What about the start up when we control a perception?5

CSGnet Archive CSG2005
1

[From Bjorn Simonsen (2005.01.06.13:57 EST)]

From Bill Powers (2004.12.18.0759 MST)

I also think I understand why some neurons/areas of the brain/groups of
neurons stabilize their frequency at a certain value.

I don't know what you mean by that. Are you referring to EEGs?

The way I understood "What about the start up when we control a perception?
[From Bruce Abbott (2004.12.21.1800 EST)] and [From Bruce Abbott
(2004.12.22.1940 EST)]" and my own knowledge is (I am sorry I will express
myself at the microscopic view. I see the advantage to express myself at the
circuit level, but I hunt for a technical statement that is based on
microscopic facts):
* (Bruce Abbot) Thus a "spike" of positive charge appears to travel rapidly
down the axon from its base toward the axon terminals.
* (Bruce Abbot) At the axon terminals, the arrival of the action potential
briefly opens calcium gates in the terminal, allowing calcium ions (Ca++) to
flood in. It is this arrival of calcium that activates a mechanism that
releases a small quantity of neurotransmitter from the axon terminals into
the synapse (the connection between neurons).
* (myself) The original electrical signal at the axon terminal is converted
to a chemical signal. A common transmitter is Acetylcholine. The higher the
frequency in the axon, the more acetylcholine is set free. When the
transmitters arrive at the other side of the synapse, they meet protein
molecules called receptors. The transmitters and the receptors react to a
third substance. This third substance has the effect that it initiates the
sodium-potassium pump. Now it comes a voltage potential into being. In this
way one of many electrical impulses run through the dendrites to the cell
body.
* When this electrical impulse reaches the cell body together with thousands
of other signals, it causes a change in voltage in the recipient cell. _If
this change in voltage is sufficient large, the sodium-potassium pump will
cause an Action potential_.
* The action potential is about 90 millivolt. Because the action potential
can't be greater than 90 millivolt, the neurone will generate more of them.
In this way the neuron emit a certain frequency.
This is an (the) explanation how a neuron emit a frequency. Let me continue
on this microscopic view and I and let me end with a comment on the circuit
level (I appreciate your comments).
I know there are thousands of neurons representing a perceptual signal or a
reference signal in PCT. But we talk about perceptual signals and reference
signals as if they are transported from one neurone to another, and I think
that is wrong. The way I understand it is that electrical signals initiates
chemical signals and they initiates new electrical signals in the dendrites.
The new electrical signal in a neuron is not the sum of the electrical
signals in coming from the dendrites. This sum just initiates the
sodium-potassium pump,- if it is great enough. It is the sodium-potassium
pump and the number of sodium-potassium channels that initiates the action
potentials and the cell frequency. _And the frequency is therefore dependent
on the number of sodium-potassium channels and the chemical environment
inside and outside the neuron.
If this explanation is correct, I understand why a reference has a certain
frequency. (I don't talk about one signal in one axon. I talk about the
principle of how a reference representing my purpose to reach for the
toothpaste tube can have the same value today and tomorrow). It is the
environment inside and outside the neuron that sets the reference value. And
the environment doesn't change. Therefore the reference value is the same.
There is also another point I will stress.
If the signals that come in from the dendrites to the cell body is not large
enough to initiate the sodium-potassium pump, there will not emit an action
potential (I still don't talk about one signal in one axon). Dependent on,
if this happens in the input function or in the output function (I don't
think I express myself correct here), I understand how the switches in your
memory model can be switched on or of. Is this nonsense?

Bjorn

2

[From Bill Powers (2005.01.06.0613 MST)]

Bjorn Simonsen (2005.01.06.13:57 EST) -

I seem to have quite a different view of neural signals from yours, but
this may be only a matter of terminology. You appear to be focusing on the
generation of a single impulse (an "action potential" either occurs or
doesn't occur). In my scheme, a neural signal is a series of impulses
occurring at a variable frequency, not (as you seem to be thinking) one
constant frequency.

Let's consider a simplified neuron with only one input. Suppose that
impulses arrive via an excitatory synapse at some low rate. For each
impulse, the post-synaptic ionic concentration rises abruptly and then
decays away exponentially because of metabolism and diffusion. This results
in a rise and fall of the potential at the axon hillock, considerably
smoothed because of the cell-wall capacitance. Some average concentration
and potential are the result.

If the rate at which incoming impulses occur increases, the rise and fall
of concentration happens more frequently -- and because of the exponential
shape of the curve, the average ionic concentration increases, and so does
the average potential at the axon hillock.

Now let's consider the effect of the average potential on the firings of
the cell.

If the average is below the threshold for generating an impulse, no output
impulse will occur. This means that there will be no output impulses even
though input impulses are arriving at some slow rate.

When the average potential reaches some minimum value, the cell will begin
to generate impulses at a low rate, probably synchronized to a submultiple
of the input rate because of the small fluctuations in the average
potential that occur with each incoming impulse. The potential at the axon
hillock will rise abruptly as the impulse is generated, and the
calcium-potassium pump will rapidly restore the potential to a negative
value below the threshold for firing (in one millisecond or so). The
potential will gradually rise again after the firing, eventually reaching
the level where the cell fires again (remember that the incoming impulses
are biasing the potential).

When the incoming frequency is low, the rise toward the firing threshold
after an output impulse is generated is slow and there is a long interval
between output impulses, meaning that the output frequency is low. As the
input frequency increases, less time is needed for the potential at the
axon hillock to reach the firing threshold again, and the frequency of
output impulses increases.

The result is to create a relationship between the input frequency and the
output frequency. To a first linear approximation, it would look like this:

          fast
  output | *
              > *
  frequency | *
              > *
          slow> *

···

----*---------------------------
              0 slow fast

                     input frequency

Perhaps now you can see why I am puzzled by your statement,

>>I also think I understand why some neurons/areas of the brain/groups of
>>neurons stabilize their frequency at a certain value.

If the Functions in an HPCT diagram are constructed from neurons with
characteristics somewhat like the above, I see nothing in the relationship
that says the input frequency would be stabilized at a certain value, so
the output frequency wouldn't be stabilized at a certain value, either.

Perhaps you're thinking of reference signals as being stabilized at
specific values. If so, this is a misunderstanding of how reference signals
are generated in the HPCT model. Reference signals are adjusted by the
outputs of a higher level of control system at all but the highest level.
They are not constant; they are continuously adjustable and in fact depend
on error signals in higher-order systems. They are adjusted to counteract
disturbances of the perceptions in higher-order systems. It's possible that
when no disturbances occur the reference signals would remain the same for
some time, but the moment the higher-order perception is disturbed (or its
reference signal changes), the reference signals sent to lower-order
systems would change.

I think there's some confusion between signals and functions in general,
not just in your post. The form of a function depends on both the
distribution of synapses and the shapes of the input-output frequency
curves of individual neurons, as well as on their interconnections. We can
expect the forms of functions and the connections among nerve cells to be
constant or change on a slow time scale. But the frequencies of neural
signals entering and leaving these functions change on a time scale that
goes down to milliseonds.

As a result, we can expect that the KINDS of perceptions we control will
change only very slowly. However, the MAGNITUDES of those perceptions are
continually and often rapidly changing. Likewise, we can expect that the
FORM of an output function, as well as the FORM of what Martin Taylor calls
the Reference Input Function (at the input to a comparator), will change
only slowly, while the SIGNAL FREQUENCIES involved can change continually
and rapidly.

So what changes slowly, and can appear constant over relatively long
periods of time, is the FORM of input and output functions -- that is, the
kinds of perceptions we have and the way in which we control them -- and
the connections from the outputs of some functions to the inputs of others.
But the magnitudes of the perceptual signals, error signals, and output
signals (which become reference signals at lower levels) change on a time
scale that is measured in seconds or fractions of a second.

Best.

Bill P.

3

[From Bjorn Simonsen (2005.01.06,22:26 EST)]

From Bill Powers (2005.01.06.0613 MST)

Thank you for your comments.

I seem to have quite a different view of neural signals from yours, but
this may be only a matter of terminology.

It seems as if it is difficult for me to express a simple question. Let me
try now. In B:CP, section: The Basic Spinal Feedback Loop, page 84 you
write:
"..... The reference signal is actually the composite of many converging
neural currents, some negative and some positive, but since the other input
to the neuron is inhibitory, there can be no effect on the motor neurone
unless the net reference current is positive. Therefore the reference signal
is assigned a net positive value, greater than or equal zero"

Then I think, if I sit on a high chair at a pub and will stretch my right
calf horizontally, there must be a reference indicating that purpose. This
reference must have a value. If I do the same tomorrow the value must be the
same. I know there are many output signals going into Martin's RIF, but the
reference signal going into the comparator at the fourth level must be the
same (about the same) tomorrow as it is today if I stretch my calf in the
same way. How can we explain that it is the same tomorrow as it is today?

If the rate at which incoming impulses occur increases, the rise and fall
of concentration happens more frequently -- and because of the exponential
shape of the curve, the average ionic concentration increases, and so does
the average potential at the axon hillock.

For me it looks like you are saying that the sum of impulses that arrive
through the thousands of dendrites continue through the cell body and
through the axon. If that is what you think? If that is correct, I am wrong.
I explained how the impulses of action potentials were formed in my last
mail.

The result is to create a relationship between the input frequency and the
output frequency. To a first linear approximation, it would look like this:

          fast
  output | *
              > *
  frequency | *
              > *
          slow> *

···

----*---------------------------
              0 slow fast

                     input frequency

Here you say that there is a linear approximation between the input
frequency and the output frequency. How can you add the thousands of
different frequencies in the dendrites and get an output frequency? I am
sorry I am not clever at low current. Remember I wrote above and in my last
mail how I thought a series of action potentials were formed. If it is
correct what you say, I am wrong. And maybe I can learn something new. What
do you say Bruce Abbot?

If the Functions in an HPCT diagram are constructed from neurons with
characteristics somewhat like the above, I see nothing in the relationship
that says the input frequency would be stabilized at a certain value, so
the output frequency wouldn't be stabilized at a certain value, either.

Do I understand you correct if I interpret what you write in that manner
that the reference value, when I stretch my calf tomorrow; is different from
what it is today?

Bjorn

4

[From Bill Powers (1005.01.07.0155 MST)]

Bjorn Simonsen (2005.01.06,22:26 EST)--

From Bill Powers (2005.01.06.0613 MST)

Thank you for your comments.
>I seem to have quite a different view of neural signals from yours, but
>this may be only a matter of terminology.

It seems as if it is difficult for me to express a simple question. Let me
try now. In B:CP, section: The Basic Spinal Feedback Loop, page 84 you
write:
"..... The reference signal is actually the composite of many converging
neural currents, some negative and some positive, but since the other input
to the neuron is inhibitory, there can be no effect on the motor neurone
unless the net reference current is positive. Therefore the reference signal
is assigned a net positive value, greater than or equal zero"

Then I think, if I sit on a high chair at a pub and will stretch my right
calf horizontally, there must be a reference indicating that purpose. This
reference must have a value. If I do the same tomorrow the value must be the
same. I know there are many output signals going into Martin's RIF, but the
reference signal going into the comparator at the fourth level must be the
same (about the same) tomorrow as it is today if I stretch my calf in the
same way. How can we explain that it is the same tomorrow as it is today?

I'd state it the other way: if the reference signal is about the same
tomorrow as it is today, the perception of stretching the calf will be
about the same. The actual stretch may or may not be the same, depending on
disturbances. The magnitude of the reference signal determines the
magnitude of the perceptual signal via the control action. I assume that as
the reference signal changes from zero to maximum, the perceived stretch of
the calf will vary from zero to maximum.

I can't picture what you mean by "stretching the calf" (the calf of the leg
is the rounded muscle at the back of the lower part of the leg that moves
the foot to point the toes downward when it contracts).

Anyway, what determines the reference signal is the effect of stretching
the calf on some higher perception. For example, if you want to raise your
body higher while standing, you can raise the reference signal for the
sensed tension in the calf muscles, which will lift you onto your toes.

For me it looks like you are saying that the sum of impulses that arrive
through the thousands of dendrites continue through the cell body and
through the axon. If that is what you think?

No. They cause a change in chemical concentrations inside the cell body,
which in turn determines the rate of firing of the cell. In general there
is no correspondence between one input impulse and one output impulse. Only
for special neurons is there such a correspondence.

>The result is to create a relationship between the input frequency and the
>output frequency. To a first linear approximation, it would look like this:

          fast
  output | *
              > *
  frequency | *
              > *
          slow> *
               ----*---------------------------
              0 slow fast

                     input frequency

Here you say that there is a linear approximation between the input
frequency and the output frequency. How can you add the thousands of
different frequencies in the dendrites and get an output frequency?

Consider a simple case in which all the inputs are equal.

1. Let each input change the average potential at the axon hillock by 0.001
millivolt per impulse per second. Then 1000 inputs will change the average
potential by 1 millivolt per impulse per second.

2. Let an average change of 1 millivolt at the axon hillock change the
output frequency by 10 impulses per second.

3. Then an increase in frequency of 1000 inputs of 10 impulses per second
each will cause an increase in the output frequency of 100 impulses per second.

>If the Functions in an HPCT diagram are constructed from neurons with
>characteristics somewhat like the above, I see nothing in the relationship
>that says the input frequency would be stabilized at a certain value, so
>the output frequency wouldn't be stabilized at a certain value, either.

Do I understand you correct if I interpret what you write in that manner
that the reference value, when I stretch my calf tomorrow; is different from
what it is today?

It could be, if you intend to stretch it more tomorrow than you intend to
stretch it today. In fact you can change the degree of stretch from small
to large and back to small in a few seconds. The perceived stretch of the
calf does not cause the change in reference signal for stretch of the calf;
it's the other way around. First the reference signal is set; then the
perception is caused (by muscle action) to match it.

Best.

Bill P.

5

[From Bjorn Simonsen (2005.01.07,14:35 EST)]

From Bill Powers (1005.01.07.0155 MST)

I am enthusiastic about travelling in the time.

I can't picture what you mean by "stretching the calf" (the calf of the leg
is the rounded muscle at the back of the lower part of the leg that moves
the foot to point the toes downward when it contracts).

Sorry. My fault. I had to use (bad) dictionary. I should have used Merriam -
Webster.

For me it looks like you are saying that the sum of impulses that arrive
through the thousands of dendrites continue through the cell body and
through the axon. If that is what you think?

No. They cause a change in chemical concentrations inside the cell body,
which in turn determines the rate of firing of the cell. In general there
is no correspondence between one input impulse and one output impulse. Only
for special neurons is there such a correspondence.

I liked your "No". And I liked your explanation. Especially I liked your
last but one sentence. And I stopped at nr. 2.

1. Let each input change the average potential at the axon hillock by 0.001
millivolt per impulse per second. Then 1000 inputs will change the average
potential by 1 millivolt per impulse per second.

OK

2. Let an average change of 1 millivolt at the axon hillock change the
output frequency by 10 impulses per second.

_Here is a central point_.
Can we instead of that say:
2. Let an average change of 1 millivolt at the axon hillock change the
output frequency by _12_ impulses per second.
3. Then an increase in frequency of 1000 inputs of 12 impulses per second
each will cause an increase in the output frequency of 120 impulses per
second.

If we can say both (either - or), there is something else than the incoming
signals that define the output frequency. And this something else can be the
number of sodium-potassium channels and the concentration of sodium and
potassium ions in the axon.
If we can say both, do you think "the something else" could be something
else?

Do I understand you correct if I interpret what you write in that manner
that the reference value, when I stretch my calf tomorrow; is different

from

what it is today?

It could be, if you intend to stretch it more tomorrow than you intend to
stretch it today.

No. I intend to stretch it just the same to morrow.
Then I suppose the reference value will be (about) the same. And my main
(main, main, main) point is that it is the "number of sodium-potassium
channels and the concentration of sodium- and potassium ions that set the
reference value" (Look at the quotes. They shall tell you I am talking some
figurative).

Do I say anything rational?

Bjorn

6

[Martin Taylor 2005.01.07.10.12]

[From Bjorn Simonsen (2005.01.07,14:35 EST)]
>From Bill Powers (1005.01.07.0155 MST)
...

1. Let each input change the average potential at the axon hillock by 0.001
millivolt per impulse per second. Then 1000 inputs will change the average
potential by 1 millivolt per impulse per second.

OK

2. Let an average change of 1 millivolt at the axon hillock change the
output frequency by 10 impulses per second.

_Here is a central point_.
Can we instead of that say:
2. Let an average change of 1 millivolt at the axon hillock change the
output frequency by _12_ impulses per second.
3. Then an increase in frequency of 1000 inputs of 12 impulses per second
each will cause an increase in the output frequency of 120 impulses per
second.

You mean 144, don't you?

If we can say both (either - or), there is something else than the incoming
signals that define the output frequency.

Why do you say this? As I read Bill's and your messages, you both
start with the assumption that the output frequency is defined by the
incoming signals. The only difference I can see is that you
substituted a different sensitivity value in your assumption 2.

And this something else can be the
number of sodium-potassium channels and the concentration of sodium and
potassium ions in the axon.
If we can say both, do you think "the something else" could be something
else?

Are you actually asking a question about how a neuron might change
its sensitivity from Bill's assumption (1 mv -> 10 impulses) to yours
(1 mv -> 12 impulses)? If so, it seems a rather different question
from the kind of topic that has linked this thread so far, and you
haven't put the question so that I can understand its implications.
As I read it, you are making an assertion that both Bill's "2" and
your "2" are possible states of the same neuron, and asking how this
can be. I would have thought that went rather beyond the scope of the
simplified neuron Bill was using as an illustrative model, but maybe
not.

Maybe I'm just stupid so early in the morning -- I prefer 22:17 to 10:17 :slight_smile:

Martin

7

[From Bill Powers (2005.01.07.0750 MST)]

Bjorn Simonsen (2005.01.07,14:35 EST)–

Can we instead of that
say:

  1. Let an average change of 1 millivolt at the axon hillock change
    the

output frequency by 12 impulses per second.

  1. Then an increase in frequency of 1000 inputs of 12 impulses per
    second

each will cause an increase in the output frequency of 120 impulses
per

second.

Yes. However, the question is what could cause such a change in the
factor relating output frequuency to hillock potential. We also have to
ask how rapidly the number of active channels, or the concentration of
ions just outside the cell, can change.
I doubt that the number of sodium-potassium channels would have a large
effect on the input-output relationship. The number of channels
determines how long it takes for conditions to be reset after an impulse
is initiated. With fewer channels or lower ion concentrations, the reset
phase would last longer (the voltage breakdown pulse would be smaller but
would last longer), so the chief effect would be on the maximum impulse
rate that the cell could produce when driven to its limits. Below the
maximum rate, the output frequency of the cell would simply be a function
of all the input frequencies.
However, for a definitive answer we would need a proper model of a
neuron. There would probably be some influence on the input-output
conversion factor as well, but there’s no way to say whether it would be
significant. This is a quantative question that can’t be settled by
qualitative arguments.
The second question you raise concerns what could change the number of
active calcium-potassium channels. Undoubtedly that number could change,
and also the concentrations of calcium, potassium, and sodium can change
because of factors external to the cell. My feeling, however, is that
these chemical effects would be much more constant than the fluctuations
in frequency of the input signals reaching dendrites of the cell. A
signal can change from maximum to minimum in a few dozen milliseconds,
but I doubt that chemical concentrations outside the cell can change
significantly in less than a few seconds.
It seems to me that you’re trying to make a single neuron into a complete
control system, with its own perceptual signal, reference signal,
comparator, and output function. I don’t object to that in principle, but
I don’t know of any evidence to support such a model. I do know that in
spinal control systems, the input function, comparator, and output
function are implemented as *different cells in different places,*with reference signals acting not through changing the number of ion
channels, but through signals reaching the dendrites of spinal motor
neurons. Those signals arrive via the output axons of other nerve cells
and have their effects through neurotransmitters.

No. I intend to stretch it just
the same to morrow.

Then I suppose the reference value will be (about) the same. And my
main

(main, main, main) point is that it is the "number of
sodium-potassium

channels and the concentration of sodium- and potassium ions that set
the

reference value" (Look at the quotes. They shall tell you I am
talking some

figurative).

When you put it that way, I have serious doubts. I think reference values
are set by neural signals arriving from axons of other cells. That is
certainly the case for spinal control systems and for at least a few
brain-stem control systems where the signals have been traced. Those
reference signals have an excitatory effect in some neural control
systems, an inhibitory effect in others (the perceptual inputs to the
same cells have the opposite effects, with the sign of the effect being
determined by the kind of neurotransmitter that is released at a
synapse). Renshaw cells, interneurons with short axons, are known to
function as sign inverters: when activated by excitatory inputs, they
generate inhibitory neurotransmitters at the ends of their
axons.

I don’t mean to rule out effects that are produced by changes in
extracellular ion concentrations or in number of active ion pumps, but we
need a good model (or direct experimental measurements) before we can say
just what those effects would be. Furthermore, we would have to account
for those changes – trace them back to what is causing them – before we
could know how they fit into the overall model. Can neural control
systems set the reference levels of other neural control systems by
altering the number of active ion pumps in the target systems? Or
are changes of that sort brought about in other ways, on a slower time
scale, by biochemical processes of the body rather than by behavioral
control systems? My present vote is for the latter proposal. Of course
facts are not determined by voting.

Best,

Bill P.

8

[From Bjorn Simonsen (2005.01.07,20:50 EST)]
Martin Taylor 2005.01.07.10.12
You mean 144, don't you?
The way I understood Bill was:

1. Let each input change the average potential at the axon hillock by 0.001
millivolt per impulse per second. Then 1000 inputs will change the average
potential by 1 millivolt per impulse per second.

When the transmitters arrive at the other side of the synapse, they meet
protein molecules called receptors. The transmitters and the receptors react
to a third substance. This third substance has the effect that it initiates
the sodium-potassium pump. Now it comes a voltage potential into being. In
this way one of many electrical impulses run through the dendrites to the
cell body.

Bill said that _this_ impulse changed the average potential at the axon
hillock by 0.001 millivolt per second. 1000 such inputs changes the average
potential at the axon by 1 millivolt per second.

2. Let an average change of 1 millivolt at the axon hillock change the
output frequency by 10 impulses per second.

Here Bill random (some) says that the average change of 1 millivolt at the
axon hillock opens the sodium-potassium channels and form an action
potential with a frequency 10 impulses per second.

In 3. Bill increases _this_ impulse (above) to 10 impulses per second. The
1000 dendrites then form average potential at the axon hillock by 10
millivolt per second. I follow Bill with this increase (10).

3. Then an increase in frequency of 1000 inputs of 10 impulses per second
each will cause an increase in the output frequency of 100 impulses per
second.

Here he multiplied 10*10. And I multiplied 10*12

I guess this was superfluous, now nearer sunset?

If we can say both (either - or), there is something else than the

incoming

signals that define the output frequency.

Why do you say this? As I read Bill's and your messages, you both
start with the assumption that the output frequency is defined by the
incoming signals. The only difference I can see is that you
substituted a different sensitivity value in your assumption 2.

No. The incoming signal reaches the axon hillock, and if this potential is
great enough it opens the sodium-potassium channels. And depending on
different circumstances, the sodium ions and potassium ions transport forms
an action potential, and a new action potential, and a new action potential.
In this way the incoming signal is responsible for an outgoing frequency
formed by the environmental conditions (?)

Are you actually asking a question about how a neuron might change
its sensitivity from Bill's assumption (1 mv -> 10 impulses) to yours
(1 mv -> 12 impulses)?

No.

As I read it, you are making an assertion that both Bill's "2" and
your "2" are possible states of the same neuron, and asking how this
can be.

No, I don't ask how this can be. I just asked him if the environmental
circumstances around the neuron is responsible for the output frequency. If
that is correct, the output frequency in some way is independent of the
incoming signals.
I think Bill describes this well (with many questions(as I do))in his [From
Bill Powers (2005.01.07.0750 MST)].

Bjorn

9

[From Bjorn Simonsen (2005.01.07,20:58 EST)]

From Bill Powers (2005.01.07.0750 MST)]

Thank you for your comments. I am pleased.

Can neural control systems set the reference levels of other neural control

systems

by altering the number of active ion pumps in the target systems? Or are

changes

of that sort brought about in other ways, on a slower time scale, by

biochemical

processes of the body rather than by behavioral control systems? My present

vote

is for the latter proposal.

My too. I will in long time in the future, when I make simulations just set
a reference value for the level I control and see what happens.

Bjorn