Feldman et al., 2014: Motor Control and Position Sense

CSGnet Archive CSG2014
1

[From MK (2014.12.17.1800 CET)]

Chapter 2: Motor Control and Position Sense: Action-Perception
Coupling by Feldman, Ilmane, Sangani and Raptis in Progress in Motor
Control, edited by Levin. Advances in Experimental Medicine and
Biology. Volume 826, 2014.
DOI 10.1007/978-1-4939-1338-1

Excerpt:

"Some emerging theories of action and perception have not integrated
the notion of referent position control. The sensory prediction theory
(SPT) suggests that the brain predicts the proprioceptive consequences
of the intended movement and that these predictions are delivered to
motoneurons via descending pathways (Adams et al. 2013). Motoneurons
are activated or not depending on the difference ("predictive error")
between the predicted and the actual proprioceptive feedback. The SPT
is reminiscent of a similar theory proposed several decades ago by
Powers (1973) with the complementary assumption that the predictive
error is eliminated by the neuromuscular system that works as a linear
closed-loop servo-controller. To be functional, such a controller
would have to have a physiologically unrealistic high gain, like in
the servo-control hypothesis by Merton (Merton 1953 ; see also
criticisms of the theory by Fowler and Turvey 1978). In any case, by
assuming that the nervous system preprograms the sensory consequences
of the motor outcome, SPTs do not help explain how PS is formed since
these sensory consequences carry ambiguous information about positions
of body segments (see Introduction). In addition, SPT seems to
misrepresent how motor actions are controlled. Consider, for example,
isometric torque production. In terms of SPTs, this is achieved by
pre- programming of the sensory signals ("afference copy") associated
with the required torque resulting from activation of motoneurons that
function depending on the error in prediction of these signals. After
several trials, the sensory signals associated with the required
torque could be identified correctly and reproduced with a minimal
error. Then, paradoxically, motoneurons that function depending on
this error would barely be activated to generate the required torque.
Therefore, not only PS but also motor control is inexplicable in the
framework of SPTs."

M

2

[From Rick Marken (2014.12.17.1545)]

···

MK (2014.12.17.1800 CET)–

MK: Chapter 2: Motor Control and Position Sense: Action-Perception

Coupling by Feldman, Ilmane, Sangani and Raptis in Progress in Motor

Control, edited by Levin. Advances in Experimental Medicine and

Biology. Volume 826, 2014.

MK: Excerpt:

MK: "Some emerging theories of action and perception have not integrated

the notion of referent position control.

RM: This is another great find, Matti. But kind of confusing. It sounds like Feldman et al like the idea of “referent position control” which sounds like control theory. But then they go on to criticize control theory. Maybe “reference position control” means control of position as an output rather than an input. Is that it?

MK: [Quoting Feldman] "The SPT is reminiscent of a similar theory proposed several decades ago by Powers (1973) with the complementary assumption that the predictive
error is eliminated by the neuromuscular system that works as a linear

closed-loop servo-controller".

RM: It’s not really predictive error; it’s current error. But he’s got the essentials.

MK: [Quoting Feldman] "To be functional, such a controller

would have to have a physiologically unrealistic high gain, like in

the servo-control hypothesis by Merton (Merton 1953 ; see also

criticisms of the theory by Fowler and Turvey 1978).

RM: I would love to see on what basis Feldman claims that a feedback control system “would have to have unrealistic high gain”. Perhaps Richard Kennaway or Bruce Abbott knows?

RM: And it’s kind of fun (“fun” if you are a masochistic control theorist, as I am;-) to see Feldman drag out the Fowler/Turvey article criticizing Powers’ theory. I know that Bill wrote a detailed (and trenchant) rebuttal to the Fowler/Turvey hatchet job because he shared it with me when he wrote it back in 1980. Unfortunately, Bill never published the rebuttal (and never tried to; it must be in the archives!) since the Fowler/Turvey paper was published in a book rather than a journal. But I did publish my own little rebuttal and hid it away in my “Nature of Behavior” paper which is reprinted in “Mind Readings” so you can read it (apparently Feldman didn’t) on pp. 21-23 of that book.

MK: [Quoting Feldman] “In any case, by assuming that the nervous
system preprograms the sensory consequences of the motor outcome,
SPTs do not help explain how PS is formed since these sensory consequences
carry ambiguous information about positions of body segments”.

RM: That’s what Fowler/Turvey said about Powers’ control model and it’s what I showed to be false.

MK: [Quoting Feldman] "In addition, SPT seems to

misrepresent how motor actions are controlled".

RM: I think we all know what’s wrong with that claim.

MK: [Quoting Feldman] "Consider, for example,

isometric torque production. In terms of SPTs, this is achieved by

pre- programming of the sensory signals (“afference copy”) associated

with the required torque resulting from activation of motoneurons that

function depending on the error in prediction of these signals."

RM: And again we know that there is no “pre-programming” of sensory signals in the control model of behavior. It’s just error driven output having an effect on sensory signals until the error goes to zero.

MK: [Quoting Feldman] " After several trials, the sensory signals
associated with the required\ torque could be identified correctly and
reproduced with a minimal error.

RM: No they couldn’t. The torque required to bring the sensory signals to the reference state can’t possible be learned (or known) because of the effect of disturbances on the variable represented by the sensory signals. The torque required will always be differnet depending on prevailing circumstances (disturbances).

MK: [Quoting Feldman]Then, paradoxically, motoneurons that function
depending on this error would barely be activated to generate the required torque.

Therefore, not only PS but also motor control is inexplicable in the

framework of SPTs."

RM: This article and Feldman’s work in general seems to be an excellent example of how much an understanding of physiology can contribute to a misunderstanding of control.

RM: But it is a great source of ideas for research testing a control theory model of “motor control”.

Best

Rick

–
Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

3

[From Bruce Abbott (2014.12.17.1935 EST]

MK (2014.12.17.1800 CET) --

Chapter 2: Motor Control and Position Sense: Action-Perception Coupling by
Feldman, Ilmane, Sangani and Raptis in Progress in Motor Control, edited by
Levin. Advances in Experimental Medicine and Biology. Volume 826, 2014.
DOI 10.1007/978-1-4939-1338-1

Excerpt:

"Some emerging theories of action and perception have not integrated the
notion of referent position control. The sensory prediction theory
(SPT) suggests that the brain predicts the proprioceptive consequences of
the intended movement and that these predictions are delivered to
motoneurons via descending pathways (Adams et al. 2013). Motoneurons are
activated or not depending on the difference ("predictive error") between
the predicted and the actual proprioceptive feedback. The SPT is reminiscent
of a similar theory proposed several decades ago by Powers (1973) with the
complementary assumption that the predictive error is eliminated by the
neuromuscular system that works as a linear closed-loop servo-controller. To
be functional, such a controller would have to have a physiologically
unrealistic high gain, like in the servo-control hypothesis by Merton
(Merton 1953 ; see also criticisms of the theory by Fowler and Turvey 1978).
In any case, by assuming that the nervous system preprograms the sensory
consequences of the motor outcome, SPTs do not help explain how PS is formed
since these sensory consequences carry ambiguous information about positions
of body segments (see Introduction). In addition, SPT seems to misrepresent
how motor actions are controlled. Consider, for example, isometric torque
production. In terms of SPTs, this is achieved by
pre- programming of the sensory signals ("afference copy") associated with
the required torque resulting from activation of motoneurons that function
depending on the error in prediction of these signals. After several trials,
the sensory signals associated with the required torque could be identified
correctly and reproduced with a minimal error. Then, paradoxically,
motoneurons that function depending on this error would barely be activated
to generate the required torque.
Therefore, not only PS but also motor control is inexplicable in the
framework of SPTs."

BA: Thanks, Matti. I'm currently working on a book chapter comparing
Powers' model of "motor control" with Feldman's "equilibrium point"
hypothesis, and will certainly add reference to this.

BA: In case anyone is interested, here is a link to a fairly extensive
presentation and discussion of the Powers proposal, provided by Fowler and
Turvey (1978) and cited above by Feldman et al. (2014):
http://books.google.com/books?hl=en&lr=&id=5qs5AwAAQBAJ&oi=fnd&pg=PT7&dq="
Motor+Control+and+Learning%22+Stelmach&ots=ddQOAG3fp_&sig=il3Oo9Nwe6iUy6mbXd
2bODYwAAo#v=onepage&q=Powers&f=false

BA: The discussion of Powers begins on page 26. Fowler and Turvey do a
fairly good job of presenting the individual control system and showing how
such systems are arranged to provide hierarchical control. Despite this they
appear to have had a limited understanding of the way a control system
works. On page 27, for example, they state that the reference signal "is
compared periodically with the actual perceptual state of affairs" (i.e.,
the perceptual signal). Periodically? Later, Fowler and Turvey question
whether such a system could learn to generate the proper error-correcting
outputs when "the error signal E does not specify to the subject what he
must do to correct it." They refer to the error signal as providing
"quantitative knowledge of results" and state that quantitative information
about the change in E is only provided by "successive loops around the
control system," whereas "the individual control system of the type that
Powers describes uses only the current value of E to guide its behavior."
This nonsense is the analysis by Fowler and Turvey that Feldman cites as
providing additional criticisms of the theory!

Bruce

4

[From Rick Marken (2014.12.17.1800)]

···

Bruce Abbott (2014.12.17.1935 EST) –

MK: Chapter 2: Motor Control and Position Sense: Action-Perception Coupling by

Feldman, Ilmane, Sangani and Raptis in Progress in Motor Control, edited by

Levin. Advances in Experimental Medicine and Biology. Volume 826, 2014.

DOI 10.1007/978-1-4939-1338-1

BA: Thanks, Matti. I’m currently working on a book chapter comparing

Powers’ model of “motor control” with Feldman’s “equilibrium point”

hypothesis, and will certainly add reference to this.

BA: In case anyone is interested, here is a link to a fairly extensive

presentation and discussion of the Powers proposal, provided by Fowler and

Turvey (1978) and cited above by Feldman et al. (2014):
[Information Processing in Motor Control and Learning - Google Books

Motor+Control+and+Learning%22+Stelmach&ots=ddQOAG3fp_&sig=il3Oo9Nwe6iUy6mbXd

2bODYwAAo#v=onepage&q=Powers&f=false](Information Processing in Motor Control and Learning - Google Books)

BA: The discussion of Powers begins on page 26…

RM: It’s great that you are doing this, Bruce. It think it would be very useful if you could get a hold of the critique of that Fowler/Turvey article that Bill wrote back then to use as a reference. It’s surely in the Powers archive at Northwestern. Maybe you could ask Allie about it (or, if Allie is listening in, maybe she could start searching for it). The paper was written in about 1979 or 1980 but it could have been written as late as 1982. Since there was no email back then Bill must have sent me the copy I read by snail mail because I remember reading it in my home in Mpls but I might have read it at his place. But I’m sure there was a hard copy and it was a pretty strong paper. I’m pretty sure it has the names “Fowler and Turvey” in the title.

Best

Rick

–

Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

5

[From MK (2014.12.18.2230 CET)]

Rick Marken (2014.12.17.1545)--

But kind of confusing. It sounds
like Feldman et al like the idea of "referent position control" which sounds
like control theory. But then they go on to criticize control theory. Maybe
"reference position control" means control of position as an output rather
than an input.

It is less confusing in context, here's a temporary copy of the
chapter in question.

http://pct.loopgain.com/Feldman, 2014 - Motor Control and Position Sense.pdf

M