Reentrancy in feature structures

[Ash Asudeh]

Martin,

I think I may have chosen unfortunate terminology in my message: when I
said type identity, I didn't mean sort identity, but rather non-token
identity. What I had in mind with my terminology was something like:

type identity (F type= G):
| a         |
| F  | b   ||
|    | H c ||		
| G  | b   ||
|    | H c ||		

token identity (F token= G):
| a      |
| F  [1] |
| G  [1] |

Even on the construal of type identity as sort identity, I don't think
it's promising for raising to have just sort identity. The problem is, you
need a way of specifying that the very thing that is (e.g.) the raised
subject is also the argument of a lower clause. Otherwise, I think you
would end up throwing away the overt argument and perhaps postulating a
null argument where it belongs in the downstairs clause, or some such
thing.

Maybe I'm not being imaginative enough, though.

Ash

On Mon, 15 Jul 2002, Martin Jansche wrote:

>
> On Mon, 15 Jul 2002, Ash Asudeh wrote:
>
> > First, I don't think agreement and reflexive binding are the only
> > places where reentrancy is relevant. Another phenomenon is raising
> > and also possibly control (depending on how it's done). Note that
> > type (i.e. substructure) identity is insufficient for raising.
>
> Is the last point a general claim about sorted feature structures (in
> which case I'd suggest it's false), or is it a claim about a
> particular sort inventory/hierarchy?  Surely you could set up your
> sorts in such a way that type identity is all you need for raising.
>
> - martin
>