Tricky business

[From Bruce Abbott (960118.0910 EST)]

John Anderson (960118.0745) --

It seems to me that finding the various components of control systems in real
nervous systems is tricky business, even in very simple ones. For example,
consider the gill withdrawal circuit of the marine snail Aplysia, . . .

John, I looked into this (briefly) about four or five months ago. I wonder
if there is something missing from the diagram, as there is no indication of
muscle spindles or other sensors of muscle length/force. As currently
reported, the diagram's muscle-contraction element is a pure S-R mechanism,
lacking feedback about the state of the muscle.

There IS feedback in the larger system comprising the mantle
touch-receptors, conducting neurons, motor neurons, and gill withdrawal
muscle. The diagram should look something like this:

                          ref (touch)
                           >
             +--------->[comp]--------------------+
             > >
             > v
       [touch recept] [muscle]
             ^ |
             +----[contact]---[mantle position]<--+
                      ^
                      >
                 disturbance

In this system what is being controlled is the perception of contact with
the mantle. The actual system probably responds more readily to sudden
changes than to gradual ones; this effect would be mediated at the
receptor/sensory neuron level. I haven't represented this explicitly in the
diagram. If the touch reference is at some relatively low value, light
stimuluation would not produce gill-withdrawal. Stronger stimuluation would
produce an error which would then produce muscle contraction; muscle
contraction would withdraw the mantle, pulling the mantle/gill/siphon system
out of contact with the source of touch-stimulation, reducing the error to
zero. If the muscle reacts strongly, so that it contracts rapidly and
fully, but relaxes only slowly, the system as diagrammed would oscillate
rather than stabilizing at some mantle position that produces light touch
stimulation; this oscillation appears to characterize the behavior of the
actual system if the reopening of the mantle puts the mantle back into
contact with the source of stimulation. What pulls the mantle open? Is it
merely the relaxation of the gill-withdrawal muscles? What produces the
gradual re-opening, as opposed to the sudden withdrawal? Again, I'm not
sure that the diagrams usualy given for aplysia are complete enough for a
full control-system analysis.

The habituation-sensitization mechanisms seem to affect the mantle-contact
control system by altering the output gain: the strength of gill-muscle
contraction under a given rate of sensory-neural stimulation. This appears
to be accomplished at the axonal-axonal junction at the sensory neuron
terminals.

How's that for starters?

Regards,

Bruce

[From John Anderson (960119.0800 EST)]

Bruce Abbott (960118.0910 EST) --

John Anderson (960118.0745) --

It seems to me that finding the various components of control systems in

real

nervous systems is tricky business, even in very simple ones. For example,
consider the gill withdrawal circuit of the marine snail Aplysia, . . .

John, I looked into this (briefly) about four or five months ago.

I can't locate your posts on this; which ones are they?

There IS feedback in the larger system comprising the mantle
touch-receptors, conducting neurons, motor neurons, and gill withdrawal
muscle. The diagram should look something like this:

                         ref (touch)
                          >
            +--------->[comp]--------------------+
            > >
            > v
      [touch recept] [muscle]
            ^ |
            +----[contact]---[mantle position]<--+
                     ^
                     >
                disturbance

In this system what is being controlled is the perception of contact with
the mantle.

      <snip>

The habituation-sensitization mechanisms seem to affect the mantle-contact
control system by altering the output gain: the strength of gill-muscle
contraction under a given rate of sensory-neural stimulation. This appears
to be accomplished at the axonal-axonal junction at the sensory neuron
terminals.

The tricky business my original post referred to is the business of
identifying the various components of the control system. Here's the diagram
I drew again:

                      head---<(FI)>---tail
                                >
                                >
---------- ----- ------

mantle | | |
   \ | ^ |
    >-------(SN)--------<(IN)--------|--
   / | | | |
       > > > >

---------- | ^ ^
                       ----------------<(MN)
                                         >
                                         >
                                      ___^___
                                     > gill |

_Introduction to Modern Psychology_ says that FI "occupies a position
corresponding to that of the reference signal", and that IN "fits the
description of a comparator", and superficially it does look that way. IMP
then goes on to explain habituation and sensitization as due respectively to
decreases and increases in the reference signal. But this is easily shown to
be inconsistent. If

p = SN activity
o = MN activity = -k*e = -k*(r - p), negative so that feedback is negative
r = FI activity = 0

then when the mantle is touched,

o(0) = -k*(0 - p) = k*p > 0

so that MN is active and the gill is withdrawn. However, if the head or tail
is shocked so that FI is active and r > 0, then

o(r > p) < 0, so that MN is not active,
        or
o(0 < r < p) > 0, but less than o(0), so that MN is less active than when r =
0.

Neither of these predictions are consistent with experiment. As you note,
predictions _are_ consistent with experiment if habituation and sensitization
are ascribed to changes in the output gain of the system.

But what are the components of the control system? The motor neuron MN is the
output function, but the sensory neuron SN seems to be the input function
_and_ the source of the error signal. Where is the comparator?

Sorry if you already covered this in your previous postings.

John

[From Bruce Abbott (960122.0900 EST)]

John Anderson (960119.0800 EST) --

Bruce Abbott (960118.0910 EST)

John, I looked into this (briefly) about four or five months ago.

I can't locate your posts on this; which ones are they?

There aren't any; I didn't mean to imply that I had done anything more than
a bit of research.

The tricky business my original post referred to is the business of
identifying the various components of the control system. Here's the diagram
I drew again:

                     head---<(FI)>---tail
                               >
                               >
---------- ----- ------

mantle | | |
   \ | ^ |
    >-------(SN)--------<(IN)--------|--
   / | | | |
       > > > >

---------- | ^ ^
                      ----------------<(MN)
                                        >
                                        >
                                     ___^___
                                    > gill |

But what are the components of the control system? The motor neuron MN is the
output function, but the sensory neuron SN seems to be the input function
_and_ the source of the error signal. Where is the comparator?

This is one of those systems in which the reference signal and comparator
function are implicit rather than explicit. To produce a contraction of the
mantle, the sensory neuron must be sufficiently stimulated to produce a
train of neural pulses AND the combined excitation and inhibition on the
motor neuron must on balance be sufficiently excitatory to produce impulses
to the mantle muscle. If it requires a given amount of sensory stimulation
before motor output begins to occur, this threshold becomes the defacto
reference level for the system; there is no explicit reference signal. The
output of the sensory neuron itself is the error signal. The sensory neuron
itself provides the input function, reference, comparator, and error signal.
The output function is handled by the two synapses with the sensory neuron,
the interneuron, and the motor neuron.

I haven't thought through whether the sensitization/habituation system might
also be conceived of as a control system, but it seems a possibility.

Regards,

Bruce