Object Shift

[Andrew Carnie]

From: LANOONAN at IRLEARN.UCD.IE

a belated reply...

Last month there was a posting from Andrew on object shift in Irish,
where he presented an analysis of his and Heidi Harley's that
hypothesises an Aspectphrase and an AGRoP internal to a Larsonian
VP layer. Here are a couple of comments and issues I'd like to raise.
(At first I should point out that I have also argued for a
Split-VP-shell with an internal AGRoP to account for Irish word order
in infinitives in both my thesis (1992) and in a WCCFL (1994) paper
('VP internal and external AGRoP: evidence from Irish').)

Andrew mentioned that there are two environments where we find overt
object shift - in infinitives such as (1)

  (1)    Is maith liom an leabhar a le/amh

and in perfective constructions, such as the following.

  (2)   Ta/ se/ tar e/is an leabhar a le/amh

It is proposed that the perfective element 'tar e/is' heads an inner
functional aspectual projection (a la Travis) and that the substantive
verb (bi/ (ta/)) heads an upper Larsonian VP shell and can thus raise
to AGRs without violating the HMC.

Note, however, that the element 'tar e/is' is generally a preposition
that takes various different complements (NPs, small clause). In the
example below it takes a VN-complement where the external argument,
appearing with the preposition 'do', follows 'tar e/is':

  (3) tar e/is dhom an leabhar a le/amh.
      after   to-me the book   aL read     'after I read the book'

Since the external argument here follows the element 'tar e/is', 'tar
e/is' could not constitute an 'inner aspectual head', but is clearly
used as a preposition taking an infinitival complement.

The question now arises concerning example (2), whether there is
any compelling reason to analyse 'tar e/is' as an aspectual head as
opposed to an aspectual preposition taking an infinitival complement
clause here. Compare the Hiberno English 'perfective' construction
(presumably arisen from the Irish construction) involving the
preposition 'after' followed by a gerundive clause: 'I'm after
getting money from the bank'; or other prepositions taking infin-
itive complement clauses, like the in aspectual 'about' in 'I'm
about to leave'. Is there any evidence/reason to assume that 'tar
e/is' in (2) is any different from 'after' and 'about' in these
examples?

Note that if it were the case that tar e/is is a preposition taking
an infinitival complement clause in both constructions, then the
second instance of overt object shift mentioned above reduces to the
first, so that there is a single environment of overt object shift,
namely in infinitives.

I have a further comment on the issue of perfectives: note
that there is another perfective construction - traditionally called
the perfective passive - involving the auxiliary (or substantive verb)
'bi/' and a perfective participle form of the verb (termed verbal
adjective in traditional grammar). In this type of construction the
internal argument appears in nominative  Case and the subject in a PP
headed by the possessive preposition 'ag' (hence the term 'perfective
passive').

  (4) Ta/  an leabhar le/ite agam.
      is   the book   read   at me

It should be noted that aspectually these constructions relate more
to the English perfective construction (in fact to the Hiberno
English construction: She has that book read') than to English
passives. I have argued in various places (e.g. my thesis and in a
another WCCFL paper (1992)) that perfective and passive participles
(I assume them to be one and the same element) fail to project an
upper VP shell containing the external argument. In fact, they only
project an aspectual projection containing the internal VP shell (thus
similar to what Andrew noted, but the head here is the participial
morphology (or - the head where a participle checks its aspectual
features) As a result, participles are universally like unaccusatives
(lacking an external argument). Taking a configurational, rather than
a lexical approach to accusativity, I argue that in order to exhibit
accusative perfective structures (as e.g. English does) a language
must have some 'extra'mechanism to provide accusativity. In English,
German, French and related languages this is accomplished by
transitive properties of the auxiliary 'have'. In particular, I argue
that 'have' accomplishes this precisely by acting as a 'filler' verb
heading the upper VP shell in transitive/unergative verbs, taking a
possessor argument in its specifier, thus providing a syntactic
geometry for accusativity.  Leaving the precise details of my
analysis aside, the crucial fact here is that auxiliary (and main
verb) 'have' is lacking in the lexicon of Irish. The  only
Irish auxiliary is 'bi/', which has an argument structure just
like English auxiliary 'be'; it takes as a single argument a
predicative complement whose argument raises to get nominative Case.
I thus propose to reduce the absence of accusative perfectives to the
absence of auxiliary 'have' in Irish. As a result, the perfective
construction exhibits an ergative pattern, where the external
argument appears in a PP headed by the possessive preposition
ag, and the internal argument gets nominative Case (an ergative/
accusative split with perfective aspect is not uncommon in languages -
see e.g. Hindi (Mahajan 1990)).

Going back to Carnie & Harley's proposal about 'bi/', note that what I
have claimed about 'have' is very similar to what they seem to
suggest about 'bi/': a 'filler' verb for the higher VP shell.
However, bi/ has none of the properties of 'have'. In particular it
neither takes an external argument, nor should it function as a
filler verb for the outer (external argument) VP shell. Their
analysis is thus quite incompatible with my account for the lack of
transitive perfectives and transitive possessive constructions in
Irish.
Any reactions to this .... ?

cheerio,

maire noonan.

------- End of Forwarded Message