Headline on the front page of the New York Times today:
Exclusive Golf Course Is Also Organic, So a Weed or Two Get In
My first reaction — really, why I noticed the head in the first place — was that I would have written gets (sg.) rather than get (pl.), and I’m still inclined that way, though I’m not willing to say that get is unacceptable or non-standard. I do have a hypothesis about where the plural might have come from.
I have no record of complaints about either one of the variants, and MWDEU doesn’t seem to take up subj-verb agreement with subjects of the form
(1) a/one Nsg or two Øpl
(where Øpl indicates an elliptical N that’s understood as plural).
Here’s how you predict a plural verb for subjects of the form (1):
(a) syntactically and semantically, (1) is just a disjunction (with or) of two NPs, the first (NP1) singular, the second (NP2) plural;
(b) ordinarily, a disjunctive subject NP1 or NP2 in combination with a V is understood as a disjunction of the two NPn+V combinations — that is, it’s understood like [NP1+V] or [NP2+V] (disjunction distributes semantically over the disjuncts);
(c) hence, disjunction distributes syntactically over the disjuncts as well, so that each disjunct separately governs the form of the V;
(d) but when the disjuncts end up requiring phonologically distinct forms of the V, the conflict is resolved by some principle, and Agreement With the Nearest (NP2 in this case) is the principle most often invoked in the handbooks (there’s a considerable literature on the matter, which is taken up by MWDEU, and in several Language Log postings).
Applying all this to (1): it’s just an ordinary disjunction, so disjunction distributes semantically and syntactically over the two disjuncts, a/one Nsg and two Øpl, with the result that the V GET must have a form that agrees with each disjunct separately. There’s then a conflict, since the first disjunct requires a singular form of GET, the second a plural form, and for GET in the present tense these are phonologically distinct — a conflict that can be resolved by Agreement With the Nearest, in favor of the plural form. Which is why the headline has get instead of gets.
The main problem comes at the very beginning of the argument.
NPs of the form (1) are not transparently ordinary disjunction, but look like instances of “approximative or“, as in “Only five people or so get in” ‘About five people get in’. (Or “Only five or so people get in” or, for that matter, “One or two weeds get in” — though in these cases plurality is determined on the head N of the subject by principles of determiner-noun agreement within the NP, which, yes, require Agreement With the Nearest; and then the V agrees with this head N.)
In defense of the claim that it’s not ordinary disjunction, note that either can’t be used with the first disjunct, nor can the disjuncts be reversed:
(2) *Either a weed or two get in.
(3) *Two (weeds) or one weed get in.
Approximative disjunction is one case, of many, where the apparent syntax of some construct is not concordant with its semantics. In such cases, we need to say that there’s a construction (a pairing of form with meaning) in addition to the ordinary one: approximative disjunction in addition to garden-variety disjunction.
There’s a lot more to be said about approximative disjunction and the forms it takes, but for my purposes here we can go on to asking how things like (1) function for the purposes of subject-verb agreement.
My intuition is that in this case the first disjunct is central, and that the second acts like an afterthought or parenthetical, as in
(4) (i) A/One weed, or two, gets in.
vs. (ii) ?A/One weed, or two, get in.
where (ii) has an Agreement With the Nearest going into a parenthetical, something that is very common but strikes many people as a performance error; I have piles of uncomfortable examples from real life (and so do the handbooks).
But so far that’s only a leading idea, which has to be fleshed out.