Pronouns Postpose at PF

1 Introducing Pronoun Postposing

The phenomenon of weak pronoun postposing in Irish (1) and Scottish Gaelic (2) (Chung and McCloskey 1987, Duffield 1995, Adger 1997, Doyle 1998, 2002, McCloskey 1999) is unusual in that it involves the rightward placement of a prosodically and informationally light element.

1. 1.

   1. a. Bhris  sé  an  chathaoir leis  an  ord     aréir.
      broke he the chair     with the hammer last night
      'He broke the chair with the hammer last night.'

   2. b. Bhris  sé  leis  an  ord     aréir     í.
      broke he with the hammer last night it-FEM
      'He broke it with the hammer last night.'

2. 2.

   1. a. Chunnaic mian  t-each anns a'   gharradh an dè.
      saw      I   the horse  in the    garden   yesterday
      'I saw the horse in the garden yesterday.'

   2. b. Chunnaic mi anns a'   gharradh an dè     e.
      saw       I   in   the garden   yesterday it-MASC
      'I saw it in the garden yesterday.'

Treating this movement at face value, as a rightward and syntactic movement (as do Chung and McCloskey (1987)), makes it theoretically extremely unusual. Most of the authors mentioned above have instead queried either the rightwardness of the movement (Duffield 1995, Kayne 2000) or its syntactic nature (Adger 1997, McCloskey 1999). In this squib, I offer a new argument for the latter position.

The argument comes from the interaction of pronoun postposing with ellipsis. I show that the derivational point at which ellipsis takes place must precede pronoun postposing (within any local domain) and further that ellipsis has the character of a PF deletion operation. On the assumption that syntactic and PF processes cannot interleave within a single local domain, it follows that pronoun postposing is postsyntactic. I focus on examples from Scottish Gaelic in this squib.

2 Scottish Gaelic Is Not Pro-Drop

Since we are going to look at operations that delete pronouns, we must first be sure that Scottish Gaelic is not, for the relevant cases, pro-drop. This can easily be seen from the following examples:

1. 3. Chunnaic *(mi) an  t-each.
   saw        I   the horse
   'I saw the horse.'

2. 4. Chunnaic an t-each *(mi).
   saw        the horse    I
   'The horse saw me.'

The same pattern is replicated with the other pronouns, with one exception. The only cases where a pronoun can be dropped in a simple finite clause are the first person singular and plural conditional. In these cases, pronouns are obligatorily dropped (see Hale and McCloskey 1984 for Irish and Adger 2000 for Gaelic).

(5)

1. a. Bhuailinn    (*mi) an t-each.
   hit-1SG.COND   I   the horse
   'I would hit the horse.'

2. b. Bhuaileamaid (*sinn) an  t-each.
   hit-1PL.COND     we   the horse
   'We would hit the horse.'

3 The Target of Ellipsis

As shown by McCloskey (1991), Irish has a process that targets the complement of the position where the finite verb is situated. This is found in responses to questions. The same phenomenon is found in Gaelic.

1. 6. An d'fhuair thu  deoch fhathast?
   Q   got      you drink  yet
   'Have you had a drink yet?'

2. 7.

   1. a. Fhuair.
      got
      'Yes.'

   2. b. Cha d'fhuair.
      NEG got
      'No.'

Specifically, McCloskey shows that these constructions involve ellipsis of the category that is complement to the position of the finite verb. That is, on the assumption that the verb is in T, the responsives are formed by eliding the FP complement of T, whose specifier the subject fills. (McCloskey (2001) argues that FP is not VP or vP in Irish, but this is irrelevant for our purposes here. I will assume it is VP with a VP-internal subject, for concreteness.)

(8)

There are two broad approaches to the analysis of ellipsis. One idea is that ellipsis is a simple PF deletion process (see, e.g., Merchant 2001); the other is that it involves a null pronominal element in the syntax, with semantic procedures to allow resolution of the ellipsis (see, e.g., Lobeck 1995). As noted by Merchant (2001:71), the latter approach cannot be correct for the structures under investigation here. If it were, the structure would look as follows, where represents a null pronominal VP...