Kent Mcclelland's paper

[Avery Andrews 960210]

I haven't finished reading Kent's paper yet, so maybe I should keep
quiet until I have, but I'll speak up now in case I've missed something
that somebody will point out to get me back on the track quickly.

1. I think it's somewhat overoptimistic about the PCT model of the
nervous system, which has the big problem that we (PCT workers in
particular, and everybody else, as far as I can see) don't know much
at all about how the perceptual systems work. As was discussed
a few years ago, we don't know how to build a fridge-door recognizer,
so we don't know how to build a system to control the degree to
which the fridge-door is open. Even worse for `hi level' things like
language.

2. I don't like using the word `accomodation' to refer to situations
where different controllers merely have non-conflicting goals, such
as the housemates A and B, where A wants fresh paint and clean windows,
and B wants nice grass and tidy garden-beds, and there's plenty of time
and money for both references to be satisfied. In this case the
references are just independent of each other, & there's not much
interesting to say, I think.

What is interesting is when references start out at least somewhat
incompatible, and get harmonized. For example Cindy & I might want
to move a table (back) into position; we would presumably agree
on its orientiation (square with the room), so there'd be identical
references for that variable, but we wouldn't necessarily have
*identical* references for its position. But we wouldn't take it
to some place between our different references and then just stand
there pulling (and maybe shouting at each other), tho I've certainly
seen children behaving this way. We'd either just put it down
somewhere, or, if both of us really cared about the location, we'd
have a negotation, put it where we agreed, and walk away. Since
the control stops, the references must have changed to become the
consistent (I think we maybe need a some terminological development
to deal with different systems' perceptions of the same CEV),
and I'd rather see `accomodation' used to refer to situations where
people harmonize previously divergent reference-levels (this is
consonant with the use of this term in sociolinguistics, where it
refers to a proces whereby people mutually adjust their pronunciation,
etc. towards each other's dialect).

Avery.Andrews@anu.edu.au