[Avery Andrews 960430]
(Rick Marken 960425.1445)
It seems to me that Rick's posting might provide the explanation for
why language use is so closely tied to awareness (the ability to talk
cogently is a pretty good guide to whether people are in a normal
conscious state). A control hierarchy such as the one in Rick's
spreadsheets or the various kinds of arm-models that there are available
provides `error handlers' for a fixed number of control systems; each
error handler takes an error signal as input, perhaps combines it with
some other stuff, massages it (by slowing, for example), and sends off
the (typically vector-valued) result as contributions to the reference
levels of other (lower level) control systems.
A fixed structure of error-handlers is not however sufficient to cope
with the survival problems afforded by nature. Reorganization provides
one mechanism whereby new perceptions and error-handlers can be
developed, or old ones adapted; it's probably sufficient to explain
how reptiles (but not amphibians) can adapt to having the flexor and
extensor nerves to their legs swapped. But reorganization is clearly
rather slow, taking place over periods measured in days (learning to
stay on you new boogie-board in the surf) to decades (becoming a top-
grade Tai Chi master), while big land animals at least face all sorts of
complicated problems that have to be sorted out in time periods measured
in minutes to seconds.
So the obvious conclusion is that awareness is what it feels like to be
the technology that evolution came up with to solve these hard problems
fast enough (this of course leaves untouched the philosophical problem of
why it feels like anything to be this technology, but I don't think
anyone is ever going to get anywhere with that problem).
So what about language? Language involves the communication of what
might be called `propositional knowledge', built out of, among other
things, Categories, Relationships and somewhat mysterious entities
I'll call Indexes (representing individuals). See the document
http://www-csli.stanford.edu/users/andrews/pctsem.txt
for some discussion of this.
Now with one Category, DOG, and and Relationship, BITE, you can
construct an indefinite range of propositions:
BITE(x,x),DOG(x) (a dog bites itself)
BITE(x,y),DOG(x),DOG(y) (a dog bites another dog)
BITE(x,y),BITE(y,x),DOG(x),DOG(y)
BITE(x,x),BITE(y,y),DOG(x),DOG(y)
...
Even if we stop at the subitation limit of around seven indexes, that's
a fair number of potential perceptions, each of which can be involved
a reference level (something you want to see, prevent from happening,
etc.). Add more concepts to the mix and the numbers increase
spectacularly, & there's clearly no way at all that a fixed hierarchy
of the kind we actually know how to model can deal with it.
So on this story, the evolution of awareness would be a precondition for
the evolution of human propositional language, because the fixed
hierarchies can't deal with the astronomical numbers of perceptions
provided by propositional systems.
Avery.Andrews@anu.edu.au