[From Rick Marken (970413.2230 PDT)]
John Anderson (970413.1730) --
Would anyone like to comment from a PCT perspective on the attached
forwarded post?
Sure:
Measurments of reaction times of masters in martial arts show that the
brain would be far to slow to calculate a response and to proceed it.
Reaction times are a rough measure of the transport lag in a control loop
that is controlling (by means of the "response" or some variable related
to the response) a variable that is disturbed by the "stimulus". Whether a
particular reaction time is relevant to a martial arts expert's ability
to control a particular variable depends on which controlled variable and
transport lag you're talking about.
At least there must be supplementary centres independent from the
brain which provide fast information processing in the humen
physiological system.
This assumes that a response (like moving the head to the left) that
compensates for a disturbance (like a swift kick from the right) must be
calculated between the onset of the disturbance and the onset of the
corrective response. But this is not the way behavior works. People
continuously act to control perceptions (like the rate at which images --
like that of the foot moving towards your head -- are expanding). These
perceptions are the simultanous result of variations in disturbances (foot
movement) and responses (movement of the head relative to the oncoming foot,
for example). In order to compute the response that keeps the perception in
its reference state the controller would have to determine, continuously, the
exact contribution of the disturbance to the controlled perception. This
would, indeed, be a very complex computation. But it is not necessary; the
control loop can (and does) produce the appropriate response by continuously
generating responses that are based on the difference between perception
and reference.
Of course, transport lag puts a limit on the frequency of disturbances that
can be successfully resisted. But the transport lag in lower level systems
is quite short (on the order of 10 to 50 msec) so these systems can deal
with rather high frequency (fast) disturbances -- like the rapidly explanding
image of a foot that _would_ be produced if the head were not moving away
at nearly the same time.
An easier theory is that years and years of practice allow martial arts
masters and fighter pilots to build a model of their opponent in their
own brain
This doesn't seem much easier. A model of the opponent that would actually
allow one to compute (in advance) the "correct" outputs would be rather
amazingly complex. It would, for example, have to predict the detailed
trajectories of the movement of all limbs from any position that the
opponnet might be in. I have the feeling that this would tax the
computational capabilities of the fastest super computers.
The basic flaw in this exchange about reaction time and skilled performance
is the notion that human systems are open loop; that a stimulus is followed,
after some delay (RT), by the appropriate response. What is ignored is that
the actual cause of the response -- the sensory effect of the stimulus -- is
being affected by the stimulus AND THE RESPONSE at the same time. That is,
the situation is:
S -->s-->| Human|-->R
^ (RT) |
ยทยทยท
>
--------------
where S is the stimulus, s is the controlled sensory effect of the stimulus
(and response) and R is the response. The RT delay occurs IN THE LOOP!
NOT
S -->s-->| Human|-->R
(RT)
whre the RT delay is time spent producing the R that will correct for S
_in the future_.
Actually, these two discussants have proposed the two main S-R models
of control: information processing and model based output generation.
This is interesting because it shows that people will come up with
the same kinds of explanations of behavior when they think of behavior
in terms of the same underlying model -- S-R (or input-output or whatever
you want to call it).
Best
Rick