[From: Avery Andrews (920318:130615)
Bruce Nevin (920317.0914.28)
A Jackendovian semantic structure is basically just a sentence with
explicit & unambiguous head-argument relationships, plus some
semantic typing. `John betrayed Mary' might look like:
[event BETRAY([person JOHN], [person MARY])]
where the BETRAY should be seen as an abbreviation for
or pointer to a semantic decomposition of the meaning of the word
into more primitive elements, and the lowercase words after the
brackets indicate the ontological categories of the constitutents.
The verb betray might have a lexical entry like this:
Category: V
Form: /betre:/
Meaning: [event Does-sth-to(X,Y)
Y falls into the power of Y's enemies]
(J has a fairly elaborate theory of verbal lexical entries, which I'm
fudging here_.
the noun `betrayal' would have the same meaning, but a different
Form & Category, leading to differences in how the arguments get
expressed.
I agree with a lot of what you say about semantic metalanguages,
but it seems to me that there's a lot to be said for at least
including fully explicit indication of coreference, predicate-argument
structure dependencies, etc. One of the tasks I see of semantics
is in fact to explain inferenceing abilities (the ones we actually have,
like simple cases of modus ponens and universal instantiation), and
I don't see how this is possible without adding the above adornments
to the semantic metalanguage (I think that some of the Wierzbickians
around here are beginning to get some of this message, at last).
Jackedendoff also wants to get an explicit representation of various
patterns he sees, such as oppositions in the system of spatial
concepts expressed by prepositions:
in into out of
on onto off of
He would certainly agree that the components of the semantic
representational system have to be grounded in perception - actually
doing this is one of his major interests, especially for prepositions,
and has published some stuff with Barbara Landau on this (I can't
recall where off hand).
I see two major weaknesses in his approach. The first is that
he has nothing much to say about how linguistically derived information
actually gets integrated from what is coming from the senses, and what
is available from memory. I don't see any problems of principle here,
and assume that he just hasn't gotten that far yet.
The second is that he seems to underplay the social aspect
of meaning. For example, we on CSGNet can all refer to & say true
things about plutonium, but I doubt that many of us have perceptual
abilities that would enable us to distinguish it from enriched uranium without
killing ourselves (well, I'm sure Bill could figure out how to do it,
but I bet he'd spend a lot of time in the library before trying!!).
The standard philosophical story about how this
is possible is that we can talk about plutonium because we are
properly plugged into a social system that contains people who actually
can tell plutonium from other things, and this story sounds right to me.
This second weakness implies that at least some words will contain
`irreducibly symbolic' components that actually can't be cashed out
in terms of perceptions, or at least perceptions of normal speakers
(non-experts). I suspect that even in the minds of real experts about
solid subjects, theoretical terms aren't really perceptually grounded
either. What is the perceptual grounding of `electron'? What is
perceptually grounded is the description of various experimental
setups and their outcomes, the best explanation for which is the
existence of electrons. (This puts me in bed with those philosophers
of science who think there really is a difference between language-of-theory
and language-of-observational after all). So I see the term `electron'
as having a rather complicated and indirect connection to perception, &
suspect that it would be wrong to try to `expand' it into anything like
perceptual terms. I'd sort of like to try to go on about this at greater
length sometimes, but no more time, now.
If I get around to reading a reasonable amount of recent Harris, perhaps
I'll insist that you read some Jackendoff!!
Avery.Andrews@anu.edu.au