Agentivity and intentionality

[Scott Delancey]

OK, this isn't that complicated.  I'm certainly not arguing that the
concept of intention plays no part in the syntax of any language.
As Tom points out, that's not a likely notion to be coming from someone
who works on Tibetan.

But it's abundantly clear that intentionality is not part of the
semantic role (or whatever you choose to call them) Agent, and I
honestly have a very hard time understanding how anyone could suppose
that it is.

If that were the case, then given a sentence like:

        Oh, shit!  I broke it!

we would have to conclude either a) that it entails an attribution of
intentionality to the subject, or b) that the subject is not an Agent.
(a) is obviously empirically false.  (b) is a recipe for disaster.
If _I_ in that sentence is not an Agent, what is it?  There are only
two solutions here.  One is to posit a new semantic role here--but
casually adding a new role every time we encounter an analytical problem
will quickly render any theory of semantic roles completely vacuous.
The other is to assign it to some other role--maybe Dative, because
it's a conscious participant?  The problem is that there seem to be
vanishingly few languages that would treat this argument that way.
Lots of languages, of course, mark the "subjects" of predicates like
'like', 'need', etc., as something other than Agents, which gives us
concrete evidence that they *are* something other than Agents.  And
so we would expect that, if arguments like my _I_ here are not Agents,
some substantial set of languages would reflect that in their surface
marking.  Not true, as far as I know.
   So identifying this argument as something other than an Agent
amounts to abandoning any aspirations to a principled account of the
relation between semantic roles and surface grammatical relations.
Definitely not a good idea.

Scott DeLancey