[From Bill Powers (930907.0830 MDT)]
Hans Blom (930907) --
Just a short question. Do you think that, physiologically, it
might be possible that effector information might be available
to higher levels of the nervous system? Does the information
that the brain has available include what I call the output of
the organism, i.e. its actions? If not, physiology would rule
against my "input-output" identification scheme.
If by "effector information" you mean something other than
sensory information, my answer would be no. However, in the
vertebrate system, the states of effectors, muscles, are
represented in two dimensions by sensory signals: the tendon
signals that represent applied force, and the length or spindle
signals that represent muscle length plus its first derivative.
At what looks like the second level (or maybe the third), the
angles of joints are represented as sensory signals.
So the first aspect of your question can be answered yes. There
are perceptual signals very closely related to effector
activities, although of course they don't measure the actual
state (of contraction) of the effector or immediate effects of
output on the world (one exception: touch and pressure sensors,
which directly measure, by Newton's laws, both the applied force
and the reaction force when the skin is pressed against an
object).
Are these sensory signals available to higher-level systems?
Again, the answer seems to be yes. The signals reaching the
spinal comparators from tendon receptors, and possibly stretch
receptors (I don't remember), branch just as they enter the
spinal cord, with one branch reaching the motor neuron
(comparator) and the other ascending the spinal cord toward other
(and higher) systems. The ascending branch is known to reach the
motor cortex in the brain, which in turn sends perceptual signals
to, among other places, the midbrain where some third-order
systems are (parts of the motor cortex seem to be "misplaced"
second-level systems). Control of directed force is control of
signals in these upgoing branches.
In general, effector actions are most directly represented by
kinesthetic sensory signals. However, more global aspects of
effector activities are monitored by other senses: the visual
sense is often overlooked. Churchland (-man?)'s crab, for
example, is designed to sense the position of an object, convert
it through sensory-motor mapping into actions, and thus move a
claw toward the object. The claw itself, apparently, is invisible
to the crab. If the claw were also represented visually, getting
the claw to the object would become much simpler! In the Little
Man model, the controlled variable at the highest level is
explicitly the finger-target distance as seen, with a reference
level that can be zero (finger on target), non-zero (finger some
specific distance from target in x-y-z) or continually changing
(finger describing a circle around the target). If you think of
the whole arm as the effector, its visual position is sensed.
Your input-output scheme can be used if you use a hierarchical
model. The immediate effects of outputs can produce sensory
signals that become inputs to higher systems, along with other
sensory signals representing more remote effects. So a higher
system could perceive the relationship between a sensed output
and a more remote sensed effect of that output. We have proposed
some such arrangement for systems that detect reversals in the
sign of external feedback loops: the relationship between
immediately sensed effects of outputs and more indirect effects
reverses, a detectable circumstance. The cursor normally moves up
when the mouse moves toward the positive y axis: both elements of
that relationship are perceived. It is easy to perceive when the
relationship reverses, although the detection takes time (roughly
0.4 second, according to Marken's measurements).
As to forgetting, I don't think it is a separate problem from
leaning. If learning is reorganization, then the random walk that
converges to one kind of organization can just as easily cause
that same organization to change again, converging to some
different organization. Of course there are many kinds of
phenomena, probably unrelated, that are called forgetting.
Forgetting a telephone number is probably quite different from
forgetting how to do a cat's cradle with a yo-yo. Common-language
words like forgetting were not designed to differentiate
theoretically-different phenomena.
ยทยทยท
----------------------------------------------------------------
Best,
Bill P.