On the non-control of angles (Re: Re-organizable analog...)

[From Matti Kolu (2013.07.26.1945CET)

Adam Matic 2013.07.26 1515 CET--

Muscles can be controlled in two dimensions independently - angle and
tension.

Tension is a label that can be applied to low-order signals, "angle"
is typically a high-order signal.

A changing "joint angle" is a possible *perceived consequence* of
muscles contracting. In humans, the perceptual functionality for
perceiving this constructed variable that we refer to as "joint angle"
is likely located in the upper levels of the hierarchy.

I have a hard time understanding how lower-order systems would know
anything about "angles". They "know" only what they can perceive --
they don't know anything about the environment or the possible
disturbances involved.

It is possible to experience a lot of tendon stretch, or tension,
without any significant change in the *high-level perception* of
"joint angle".

As you are modeling control systems in a virtual environment, which by
definition is artificial, and disturbance-free unless disturbances are
specifically added, you are perhaps taking for granted the
availability of easily accessible coordinate points, in which case
creating an "angle reading" is a matter of doing some simply
trigonometry.

It seems to me that it's only in a virtual environment of that kind
you can get away with calling an "angle" a low-level perceptual
function.

Matti

[Martin Taylor 2013.07.26.14.09]

[From Matti Kolu (2013.07.26.1945CET)

Adam Matic 2013.07.26 1515 CET--

Muscles can be controlled in two dimensions independently - angle and
tension.

Tension is a label that can be applied to low-order signals, "angle"
is typically a high-order signal.

A changing "joint angle" is a possible *perceived consequence* of
muscles contracting. In humans, the perceptual functionality for
perceiving this constructed variable that we refer to as "joint angle"
is likely located in the upper levels of the hierarchy.

I have a hard time understanding how lower-order systems would know
anything about "angles". They "know" only what they can perceive --
they don't know anything about the environment or the possible
disturbances involved.

Do you know that there are no sensors of joint angle in the body? If there are none, then as you say, joint angle perception is based on relationships of other sensor signals, though they could still be low-order. But if joint angle sensors exist, wouldn't their outputs be lowest-order perceptions (not just "low-order")?

Martin

[From Adam Matic 2013.07.26 2030 CET]

···

AM: Muscles can be controlled in two dimensions independently - angle and

tension.

MK: Tension is a label that can be applied to low-order signals, “angle”

is typically a high-order signal.

A changing “joint angle” is a possible perceived consequence of

muscles contracting. In humans, the perceptual functionality for

perceiving this constructed variable that we refer to as “joint angle”

is likely located in the upper levels of the hierarchy.

I have a hard time understanding how lower-order systems would know

anything about “angles”. They “know” only what they can perceive –

they don’t know anything about the environment or the possible

disturbances involved.

It is possible to experience a lot of tendon stretch, or tension,

without any significant change in the high-level perception of

“joint angle”.

AM: I’m not sure I understand the problem. It is also possible to experience no significant change in tension while changing the angle of some joint, like if you move your hand repeatedly up and down - the angle is changing, but the tension is pretty much constant.

But you did get me thinking. As you say, it quite easy to represent angles in a virtual environment, but when it comes to how real neurons calculate or represent angles, it gets a bit complicated. If first order perceptions represent muscle and tendon stretch, and tendons and muscles can stretch independently of joint angle, then the angle is probably a second order function which includes tension receptor signals from both flexor and extensor muscles and their tendons. If you are changing the tension of say, biceps, the triceps will tense up too. A signal representing relative stretch of flexor and extensor muscles and tendons could represent joint angle. Not in degrees, but a variable that correlates in some way with what an external observer would call ‘angle’.

In Bill’s arm models, first order systems control the velocity by changing muscle force. Second order systems control position (angle) by changing velocity, but this could also be done in a single order control system. Models do behave quite similar to how human arms behave, but finding out how exactly is this done in human control systems requires some low-level meddling with spinal cord neurons, specifically with the PCT model in mind.

Adam

[From Matti Kolu (2013.07.27.22.00 CET)]

Martin Taylor (2013.07.26.14.09)--

Do you know that there are no sensors of joint angle in the body? If there
are none, then as you say, joint angle perception is based on relationships
of other sensor signals, though they could still be low-order.

No. But I have, due to some extremely peculiar circumstances, a
resemblance of an idea of what kind of perceptual variables are
constructed when the highest levels in the human hierarchy are formed.
I don't see myself being able to describe this, at the moment tacit
understanding, in any rigorous way during the foreseeable future. But
it does mean that, when it comes to human beings, I'm able to treat
certain lines of thought as either implausible or very unlikely.

The concept of "joint angles" is one such thing. The phrase is useful
and innocuous when dealing with "stick figure models with joints" in
virtual environments, or when working with robots, but I suspect that
the term is highly misleading when the discussion concerns the human
perceptual hierarchy. The expression seems to be "carried over" a bit
too readily from discussions about models in virtual environments to
descriptions that concern plausible organizations of human control.

The careless adoption of "angle" in situations where another phrase
would do, risks sending future control scientists down extremely
unproductive lines of thought. Before asking ourselves whether there
are joint angle sensors in the body, or how joint angles are
controlled, we should make sure that the framing of the questions is
correct.

What are the assumptions underlying those questions? Doesn't "angle"
presuppose the existence of a certain kind coordinate system, that
when -- if you take the empathetic / inside view of the control system
-- doesn't actually seem to be there?

Matti

[Martin Taylor 2013.07.27.16.37]

[From Matti Kolu (2013.07.27.22.00 CET)]
  Doesn't "angle"
presuppose the existence of a certain kind coordinate system, that
when -- if you take the empathetic / inside view of the control system
-- doesn't actually seem to be there?

Why? all it requires is a sensor that provides a signal that changes monotonically with change in what an external observer would call the angle of the joint. No coordinate system is implied, just that more flexure of, say, the elbow, means more (or less) output from that sensor. If there's no such sensor, then a second-level system that perceives the ratio of stretch from the two opposed muscle/tendons would serve the same function. No coordinate system in that case, either.

Martin

[From Matti Kolu (2013.07.28.0230 CET)]

Why? all it requires is a sensor that provides a signal that changes
monotonically with change in what an external observer would call the angle
of the joint.

Hm. I don't think I can object to that.

I've been looking at an abbreviated and slightly outdated physiology
book. The authors have conveniently left out basically everything that
deals with the sensory innervation of the joints, giving the
impression that there's little going on there. I should have known to
switch to a more extensive source, sooner.

Matti

[From Matti Kolu (2013.07.28.0435 CET)]

Matti Kolu (2013.07.28.0230 CET)--

I should have known to switch to a more extensive source, sooner.

To the authors' defence, it's seems to be hard to sum up what's going
on in the joints in a concise way. When reading reviews on
mechanoreceptors and joint angles, it often sounds like this:

"The pursuit of the neural mechanisms of joint angle perception has a
long, rich history. In the 1960s, there was near universal agreement
that joint receptors are responsible for joint angle perception
(Goodwin et al. 1972b). What could be more obvious? The joints are
richly innervated with sensory receptors, and they are ideally located
to signal joint angle. Signals from the muscles and skin are indirect
and fraught with complications. However, almost immediately the
evidence weighed in heavily against joint receptors. In 1969, Burgess
and Clark showed that the vast majority of joint receptors respond
only at the extremes of extension and flexion and that they signal
joint angle poorly at intermediate angles (Burgess and Clark 1969;
Burke et al. 1988; Clark and Burgess 1975). In 1979, Clark et al.
showed that anesthetizing the receptors of the knee joint capsule and
the surrounding skin had no effect on the human ability to judge and
discriminate joint angle. In the meantime, Goodwin et al. (1972a) had
shown that vibratory stimuli (that activate muscle spindles strongly)
applied to the tendons of biceps or triceps muscles alter the
perceived elbow angle by as much as 40�. Thus within a decade the
joint receptor lost its central role almost completely (Proske et al.
1988) and joint angle perception seemed to depend almost exclusively
on muscle spindles (Matthews 1982). Note, however, that Ferrell has
repeatedly presented evidence that the joint receptor plays a role,
albeit a tertiary role (e.g., Ferrell et al. 1987). "

Johnson K. Closing in on the neural mechanisms of finger joint angle
sense. Focus on "Quantitative analysis of dynamic strain sensitivity
in human skin mechanoreceptors". J Neurophysiol. 2004
Dec;92(6):3167-8.

http://jn.physiology.org/cgi/pmidlookup?view=long&pmid=15548632

Matti

[From Matti Kolu (2013.07.28.1840 CET)]

Adam Matic 2013.07.26 2030 CET]

But you did get me thinking. As you say, it quite easy to represent angles
in a virtual environment, but when it comes to how real neurons calculate or
represent angles, it gets a bit complicated.

Here's a partial explanation for why *I* am constantly getting
confused about what you guys are writing. When perceiving what we'd
consider to be an ordinary action -- such as reaching for a cup of
coffee -- I am currently perceiving it primarily as changes of
relationships in, and transformations of -- for a lack of better term
-- "topological surfaces".

(That's an colloquial expression -- I don't know anything about topology yet).

This is true regardless of whether I am observing my own actions or
the actions of another person. As I have an alternative and
idiosyncratic way of seeing for example the relationship between
"forearm" and "upperarm" I get confused when such a relationship is
described in terms of "joint angle" because *for me* that's not at all
what's going on. Such a description strikes me as false. But that's
based on feeling, not on technical analysis.

So I do see that something like this:

A signal representing relative stretch of flexor and extensor muscles and tendons
could represent joint angle. Not in degrees, but a variable that correlates in some
way with what an external observer would call 'angle'.

Is possible.

I've attached two images to attempt to illustrate what I'm getting at.
In the first image only the A) section is relevant -- these are just
the first tiny parts of a sketch.

Matti

Content-Type: image/png; name="1. surfaces-vs-angles-sketch.png"
Content-Disposition: attachment; filename="1. surfaces-vs-angles-sketch.png"
X-Attachment-Id: f_hjogolas0

Content-Type: image/png; name="2. metromode-partial-squat-angle-overlay.png"
Content-Disposition: attachment;
  filename="2. metromode-partial-squat-angle-overlay.png"
X-Attachment-Id: f_hjogovje1

[From Matti Kolu (2013.07.29.1250 CET)]

[From Matti Kolu (2013.07.28.1840 CET)]
I am currently perceiving it primarily as changes of
relationships in, and transformations of -- for a lack of better term
-- "topological surfaces".

Okay. So this is related to what's in vision and computer vision
research referred to as the shape-from-shading problem. The next
question is to which degree it has been tied to developmental
psychology.

Because the ability to do the discrimination described in the earlier
attached image "1. surfaces-vs-angles-sketch.png" isn't there from the
beginning. The associated perceptual functionality should typically be
constructed at a specific developmental step -- which should coincide
or be identical to one of the levels described in Plooj's works.

Furthermore, we are not talking about something that simply goes from
off to on. There are a bunch of "parameters" -- ("sensitivity",
"simultaneous complexity", speed of change in the "transformations" of
the delimited surfaces?) -- involved that are slowly improved, once
the initial functionality has been constructed.

If you can devise a test for this functionality, and the associated
parameters -- which won't be easy, as we are dealing with young
children -- you'll have a measure for a specific type of perceptual
development. Perhaps it has already been done? The test can be used in
a non-intuitive way as a proxy for a whole bunch of developmentally
interesting stuff.

If you test a large enough population you should find individuals who
are not able to this discrimination or who do it really poorly. This
will show up in ways that aren't immediately obvious. Among other
things, it'll have a subtle effect on measurements related to "motor
control" (what a weird expression!) -- they'll be able to perform most
tasks just as well as anybody else, but as they are performing the
tasks they are actually not controlling for certain things (they can't
perceive them) which means that the externally viewed postures and
limb trajectories should deviate from what you'd expect if their
perceptual hierarchies were constructed in more common ways.

Matti

[Martin Taylor 2013.07.30.10.44]

[From Matti Kolu (2013.07.28.1840 CET)]

...

I've attached two images to attempt to illustrate what I'm getting at.
In the first image only the A) section is relevant -- these are just
the first tiny parts of a sketch.

Matti

In the second panel showing the semi-squatting woman, you ask "Does a young child � or another primate � perceive things in terms of "joint angles"? If not � what are *they* perceiving?"

I rather suspect you are confusing "perceptual signals" or "perceptual functions" with "conscious perception". It takes a bit of training to be able consciously to perceive many of our lower-level perceptual signals. It's part of what a top-class athlete or, especially, a ballet dancer, has to be able to do well, and what a novice has to learn to do. If you play golf, can you perceive what the pro tries to get you to perceive about when to feel what tensions in which muscles, how and when to rotate your shoulders, when your head does and does not move? It's difficult.

So the answer to your question is "we won't know what they are perceiving unless we do some physiological testing to see what signals are there that might be available to become incorporated into perceptions". It's harder to know which of those signals are actually used as perceptions possibly subject to control. Whether "joint angle" is among them, either directly sensed or as a low-level difference or ratio of muscle tensions, is a specific question that could then be answered. But it has nothing to do with whether the person consciously perceives any joint angles.

From a subjective point of view, when somebody says "hold that pose" and then gently tries to push you around, you probably can maintain your joint angles, which are what determine your pose. The "Test for the Controlled Variable" would suggest that joint angles are controlled perceptions in that situation. When all the chorus ballerinas mirror each other's pose very accurately, they surely are controlling perceptions that are functionally equivalent to joint angles.

Whether you could get a small child or another primate to want to "hold that pose" could be a problem. If they could, but didn't want to, it would be hard to determine whether or not they were controlling joint angle when they move around.

Martin

[From Rick Marken (2013.07.29.1250)]

Matti Kolu (2013.07.28.1840 CET)

Here's a partial explanation for why *I* am constantly getting
confused about what you guys are writing. When perceiving what we'd
consider to be an ordinary action -- such as reaching for a cup of
coffee -- I am currently perceiving it primarily as changes of
relationships in, and transformations of -- for a lack of better term
-- "topological surfaces".

I'm confused myself. I don't see what your way of perceiving things
has to do with the subject of this thread, which is "On the
non-control of angles". The title implies that we cannot control an
angle, such as the angle at the elbow between upper arm and forearm.
But I can control that angle and so can everyone I've asked to do it.
So maybe we should start all over again with you explaining why you
think we can't control angles.

Best regards

Rick

···

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

[From Matti Kolu (2013.07.29.2340 CET)]

Rick Marken (2013.07.29.1250)--

I don't see what your way of perceiving things
has to do with the subject of this thread, which is "On the
non-control of angles".

Can you show me a way to represent or describe the relationships
between N continuously changing surfaces, delimited by the distinction
(D) that I pointed at in "1. surfaces-vs-angles-sketch.png", using
angles? Pretend that I'm from the future and that I have access to 30
months of data -- collected from the organism's point of view --
giving an idea of what the introduction of the perceptual
functionality related to D has on the variables that the organism,
following the introduction of the ability to discern D, is now able to
control for.

Matti

[From Rick Marken (2013.07.29.1500)]

Matti Kolu (2013.07.29.2340 CET)--

RM: I don't see what your way of perceiving things
has to do with the subject of this thread, which is "On the
non-control of angles".

MK: Can you show me a way to represent or describe the relationships
between N continuously changing surfaces, delimited by the distinction
(D) that I pointed at in "1. surfaces-vs-angles-sketch.png", using
angles?

RM: No.

Now could you answer my question: What does your way of perceiving
things have to do with the subject of this thread: non-control of
angles?

Thanks

Best

Rick

Pretend that I'm from the future and that I have access to 30

···

months of data -- collected from the organism's point of view --
giving an idea of what the introduction of the perceptual
functionality related to D has on the variables that the organism,
following the introduction of the ability to discern D, is now able to
control for.

Matti

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

[Martin Taylor 2013.07.29.19.27]

[From Matti Kolu (2013.07.29.2340 CET)]

Can you show me a way to represent or describe the relationships
between N continuously changing surfaces, delimited by the distinction
(D) that I pointed at in "1. surfaces-vs-angles-sketch.png", using
angles? Pretend that I'm from the future and that I have access to 30
months of data -- collected from the organism's point of view --
giving an idea of what the introduction of the perceptual
functionality related to D has on the variables that the organism,
following the introduction of the ability to discern D, is now able to
control for.

From that standpoint, it would be very difficult to figure out what the person was controlling for, unless the data included measures of disturbances to possibly controlled variables that had been introduced to see whether those variables were in fact controlled. From outside, without such disturbances, all you know is that it seems as though the person might possibly be controlling for just obscuring one surface with another, but you couldn't actually be sure that the effect wasn't a side effect of controlling for something quite different such as holding the barbell at a particular height off the floor. In order to achieve this reference value for the controlled perception of the degree to which the knee overlaps the foot (or of keeping the barbell at that height), the outside observer might note that the person maintains some joint angles at very specific values.

That outside observer might then hypothesize that somewhere in the hierarchy through which the person controls the visible overlap or the barbell height there are control units that control joint angle perceptions. But the outside observer wouldn't know whether that was true, any more than Rick and I know whether in my tracking study the subject controls for the angle between the cursor and target or the horizontal disparity between cursor and target. You just can't tell by looking at what someone is doing. That's another PCT mantra, along with "All behaviour is the control of perception" -- "You can't see what someone is doing by looking at what they are doing". Even from inside, it's often hard to tell. Rick and I both believe that subjectively it feels as though we are controlling angle and not displacement, but so far it has not been demonstrated because we have not disturbed both possibilities independently.

I can see no "versus" in your sketches. You might get away with exchanging "vs" with "by means of", but it would only be a hypothesis.

Martin

[From Matti Kolu (2013.07.30.1230 CET)]

Rick Marken (2013.07.29.1500)--

What does your way of perceiving
things have to do with the subject of this thread: non-control of
angles?

If you look carefully you will note that the delimiter D creates
distinct surfaces in 3d-space that allows you to -- among other things
-- to perceive what's commonly referred to as:

1."The separation between the body and the environment"
and (depending on the observer's perspective)
2. "Different body parts".

What you refer to as "upper arm" and "forearm" could be seen as
referring to a *multitude* of "surfaces" in 3d-space delimited -- or
characterized -- by D.

When I asked you if you could show me a way to represent or describe
relationships between N surfaces, delimited by D, using angles, you
answered no.

Hence the non-control of angles.

Matti

[From Matti Kolu (2013.07.30.1315 CET)]

Martin Taylor 2013.07.30.1044--

It takes a bit of training to be
able consciously to perceive many of our lower-level perceptual signals.

This is normally the case. I think there are some posts in the
archives where Powers describes how he needed decades of careful
observation to come up with the proposed hierarchy, which many of us
probably consider to be rough starting point, if that.

But to quote the English physician William Harvey: "Nature is nowhere
accustomed more openly to display her secret mysteries than in cases
where she shows tracings of her workings apart from the beaten paths."

Here is a partially marked up image of the squatting woman according
to what I find most pertinent about it:

There was little effort involved in this. I didn't have to force
myself to only pay attention to certain aspects of the image. The
things the red lines represent stand out extremely clearly to me --
not only in static pictures, but in real life.

What would you say that my markings represent? Would you say that they
are related to the distinctions I tried to illustrate in the sketchy
drawing that was attached earlier?

Note the area of the image where most of the markings are found.
There's really nothing special about fabrics per se. But due to the
flexibility of textiles they naturally end up with wrinkles and folds
that create easily-discernable distinctions that I'm currently
perceiving *everywhere*.

Here's another partially marked-up example:

Think of a uniformly colored flag in the wind. Or curtains by an open
window. A single-colored dress draping around the body of a woman as
she's walking. The changing nature of those surfaces make them
extremely noticable.

Note the also the lack of wrinkles and folds around the woman's thigh
area. Look at the leg closest to you as a viewer. There are two types
of fabric used in her workout pants. The shinier material creates an
edge when it meets the less shiny material. This does *not* -- at
least not in this particular image -- create the distinction that I
find relevant.

This is not about simple edge detection.

"Logically" and for the sake of consistency there should probably be a
red outline around the squatting woman's body. The lack of this
outline is actually important: the distinction is easiest to make when
two plain surfaces are contrasted with each other. That's one of the
reasons for why the sketch includes the part where I tell you to stand
on a towel. (Although the towel will be changed to something like a
fine-threaded pillow cover in future revisions -- the rought texture
of most towels might distract from the point I'm trying to make.) The
towel, or pillow cover, is there to make the contrasting background as
uniform as possible.

Matti

[Martin Taylor 2013.07.30.09.45]

[From Matti Kolu (2013.07.30.1315 CET)]

Martin Taylor 2013.07.30.1044--

It takes a bit of training to be
able consciously to perceive many of our lower-level perceptual signals.

...

Here is a partially marked up image of the squatting woman according
to what I find most pertinent about it:

Dropbox - metromode-with-partial-d.png - Simplify your life

I don't see any markup. I see two apparently identical images.

If I interpret you correctly, you are asking about the details of the variety of perceptual functions that provide signals that correspond to the different natures of surfaces and their distances from the eye. If this is the case, I'm with Rick in wondering why the word "angles" is in the topic line. I would add that I wonder why the word "control" is there, too. Would I be right if I suggested that a subject line more aligned with your interest would be "Perceptual functions for surface perception?" ?

I can't tell you how such perceptual functions might be constructed. I believe it is still an ongoing problem in computer vision, and since there's a good back and forth interplay between research in organic and inorganic perceptual systems, I imagine it's an unsolved problem for human perception as well. Wrinkles and other texture probably are important, and if you solve the problem you may have made an advance.

Martin

[From Matti Kolu (2013.07.30.1700 CET)]

Martin Taylor 2013.07.30.09.45--

Would
I be right if I suggested that a subject line more aligned with your
interest would be "Perceptual functions for surface perception?" ?

Yes, I should have put in more thought into the title. Although in my
mind, "surface perception" is directly related to not angles, but what
has been referred to as "joint angles". I don't think I'll be able
describe that connection in a technical manner within the nearest 15
to 20 years, and as this is primarily a technical and scientific
forum, I'll refrain from make any further comments along those lines
until my understanding of the involved subjects is dramatically
improved.

I can't tell you how such perceptual functions might be constructed.

At this point in time I wouldn't be able to understand such a function
even if it was described to me. This brings us back to the beginning:
I noted that there is an unfortunate lack of material dealing with the
issue of constructing perceptual functions in general.

Matti

[From Rick Marken (2013.07.30.0920)]

Matti Kolu (2013.07.30.1230 CET)--

RM: What does your way of perceiving
things have to do with the subject of this thread: non-control of
angles?

MK: When I asked you if you could show me a way to represent or describe
relationships between N surfaces, delimited by D, using angles, you
answered no.

Hence the non-control of angles.

RM: My "no" simply meant that I didn't understand what you were saying
and therefore could not do what you requested.

But I think I know what your problem might be. I wrote a nice long
post on it and then accidentally deleted it. So I'll just say that you
seem to be conflating the point of view of the observer and behaving
system when you say angle cannot be controlled (I think Martin made
this point as well). There is no question that people can control the
angle of their arm at the elbow; if I ask someone to keep their arm
angle at 90, 60 or 30 degrees they can do it, protected from
disturbances like changes in the orientation of the arm. So from the
observer's perspective angle is a controlled variable. But angle is a
visual variable and people can control joint angle without looking at
the joint. So from the perspective of the behaving system angle is
unquestionably not being controlled; what is controlled is some
proprioceptive perception (exactly what perception would have to be
determined by test).

So from the observer's perspective angle is a controlled variable;
from the behaving system's perspective a proprioceptive perception is
a controlled variable. This is an ambiguity that exists when talking
about control that Powers tried to resolve by calling the controlled
variable from the observer's perspective a controlled quantity and
reserving the term controlled variable for the perceptual variable(s)
controlled by the behaving system.

Hope that helps

Best regards

Rick

···

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