April 9, 2004		
Yoshiyuki Moriyama 
John Whitman                                   Department of Linguistics
Cornell University                           University of Massachusetts

                     Null Objects in Japanese Revisited

It has been known for many years that unpronounced arguments in languages
such as Chinese, Japanese, and Korean give rise to sloppy identity
readings, as in (1b) from Japanese. 

(1) a.  Mary wa  zibun no  kuruma o   aratta.
        Mary TOP self  GEN car    ACC washed
       'Mary washed self's car.'

    b.  John mo   [e] aratta.
        John also     washed
          (i) 'John washed it (Mary's car) too.'
         (ii) 'John washed self's car too.'

Accounts of these readings break down into two types. One approach
attempts to assimilate them to ellipsis phenomena such as VP Ellipsis in
English (Whitman 1988, Huang 1988, Otani and Whitman 1991). The
alternative approach attempts to describe the empty arguments in (1b) as a
special type of empty pronoun (Xu 1989, Hoji 1998, Tomioka 1998). Kim
(1999) represents an intermediate position. 

The arguments marshaled by Hoji, Kim, and Tomioka and others show that
(1b) is not an instance of VP Ellipsis. But they do not show that ellipsis
is not involved. There are numerous problems with the analysis of the
empty argument in (1b) as a subspecies of pro (aside from the initial
problem of accounting for the sloppy reading in (1b) itself. 

First, it is clear that some type of parallelism is required for the
sloppy reading to obtain. (1a) continued as (2) below does not allow the
sloppy reading. 

(2)  Atode     John wa [e] notta.
     Afterward John TOP    rode
       (i) 'Afterward John rode (in it).'
      (ii) NOT 'Afterward John rode in self's car.'

Second, the empty argument can in some cases be indefinite, as pointed out
by Hoji:. 

(3)  John mo   [e] aratta. Demo dare no  kuruma ka wakar-anai.	(Hoji 1998)
     John also     washed  but  who  GEN car    Q  know-not
    'John washed a car/some cars too. But I don't know whose car(s).'

Indefinite interpretation of the empty argument is rather precisely
restricted: 

(4) a.  Mary wa  san-satu no  hon  o   yonda.
        Mary TOP 3-CL     GEN book ACC read
       'Mary read three books.'
    b.  John mo   [e] yonda.
        John also     read
          (i)  'John read them (the same three books that Mary read) too.'
         (ii)  'John read three books too.'
        (iii)  'John read (some) books too.'

The 'indefinite' readings of the empty argument (4b ii and iii)
countenance John reading an unspecified number of books or John reading
the same number of books that Mary read. Even if we allow for the
possibility of an indefinite pro, it is not clear how the interpretive
restrictions on this pro are to be established. 

In this paper we argue that the null arguments in (1b), 3, and (4b) all
involve ellipsis - NP Ellipsis (nŽe N' Ellipsis). Our analysis is built on
three premises, only one of which is likely controversial: 

(5) a. Japanese (and Chinese, Korean?) has NP Ellipsis 
       (Saito and Murasugi 1991).
    b. These languages have pro.
    c. These languages have a null indefinite determiner INDEF.

On the covert NP Ellipsis analysis, (1b), (3), and (4b) are interpreted 
as follows:

(1b) John mo   [DP pro [NP e]] aratta.
     John also                 washed
    'John washed his too.'

(3)  John mo   [DP INDEF [NP e]] aratta.
     John also                   washed
    'John washed one/some too.'

(3)  John mo   [DP INDEF [NP e]] yonda.
     John also                   read
       (i) 'John read one/some too.'
      (ii) 'John read three too.'