Adverb position and verb raising

[alex.alsina at trad.upf.es]

The recent message by Aaron Broadwell has raised an interesting
question.

As pointed out by Aaron, one of the fundamental ideas driving the verb
movement analysis in languages like French and Hebrew is the idea that
adjuncts are external to the XP that they modify. In the case of
adjuncts of clauses, this is taken to mean that adjuncts are adjoined to
the VP. This idea is very appealing because it preserves a certain
amount of isomorphism between syntax and semantics: the scope-taking
relation at the semantic level matches the sisterhood relation at the
syntactic level. In other words, the adjunt takes scope over the
constituent that is its syntactic sister. The idea seems to work quite
well in English - I don't know if there are any major problems with it.
The next step in the reasoning is: if it is right for English, it has to
be right for all languages. (As a footnote, let me say that, to be fair
with the underlying philosophy, this is just an instance of the more
general principle: if it is right for one language, it has to be right
for all languages. The only thing is that, misquoting Orwell or someone,
all languages are equal, only some are more so. So, principles that work
for English have a way of acquiring universal status, in spite of
problems raised by other languages, whereas a putative universal
principle proposed on the basis of a language other than English will
not be looked upon as a universal if it does not work for English.)

French and Hebrew pose a problem for the idea that adjuncts are always
adjoined to XP because they allow adjuncts to appear separating the verb
and its complement. Solution: the verb in these languages is not in the
head position of VP, but in I; the I has a VP sister that contains an
initial adjunt. The result is that the verb precedes the adjunct, which
precedes the complement of the verb, and we haven't had to give up the
idea that adjuncts at the clause level are sisters of the VP. The VP
itself is headless or has an empty head, depending on one's
inclinations, but we can live with that.

So, to go back to Aaron's question, can we claim that, in all languages
in which an adjunct can appear between the verb and its complement, the
verb is in I? Is anything gained by making this claim? Without hoping to
give a definitive answer to these questions, I would like to suggest
that, on the basis of Catalan, another Romance language, the answer is
no. Like French, Catalan allows adjuncts to appear between the verb and
its complement, but, unlike what happens in French, there is no contrast
in this respect between finite and nonfinite verb forms: they can all be
followed by an adjunct and complements in this order:

(1) a.  Llegia   sovint aquest poema.
        she read often  this   poem

    b.  Ha      llegit sovint aquest poema.
        she has read   often  this   poem

    c.  Li agrada llegir  sovint aquest poema.
        she likes to read often  this   poem

In order to preserve the VP-adjoined adjunct idea, we would have to say
that nonfinite clauses also have an I position, in which infinitives
like "llegir" appear in (1c), and that nonfinite forms appear in I even
when there already is an auxiliary there, as in (1b). In (1b) both the
auxiliary "ha" and the main verb "llegit" would have to be in I,
somewhat stretching the idea that I is a preterminal node and should
only dominate lexical items.

And the end result would be that verbs would always be outside their VP
in Catalan (and Spanish, which is similar enough in this respect). In
other words, VPs would always be headless (or headed by a null verb) in
such a language. I'm not sure how happy we would be with this idea.

In some frameworks there seems to be a greater commitment than in other
frameworks to the idea that clause-level adjuncts are always adjoined to
XP. It is clear that LFG, as a framework, is not committed to such an
idea and, in general, is less inclined than frameworks such as GB/MP
towards the idea that grammatical relations and properties of argument
structure should have a fixed universal representation in phrase
structure terms. From Joan's message, it appears that the VP-adjoined
adjunct hypothesis has not been adopted in some recent LFG proposals. In
such proposals, adverb position would not be a good diagnostic for VP
boundaries, since they allow the possibility of adjoining adjuncts to V
and V'.

Another, related, question is whether we want to rule out the flat
ternary structure [V Adv NP] in favor of the binary structures that Joan
sketches in her message. It would be good to study the arguments on
either side.

Alex Alsina