[Avery Andrews 930693.1016]
(Tom Anastasio, via Gary Cziko 930505.1925)
for Gary to forward to Tom...
>No one will argue against the idea that a driver uses negative feedback to
>steer a car down the road.
But, in effect, they do, although this is hard to believe. At the
lowest levels, there seems to be no serious problems, but when you
get into the higher levels of the `organization of behavior' (in scare
quotes because `org. of beh.' is not a real concept from the point of
view of pct), the wheels seem to come off.
For example, there's a mostly nice and useful article by Abbs and
Winstein (1990: `Functional Contributions of Rapid and Automatic
Sensory-Based Adjustments to Motor Output', in Jeannerod, M (ed)
_Attention and Performance_ XIII, 627-652, which summarizes a lot
of information about the speed of various feedback loops (refuting
claims that were commonly made by motor control people in the late
seventies and early eighties that feedback was `too slow' to be
useful for various things).
But then they go on (I don't have the page reference at the moment, &
the book is out of the library) to say that feedback as `technically
defined' to be restricted to situations where corrections are made
at the point where errors are introduced, whatever that is supposed
to mean. I think it's supposed to allow steam-engine governers and
stretch reflexes as feedback, but their example of what is supposed to
be *excluded* is `distributed compensatory responses', where, when
an effect is produced by two articulators acting jointly, one will
respond to compensate for disturbances applied to the other. For
example if someone is trying to articulate a `w' sound, and one
lip is restrained, the other will move further to produce the
desired degree of closure (Abbs and Winstein have done a lot of work
on this, and similar situations, such as precision grasping).
They are 100% explicit in wanting to define feedback so as to be
inapplicable to this kind of situation, even though the `technical
definition' corresponds to nothing in the technical literature known
to us, and doesn't make sense anyway (it rules out the thermostat,
everybody's favorite example, since a disturbance applied at the
front door, when it is opened, gets corrected in the basement, when
the furnace comes on), and it's pretty obvious how a feedback analysis
of this sort of thing would work (but surely not easy to figure out
which one is right, and show how it is physiologicallay realized!).
The effect of restricting feedback in this way is to make it irrelevant
to the higher level of behavioral organization, since almost everything
at higher levels is produced by cooperative actions of various (e.g.
people can and do control the position of the steering wheel with two
hands and two knees, acting at once or in various combinations).
There's more to say in this vein, but I think people have to check this
stuff out for themselves and draw their own conclusions, so I'll stop
here for now, except I'll make a final observation that there seems
to be a big gap between physiologists, who are pretty cluey about
feedback, and motor control people, who are not. For example,
the Schmidt (1988) textbook on motor control devotes a fair amount of
space to feedback, but you couldn't possibly learn how simple feedback
systems work from studying it, and he seems to have bought the restricted
`technical definition' too (without attributing it to a source I
can check).
Something else to think about is that shortly after BCP appeared, the
motor control community went holus bolus for `motor programming theory',
which at the outside denied that feedback had any significant
involvement in the production of ordinary behavior - look at the
paper by R.A. Schmidt in Stelmach (ed) (1980) _Tutorials in Motor Behavior_
North-Holland (1980) for some discussion. This paper
seems to me so far to be the beginning of the retreat from the initial
extreme position of motor programming theory - what I find
incomprehensible is that anyone was able to take this position seriously
at all in the mid seventies, when not only did Bill's stuff exist in
an accessible form, but also N. Bernstein's stuff was quite well known
and often cited, but not, it would seem, understood.
Avery.Andrews@anu.edu.au