by Heidi Harley at March 11, 2007 01:17 PM