A motor control study supporting hierarchical control?

CSGnet Archive CSG2013
1

[Adam Matic, 2013.1.21.00.20 GMT+1]

This paper seems to be directly supporting the pct model of motor control. They don’t cite any PCT works or papers, and don’t seem to be familiar with control theory.

Nabil Ilmane, Samir Sangani, Anatol G. Feldman (2013).

Corticospinal control strategies underlying voluntary and involuntary wrist movements. Behavioural Brain Research 236, 360-358

Attached is the whole paper, I’ll just copy parts of the conclusion.

N. Ilmane et all: "Our analysis shows that descending systems are actively

involved in both voluntary and involuntary changes in the wrist

position but in fundamentally different ways."

AM: Motor paths decending to the spine - in PCT, these would be the second(?) level error pathways that set references for spinal reflex loops.

NI: "Descending systems reset the spatial thresholds of reflexes when voluntary changes in position are made, but maintain these thresholds when positional changes are produced involuntarily, following external perturbations.

The differences in the control strategies are related to the necessity to reset spatial reflex thresholds to a new posture and thus convert the posture-stabilizing mechanisms from those resisting to those assisting self-initiated motion. In contrast, descending systems may rely on these mechanisms to generate involuntary responses to perturbations by maintaining the same spatial thresholds of reflexes. "

AM: This would sound so much better in PCT - lingo. They are saying that the second level system is maintaining the reference value, while the first level system is working independently, controlling its perceptual signal.

NI: “These results substantially constrain the choice between different theories of motor control.”

AM:

Interesting, isn’t it? I’m trying to figure out what their data represents, perhaps it could be mapped to a functional model.

Adam

Illmane et all (2013). Corticospinal control strategies, wrist movement.pdf (928 KB)

2

[From Rick Marken (2013.01.21.1245)]

Thanks Adam. Sounds great. I’ll look at it ASAP.

Best

Rick

···

On Sun, Jan 20, 2013 at 3:31 PM, Adam Matic adam.matic@gmail.com wrote:

[Adam Matic, 2013.1.21.00.20 GMT+1]

This paper seems to be directly supporting the pct model of motor control. They don’t cite any PCT works or papers, and don’t seem to be familiar with control theory.

Nabil Ilmane, Samir Sangani, Anatol G. Feldman (2013).

Corticospinal control strategies underlying voluntary and involuntary wrist movements. Behavioural Brain Research 236, 360-358

Attached is the whole paper, I’ll just copy parts of the conclusion.

N. Ilmane et all: "Our analysis shows that descending systems are actively

involved in both voluntary and involuntary changes in the wrist

position but in fundamentally different ways."

AM: Motor paths decending to the spine - in PCT, these would be the second(?) level error pathways that set references for spinal reflex loops.

NI: "Descending systems reset the spatial thresholds of reflexes when voluntary changes in position are made, but maintain these thresholds when positional changes are produced involuntarily, following external perturbations.

The differences in the control strategies are related to the necessity to reset spatial reflex thresholds to a new posture and thus convert the posture-stabilizing mechanisms from those resisting to those assisting self-initiated motion. In contrast, descending systems may rely on these mechanisms to generate involuntary responses to perturbations by maintaining the same spatial thresholds of reflexes. "

AM: This would sound so much better in PCT - lingo. They are saying that the second level system is maintaining the reference value, while the first level system is working independently, controlling its perceptual signal.

NI: “These results substantially constrain the choice between different theories of motor control.”

AM:

Interesting, isn’t it? I’m trying to figure out what their data represents, perhaps it could be mapped to a functional model.

Adam

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

3

[From Bill Powers (2013.01.22.1241 MST)]

[Adam Matic, 2013.1.21.00.20 GMT+1]

This paper seems to be directly supporting the pct model of motor control. They don't cite any PCT works or papers, and don't seem to be familiar with control theory.

BP: There have been communications back and forth between Anatol Feldman (third author) and us for some time, though after about a year they reached an impasse and a few months ago ceased. Feldman does not think PCT or control theory in general has anything to do with his Equilibrium Point Theory of motor control or more recent versions of it known as Threshold Control Theory (which, I'm afraid, I tactlessly said was conceptually identical to PCT). So if you communicate with him, don't expect a friendly reception.

Best,

Bill P.

···

At 12:31 AM 1/21/2013 +0100, Adam Matic wrote:

4

[From Rick Marken (2013.01.22.1210)

Bill Powers (2013.01.22.1241 MST)]

AM: This paper seems to be directly supporting the pct model of motor control. They don’t cite any PCT works or papers, and don’t seem to be familiar with control theory.

BP: There have been communications back and forth between Anatol Feldman (third author) and us for some time, though after about a year they reached an impasse and a few months ago ceased. Feldman does not think PCT or control theory in general has anything to do with his Equilibrium Point Theory of motor control or more recent versions of it known as Threshold Control Theory (which, I’m afraid, I tactlessly said was conceptually identical to PCT). So if you communicate with him, don’t expect a friendly reception.

RM: Thanks Bill. That will save me some reading. I couldn’t understand what the heck they were doing anyway.

Best

Rick

···

Best,

Bill P.

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

5

I[From Adam Matic (2013.1.22.21.30 cet)]

···

Bill Powers (2013.01.22.1241 MST)

BP: There have been communications back and forth between Anatol Feldman (third author) and us for some time, though after about a year they reached an impasse and a few months ago ceased. Feldman does not think PCT or control theory in general has anything to do with his Equilibrium Point Theory of motor control or more recent versions of it known as Threshold Control Theory (which, I’m afraid, I tactlessly said was conceptually identical to PCT). So if you communicate with him, don’t expect a friendly reception.

AM: Ah, well… at least he’s on the same side regarding internal model theories. Thank you.

[Rick Marken (2013.01.22.1210)

RM: Thanks Bill. That will save me some reading. I couldn’t understand what the heck they were doing anyway.

AM: They lost me somewhere around “TMS conditioning” and H-reflexes. I figure they’re manually setting reference values for writs rotation, but I couldn’t explain exactly how they’re doing it.

So, I was thinking about using muscle activation measures (EMG). I’ve read somewhere that you (BIll?) used EMG’s to prove that there really is muscle activation in the ‘keep arm elevation’ task that demonstrates the hierarchical relationship of control loops. Can’t find where I’ve read that.

Just theoretically, how complex would a model of that have to be? Could EMGs be recorded and then compared to simulated EMGs in an identical situation?

Adam

6

[From Bill Powers ()2013.01.23.1620 MST)]

Adam Matic (2013.1.22.21.30 cet) --

So, I was thinking about using muscle activation measures (EMG). I've read somewhere that you (BIll?) used EMG's to prove that there really is muscle activation in the 'keep arm elevation' task that demonstrates the hierarchical relationship of control loops. Can't find where I've read that.

Just theoretically, how complex would a model of that have to be? Could EMGs be recorded and then compared to simulated EMGs in an identical situation?

Yes, I did this (crudely) at the VA Research Hospital in Chicago, first just to measure the true reaction time to a sensory or mechanical disturbance. All the measures I had seen required a hand to move enough to cause a lever to be depressed and close a contact, which added time to the actual nervous system lag in the task. But I also saw that when a disturbance was applied to an outstretched arm, there was an active resistance to the force, not just the resistance one would expect from muscle-tendon elasticity. This proved that the control system was active and working, which is not always easy to see when the only thing that happens is that the arm does NOT move much. If the resistance had been totally elastic there would have been no EMG response.

Anatol Feldman used EMG measures to estimate the amount of muscle activation that was happening in a continuous task. I don't recommend this because the EMG electrodes probably pick up the activity in just the nearest muscle fibers.

At the VA hospital I was trying to define levels, and wanted to see the absolute minimum delay time, since delay times vary and there can be complications that artificially slow the reaction time (such as indistinct stimuli). I figured that in order for two reaction times to represent different levels, the fastest reaction at a higher level should be slower than the fastest reaction at the lower level. This got rid of a lot of variability.

Best,

Bill P.

7

[From Chad Green 2013.01.24.1058 EST)]

That's Feldman's loss. PCT is a gift that keeps on giving. :slight_smile:

Incidentally, I was reading an article by Fischer (2008) in which he wrote (p. 136): "Action systems generate single representations. Representational systems generate single abstractions. Abstract systems generate single principles. (There is no evidence to date of emergence of new levels beyond single principles.)" An example of a principle is the Golden Rule, or my favorite: "The most important thing is always missing." Fischer calls his approach dynamic skill theory.

HPCT indicates that system concepts follow principles. Thoughts?

Source: http://www.gse.harvard.edu/~ddl/articlesCopy/FischerCyclesCognBrain.EducBrainCUP.pdf (see also Figure 8.2, p. 130)

Reference: http://www.gse.harvard.edu/~ddl/articlesCopy/FischerBidellProofsCorrected.0706.pdf (see Figure 7.3, p. 323)

Best,
Chad

Chad Green, PMP
Program Analyst
Loudoun County Public Schools
21000 Education Court
Ashburn, VA 20148
Voice: 571-252-1486
Fax: 571-252-1633

"If you want sense, you'll have to make it yourself." - Norton Juster

Adam Matic <adam.matic@GMAIL.COM> 1/22/2013 4:20 PM >>>

I[From Adam Matic (2013.1.22.21.30 cet)]

Bill Powers (2013.01.22.1241 MST)
BP: There have been communications back and forth between Anatol
Feldman(third author) and us for some time, though after about a year
they reached
an impasse and a few months ago ceased. Feldman does not think PCT or
control theory in general has anything to do with his Equilibrium Point
Theory of motor control or more recent versions of it known as Threshold
Control Theory (which, I'm afraid, I tactlessly said was conceptually
identical to PCT). So if you communicate with him, don't expect a friendly
reception.

AM: Ah, well.. at least he's on the same side regarding internal model
theories. Thank you.

[Rick Marken (2013.01.22.1210)
RM: Thanks Bill. That will save me some reading. I couldn't understand
what the heck they were doing anyway.

AM: They lost me somewhere around "TMS conditioning" and H-reflexes. I
figure they're manually setting reference values for writs rotation, but I
couldn't explain exactly how they're doing it.

···

____

So, I was thinking about using muscle activation measures (EMG). I've read
somewhere that you (BIll?) used EMG's to prove that there really is muscle
activation in the 'keep arm elevation' task that demonstrates the
hierarchical relationship of control loops. Can't find where I've read that.

Just theoretically, how complex would a model of that have to be?
Could EMGsbe recorded and then compared to simulated EMGs
in an identical situation?

Adam

8

[From Kent McClelland (2013.01.28.1100 CST)]

Adam Matic (2013.1.21.00.20 GMT+1)

Here's the abstract of another recent article that supports the PCT view of motor control (as hierarchically organized and tightly linked to perception). In this case, they do cite an article by Bill Powers.

Perhaps you've seen it already, Adam?

Kent

Motor-Sensory Confluence in Tactile Perception

The Journal of Neuroscience, 3 October 2012, 32(40): 14022-14032

  � Avraham Saig*,
  � Goren Gordon*,
  �
  • Eldad Assa,
  � Amos Arieli, and
  � Ehud Ahissar

···

+
+ Author Affiliations

  � Department of Neurobiology, Weizmann Institute of Science, 76100 Rehovot, Israel
  • Author contributions: A.S. and E. Ahissar designed research; A.S. and G.G. performed research; A.A. and E. Ahissar contributed unpublished reagents/analytic tools; A.S., G.G., E. Assa, and E. Ahissar analyzed data; A.S., G.G., E. Assa, A.A., and E. Ahissar wrote the paper.

  � �↵*A.S. and G.G. contributed equally to this work.

Abstract

Perception involves motor control of sensory organs. However, the dynamics underlying emergence of perception from motor-sensory interactions are not yet known. Two extreme possibilities are as follows: (1) motor and sensory signals interact within an open-loop scheme in which motor signals determine sensory sampling but are not affected by sensory processing and (2) motor and sensory signals are affected by each other within a closed-loop scheme. We studied the scheme of motor-sensory interactions in humans using a novel object localization task that enabled monitoring the relevant overt motor and sensory variables. We found that motor variables were dynamically controlled within each perceptual trial, such that they gradually converged to steady values. Training on this task resulted in improvement in perceptual acuity, which was achieved solely by changes in motor variables, without any change in the acuity of sensory readout. The within-trial dynamics is captured by a hierarchical closed-loop model in which lower loops actively maintain constant sensory coding, and higher loops maintain constant sensory update flow. These findings demonstrate interchangeability of motor and sensory variables in perception, motor convergence during perception, and a consistent hierarchical closed-loop perceptual model.

On Jan 20, 2013, at 5:31 PM, Adam Matic wrote:

[Adam Matic, 2013.1.21.00.20 GMT+1]

This paper seems to be directly supporting the pct model of motor control. They don't cite any PCT works or papers, and don't seem to be familiar with control theory.

Nabil Ilmane, Samir Sangani, Anatol G. Feldman (2013).
Corticospinal control strategies underlying voluntary and involuntary wrist movements. Behavioural Brain Research 236, 360-358

Attached is the whole paper, I'll just copy parts of the conclusion.

N. Ilmane et all: "Our analysis shows that descending systems are actively
involved in both voluntary and involuntary changes in the wrist
position but in fundamentally different ways."

AM: Motor paths decending to the spine - in PCT, these would be the second(?) level error pathways that set references for spinal reflex loops.

NI: "Descending systems reset the spatial thresholds of reflexes when voluntary changes in position are made, but maintain these thresholds when positional changes are produced involuntarily, following external perturbations.
The differences in the control strategies are related to the necessity to reset spatial reflex thresholds to a new posture and thus convert the posture-stabilizing mechanisms from those resisting to those assisting self-initiated motion. In contrast, descending systems may rely on these mechanisms to generate involuntary responses to perturbations by maintaining the same spatial thresholds of reflexes. "

AM: This would sound so much better in PCT - lingo. They are saying that the second level system is maintaining the reference value, while the first level system is working independently, controlling its perceptual signal.

NI: "These results substantially constrain the choice between different theories of motor control."

AM:
Interesting, isn't it? I'm trying to figure out what their data represents, perhaps it could be mapped to a functional model.

Adam

9

[From Richard Kennaway (2013.01.28.1745 GMT)]

[From Kent McClelland (2013.01.28.1100 CST)]

Adam Matic (2013.1.21.00.20 GMT+1)

Here's the abstract of another recent article that supports the PCT view of motor control (as hierarchically organized and tightly linked to perception). In this case, they do cite an article by Bill Powers.

My university doesn't subscribe to the journal. If you have the full version, can you send me a copy?

-- Richard

···

--
Richard Kennaway, R.Kennaway@uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

10

[From Adam Matic 2013.1.28.1840 cet]
Kent McClelland (2013.01.28.1100 CST)

Here's the abstract of another recent article that supports the PCT view of motor control (as hierarchically organized and tightly linked to perception). In this case, they do cite an article by Bill Powers.

Perhaps you've seen it already, Adam?

• Avraham Saig*,
• Goren Gordon*,
• Eldad Assa,
• Amos Arieli, and
• Ehud Ahissar

AM: This looks great. I remember reading prof. Ahissar's papers on rat whisker (vibrissa) system and movement and waiting for a PCT reference to jump out, but he never mentioned it explicitly (not that I could find anyway).
By the way, my M.A. research will most likely be on motor control, so this is very useful. I'm in the process of collecting various papers and theories of motor control. So, thank you, don't hesitate to send another paper if you find it. I'll go read this one in detail.
@Richard Kennaway, here is a link to the article:
<http://webout.weizmann.ac.il/neurobiology/labs/ahissar/new_pages/pdf/JNS1012.pdf&gt;http://webout.weizmann.ac.il/neurobiology/labs/ahissar/new_pages/pdf/JNS1012.pdf

Adam

11

Kent

···

[From Kent McClelland (2013.01.28.1325 CST)]

Adam Matic (2013.1.28.1840 cet)

Warren Mansell of the University of Manchester keeps an eagle eye out for any new publications related to PCT, and this citation has already been posted on the “biology & neuroscience” page of pct
web.org, the website he maintains, which is a good source for many different things PCT-related. However, I haven’t seen anything listed there that’s more recent than this paper.

[From Adam Matic 2013.1.28.1840 cet]

Kent McClelland (2013.01.28.1100 CST)

Here’s the abstract of another recent article that supports the PCT view of motor control (as hierarchically organized and tightly linked to perception). In this case, they do cite an article by Bill Powers.

Perhaps you’ve seen it already, Adam?

• Avraham Saig*,

    • Goren Gordon*,

    • Eldad Assa,

    • Amos Arieli, and

    • Ehud Ahissar

AM: This looks great. I remember reading prof. Ahissar’s papers on rat whisker (vibrissa) system and movement and waiting for a PCT reference to jump out, but he never mentioned it explicitly (not that I could find anyway).

By the way, my M.A. research will most likely be on motor control, so this is very useful. I’m in the process of collecting various papers and theories of motor control. So, thank you, don’t hesitate to send another paper if you find it. I’ll go read this
one in detail.

@Richard Kennaway, here is a link to the article:

http://webout.weizmann.ac.il/neurobiology/labs/ahissar/new_pages/pdf/JNS1012.pdf

Adam

12

[From Rick Marken (2013.01.28.2100)]

RM: Thanks for the link Adam. I read it but could not really make heads or tails of it. Too many equations. Could someone please explain to me what they found in nice simple terms that even I could understand. Based on the little I could understand it doesn’t seem like it really has that much to do with PCT. But then when they talk about their model of what’s going on they talk about levels of perceptions and controlled variables. So maybe it is relevant to PCT and I just don’t understand how it is. Help me out here all you smart people out there.

Best

Rick

···

Adam Matic 2013.1.28.1840 cet

AM: This looks great. I remember reading prof. Ahissar’s papers on rat whisker (vibrissa) system and movement and waiting for a PCT reference to jump out, but he never mentioned it explicitly (not that I could find anyway).

By the way, my M.A. research will most likely be on motor control, so this is very useful. I’m in the process of collecting various papers and theories of motor control. So, thank you, don’t hesitate to send another paper if you find it. I’ll go read this one in detail.

@Richard Kennaway, here is a link to the article:

http://webout.weizmann.ac.il/neurobiology/labs/ahissar/new_pages/pdf/JNS1012.pdf

13

[From Adam Matic 2013.2.29.16:40 cet]

@ Kent

I guess I missed that one on pctweb. I’ll be checking it for updates more regularly.

@Rick

The paper is relevant because they are refuting feedforward, open-loop models. Their mention of PCT is marginal, placed together with theories such as optimal control, but their results seem consistent with what PCT would predict.

The person is sitting on a chair, with eyes closed. There are two poles on each side. The participant’s are trying to figure out which pole is more posterior. They did it by spreading their arms and moving them toward the poles a few times until they figure out which one is closer.

Their main point is that hand behavior looks more like behavior of a closed loop system then an open loop system. They’re using Bayes probability calculations to model participants’ improvement in discrimination of distances. I’m afraid I can’t explain what that is all about, or if it’s relevant for PCT. I dont’ see how brain would do probability calculations, but perhaps something analogous is happening. They do have a nice fit.

I’m also a bit confused when they talk about the hierarchy. Like this one:

"Using mathematical modeling we showed that our data fit a scheme of hierarchical closed loops in which lower loops

control sensation reliability and higher loops control sensory update flow. Constant sensation reliability was obtained here by keeping constant velocity and interhand difference, and constant update flow was kept by decreasing cycle duration."

What they mean is, I assume, that the person tried to move both hands at equal speed toward the poles. That was what the lower loop did. The higher loop was doing the repeating of ‘measurments’ of distances by using the lower loop’s data and ‘bayesian inference’. They assume the input for the higher loop was some ‘amount of information’, the reference was a constant value. and the behavior was decreasing duration of each cycle until the position could be discriminated.

Adam

14

[From Rick Marken (2013.01.29.1700)]

Adam Matic (2013.2.29.16:40 cet)–

AM: The paper is relevant because they are refuting feedforward, open-loop models. Their mention of PCT is marginal, placed together with theories such as optimal control, but their results seem consistent with what PCT would predict.

RM: I wonder about that. I’ll have more time to deal with this tomorrow and, hopefully, present a more detailed critique. But my impression is that the results have very little to do with PCT. I think it’s worth looking at this study more closely because I’m wondering whether it is better to have people refer to PCT and get it wrong or just not refer to it. I’m thinking, for example, of the work of Carver and Scheier. Thee guys carried on a whole program of research on “self-regulation” (maybe it’s still going on) that was ostensibly based on PCT. And yet their research clearly assumed that there subjects were S-R systems. That is, their research was based on an S-R model rather than a control model. The same seems to be true of this Saig et al paper. They are talking “closed loop” but what they are saying sounds to me kind of mishuga.

Ah well, let’s see if I can come up with something coherent tomorrow.

Best

Rick

···

The person is sitting on a chair, with eyes closed. There are two poles on each side. The participant’s are trying to figure out which pole is more posterior. They did it by spreading their arms and moving them toward the poles a few times until they figure out which one is closer.

Their main point is that hand behavior looks more like behavior of a closed loop system then an open loop system. They’re using Bayes probability calculations to model participants’ improvement in discrimination of distances. I’m afraid I can’t explain what that is all about, or if it’s relevant for PCT. I dont’ see how brain would do probability calculations, but perhaps something analogous is happening. They do have a nice fit.

I’m also a bit confused when they talk about the hierarchy. Like this one:

"Using mathematical modeling we showed that our data fit a scheme of hierarchical closed loops in which lower loops

control sensation reliability and higher loops control sensory update flow. Constant sensation reliability was obtained here by keeping constant velocity and interhand difference, and constant update flow was kept by decreasing cycle duration."

What they mean is, I assume, that the person tried to move both hands at equal speed toward the poles. That was what the lower loop did. The higher loop was doing the repeating of ‘measurments’ of distances by using the lower loop’s data and ‘bayesian inference’. They assume the input for the higher loop was some ‘amount of information’, the reference was a constant value. and the behavior was decreasing duration of each cycle until the position could be discriminated.

Adam

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

15

[From Rick Marken (2013.01.31.1100)]

OK, I admit defeat. I just can’t get my arms around the Saig et al “Motor sensory” paper. At the most basic level I just find it difficult to follow because it seems to be rather poorly written; For example, it often introduce variables out of nowhere (delta t and delta H for example) and I had to search all over the paper to figure out what the heck they were talking about.

As best as I can tell, they are trying to determine whether motor outputs contribute to the accuracy of sensory discrimination. I think the subjects were trying to tell which pole was farther back by moving their arms (and fingers) back from straight out until then made contact with the pole. I guess the idea was that the subjects were using the time difference of movement to “enhance” the sensory basis of the judgment of which pole was further back. They tested this by fitting their model (Equation 5) to the data in Fig 3B. Their model is an equation that predicts “whisking cycles” as a function of delta t (the time to move from horizontal to a pole?). Anyway, their model fitting is really just an exercise in curve fitting; this is not PCT modeling, where we actually build working simulations to see if they produce the behavior seen in the experiments.

Ignoring for a moment all the data aspects of this paper, their verbal descriptions of what they are studying reveals a very misleading understanding of what PCT is about (to the extent that they think they are studying control). For example, in the introduction the authors say the following about closed loop control : “A typical behavior of closed loops is convergence toward a steady state that is characterized by gradual changes of loop variables until a steady behavior is reached”. This is not even close. The typical behavior of a control system is to keep perceptions tracking references; when there is good tight control the convergence of perception to reference is nearly instantaneous. The “typical behavior” being described by Saig et al is that of a “dynamical system”, such as a pendulum, that gradually returns to a “steady state” (perpendicular to the ground in the case of the pendulum) after a disturbance (push on the bob in the case of the pendulum). These dynamical systems are no closed loop nor do they control. But they have been used as stand-in’s for closed-loop control systems, probably because such behavior can be understood in terms of open-loop urve fitting models (as is done by Saig et al). When said talks about controlled variables he is really just talking about variables that seem to reach a constant, steady state value (as do the subject’s hand movements in their study). But there are not controlled variables in the PCT sense; controlled variables are variables that remain in reference states, protected from the effects of disturbances by the actions of the system. Saig et all introduce no disturbances to test whether the variables they consider to be controlled are actually controlled. This is because they have no idea what control is.

So my conclusion is that this paper, that is presumably “supportive” of PCT is, at best, irrelevant to PCT and, at worst, completely misleading about the nature of purposeful (control) behavior. In my estimation this paper is worth less than nothing.

Best

Rick

···

On Tue, Jan 29, 2013 at 5:05 PM, Richard Marken rsmarken@gmail.com wrote:

[From Rick Marken (2013.01.29.1700)]

Adam Matic (2013.2.29.16:40 cet)–

AM: The paper is relevant because they are refuting feedforward, open-loop models. Their mention of PCT is marginal, placed together with theories such as optimal control, but their results seem consistent with what PCT would predict.

RM: I wonder about that. I’ll have more time to deal with this tomorrow and, hopefully, present a more detailed critique. But my impression is that the results have very little to do with PCT. I think it’s worth looking at this study more closely because I’m wondering whether it is better to have people refer to PCT and get it wrong or just not refer to it. I’m thinking, for example, of the work of Carver and Scheier. Thee guys carried on a whole program of research on “self-regulation” (maybe it’s still going on) that was ostensibly based on PCT. And yet their research clearly assumed that there subjects were S-R systems. That is, their research was based on an S-R model rather than a control model. The same seems to be true of this Saig et al paper. They are talking “closed loop” but what they are saying sounds to me kind of mishuga.

Ah well, let’s see if I can come up with something coherent tomorrow.

Best

Rick

The person is sitting on a chair, with eyes closed. There are two poles on each side. The participant’s are trying to figure out which pole is more posterior. They did it by spreading their arms and moving them toward the poles a few times until they figure out which one is closer.

Their main point is that hand behavior looks more like behavior of a closed loop system then an open loop system. They’re using Bayes probability calculations to model participants’ improvement in discrimination of distances. I’m afraid I can’t explain what that is all about, or if it’s relevant for PCT. I dont’ see how brain would do probability calculations, but perhaps something analogous is happening. They do have a nice fit.

I’m also a bit confused when they talk about the hierarchy. Like this one:

"Using mathematical modeling we showed that our data fit a scheme of hierarchical closed loops in which lower loops

control sensation reliability and higher loops control sensory update flow. Constant sensation reliability was obtained here by keeping constant velocity and interhand difference, and constant update flow was kept by decreasing cycle duration."

What they mean is, I assume, that the person tried to move both hands at equal speed toward the poles. That was what the lower loop did. The higher loop was doing the repeating of ‘measurments’ of distances by using the lower loop’s data and ‘bayesian inference’. They assume the input for the higher loop was some ‘amount of information’, the reference was a constant value. and the behavior was decreasing duration of each cycle until the position could be discriminated.

Adam

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

16

[From Adam Matic 2013.1.31.2100 cet]

···

Rick Marken (2013.01.31.1100)

[…]

So my conclusion is that this paper, that is presumably “supportive” of PCT is, at best, irrelevant to PCT and, at worst, completely misleading about the nature of purposeful (control) behavior. In my estimation this paper is worth less than nothing.

AM:

After the tone of your previous post, I expected something like this, so I went over the paper a few more times and… yeah… they don’t look like they understand control theory let alone PCT. There are really no control systems modeled, of any kind.

This line sums up a lot: “Given the inherent uncertainty of sensory coding, we tried to see whether our results can be explained by a closed-loop optimal control of sensory coding (Powers, 1973)”.

Aside from the obvious mention of optimal control, there is no connection between PCT and “inherent uncertainty of sensory coding”, “sensory coding” is not at all what is controlled, and their reference is not BCP as I first thought, but “Feedback: beyond behaviorism” which they, apparently, did not fully understand. After that, they merrily jump to Bayesian inference modeling, instead of control systems modeling.

Perhaps it was a bit rushed to conclude that their results are supportive of PCT. I saw their conclusion, (hierarchy of closed loops in motor control), the verbally described model seemed vaguely possible… And they actualy said PCT is right.

Thank you for taking the time to review it.

Adam

17

[From Adam Matic 2013.2.1.0020 cet]

@Rick

By the way, do you think it would be hard to model something like their distance discrimination task, to try to compare their behavioral data to PCT model data?

I imagine about three levels or so, bottom levels control the position and velocity of arms, above them is a “contact time difference controller”. If the difference is close to 0, then velocity gets decreased in the next try. When introspecting during the task, I feel like that is what I do.

There are contact receptors… somewhere… I have no idea where to plug touch or how to connect it to other loops.

Adam

···

On Thu, Jan 31, 2013 at 9:02 PM, Adam Matic adam.matic@gmail.com wrote:

[From Adam Matic 2013.1.31.2100 cet]

Rick Marken (2013.01.31.1100)

[…]

So my conclusion is that this paper, that is presumably “supportive” of PCT is, at best, irrelevant to PCT and, at worst, completely misleading about the nature of purposeful (control) behavior. In my estimation this paper is worth less than nothing.

AM:

After the tone of your previous post, I expected something like this, so I went over the paper a few more times and… yeah… they don’t look like they understand control theory let alone PCT. There are really no control systems modeled, of any kind.

This line sums up a lot: “Given the inherent uncertainty of sensory coding, we tried to see whether our results can be explained by a closed-loop optimal control of sensory coding (Powers, 1973)”.

Aside from the obvious mention of optimal control, there is no connection between PCT and “inherent uncertainty of sensory coding”, “sensory coding” is not at all what is controlled, and their reference is not BCP as I first thought, but “Feedback: beyond behaviorism” which they, apparently, did not fully understand. After that, they merrily jump to Bayesian inference modeling, instead of control systems modeling.

Perhaps it was a bit rushed to conclude that their results are supportive of PCT. I saw their conclusion, (hierarchy of closed loops in motor control), the verbally described model seemed vaguely possible… And they actualy said PCT is right.

Thank you for taking the time to review it.

Adam

18

[From Rick Marken (2013.01.31.1545)]

[From Adam Matic 2013.1.31.2100 cet]

RM: I certainly understand the desire to see other people’s work as acknowledging and being supportive of PCT. It’s nice to feel like one has a support group; to be part of a community with a common interest. When I first got into PCT I was certainly looking everywhere for others who were just as excited as me about what we now call PCT. I found very few but those eventually became the CSG. It quickly became apparent to me that there was never going to be a lot of PCT supporters because there was great resistance to PCT. So I knew PCT was going to be a fairly lonely trip. But I am always hoping to gain supporters and so are the others who are in CSG. And there lies the rub. Because I think there is some difference of opinion among supporters of PCT about how to increase the size of the PCT community. I have no fixed policy myself. How I deal with it depends on the context.

The main point of contention seems to be what to do about people who seem to support PCT (like Saig et al) but get it wrong to some degree. One approach is to welcome them and ignore the wrong stuff as best as possible. Another is to demand correctness and be critical no matter how much people say they love and support PCT if what they love and support is something only vaguely related to PCT. As I said, I take both approaches, depending on context. But at heart I feel that the only support I enjoy comes from people who really are really excited about PCT and want to understand it. These are people – and I think you are one of them, for sure, Adam – who are really trying to learn PCT (by doing the modeling and research required) and deal with criticism in the same way I would if I were learning PCT (or anything, for that matter) for the first time. So if Saig et al are really into learning PCT – and not just finding verbal similarities to what they already believe – they will take whatever criticism they get as constructive.

Best

Rick

···

RM: So my conclusion is that this paper, that is presumably “supportive” of PCT is, at best, irrelevant to PCT and, at worst, completely misleading about the nature of purposeful (control) behavior. In my estimation this paper is worth less than nothing.

AM:

After the tone of your previous post, I expected something like this, so I went over the paper a few more times and… yeah… they don’t look like they understand control theory let alone PCT. There are really no control systems modeled, of any kind…

Perhaps it was a bit rushed to conclude that their results are supportive of PCT. I saw their conclusion, (hierarchy of closed loops in motor control), the verbally described model seemed vaguely possible… And they actualy said PCT is right.

Thank you for taking the time to review it.

19

[From Rick Marken (2013.01.31.1600)]

Yes, I think it would be very hard. For one main reason: this is a task which is designed to give the subjects very poor control of the main variable they are supposed to control, which is the relationship between their response (“left” or “right” pole) and the actual relative posterior positions of the poles. PCT is a theory of how people control. This kind of experiment is set up to produce behavior that is consistent with the probabilistic model of of behavior assumed by conventional psychology. You could probably come up with a PCT model that mimics the arm/finger movement that is observed but I wouldn’t be confident that it was really controlling in the way the subjects are – or aren’t. This is a good example of conventional research that just isn’t very useful for control analysis. I certainly believe that the behavior in this experiment is the result of a control organization in the subjects; but I think it would be very difficult to develop a good model of what’s going on.

It’s just really hard to use PCT to model the behavior in experiments that were not designed to test the PCT model. Sometimes it seems like you can: Bill Powers did it with a model of shock avoidance behavior (reported in the Feedback Model of a Rat Experiment that is reprinted in LCS I). The model worked pretty well but I know that Bill’s later work with modeling operant behavior revealed that many of the assumptions of the model (such as constant response rate) are just not what actually happened in the experiment.

So in general I would say that the only research worth looking at in terms of providing tests of the PCT model is research that is informed by an understanding of PCT.

Best

Rick

···

Adam Matic (2013.2.1.0020 cet)–

@Rick

By the way, do you think it would be hard to model something like their distance discrimination task, to try to compare their behavioral data to PCT model data?

I imagine about three levels or so, bottom levels control the position and velocity of arms, above them is a “contact time difference controller”. If the difference is close to 0, then velocity gets decreased in the next try. When introspecting during the task, I feel like that is what I do.

There are contact receptors… somewhere… I have no idea where to plug touch or how to connect it to other loops.

20

[From Bill Powers (2013.02.01.1635 MST)]

Adam Matic 2013.1.31.2100 cet 00

Rick Marken (2013.01.31.1100)

[…]
So my conclusion is that this paper, that is presumably
“supportive” of PCT is, at best, irrelevant to PCT and, at
worst, completely misleading about the nature of purposeful (control)
behavior. In my estimation this paper is worth less than
nothing.

AM:

After the tone of your previous post, I expected something like this, so
I went over the paper a few more times and… yeah… they don’t look like
they understand control theory let alone PCT. There are really no control
systems modeled, of any kind.

BP: I ran across a paper about mice or rats using vibrissae (movable
whiskers) to sense the positions of things – didn’t save it, but I think
the Saig et. al, paper is specifically about that kind of sensing. The
two arms sweeping the hands back and forth until they touch something are
like what the whiskers do.

I concur with your judgement, and Rick’s, about the relevance this paper
to PCT. Zero, or perhaps a little less.

Best,

Bill P.

···

This line sums up a lot:
“Given the inherent uncertainty of sensory coding, we tried to see
whether our results can be explained by a closed-loop optimal control of
sensory coding (Powers, 1973)”.

Aside from the obvious mention of optimal control, there is no connection
between PCT and “inherent uncertainty of sensory coding”,
“sensory coding” is not at all what is controlled, and their
reference is not BCP as I first thought, but “Feedback: beyond
behaviorism” which they, apparently, did not fully understand. After
that, they merrily jump to Bayesian inference modeling, instead of
control systems modeling.

Perhaps it was a bit rushed to conclude that their results are supportive
of PCT. I saw their conclusion, (hierarchy of closed loops in motor
control), the verbally described model seemed vaguely possible… And they
actualy said PCT is right.

Thank you for taking the time to review it.

Adam