Levels

CSGnet Archive CSG1997
1

[ From Peter Burke 6/11/97 1:20PM]

(Bill Powers 970610.1312 MDT)

Bill, thanks for your response reminding me that each level adds something
that didn't (couldn't) exist at lower levels (organizations of lower
levels). Hence, we are dealing with two (conceptually) different
hierarchies. I think this gets confused, sometimes, in the discussions, as
well as in my own thinking.
Peter

···

------------------------------------------------------------------------
Peter J. Burke Phone: 509/335-3249
Professor and Research Scientist Fax: 509/335-6419
Department of Sociology E-mail: burkep@wsu.edu
Washington State University http://burkep.libarts.wsu.edu
Pullman, WA 99164-4020
------------------------------------------------------------------------

2

[From Rick Marken (981103.0830)]

Given the current discussion of levels of perception and control,
perhaps this is a good time to ask about one of the Nature articles
that was recommended to me as having something to do with
controlled variables. Of course, I left the article at home but
I think I remember the relevant stuff.

The article that seemed particularly relevant to levels was a
study of some sea slug that likes to keep it's head pointing upward
(toward the surface of the water); the slug basically controls
the location of a statocyst (a round pebble) that moves around
in a sensor lined chamber. When the slug's head is pointing
straight up (vertically) the statocyst rests on the bottom of the
chamber; when the slug's head is at an angle with respect to up
the statocyst moves away from the bottom, stimulating sensors to
the left or right of the bottom. Under natural conditions, the
slug apparently bends its tail to turn it's body to keep it's
head pointing up (and get the statocyst back to the bottom
of the chamber) when there is a disturbance (pressure wave?)
that would tend to push the slug's head away from straight up.
The slug is apparently controlling the sensory signal coming from
the statocyst chamber, keeping that signal at the level that
corresponds to "statocyst at bottom of chamber".

The Nature article describes a preparation where the entire
statocyst chamber, including sensors and neural connections,
was mounted on a turntable that was oriented perpendicular
to the floor; the neural outputs from the statocyst chamber
were used to drive motors that turned the turntable. The article
showed that the statotcyst preparation in this artificial control
loop (the artificial part was the feedback path, via motors and
turntable, to the statocyst) was able to control the statocyst,
keeping it at the bottom of the chamber. Control wasn't great;
there were small oscillations in the state of the presumed controlled
variable, but these oscillations were probably due to the gain
and slowing charateristics of the artificial feedback loop.

The researchers never really did test properly for the controlled
variable; they didn't monitor the state of the statocyst _while_
the orientation of the preparation was disturbed; they just showed
that the preparation would return itself to a straight up
orientation when the preparation was oriented off vertical before
the feedback loop was closed. But I'm sure the preparation would
have maintained itself near straight uo if there were a small
disturbing force exerted on the turntable tending to move the
preparation away from straight up.

Now, finally, the levels. What these researchers found is
that the preparation controlled for having the statocyst on
what had been the top part of the chamber (and, thus, for
being oriented straight downward) when the preparation was
_heated_. The preparation acts like it's reference for head
orientation changes by 180 degrees when it is heated. This
fits with the behavior of the actual slug, which apparently
turns around and swims _downward_ when it finds itself in hot
water.

So there are _apparently_ two levels of control involved in this
study; control of statocyst orientation and control of body
temperature. It seems like the temperature control system
changes the reference for statocyst orientation as the _means_
of controlling temperature. There must be behavioral tests of
this hypothesis; if not, this would sure be an intersting
study. By appropriately disturbing both body temperature and
orientation one could map out the hierarchical relationship
between control of these two variables.

If orientation _is_ controlled as the means of controlling
temperature, how does this fit into the current hierarchical
model of control? Could it be that temperature is a "sensation"
while orientation is an "intensity"? Does there need to be a
difference in _type_ of variable in order for one variable to
be used as the means for controlling another?

I would love to hear Bill's thoughts about this; it's all way
too complicated for me.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

3

[From Bill Powers (91105.0420 MDT)]

Rick Marken (981103.0830)--

If orientation _is_ controlled as the means of controlling
temperature, how does this fit into the current hierarchical
model of control? Could it be that temperature is a "sensation"
while orientation is an "intensity"? Does there need to be a
difference in _type_ of variable in order for one variable to
be used as the means for controlling another?

I don't know. Is there any reason to think that a slug's hierarchy would
resemble that of a human being? Naturally, when we look at what's going on
with the slug, we see it in terms of our own perceptual organization.
Perhaps the slug knows only that when it gets too hot, using its muscles to
create a particular touch-intensity will fix the problem. We see it turn
and swim downward, but IT doesn't see that. It probably knows nothing about
spatial relationships or even transitions.

When the statocyst moves to a different place, it stimulates different
receptors. By definition, a signal coming directly out of a sensory
receptor is a first-order perceptual signal. So the temperature-control
system (order two?) is operating by specifying which of two first-order
perceptual signals is to be present.

I'd just forget about "the" control hierarchy, when dealing with a specific
real system. We can see how a temperature control system might work through
specifying reference levels for a statocyst-touch control system, and that
would be sufficient to account for the observed behavior of the slug. If
you can see how that works, why do you need to name the levels?

Sometimes I regret ever producing that list of names. It seemed like a good
idea at the time.

Best,

Bill P.

4

[From Rick Marken (981105.1050)

Bill Powers (91105.0420 MDT)--

I'd just forget about "the" control hierarchy, when dealing with
a specific real system. We can see how a temperature control system
might work through specifying reference levels for a statocyst-touch
control system, and that would be sufficient to account for the
observed behavior of the slug. If you can see how that works, why
do you need to name the levels?

I agree completely.

Sometimes I regret ever producing that list of names. It seemed
like a good idea at the time.

I think it was (and still is) a _great_ idea. Your hypothetical
hierarchy has provided some real "value added" to the basic PCT
model. For example:

1. It shows how, _in principle_, a control of input system can
account for the kind of complex behavior we have experienced "from
the inside" (we are familiar with our own control of perceptions
of relationships, like marriage, principles, like fairness, and
system concepts, like "American") and "from the outside" (we see
that others seem to be controlling for these complex perceptions,
too). So the hierarchy adds value by showing the _feasibility_
of a control model of _all_ behavior.

2. It gives people a description of the kinds of perceptual variables
that they can experience and become aware of as controlled variables.
I know that my own understanding of PCT increased significantly when
I became aware of the variable sequences, relationships, programs,
principles and system concepts that were part of my own perceptual
world. So the hierarchy adds educational value.

3. It suggests hypotheses about the kinds of variables people
might control. The design of my own little "Hierarchy of Perception
and Control" demo/experiment at:

http://home.earthlink.net/~rmarken/ControlDemo/

(which works extremely well on my Mac G3 running Netscape 4.5) was
helped by your notion that the perception of a _sequence_ is not
the same as the perception of a _motion_ (which, in the demo, is an
illusion created by showing a rapid "sequence" of configurations).
So the hierarchy adds value by suggesting research hypothesis.

Unfortunately, the hierarchy has created some problems as well.
Actually, I can think of only one problem it has created:

1. It gives people -- especially those who are not inclined toward
observation -- the impression that the perceptual hierarchy
(intensity, sensation, transition, relationship, etc.) _is_
the PCT model. This leads to the "How many angels can dance on
the head of a pin?" sort of discussions about "where the X level
should be located" or "whether there should be a level between
level Y and Z".

I think the "value added" by the hierarchy far outweighs this
"cost".

I don't regret that you produced the names of the perceptual
levels that make up the hierarchy. Once one gets out there and
starts looking at the phenomenon of control (from the outside
and/or from the inside) one quickly learns to treat the proposed
names and hierarchical arrangement of the levels as what they
are -- a helpful heuristic.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

5

Phil Runkel on 8 November 98 commenting on Marken's posting of 981105.1050
expressing gratitude to Wm. Powers for his conception of the
neural hierarchy:

                Hear! Hear!