A simple form of reorganization

[From Bruce Abbott (970714.1145 EST)]

Habituation is a form of nonassociative learning in which the response to a
stimulus weakens across repeated presentations of an innocuous stimulus.
The cellular mechanisms underlying habituation have been studied by Eric
Kandel and his associates using the sea slug, aplysia californica, chosen
because it has a relatively simple nervous system of around 20,000 central
nerve cells and because it has a defensive reflex -- the closing of its
respiratory mantle in response to touch of the siphon or mantle skin -- that
exhibits habituation to repeated touch.

The basic circuit includes sensory neurons that respond to touch,
interneurons, and motor neurons whose activation contracts the mantle
muscles. The sensory neurons have direct, excitatory synaptic connections
to the motor neurons and to the interneurons. The interneurons, in turn,
also have excitatory synaptic connections to the same motor neurons; thus
there are both direct and indirect pathways from the sensory neurons to the
motor neurons.

When the siphon skin is touched repeatedly, the initially strong contraction
of the mantle shelf weakens and may even disappear. A few touches may
produce habituation that lasts up to an hour (short-term habituation);
repeating this process a few times produces habituation that may last a week
(long-term habituation). Both forms turn out to result from decreased
release of neurotransmitter from the presynaptic terminals of the sensory
neurons, at both the direct (motor neuron) and indirect (interneuron) sites.
In addition, long-term habituation involves a decrease in the number of
synaptic junctions present at each site.

Although the (unhabituated) circuit can be viewed as an ordinary feedback
control mechanism that keeps the level of touch of the siphon skin near
zero, the _function_ of the system would appear to be that of protecting the
respiratory ogran (gill) from being injured (by, for example, the nibbles of
small fish). However, closing off the gill is expensive in terms of
muscular effort and loss of respiration, and to repeatedly respond to
innocuous stimulation (e.g., by seaweed waving in the currents) wastes
resources. By turning down the "gain" of the sensory transducers in the
presence of repeated innocuous stimulation, aplysia is able to conserve
these resources.

If what actually needs to be controlled is pain arising from attack on the
gill structure, then the siphon touch-control system can be viewed as a
feedforward mechanism in which one controlled variable (touch) is being used
in order to better control another variable (pain/gill damage) which in the
evolutionary history of aplysia is usually preceded by siphon/mantle touch.
Habituation reduces the sensitivity of the feedforward system when the
situation would seem to require that action.

These changes do not appear to be associated with persistent error in
intrinsic variables, and so would not seem to fit the classic HPCT model of
reorganization.

Sluggishly yours,

Bruce

[From Bruce Gregory 9970714.1320 EDT)]

Bruce Abbott (970714.1145 EST)

If what actually needs to be controlled is pain arising from attack on the
gill structure, then the siphon touch-control system can be viewed as a
feedforward mechanism in which one controlled variable (touch) is being used
in order to better control another variable (pain/gill damage) which in the
evolutionary history of aplysia is usually preceded by siphon/mantle touch.
Habituation reduces the sensitivity of the feedforward system when the
situation would seem to require that action.

If one assumes that discomfort is associated with the existence
of an error signal, resetting the reference level associated
with the response would reduce the error and the discomfort.

Bruce

[From Bruce Abbott (970714.1350 EST)]

Bruce Gregory 9970714.1320 EDT --

If one assumes that discomfort is associated with the existence
of an error signal, resetting the reference level associated
with the response would reduce the error and the discomfort.

The rapid retraction of the mantle quickly removes the source of stimulation
(siphon touch), so this particular control system rapidly removes the error.
Error does not persist, ergo, there is nothing to drive the classic
reorganization process.

You will need to explain to me how "discomfort" differs from "error" and how
its introduction in association with error is supposed to resolve the problem.

Regards,

Bruce

[From Bruce Gregory (970714.1510 EDT)]

Bruce Abbott (970714.1350 EST)

The rapid retraction of the mantle quickly removes the source of stimulation
(siphon touch), so this particular control system rapidly removes the error.
Error does not persist, ergo, there is nothing to drive the classic
reorganization process.

You seem to be assuming that there is no hierarchy -- that no
higher level system has a reference level for not experiencing
continuing bouts of "pain" in your terms -- error. I make no
such assumption. If your analysis were correct there would be
no reason for me not to keep sticking my hand into a
flame, since I could withdraw it promptly and reduce the error.

You will need to explain to me how "discomfort" differs from "error" and how
its introduction in association with error is supposed to resolve the problem.

You used the word pain. I used discomfort. How does pain relate
to the error signal in your view?

Bruce

[From Bill Powers (970714.1345 MDT)]

Bruce Abbott (970714.1145 EST)

The basic circuit includes sensory neurons that respond to touch,
interneurons, and motor neurons whose activation contracts the mantle
muscles. The sensory neurons have direct, excitatory synaptic
connections to the motor neurons and to the interneurons. The
interneurons, in turn, also have excitatory synaptic connections to >the

same motor neurons; thus there are both direct and indirect >pathways from
the sensory neurons to the motor neurons.

When the siphon skin is touched repeatedly, the initially strong
contraction of the mantle shelf weakens and may even disappear. A few
touches may produce habituation that lasts up to an hour (short-term
habituation); repeating this process a few times produces habituation
that may last a week (long-term habituation).

Your description of the connections from the sensory neurons to the
interneurons and muscles is not sufficient to allow modeling this
phenomenon. Also, you do not say whether there are any _other_ connections
beside those you mention -- any higher-order loops.

Functionally, what we seem to have here is a form of low-pass filter.
Widely-space touches allow the touch signals to get through unattenuated.
Spacing them closer together results in a loss of amplitude response.

However, there are many ways to get this effect -- so many that we would
need much more information about the circuitry to guess at the actual
mechanism.

Both forms turn out to result from decreased
release of neurotransmitter from the presynaptic terminals of the >sensory

neurons, at both the direct (motor neuron) and indirect >(interneuron) sites.

This doesn't tell us much unless we know what is causing the decrease in
release of neurotransmitter. Is there is fixed pool of it that can resupply
the vesicles only very slowly, so a few releases exhaust the local supply?
Is the decrease caused by a decrease in the frequency of incoming signals?
Is there a separate chemical path that affects the rate of release of
neurotransmitter at a given input signal frequency? And what activates this
separate path, if it exists?

In addition, long-term habituation involves a decrease in the number >of

synaptic junctions present at each site.

This is the only effect that looks pretty much like reorganization.

Although the (unhabituated) circuit can be viewed as an ordinary >feedback

control mechanism that keeps the level of touch of the siphon >skin near zero,

A feedback control mechanism with a fixed reference level of zero is
indistinguishable from a stimulus-response system. There is, of course,
negative feedback through the effect of muscle contraction on the stimulus,
but if the reference level can never be set to _produce_ a non-zero amount
of touch, this system can't be part of a hierarchy of control.

.. the _function_ of the system would appear to be that
of protecting the respiratory ogran (gill) from being injured (by, for
example, the nibbles of small fish). However, closing off the gill is
expensive in >terms of muscular effort and loss of respiration, and to
repeatedly
respond to innocuous stimulation (e.g., by seaweed waving in the
currents) wastes resources. By turning down the "gain" of the sensory
transducers in the presence of repeated innocuous stimulation, aplysia >is

able to conserve these resources.

I am unmoved by such "functional" arguments. Given something that happens,
that has consequences, you can always argue that the consequences were the
reason for the event, as if there were some designer trying to achieve a
design goal. I doubt very much whether Aplysia knows anything about its
resources or tries to conserve them. Or that anyone else does, even Prof.
Evolution.

If you could show that the gill-control system has a reference level (or a
gain) set by the output of a higher system that is controlling some other
variable also affected by touches, then we could figure out what goal _of
Aplesia_ was being satisfied by the observed changes. The data you have
presented don't seem enough to let us do that.

Best,

Bill P.

[From Bruce Abbott (970714.2000 EST)]

Bruce Gregory (970714.1510 EDT) --

Bruce Abbott (970714.1350 EST)

The rapid retraction of the mantle quickly removes the source of stimulation
(siphon touch), so this particular control system rapidly removes the error.
Error does not persist, ergo, there is nothing to drive the classic
reorganization process.

You seem to be assuming that there is no hierarchy -- that no
higher level system has a reference level for not experiencing
continuing bouts of "pain" in your terms -- error. I make no
such assumption. If your analysis were correct there would be
no reason for me not to keep sticking my hand into a
flame, since I could withdraw it promptly and reduce the error.

Pain? We're talking about a light touch. When painful stimulation is
involved, you do _not_ get habituation. So the example we're dealing with
is not analogous to sticking your hand into a flame, and my analysis is not
damaged by your observation that such actions do not habituate.

Years ago I bought a grandfather clock that chimes every quarter hour and
counts out the hour on the hour. The first night, Steph and I didn't sleep
well because of all the racket that clock put out, but now we don't even
notice it when we're _awake_. That's habituation.

You will need to explain to me how "discomfort" differs from "error" and how
its introduction in association with error is supposed to resolve the

problem.

You used the word pain. I used discomfort. How does pain relate
to the error signal in your view?

Pain is a perceptual signal arising from nerve fibers that respond to
certain forms of mechanical deformation or chemical reactions arising from
tissue damage. It is not an error signal.

Regards,

Bruce

[From Bruce Abbott (970714.2005 EST)]

From Bill Powers (970714.1345 MDT) --

Bruce Abbott (970714.1145 EST)

Your description of the connections from the sensory neurons to the
interneurons and muscles is not sufficient to allow modeling this
phenomenon.

What additional information do you need?

Also, you do not say whether there are any _other_ connections
beside those you mention -- any higher-order loops.

There are other connections (of course), but these are not relevant to the
phenomenon, which has been traced to changes taking place in the sensory
cell itself when it is repeatedly stimulated at relatively short intervals.

Functionally, what we seem to have here is a form of low-pass filter.
Widely-space touches allow the touch signals to get through unattenuated.
Spacing them closer together results in a loss of amplitude response.

I like that description, but it is not quite accurate: the loss of
amplitude response may persist for some time and attenuate the amplitude
even to widely spaced stimulations that by themselves would not have
produced measurable habituation.

However, there are many ways to get this effect -- so many that we would
need much more information about the circuitry to guess at the actual
mechanism.

Some details remain to be worked out, but the mechanism involves (among
other things) a chemical deactivation of calcium channels in the membrane of
the sensory cell. Calcium must flow into the cell is needed to trigger
release of the neurotransmitter.

Both forms turn out to result from decreased
release of neurotransmitter from the presynaptic terminals of the >sensory

neurons, at both the direct (motor neuron) and indirect >(interneuron) sites.

This doesn't tell us much unless we know what is causing the decrease in
release of neurotransmitter. Is there is fixed pool of it that can resupply
the vesicles only very slowly, so a few releases exhaust the local supply?

No. However, the changes that take place within the sensory cell's
presynaptic terminals during habituation include a decreased ability to
transport neurotransmitter vesicles into the active zone where it can be
available for release.

Is the decrease caused by a decrease in the frequency of incoming signals?

No.

Is there a separate chemical path that affects the rate of release of
neurotransmitter at a given input signal frequency? And what activates this
separate path, if it exists?

I don't know the details, but the process would appear to be frequency
dependent. However, it is not simple exhaustion.

In addition, long-term habituation involves a decrease in the number >of

synaptic junctions present at each site.

This is the only effect that looks pretty much like reorganization.

O.K., so let's restrict our attention to long-term habituation, if you like.

Although the (unhabituated) circuit can be viewed as an ordinary
feedback control mechanism that keeps the level of touch of the siphon
skin near zero,

A feedback control mechanism with a fixed reference level of zero is
indistinguishable from a stimulus-response system. There is, of course,
negative feedback through the effect of muscle contraction on the stimulus,
but if the reference level can never be set to _produce_ a non-zero amount
of touch, this system can't be part of a hierarchy of control.

But it can still be a significant part of a _system_ of control, yes?

.. the _function_ of the system would appear to be that
of protecting the respiratory ogran (gill) from being injured (by, for
example, the nibbles of small fish). However, closing off the gill is
expensive in terms of muscular effort and loss of respiration, and to
repeatedly respond to innocuous stimulation (e.g., by seaweed waving in the
currents) wastes resources. By turning down the "gain" of the sensory
transducers in the presence of repeated innocuous stimulation, aplysia
is able to conserve these resources.

I am unmoved by such "functional" arguments.

Yes, I know. Too bad.

Given something that happens,
that has consequences, you can always argue that the consequences were the
reason for the event, as if there were some designer trying to achieve a
design goal.

There are ways to put functional hypotheses to test. If I deduced that eyes
are for seeing, I suppose you would argue that this is poppycock. Perhaps
they are there merely to provide support for the eyelids.

I doubt very much whether Aplysia knows anything about its
resources or tries to conserve them. Or that anyone else does, even Prof.
Evolution.

Bill, you are being silly. My supposition does not depend on what Aplysia
"knows" about its resources or "tries" to do about them. However, it
doesn't matter what either you or I think is the reason the system has
evolved as it has; the facts are that touching the siphon or mantle surface
leads to retraction, and that a retracted gill is more protected from injury
than an open one. If you want to think that this gill retraction has no
function in the life of Aplysia, or that Aplysia does it to entertain
scientists, fine, I won't argue.

If you could show that the gill-control system has a reference level (or a
gain) set by the output of a higher system that is controlling some other
variable also affected by touches, then we could figure out what goal _of
Aplesia_ was being satisfied by the observed changes. The data you have
presented don't seem enough to let us do that.

The mantle must be open for the gill to do its job of respiration; the
little control system I described closes the gill temporarily in response to
touch. Repeated stimulation that is not accompanied by strong input these
or other sensory receptors leads to habituation of the response, through
mechanisms involving the attenuation of neurotransmitter release at the
sensory presynaptic terminals and the reduction of synapses. This
"reorganization" occurs in the absence of any apparent disturbance to
intrinsic variables, and does not depend on input from "higher" systems. I
don't see how the additional data you say you need would alter any of that.

Regards,

Bruce

[From Bill Powers (970714.1953 MDT)]

Bruce Abbott (970714.2005 EST)--

Your description of the connections from the sensory neurons to the
interneurons and muscles is not sufficient to allow modeling this
phenomenon.

What additional information do you need?

We need to understand more of the whole system. As you said, for example,
there are other neural pathways than the simple one you described, and they
could well be acting on the basic loop. In more complex organisms, it's
clear that there are connections from sensory nerves to muscles, some
through a single synapse and others through internuncials. But these loops
can't explain behavior all by themselves; we can't even say what they will
do without knowing how their side-effects are treated by higher systems
which contribute to the reference inputs (alpha and gamma) of the spinal
loops, and perhaps even alter the chemical environment of these loops.

Also, you do not say whether there are any _other_ connections
beside those you mention -- any higher-order loops.

There are other connections (of course), but these are not relevant to
the phenomenon, which has been traced to changes taking place in the
sensory cell itself when it is repeatedly stimulated at relatively >short

intervals.

But what is _causing_ those changes? The causes could well be outputs of
higher systems of either neural or chemical nature. There's no way to
understand a complex system until every variable has been traced to the
output of some system function, or has been determined to be independent
(originating from outside the system). The picture you present is full of
dangling variables.

Functionally, what we seem to have here is a form of low-pass filter.
Widely-space touches allow the touch signals to get through
unattenuated. Spacing them closer together results in a loss of
amplitude response.

I like that description, but it is not quite accurate: the loss of
amplitude response may persist for some time and attenuate the >amplitude

even to widely spaced stimulations that by themselves would >not have
produced measurable habituation.

You're right. The effect is more like that of an automatic gain control
with a fast "on" time and a long "off" time constant.

However, there are many ways to get this effect -- so many that we
would need much more information about the circuitry to guess at the
actual mechanism.

Some details remain to be worked out, but the mechanism involves >(among

other things) a chemical deactivation of calcium channels in >the membrane
of the sensory cell. Calcium must flow into the cell is >needed to trigger
release of the neurotransmitter.

Fine, that takes us one step toward completing the system analysis. Now
what chemical is causing the deactivation, and what it is that varies the
concentration of that chemical? And so on and so on, until you have either
closed a loop or traced the source to something outside the system. You
can't understand the system by looking at it through a keyhole.

Is there a fixed pool of [neurotransmitter] that can resupply
the vesicles only very slowly, so a few releases exhaust the local >>supply?

No. However, the changes that take place within the sensory cell's
presynaptic terminals during habituation include a decreased ability >to

transport neurotransmitter vesicles into the active zone where it >can be
available for release.

And what is causing that decrease in ability?

Is the decrease caused by a decrease in the frequency of incoming signals?

No.

OK, so it is something that directly affects the production/reception of
neurotransmitters that is the direct cause of the loss of sensitivity. Now
we have to identify that something and find out what IT is the output of.

Is there a separate chemical path that affects the rate of release of
neurotransmitter at a given input signal frequency? And what >>activates

this separate path, if it exists?

I don't know the details, but the process would appear to be frequency
dependent. However, it is not simple exhaustion.

Right, but it doesn't "just happen." Something else is making it happen.

In addition, long-term habituation involves a decrease in the number
of synaptic junctions present at each site.

This is the only effect that looks pretty much like reorganization.

O.K., so let's restrict our attention to long-term habituation, if you >like.

OK, but that admits of many different interpretations.

···

------------------------------------------

A feedback control mechanism with a fixed reference level of zero is
indistinguishable from a stimulus-response system.

But it can still be a significant part of a _system_ of control, yes?

No. Relative to an organized hierarchy, for example, it is like an
independent parasite; its behavior is entirely localized and not subject to
manipulation by the rest of the hierarchy. It's like a separate organism,
not part of a whole system.
-----------------------------------

.. the _function_ of the system would appear to be that
of ...

I am unmoved by such "functional" arguments.

Yes, I know. Too bad.

There are ways to put functional hypotheses to test. If I deduced >that

eyes are for seeing, I suppose you would argue that this is >poppycock.
Perhaps they are there merely to provide support for the >eyelids.

Since they do both, you can't argue that they do NOT provide support for
the eyelids. Eyes provide visual signals, eyes provide support for eyelids,
eyes provide signals from an actor to an audience about emotions, eyes
provide one element of crying, and so on. But seeing did not exist, none of
these "functions" existed, before eyes existed; it was the other way around.

This "functional" stuff is an awkward way to talk about purposive or
goal-driven behavior when there is actually purpose, and it's a way of
asserting purposes or goals when there are none. What is the "function" of
a ruler? Is it to draw straight lines, to measure distances, or to prop a
window open? The function of a ruler is whatever function it is made to
serve by the system using it.

I doubt very much whether Aplysia knows anything about its
resources or tries to conserve them. Or that anyone else does, even
Prof. Evolution.

Bill, you are being silly. My supposition does not depend on what
Aplysia "knows" about its resources or "tries" to do about them.
However, it doesn't matter what either you or I think is the reason >the

system has evolved as it has; the facts are that touching the >siphon or
mantle surface leads to retraction, and that a retracted >gill is more
protected from injury than an open one. If you want to >think that this
gill retraction has no function in the life of >Aplysia, or that Aplysia
does it to entertain scientists, fine, I >won't argue.

The only way I would ever agree that Aplysia retracts its siphon to keep
its gill from being injured would be if you showed me the reference signal
for degree of injury, and a perceptual signal to go with it, and the
control system that closes the loop by operating the right muscles.
"Functional" comments about siphon retraction are an opinion of the
observer and have nothing to do with how the system works or how it got
that way. The "functional" approach is subjective.

The mantle must be open for the gill to do its job of respiration; the
little control system I described closes the gill temporarily in >response

to touch.

So the function is to interrupt respiration?

Repeated stimulation that is not accompanied by
strong input these or other sensory receptors leads to habituation of >the

response, through mechanisms involving the attenuation of >neurotransmitter
release at the sensory presynaptic terminals and the >reduction of
synapses. This "reorganization" occurs in the absence of >any apparent
disturbance to intrinsic variables, and does not depend >on input from
"higher" systems. I don't see how the additional data >you say you need
would alter any of that.

It wouldn't alter this fragmentary picture of part of the system. It would
just tell us how it really works. And by the way, since this is a perfectly
reproducible and predictable phenomenon, no reorganization is involved.

Best,

Bill P.

[Martin Taylor 970715 10:05

Bill Powers (970714.1953 MDT)

The only way I would ever agree that Aplysia retracts its siphon to keep
its gill from being injured would be if you showed me the reference signal
for degree of injury, and a perceptual signal to go with it, and the
control system that closes the loop by operating the right muscles.

I take it, then, that you would not agree with any proposition that
suggests a perceptual control system or hierarchy comes to control
what it controls so as to keep some intrinsic variable near its evolved
reference level?

Martin

[Hans Blom, 970715b]

(Bruce Abbott (970714.1145 EST))

Habituation is a form of nonassociative learning in which the
response to a stimulus weakens across repeated presentations of an
innocuous stimulus.

Bruce, thanks for an interesting post. I have, however, always found
it strange that habituation is considered _nonassociative_ learning.
In my opinion, learning is always and by definition associative: it
is adjusting behavior based on perception (or imagination, maybe?) of
consequences.

If I may anthropomorphize greatly: by default, Aplysia has a "worst
case" scenario in its imagination -- being touched is being bitten --
and its reaction is initially based on this "internal model",
withdraw from "danger". The closing of the mantle cannot be
permanent, however, because Aplysia needs to breathe and eat as well.
In the face of this "conflict", the environment has to be "resampled"
periodically in order to establish whether the supposed "danger" has
gone or is not dangerous after all. In the latter case, Aplysia
performs a Test -- will opening the mantle lead to being bitten? --
and as a result its "internal model" may be readjusted: what was
initially considered to be dangerous may prove to be harmless after
all.

I know that I'm not supposed to entertain these thoughts and ought to
focus on mechanism (proximate goals, the "how" question) rather than
ultimate goals (the "why" question). But I seem incapable to resist
entertaining the (heritical) thought that "hows" serve a "why"...

Anyway, I believe that even habituation behavior must be associated
to something that is relevant to Aplysia. Whether that relevance is
observed by an individual Aplysia [which would make it feedback] or
has been "observed" by the evolutionary process [which would make it
feedforward] is another question. Or is it a matter of terminology,
where nonassociative is more or less associated with feedforward?

I would approach this problem by noticing the basic conflict between
_using_ the environment (breathing, eating) and _being used_ by the
environment (being eaten) and finding the "optimal" solution to this
conflict, where "optimal" is decidedly in terms of progeny survival.
But then, some might object, how can Aplysia be so smart that it can
perform such an extremely difficult kind of optimization? :wink:

Greetings,

Hans

[From Bruce Gregory (970715.1015 EDT)]

Bruce Abbott (970714.2000 EST)

Years ago I bought a grandfather clock that chimes every quarter hour and
counts out the hour on the hour. The first night, Steph and I didn't sleep
well because of all the racket that clock put out, but now we don't even
notice it when we're _awake_. That's habituation.

Is habituation a control phenomenon? It's not obvious to me that
it is. What disturbance is being resisted that demonstrates
control? If is not an example of control, why should HPCT be
relevant?

Bruce

[Hans Blom, 970715c]

(Bill Powers (970714.1345 MDT))

A feedback control mechanism with a fixed reference level of zero is
indistinguishable from a stimulus-response system.

Yes, that appears to be a very helpful rule. The "of zero" can be
left out, however, so that we get:

   A feedback control mechanism with a fixed reference level is
   indistinguishable from a stimulus-response system.

It works the other way around as well:

   A stimulus-response system is indistinguishable from a feedback
   control mechanism with a fixed reference level.

These two rules let me translate back and forth between SR theory and
control theory.

Greetings,

Hans

[From Bill Powers (970715.0857 MDT)]

Martin Taylor 970715 10:05--

I take it, then, that you would not agree with any proposition that
suggests a perceptual control system or hierarchy comes to control
what it controls so as to keep some intrinsic variable near its >evolved

reference level?

If you will recall, the model I propose says that there is an intrinsic
control system, with reference signals for each critical variable. These
reference signals (in whatever form they take) are the embodied purposes of
the system itself. On the other hand, "survival" is not (I presume)
represented as an embodied reference signal, so it is not a goal of the
organism.

Best,

Bill P.

[From Bill Powers (970715.0905 MDT)]

Hans Blom, 970715c--

Yes, that appears to be a very helpful rule. The "of zero" can be
left out, however, so that we get:

  A feedback control mechanism with a fixed reference level is
  indistinguishable from a stimulus-response system.

It works the other way around as well:

  A stimulus-response system is indistinguishable from a feedback
  control mechanism with a fixed reference level.

These two rules let me translate back and forth between SR theory and
control theory.

I agree with your amendment of the principle.

S-R theory, however, also needs to be written as

R = F(S,R)

rather than

R = F(S).

Best,

Bill P.

[Martin Taylor 970715 16:16]

Bill Powers (970715.0857 MDT)]

Martin Taylor 970715 10:05--

I take it, then, that you would not agree with any proposition that
suggests a perceptual control system or hierarchy comes to control
what it controls so as to keep some intrinsic variable near its >evolved

reference level?

If you will recall, the model I propose says that there is an intrinsic
control system, with reference signals for each critical variable. These
reference signals (in whatever form they take) are the embodied purposes of
the system itself. On the other hand, "survival" is not (I presume)
represented as an embodied reference signal, so it is not a goal of the
organism.

If you will recall, my comment was in regard to the following quote
from Bill Powers (970714.1953 MDT):

The only way I would ever agree that Aplysia retracts its siphon to keep
its gill from being injured would be if you showed me the reference signal
for degree of injury, and a perceptual signal to go with it, and the
control system that closes the loop by operating the right muscles.

The implication of this quote is that the gill-retraction of Aplysia is
not a special case. Equally you could substitute "...that a person
eats in order to keep the blood sugar level up..."

There is no perceptual reference nor perceptual signal for blood sugar
level. The controlled perceptions involved throughout the hierarchy in
the action "eating" nowhere relate to the level of the intrinsic variable.
Intrinsic variables are _by definition_ outside the perceptual control
hierarchy. The perceptual control hierarchy influences the intrinsic
variables in such a way as to keep them near their reference levels
_entirely_ through the side-effects of perceptual control. Reorganization
sustains those characteristics of the perceptual control hierarchy that
have "good" effects on the intrinsic variables, and tends to change those
characteristics that have neutral or bad effects on the intrinsic variables.

I am quite happy to say that one eats in order to keep the blood sugar
up, even though the perceptions being controlled have to do with perceptions
we might call "hunger" or "nice taste." You sounded as if you were
unhappy withthis kind of formulation. I see no difference in principle
between this and the fact that some control system in Aplysia acts in
a way developed through evolution that has the side effect of allowing
Aplysia to "keep its gill from being injured." Both ways, the "in order
to" reason is a side effect of controlling some perception that has been
stabilized by, in the one case reorganization, in the other, evolution.
They work.

Martin

[From Bill Powers (970715.1513 MDT)]

Martin Taylor 970715 16:16]--

If you will recall, my comment was in regard to the following quote
from Bill Powers (970714.1953 MDT):

The only way I would ever agree that Aplysia retracts its siphon to
keep its gill from being injured would be if you showed me the
reference signal for degree of injury, and a perceptual signal to go
with it, and the control system that closes the loop by operating the
right muscles.

The implication of this quote is that the gill-retraction of Aplysia >is

not a special case. Equally you could substitute "...that a person

eats in order to keep the blood sugar level up..."

There is no perceptual reference nor perceptual signal for blood sugar
level. The controlled perceptions involved throughout the hierarchy in
the action "eating" nowhere relate to the level of the intrinsic >variable.

Intrinsic variables are _by definition_ outside the perceptual control
hierarchy. The perceptual control hierarchy influences the intrinsic
variables in such a way as to keep them near their reference levels
_entirely_ through the side-effects of perceptual control.

An understandable interpretation of my words, since I didn't specifically
include biochemical control systems. However, I think "it's control systems
all the way down," and such things as blood sugar are indeed controlled
"perceptions" relative to embodied (meaning physically existent rather than
conceptual) reference signals. The controlled perceptions, as far as I
know, are completely unavailable to awareness, but they fit the basic
definition of perceptual signals except for our usual assumption that we
are talking about the nervous system and any potential link to consciousness.

Reorganization sustains those characteristics of the perceptual >control

hierarchy that have "good" effects on the intrinsic variables, >and tends
to change those characteristics that have neutral or bad >effects on the
intrinsic variables.

The reorganization system was proposed specifically to provide a link
between physiological states and neural learning. However, there are many
other non-neural physiological systems which are perfectly good control
systems -- the "life support" systems.

I am quite happy to say that one eats in order to keep the blood sugar
up, even though the perceptions being controlled have to do with
perceptions we might call "hunger" or "nice taste." You sounded as if >you

were unhappy withthis kind of formulation. I see no difference in

principle between this and the fact that some control system in >Aplysia

acts in a way developed through evolution that has the side >effect of
allowing Aplysia to "keep its gill from being injured." Both >ways, the "in
order to" reason is a side effect of controlling some >perception that has
been stabilized by, in the one case >reorganization, in the other, evolution.

They work.

Reorganization is a control process that maintains specific perceptual
signals (for example, those involving blood sugar concentrations) in a
match with specific reference signals (or the equivalent), whereas
evolution neither perceives its effects nor desires that these effects be
of any particular sort. When you say that something is done or exists "in
order to" produce effect X, you must show that at least in principle there
is a reference condition X*, that X is sensed and compared with it, and
that the resulting error drives the action or produces the condition. To do
this requires a physical mechanism inside the organism (or wherever you
propose to place it).

Intellectually, you may think you eat to keep your blood sugar up (not too
far up, of course), but that is only imagination. You are not able
consciously to sense your blood sugar concentration; all you know about it
comes from perceptions of sensations from the body, which are not specific
to blood sugar concentration. If you feel tired, you may tell yourself that
this is because your blood sugar level is low, but that is only a guess;
what you control for is not being tired. Control of blood sugar
concentration comes about at the biochemical level.

The formulation I object to is exemplified by the statement that organisms
eat in order to survive. This implies (once you understand how control
systems work) that survival is perceived, compared with a reference level
for survival, and controlled by the behavior that results from the error
(although how an organism would react to failure to survive is an
interesting puzzle). Of course a human being may think and reason about
survival using symbol-handling systems, but this is obviously not what is
meant when the organism in question is a bacterium or a rat.

Organisms survive because they control (through reorganization and in other
ways) the physical variables on which survival depends, but not because
they aim to survive. Survival is, barring accidents, a reasonably reliable
consequence of keeping all these variables near their reference levels, but
is not itself, in general, a controlled variable.

It may seem very reasonable to assume that because Aplysia reacts as it
does to touches, it is trying to protect its gills from injury -- that is
certainly one of the effects that is produced, along with others (such as
tending to asphyxiate itself). However, we can't show that any consequence
of behavior is a goal of the system unless we can show that there is a
controlled perception of that consequence; those are the only circumstances
under which we can say that the behavior occurs "in order to" produce the
consequence. If we can't do that, the best we can say is that the
consequence occurs.

The basic problem here is that people talk about purposive behavior without
any clear understanding of the difference between controlled and
uncontrolled consequences. If this difference were clearly understood, we
would use different language in the two cases and I would have nothing to
object to in that regard. As matters stand, we have people using what is
basically the language of purpose, but believing that they are somehow
avoiding attributing purpose to the system just because they aren't using
certain words.

Best,

Bill P.

[Hans Blom, 970716c]

(Bill Powers (970714.1953 MDT))

The only way I would ever agree that Aplysia retracts its siphon to
keep its gill from being injured would be if you showed me the
reference signal for degree of injury, and a perceptual signal to
go with it, and the control system that closes the loop by
operating the right muscles.

You require much stricter evidence than most scientists I know. They
would accept the proposition that Aplysia retracts its siphon to keep
its gill from being injured if a test would show that the degree of
gill injury would significantly and positively correlate with
experimentally or naturally imposed limitations on its ability to
retract its siphon.

This post seems to indicate that you cannot believe that control
operates in a particular case if you have not been able to open and
inspect the "black box" and discovered its mechanics, in particular
the "reference signal line". Most people that I know, however,
believe that control/purposive behavior can be established even if
the "black box" cannot be opened. You cannot open me; do you accept
that I -- weird mutation that I am -- show purposive behavior?

Greetings,

Hans

[Hans Blom, 970716d]

(Bruce Gregory (970715.1015 EDT))

Bruce Abbott (970714.2000 EST): Years ago I bought a grandfather
clock that chimes every quarter hour and counts out the hour on the
hour. The first night, Steph and I didn't sleep well because of
all the racket that clock put out, but now we don't even notice it
when we're _awake_. That's habituation.

Is habituation a control phenomenon? It's not obvious to me that it
is. What disturbance is being resisted that demonstrates control? If
is not an example of control, why should HPCT be relevant?

I consider habituation to be an adjustment of the perceptual
apparatus that "filters out" and discards predictable perceptions
that are unimportant for how we generate actions. It allows us to
refrain from unnecessary mental operations and physical actions. In
Bruce's example, mentally compensating the clock's ticking with an
anti-phase signal (don't take me too literally!) may allow him to
hear things that were previously drowned out by the ticking.

Paradoxically, Bruce may suddenly "hear" his clock (or at least
something unusual) if it _stops_ ticking...

Greetings,

Hans

[From Bill Powers (970716.0757 MDT)]

Hans Blom, 970716c --

You require much stricter evidence than most scientists I know. They
would accept the proposition that Aplysia retracts its siphon to keep
its gill from being injured if a test would show that the degree of
gill injury would significantly and positively correlate with
experimentally or naturally imposed limitations on its ability to
retract its siphon.

This post seems to indicate that you cannot believe that control
operates in a particular case if you have not been able to open and
inspect the "black box" and discovered its mechanics, in particular
the "reference signal line".

That's not it at all. If you want to propose that Aplysia actually has the
ability to perceive "injury" and a reference signal setting the desired
degree of injury, I will not reject that out of hand. All I want is for you
to recognize just what it is you are proposing when you say that Aplysia
reacts in order to protect its gills from injury. Of course when you put
your proposal as I suggest, you may decide that you don't really believe
(as I don't) that Aplysia is sufficiently complex to perceive and control
something as abstract as the general idea of "injury." It might perceive
pain, but I doubt strongly that it could perceive "damage." And I don't
believe that you think it could, either.

Best,

Bill P.
'

[From Bill Powers (970716.0814 MDT)]

Hans Blom, 970716d--

I consider habituation to be an adjustment of the perceptual
apparatus that "filters out" and discards predictable perceptions
that are unimportant for how we generate actions. It allows us to
refrain from unnecessary mental operations and physical actions. In
Bruce's example, mentally compensating the clock's ticking with an
anti-phase signal (don't take me too literally!) may allow him to
hear things that were previously drowned out by the ticking.

This is the same mode of explanation that I have been arguing against in
the discussion of Aplysia. You are describing what you see as plausible
consequences of habituation, and giving them causal force. But if you
subscribe to a control-system model, your proposal amounts to saying that
there is a perception of filtering out, and a reference signal for it, that
results in habituation that accomplishes the desired end (and similarly for
the other desirable consequences of habituation that you mention). I doubt
whether you really mean to make any such assertion, but if you don't make
it, what is the alternative? It is only to say that because such
consequences would be beneficial, the behavior that produces them SOMEHOW
occurs (and, presumably, would not occur if those same consequences were
not beneficial). The only recourse you have then, if you want a scientific
model, is to appeal to some indefinite process like evolution, which can
explain anything we want explained.

Best,

Bill P.