Hubel and Wiesel and the levels in the hierarchy (was: Perception in PCT...)

CSGnet Archive CSG2018
1

[Martin Taylor 2018.09.29.12.01]

A postscript to

[Martin Taylor 2018.09.29.10.47]

[Rick Marken 2018-09-28_22:26:02]

[Martin Taylor 2018.09.28.16.12]

                MT: Indeed, what we know of neurophysiology tells

us that there is no single place within the brain
where that [perceptual] value exists for some future
technology to measure.

            RM: The work of Hubel & Wiesel suggests that

perceptions, in the form of neural firing rates, are
located in afferent neurons and other single neurons in
the brain.

  It's interesting that you should mention H&W, ...



  The result of this was always that it demonstrated that the

perceptions we form depended on recoding the mix of
on-centre-off-surround, oriented edges, corners, and so on signals
that were reported along the optic nerve to the brain, and were
never located in any single place thereafter, any more than any
particular component of a Fourier Transform locates a value of a
signal at any one moment.

  In other words, what you say is simply not true. Whether at some

future time some perception or other is found to be represented in
one and only one place in the brain, it is not true that the
findings of H&W suggest that it is.
When one considers the lower levels of the Powers hierarchy, it is
not easy to reconcile them with the kinds of data that are reported
up the optic nerve to the brain. The types I mentioned in my earlier
message are all of a kind that seems to fit Powers’s “configuration”
level, though one type I did not mention (motion-sensitive oriented
edge detectors, which report velocity of an edge in a direction
orthogonal to the direction of the edge) is more like an event-level
perception. I don’t know what other types of data are directly
reported up the optic nerve, but all of the ones I know are in one
way or another differential in either space or time, or both space
and time as in the case of the motion-sensitive edge orientation
detectors.

There may be some absolute intensity detectors for brightness (or

acoustic intensity or tactile pressure or odour intensity), but if
there are, they have to work over many orders of magnitude of
brightness, and the brightness we perceive does not correspond to
their reports. Intensity is an inferred variable, not a basic one.
This does not mean that intensity control is not at the base of the
hierarchy, but if it is, a lot of processing must be done below the
perceptual base, and the values for configuration and event reported
through the optic nerve must feed directly into the higher-level
perceptual signals without going through the intensity level on the
way, though they might also branch off to contribute to the
intensity perception.

Bottom line: If we are going to try to figure out working models

for, say, program control, it is probably worth reconsidering the
actual levels in the control hierarchy. The work should be based on
what is known to contemporary neurophysiology rather than on Bill’s
personal intuition, which is the basis for the eleven-levels now
considered canonical. If that leads to a drastic revision of the
levels, so be it. If it suggests that the neurophysiologists might
be missing something neurons can do and do do, so be it. Perhaps
such a result might get neurophysiologists looking where they have
not looked, or perhaps it would mean that the analysis was wrong.
Either way results in a more scientific basis for the actual level
structure of the perceptual control hierarchy.

Martin
2

[Rick Marken 2018-09-29_13:35:13]

[Martin Taylor 2018.09.29.12.01]

MT: Bottom line: If we are going to try to figure out working models

for, say, program control, it is probably worth reconsidering the
actual levels in the control hierarchy.

RM: The exercise I am proposing requires no assumptions about the hierarchy at all. What I would like to see is a control model that can do what a person can do in my “Program Control” demo (https://www.mindreadings.com/ControlDemo/ProgramControl.html), which is to control a program, such as the one in the demo: “if the shape is circle, the next color is blue; else, the next color is red”.Â

Best

Rick

···

Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

3

[Bruce Nevin 2018-10-02_10:05:42 ET]

I’ll restate an argument that I’ve made before.

The code of the routine that produces the display on the screen is a program, and the verbal description has the form of a program. The two elements are of two kinds, a shape and a color.Â

However, the program is imagined to be running in the brain of the subject, where there is a third controlled perception: the event of a mouse click. The entire program might be stated:

If false

 if circle then blue

 else if { triangle | square } then red

 fi

then click

fi

(In the else statement you do have to specify the other shapes, because e.g. if the entire screen went red or blue it wouldn’t count.)

The subject is controlling three perceptions: shape, color, and a mouse click (an event perception). How do you rule out the possibility that the subject is practicing to recognize and control three sequences?Â

Circle - red - click

Square - blue - click

Triangle - blue - click

This predicts an improvement in performance with practice, that is, ability to control at a faster display speed. Running the program control loop each time should not vary much, but practicing these three sequences until they become fluent enough to be considered event perceptions is expected to improve performance. Comparison with other situations where control involves three sequences, either of which might occur next, is relevant.

···

/Bruce

On Sat, Sep 29, 2018 at 4:37 PM Richard Marken csgnet@lists.illinois.edu wrote:

[Rick Marken 2018-09-29_13:35:13]

[Martin Taylor 2018.09.29.12.01]

MT: Bottom line: If we are going to try to figure out working models

for, say, program control, it is probably worth reconsidering the
actual levels in the control hierarchy.

RM: The exercise I am proposing requires no assumptions about the hierarchy at all. What I would like to see is a control model that can do what a person can do in my “Program Control” demo (https://www.mindreadings.com/ControlDemo/ProgramControl.html), which is to control a program, such as the one in the demo: “if the shape is circle, the next color is blue; else, the next color is red”.Â

Best

Rick

–
Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

4

[Rick Marken 2018-10-02_12:10:55]

[Bruce Nevin 2018-10-02_10:05:42 ET]

BN: I’ll restate an argument that I’ve made before.

RM: I think your comment is in reply to what I said earlier:Â

RM: The exercise I am proposing requires no assumptions about the hierarchy at all. What I would like to see is a control model that can do what a person can do in my “Program Control” demo (https://www.mindreadings.com/ControlDemo/ProgramControl.html), which is to control a program, such as the one in the demo: “if the shape is circle, the next color is blue; else, the next color is red”.Â

Â

RM: What we know from that demonstration is that controlling a sequence (small, medium large) can be done at a faster rate of presentation than controlling what I call a “program” (if Circle then blue, else if Square then red). Here is the data from a run I just did:

Picture2.png

RM: Note that at medium speed I was able to keep the sequence in the reference state 90% of the time while the program (which I was not controlling) was on target only 50% of the time. So I was able to control the sequence pretty well at the medium speed. When I tried to control the program at medium speed I did little better than chance as indicated by the fact that the program was kept at the reference only 60% of the time, which was no better than how often the sequence was (accidentally) kept at the reference. However, at the low speed I was able to control the program (keep it at the reference 88% of the time) nearly as well as I had controlled the sequence at the medium speed.Â

RM: So these results show that 1) a person can control both sequences and what I call programs 2) the perception that is controlled when one controls a sequence is different than the perception one controls when one controls a program and 3) the perception controlled when one controls a program is at a higher level than the perception controlled when one controls a sequence, as indicated by the fact the program must occur more slowly than the sequence in order for one to be able to control it.Â

RM: What I would like to see is a model of a control system that can do what a person can do in this demo, which is to control either a sequence or a program.Â

BN: The code of the routine that produces the display on the screen is a program, and the verbal description has the form of a program. The two elements are of two kinds, a shape and a color.Â

RM: So far so good.Â

BN: However, the program is imagined to be running in the brain of the subject,

RM: According to PCT, the program that we see being controlled in the demo exists as a perceptual signal in the brain of the subject. Â

where there is a third controlled perception: the event of a mouse click.Â

RM: It’s actually the press of the space bar that is used to control the sequence and program. And while, like all actions, it is, indeed, a controlled perception, I don’t think that fact is relevant to the model that would control sequence or program; all the model has to do to control a sequence or program is produce a bar press at the appropriate time to counter the change in the sequence or program, which is the disturbance to the particular sequence or program that is being controlled.Â

BN: The entire program might be stated:

If false

 if circle then blue

 else if { triangle | square } then red

 fi

then click

fi

(In the else statement you do have to specify the other shapes, because e.g. if the entire screen went red or blue it wouldn’t count.)

RM: If you want to define the program that way that’s fine. But I think the problem for modeling this behavior is to figure out how to convert the relevant programmatic aspect of what is happening on the screen into a perceptual signal, probably binary, which has one value when the desired program is happening and a different value when it is not.

BN: The subject is controlling three perceptions: shape, color, and a mouse click (an event perception).Â

RM: No, the subject cannot control the shape or color of the displayed object; the subject’s output (bar press, not mouse click) has no systematic effect on what shape or color occurs. What we know from the data is that the subject can control either a sequence (of sizes) or a program (of shape/color contingencies) if the speed of presentation of the components of these perceptions is appropriate.

Â

BN: How do you rule out the possibility that the subject is practicing to recognize and control three sequences?Â

Circle - red - click

Square - blue - click

Triangle - blue - click

RM: I think the demo shows pretty clearly that a person can control a sequence at a faster rate than a program. So sequence and program perceptions are clearly different kinds of perceptions at different level of the control hierarchy. If you want to see the perception that I call a program as a perception of two sequences (there is no triangle in the demo) then I think you would have to explain why controlling these sequences is so much harder than controlling any other sequences, like the “small medium large” sequence in the demo. Controlling the “small medium large” sequence could also be seen as a controlling a set of sequences:

small - large – press

medium - small - press

large - medium - press

So why is it easier to control this set of sequences than this one:

Circle - red - press

Square - blue press

?

BN: This predicts an improvement in performance with practice, that is, ability to control at a faster display speed. Running the program control loop each time should not vary much, but practicing these three sequences until they become fluent enough to be considered event perceptions is expected to improve performance. Comparison with other situations where control involves three sequences, either of which might occur next, is relevant.

RM: I think it would be great if you tried this, Bruce. I predict that you will never be able to control what I call the “program” at the rate at which you can control the sequence. I believe that they are different kind of perceptions, as Bill predicted, and that you can’t “learn” your way out of this. But your proposed study would certainly test this idea.

BestÂ

Rick

Â

···

/Bruce

On Sat, Sep 29, 2018 at 4:37 PM Richard Marken csgnet@lists.illinois.edu wrote:

[Rick Marken 2018-09-29_13:35:13]

[Martin Taylor 2018.09.29.12.01]

MT: Bottom line: If we are going to try to figure out working models

for, say, program control, it is probably worth reconsidering the
actual levels in the control hierarchy.

RM: The exercise I am proposing requires no assumptions about the hierarchy at all. What I would like to see is a control model that can do what a person can do in my “Program Control” demo (https://www.mindreadings.com/ControlDemo/ProgramControl.html), which is to control a program, such as the one in the demo: “if the shape is circle, the next color is blue; else, the next color is red”.Â

Best

Rick

–
Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

–
Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery