adverb position and verb raising

[Danièle Godard]

Since French is mentioned as a language showing that the V has a different
syntactic position depending on finiteness, it is necessary to point out
the fragility of empirical arguments based on the distribution of French
adverbs.

(i) The Catalan data offered by Alex Alsina have parallels in French: an
adverb such as 'souvent', which can left-adjoin to VPinf, can also occur
between a Vinf or a Vpart. and its complement.
The answer that has been given is the following: there are other functional
projections where the Vnon-fin can move.
For Pollock 89, the Vinf in (1b) moves to Agr:

(1)a.  Paul se rejouissait de souvent aller au Japon
       Paul was happy      of often   go(ing) to J.

  b.   Paul se rejouissait d'aller souvent au Japon

  c.   Paul est allé souvent au J.
       Paul is  gone often   to J.

conclusion: Infl is not sufficient; FP have to be multiplied.
(see Iatridou, LI 1990).

(ii) There are adverbs which occur postverbally, but cannot left-adjoin to
the VP. Since they can be VP-final, they will be analyzed as  adjoined to
the right of the VP. Clear cases are 'resultative' adverbs, which are
predicates of an event resulting from the action associated with the V
(Geuder, 2000).

(2)a.  Paul avait resolu de ranger les livres horizontalement
       Paul had   decided to put the books horizontally
       (note: the adverb is not obligatory with 'ranger', which
       means 'to put st in place')

   b.  * Paul avait resolu d'horizontalement ranger les livres

   c.  Paul avait resolu de ranger horizontalement les livres

   d.  Paul rangera les livres horizontalement
       Paul will-put the books  horizontally

   e.  Paul rangera horizontalement les livres
       Paul will-put horizontally   the books

To get (2c) from (2a), or (2e) from (2d), one will say that the complement
NP is moved (extraposed) to the right, over the adv.

conclusion: not only the V's, but the complements must have different
positions.
(as recognized by Pollock, LI 89; interestingly, G. Bouma in his recent
message to the lfg list notes that an account of adverbs in Dutch and
German also requires BOTH V raising and NP scrambling).

(iii) Going back to adv's such as 'souvent/frequemment' which can occur to
the left of VPinf, we observe that they can also occur at the end of the
VP; in other words, they can be right adjoined:

(3)a.  Il se rejouissait d'aller au J. tres souvent
       He was happy      of go   to J. very often
   b.  Il se rejouissait d'aller tres souvent au J.
   c.  Il se rejouissait de tres souvent aller au J.

To get (3b), we can either start from (3a), and move the complement PP over
the right adjoined adverb, or from (3c) and move the V over the
left-adjoined adv.

conclusion: the analysis of adverbs as adjuncts to VP leads to spurious
structural ambiguities.
(Abeille-Godard, WCCFL 94)

(iv) If we accept the hypothesis that an adjunct must have scope (at least)
over the head (what it is adjoined to), there is a test for adjunction to
VP: an adverb adjoined to VP has scope (can have scope) over a conjunction
of VP. While most adverbs which occur before Vinf can have scope over such
a conjunction, this is not always the case in French. There is a class of
adverbs which cannot (the 'bien' class of degree-manner adverbs):

(4)a.  Il reve de frequemment aller au cinema et faire des
       rencontres interessantes
       he dreams of frequently go to the movies and meet
       interesting people
       (wide scope possible)

   b.  ?? Il reve de beaucoup aller au cinema et partir en voyage
          He dreams of a-lot go to the movies and go away traveling
       (no wide scope)

   c.  Il reve de beaucoup aller au cinema et de (beaucoup) partir
       en voyage

If the complement is a conjunction of VP[de], 'de' has to be repeated. It
is not repeated in (4a), because the adverb frequemment is adjoined to VP.
On the other hand, the adverb is adjoined to the lexical Vinf in (4b), not
to the VP, and the 'de' should be repeated.

(4a)             VP[de]
              /          \
          Adv               VP
           |           /        \
                      VP          VP
     de+freq.  [aller au cinema] [et faire des rencontres ]


(4c)                VP[de]
       	      /                    \
          VP[de]                 VP[de]
        /         \             /      \
       V           PP          V       PP
     /    \                    |        |
   Adv      V       |          |        |
    |       |       |          |        |
de+beaucoup aller au cinema  et de+partir en voyage


(4b)            * VP[de]
              /             \
        VP[de]                 	VP
       /        \                |
      V           PP             |
    /     \                      |
   Adv      V      |             |
    |       |
de+beaucoup aller au cinema [et partir en voyage]


conclusion: adjunction to lexical V must be allowed; occurrence to the left
of Vinf (or another cat) is not in itself sufficient evidence in favor of
adjunction to VP (or another phrase).
(Abeille-Godard, 1997, in D. Forget, P. Hirschbuhler, F. Martineau et M-L.
Rivero (eds), Negation and Polarity, 1-27. Amsterdam: J. Benjamins;
Abeille-Godard, 2001, in J. Camps and C. Wiltshire (ed). Syntax, semantics
and L2 acquisition in Romance. Amsterdam: J. Benjamins).

(v) The piece of data that has been said to show evidence of fin V movement
is the following: adverbs seem to adjoin to the left of VPinf, but not to
the left of VPfin.

(5)a. Paul se rejouit de frequemment aller au J.
   b. ?? Paul frequemment va au J.

However, this is more an illusion than a solid piece of evidence.
The truth of the matter is that we must distinguish between adverbs with an
'incident' prosody, and adverbs with an 'integrated' prosody (something
like that has always appeared necessary, but has not been systematically
taken into account). If we make this distinction seriously, we see that
there are very few adverbs in French which can be adjoined to Sfin with an
integrated prosody: with a few exceptions, they are incidents. In other
words, the distinction is not between VPinf and VPfin, it is between Sfin
and VPfin on the one hand (projections of Vfin, if one accepts that V is
the head of S) and VPinf, on the other. Now, if we distinguish between
syntax and prosody, there is no reason to say that incident adverbs are not
adjuncts: they are adjuncts with an incident prosody, as opposed to
adjuncts with an integrated prosody. Adverbs in a general way are incident
adjuncts when the head is Sfin or VPfin, and they are integrated adjuncts
when the head is VPinf. If we note incident prosody with a comma, we have
the following:

(6)a. Frequemment, Paul va au J.
     Frequently, Paul goes to J.
   b. Paul, frequemment, va au J.
   c. ?? Frequemment Paul va au J.
   d. Paul frequemment va au J.

(7)a. ?? Paul se rejouit de frequemment, aller au J.
   b. Paul se rejouit de frequemment aller au J.

(Bonami-Godard-Kampers, in prep, in Corblin-de Swart, handbook of French
semantics).
If the contrast in (5) points towards a difference in prosody rather than
syntax, there is no evidence in favor of V movement based on adverb
position in general.

In conclusion: What is wrong with having postverbal adverbs at the same
level as complement NPs and PPs? Certainly, one has to work on the
syntax-semantics interface, but, at least, there is a chance to have a more
adequate analysis.

Daniele Godard and Anne Abeille