RNR and V-coordination

We have just finished reading Bruening 2018 here at UW (“The Lexicalist Hypothesis: Both Wrong and Superfluous”, Language) and found an interesting bit of data in it. Consider the following (from p.28):

(1a) Paul read and annotated two linguistics books.
(1b) Paul read, and Mary annotated, two linguistics books.

The claim in the paper is that in (1a) there are exactly two books involved whereas in (1b) there could two or four — that is the books that were read and the books that were annotated might not be the same. Some, but not all, of us agreed with those judgments, but let’s take them as given for the sake of argumentation.

Bruening’s argumentation around this example turns on a misunderstanding of what V coordination means in HPSG. (Specifically, he seems to assume that the category V must be lexical, and that nothing with more than one word in it can still be a V.)

This did lead us to wonder about the current analysis of V coordination and right node raising in the ERG. My understanding is that the ERG assimilates V coordination to RNR because otherwise there would always be systematic ambiguity where both analyses are available. Given that, is there a way to capture the contrast in ambiguity in (1a,b)?

I’ve always assumed that the ERG presupposes some sort of
underspecified account of distributivity in the semantics, so that
we can get the 12 pizzas reading of as well as the 4 pizzas in
total reading from the MRS for:

  1. three boys ate four pizzas

For the sake of this discussion, I don’t think it matters exactly
what that account is. Whatever it is, there should also be a four
books reading for:

  1. Paul and Mary read two books.

and (I suppose) for

  1. Paul read, and Mary read, two books

however that’s constructed syntactically. It would seem bizarre
to allow distributivity with the 2 and not with 3.

Under such an account, I can imagine allowing the four books
reading for:

  1. Paul read and annotated two books.

but I can also imagine claiming that “read and annotated” is a
composite event that can’t be split up in such a way as to allow
the four books reading.

This might be the sort of evidence that might distinguish between
possible underspecified distributivity accounts.

All best,


This know this isn’t the main point of this thread, but “read and annotated” is a problematic example for getting judgements, because you can hardly annotate something without reading it. “read and bought” might be better, because you can read something without buying it, and buy something without reading it.

1 Like

Thanks, Ann. Agreed that this might be useful evidence around those questions.

p.s. Guy’s point about the example is well-taken. (The example was as given in the original paper.)