Gerunds (was Re: Filler-gap mismatches)

[Malouf R.]

Hi,

Yehuda N. Falk writes:
 > Well, I think we agree that the issue is ultimately empirical. The problem
 > at this point is the interpretation of the empirical data. I don't know
 > enough about the Arabic construction to say anything, although as your
 > disagreement with Joan about the constituency of the Dagaare mixed category
 > shows the question of phrasal coherence depends on what constituency one
 > hypothesizes.

Exactly.  Interpreting facts about phrase structure is so dependent on all the
other assumptions you make, that I'd be surprised if there are any really
convincing factual arguments to be made one way or the other.  Just as in the
case of filler-gap mismatches, there are theoretical and methodological
reasons to prefer one analysis over another, but not much in the way of
knock-down empirical arguments.

 > I have been looking recently at the Hebrew action nominal, and will be
 > presenting an analysis at the LFG conference (and the Israeli theoretical
 > linguistics conference). The Hebrew action nominal is morphologically a
 > noun and its phrase has the distribution of NP. My analysis is NP-over-VP,
 > with the VP cohead adjoined to NP. As I understand lexical coherence, the
 > Hebrew construction does not display it. In a (very small) nutshell, here
 > is how the construction works: The arguments can be realized either like
 > verbal arguments (except that the SUBJ looks like a possessor) or like
 > nominal arguments. The nominal-type realization of arguments is by far the
 > preferred one. Regardless of the realization of arguments, the nominal can
 > take either adjectival or adverbial modification. Given the constituency
 > that I argue for, the construction *does* display phrasal coherence: when
 > there are both an accusative OBJ and an adverb, they must be adjacent (and
 > in my analysis, form a VP).

Interesting!  I look forward to seeing the paper.  But from your synopsis, it
sounds like the Hebrew action nominal does in fact obey the lexical coherence
constraint.  It combines with its subject like a noun does and its objects
either like a verb does or like a noun does.  That's what the lexical
coherence constraint predicts you'll find.

Now, what I don't predict is the behavior of the modifiers.  It's not that I
predict you won't find this behavior, it's just that I don't make any
predictions about modifiers at all.  So at worst to account for this
construction I'll need to say something extra. Of course, there are many many
properties of particular gerund constructions which are not predicted by
either approach, so whether or not the failure to predict this particular
property is a serious problem depends on your point of view.  For instance,
Italian shows a similar distribution of adjectival and adverbial modifiers in
nominalizations, but I've argued that this actually follows from completely
general properties of the syntax of Italian NPs. The inventory of
constructions in Italian is such that some modification possibilities are
licensed and others aren't.  Since it doesn't really have anything to do with
gerunds specifically, the failure of an analysis of gerunds to predict it is
no embarrassment.  I don't know whether a similar argument could be made for
Hebrew, though.

 > So I think at best the direction that the empirical evidence will point
 > when it is better understoof is not yet clear, but it seems to me that
 > Hebrew works better under an LFG-type analysis.

Could well be.  And there are other approaches out there which are getting
hold of other aspects of the problem (e.g., Lizanne Kaiser's dissertation
presents an analysis which works great for Korean nominalizations). But I
don't think we're yet at a point where we can say the facts really point in
the direction of a Right Answer.

--
Rob Malouf
malouf at let.rug.nl