Geese: Mush from outside, Science on inside. -1

[From Dag Forssell (951022 2100)] Part 1 of 2

[Shannon Williams (951021)]

   What is the primary difference between the concept of alchemy

     and chemistry? (Do chemists deal with different substances
     than alchemists? Or do they mainly understand these substances
     differently?)

Today, in my capacity as deputy archivist and PCT advocate, I shall
come out of lurking briefly to welcome and support Shannon.
Welcome! Considering your comment above, I must ask you if you have
read the demodisk texts (available from the WWW server and from
me), particularly my little essay on alchemy and discussion of
alchemy in my book _Management and Leadership: Insight for
Effective Practice_. I will be very pleased to feed your PCT-study
habit.

   I am arguing that if you believe in an earth-centered

     universe, you will not visualize Kepler's laws. And even when
     the laws predict correctly, you will not be able to apply them
     to your comprehension.

   If you believe that you can intuitively understand another

     being's behavior just by observing it, you will not be able to
     visualize PC. Your objective data could be useful to a PC
     experimenter, but he could not build upon your theory or
     analysis.

Shannon, this comment is pregnant with insight that is right on.
While Bruce can articulate PCT it appears to me most unlikely that
he visualizes it yet. While Bruce has constrained his claims to
see value in conventional research by limiting what kind of
research, under what conditions it was done and how detailed the
observations were, he continues to debate eloquently.

I am reminded of discussion long ago about such value. Value cannot
be denied, and we should concede this to Bruce, but the next
question is how much. The analogy was mentioned on this net of a
barrel full of peanut shells. There is undoubtedly some whole
peanuts in there, but is it worthwhile to sift through all the
shells to find them?

···

--------------------------------
[Bruce Abbott (951021.1600 EST)] in response to Shannon

[Emphasis in the form of CAPS added by Dag]

   All of which may be true but is irrelevant to the position I

     have been advancing. Whether or not ethologists, for example,
     have viewed behavior in control-system terms, they have done
     a magnificent job, in my opinion, of providing germane
     observations AND YES, EVEN CONCLUSIONS (E.G., ABOUT THE
     FUNCTIONAL SIGNIFICANCE OF MANY BEHAVIORAL PATTERNS), that a
     control-systems scientist would be better off knowing than
     ignoring when beginning a study of, for example, termite
     control systems.

...

   Secondly, although the PC experimenter would not wish to

     "build upon the theory" of the ethologist (assuming the
     ethologist is, for example, relying on an S-R type of
     interpretation), there is no necessary reason why a PC
     experimenter could not build upon HIS ANALYSIS, IF HIS
     ANALYSIS WERE BASED ON THE RIGHT KIND OF INFORMATION. How
     about the observation that the egg-retrieval motion of the
     greylag goose continues, once begun, even if the egg is
     removed during the process? ETHOLOGISTS HAVE CONCLUDED from
     such observations that this motion, once begun, operates
     independently of the tactile feedback from the egg. They have
     also noted that the side-to-side motion of the bill during
     this behavior, which normally occurs during the retrieval
     motion and tends to keep the egg from rolling off the bill,
     does depend on such feedback, because it is absent once the
     egg has been removed. THESE SEEM TO BE SOUND INTERPRETATIONS
     TO ME, and I'M SURE I COULD WRITE A VERY FINE COMPUTER MODEL
     OF THIS EGG-RETRIEVAL SYSTEM THAT WOULD PERFORM EXACTLY AS THE
     GOOSE DOES. Are you sure you don't want to reconsider your
     position? (;->

Bruce, you see value in egg-rolling behavior research because you
have yet to see things from the PCT Neckar cube perspective as a
matter of course. Perhaps Bill mentioned egg-rolling because he
remembered discussing it before. I shall copy a thread from June of
1991 below in its entirety. Perhaps if you read this carefully you
will see the futility of your argument -- of seeing value in
egg-rolling research -- if you read it carefully.

Bruce, You have abstained from commenting on my posts on _Science
or Mush_ and _I've got PCT in my bones_ last July even though you
promised to, but have made reference to them. I submit your
argument in the last paragraph above is an excellent example of
mush, looking for the magic of behavior instead of trying to model
the physics of it. A computer model, as I have noted before, which
invokes magic in the form of statistical probabilities, is not a
physical model. I am not disputing your ability to write a
computer model, just saying that it would still be mush.

Your argument that not only the observations themselves [which Bill
P. has agreed in principle can have value] but that the analysis
and conclusions drawn from traditional research also have value
would be funny if it was not so tragic. There is a critical
difference between observing behavior from the outside and creating
a physical model of what goes on on the inside.

Here is the thread. Any lurker can draw his/her own conclusions.

From CSG archives:

-------------------------

Subject: Egg-Rolling

[from Gary Cziko] Thu, 13 Jun 91 14:21:14 -0500

Along the lines of the stumped rat, here is Lorenz's description of
the egg-rolling behavior of the greylag goose. What is so
interesting about this is that there does seem to be what appears
to be a combination of control of motor output (drawing the egg
back toward the nest) and control of perception (keeping the egg
balanced on the underside of the bill).

"The fixed motor pattern used in the action consists of a
stretching forward of the neck, bending the head downward so as to
touch the egg with the underside of the bill, and then rolling it
toward the nest by means of a slow bending of the neck.
Concomitant compensatory movements of head and bill to each side
keep the egg in balance and prevent its deviating from the intended
path. The fixed motor pattern can be isolated by deftly snatching
away the egg after the movement has been released. The movement
then continues to run smoothly all the way through to the nest cup,
staying strictly within the median, that is, along the bird's plan
of symmetry. Once the movement has been released, it can only run
its way through to the end and can be changed neither in its
coordination , nor in its strength. If one offers the goose an
object much too large, such as a huge cardboard easter egg, the
movement literally "jams"; the goose proves unable to move the
object in any other than the prescribed way--for instance by
walking backward. If the object is not heavy enough, it is lifted
off the ground; if it is too heavy by even only a slight amount,
the movement fails to budge it. This is remarkable because a
goose's neck is capable of producing a prodigious amount of power,
for instance, enough to pull a table cloth loaded with a complete
tea set off a table, or, in a more teleonomic way, to tear heavily
rooted plants out of the bottom of a pond. However, the power at
the disposal of the fixed motor pattern is strictly measured to
serve its single function.

"As can easily be demonstrated, the movements to each side, which
during the whole procedure keep the egg balanced on the underside
of the bill, are elicited by tactile stimuli emanating from the
object. Whenever the egg deviates to on side, the bill immediately
follows it and guides it back into the right direction. It is
possible to make the egg "run on rails' during the rolling process
by arranging a bundle of reeds obliquely across its path. Then the
movement tries to overtake the egg in order to correct the "wrong"
direction and sometimes succeeds at the moment when the pressure of
the bill acts at a right angle to the obstacle. If the goose
"rolls"a square object that facilitates the establishment of a
stable contact with both branches of the mandible and thus not
diverging from the straight line either to the right or to the
left, the balancing movements cease altogether and the fixed
pattern alone predominates, just as it does when the object being
rolled is removed altogether." [Lorenz, Konrad Z. (1981). _The
foundations of ethology_. New York: Simon & Schuster. (pp.
236-237)]

So if Lorenz's description is accurate, disturbances are
compensated for side to side but not far to near. Can the far to
near movement be explained as anything but the unreeling of a fixed
behavior pattern which is not sensitive to feedback control?

I would really appreciate some help with this one since in my book
I'd like to critique Lorenz's idea of innate behaviors, but
examples like this are making that tough to do, particularly since
Lorenz does reveal instances of when perceptual control appear to
be involved, as in the side-to-side movements described here.

--Gary

Subject: Re: egg-rolling

From Tom Bourbon -- Thu, 13 Jun 91 17:20:55 CDT

Gary Cziko [910613n], in the case of the egg-rolling critters,
instead of control of output, what about control of perceptions
around a fixed reference to "feel it coming straight back, but not
any harder than this." That would account for the corrections of
lateral deviations and would explain the relatively feeble efforts,
compared to the forces we know the animal can generate.

Tom Bourbon

Subject: Open Loop Goose

[From Rick Marken] Thu, 13 Jun 91 18:06:23 -0700

Gary Cziko (910613)

I like the Lorenz description of the egg-rolling goose but I don't
think I understand exactly what is going on. I get an image of a
goose backing toward a nest balancing an egg under its bill. The
duck compensates for the odd movements of the egg -- keeping it
rolling in a straight path. So the duck seems to be controlling for
the perception of the force exerted by the egg against its bill. If
the egg is removed after it starts rolling it to the nest it just
continues until it gets to the nest? Is that right? And if the egg
is too big it has trouble moving it at all and it just stays in one
place? What does Lorenz (that quaint little Nazi asshole) think is
the "fixed motor pattern"? I doubt that the muscles that move the
duck to the nest are flexing in the same way on each egg rolling
occasion. Is it the path taken (straight to the nest)? if so, there
were no disturbances applied to test this.

I do think this is a very interesting example and. I am sure, if
you could get a hold of one of these geese you could quickly show
which perceptual variables it was controlling. It is interesting
because a behavior pattern does seem to be running off open loop.
But I'm sure you would not be getting whatever consistency is
seeing if, indeed, it was running open loop. I've never seen a
goose or duck that bumped around the environment like one of those
toy robots, which does move open loop.

I tentatively go with Tom Bourbon's description of the controlled
perceptual variables in egg rolling ( I especially like the "not
too hard" reference to explain the goose's reluctance to push too
hard -- a good reference to have when you are pushing eggs around
with your nose). But I still would like to get a better concept of
what the bird is doing and what Lorenz thinks the bird is doing.

Thanks Rick

Subject: geese

[From Bill Powers (910613.1930)]

------------------------------------------

Gary Cziko (910613) --

   Lorenz's description of the egg-rolling behavior of the

     greylag goose.

Remember that ALL behavior is a process of controlling perceptions.
Even bending the goose's neck. The question shouldn't be "are some
behaviors designed for controlling perceptions and are some simply
innate?" but "Are some control systems learned and are some
innate?" The goose may have to learn how to keep the egg rolling in
a straight line, because the shape of the egg and the terrain can't
be predicted. But given that the lateral control system is designed
to work in a variable environment, there's evidently no need to
control the egg for distance: if the right force is applied to the
egg in the right direction, that is normally sufficient (often
enough) to bring the egg closer. So the inherited reference signal
"bring the egg closer" can be achieved by an inherited control
system using a pattern of variation of reference signals for
lower-order neck-bending systems. From Lorenz' description, I would
guess that the main control system senses the force applied by the
beak, moving the neck until sensed force equals reference force,
with velocity feedback limiting the speed. When the object is
heavy, the required force is reached before the object moves. When
it is light, the object has to be accelerated to make the force
reach the required level -- this lifts it into the air and probably
throws it. It's clearly not a beak-position control system in the
radial direction. There could be some evolutionary advantages in
this seemingly stupid behavior. It would keep objects both heavier
and lighter than normal eggs from being pulled into the nest.

Note that if you remove the egg, the motion will persist (with
negative velocity feedback) until a limit is reached, because the
sensed force never does reach the reference level.

The force vector is evidently always aimed at the nest, so the
directional aspect of the control system is clearly working (else
what does "toward the nest" mean?).

If you don't accept ANY description of behavior as a description of
emitted output, you can usually find a reasonable control system
that will entail producing the observed actions. Of course that's
just the start. You have to apply the Test to see if I have guessed
right, or more likely, whether a more complex set of controlled
variables is involved. Now that I've provided a testable hypothesis
about the goose, you, of course, are now required to do the
experiment. Unless you think that answering this question is not
worth the effort.

-------------------------

Sorry to go on so long, but you guys pulled a whole lot of
triggers. It's not MY fault.

Subject: Re: Egg-Rolling

Jay Mittenthal Fri, 14 Jun 91 16:38:05 -0500

sorry, Gary, no good ideas. what especially surprised me about the
goose example is that its neck doesn't deliver enough power to roll
egg-shaped objects slightly heavier than eggs. Perhaps naively,
that isn't what I would have expected of a control system.

best, Jay

Subject: Cooked Goose

[From Rick Marken (910614.1500)]

Gary (910614)

   But I still have a problem figuring out what THE TEST would

     look like to see what was being controlled by the goose's far
     to near movement of the head/neck.

What are the "near to far movements? What variable does Lorenz
think is not controlled? What variables (results of the gooses
muscle contractions) do you think MIGHT be controlled? I'm asking
because I was telling the truth -- I really don't understand parts
of the Lorentz description. It would really help if you would give
me your own description of what is going on and what the "open
loop" motor program might be. Once you have a reasonable idea about
what variable MIGHT or MIGHT NOT be controlled then you can think
of a test more easily, I think.

Thanks Richard S. Marken

Subject: 1 Re: cooked goose; 2 "Big deal"

From Tom Bourbon --Sat, 15 Jun 91 01:44:46 CDT

Gary Cziko [910615]:

you asked again whether the front-back movements of the goose's
neck might not be open loop. Like Rick, I don't quite get the point
of what Lorenz said. It has been many years since I read that work.
Could you refresh us on it?

"Big deal!" That was the remark from one of my faculty colleagues
when one of my thesis students presented his research proposal to
our faculty today -- at least he presented it to the portion of the
faculty that will even attend a proposal on PCT. The remark came
after the faculty member, a cognitive scientist, had asked the
student, Wade Harman, if he was telling us that the model "merely
predicted a person would do what she did the first time?"

Wade's task, which he programmed himself, is a "4-D" tracking task.
The target is an arrow head, like the emblems on Star Trek
uniforms. It moves around the screen, driven by random waveforms
that move it in X, Y, Z (size, or apparent distance), and A (or
angle relative to the vertical). The participant uses a mouse to
control the cursor in X and Y, and a joystick to control A and Z.
People learn the task relatively quickly and perform it reliably,
which prompted some of my colleagues to inform Wade that it must be
easy, therefore predicting that people would continue to perform
well was trivial. (No, they did not get the point!)

Wade models the participant as four independent control loops,
deriving a reference signal (mean cursor-target distance) for each
dimension of movement and deriving an integration factor for each
loop by the tried-and-true procedure. Correlations between the four
models and the person's movement of the cursor in each dimension
run from .99+ down to some low .89s, for a few people who have some
"problems" doing the task.

During the proposal, Wade was demonstrating the next step in the
run, which was purely for demonstration purposes: The four models
were moving the cursor to track the target, which was driven by a
new set of random numbers for the second run -- he was showing the
performance the models predicted for the up-coming run. I was
watching in amazement, even though I have seen it run many times,
when my cognitive colleague remarked, "This is all just circular
reasoning -- you are predicting this with parameters you derived
from what she did a while ago." Of course, he was right, but why do
I still believe he missed the point?

Wade gets to go ahead with his project, but only because the
faculty knows better than to say no for no good reason.

Rick, several years ago, you had a "coming out," in which you told
your faculty colleagues you would not go on playing the role of
psychologist, teaching people things you did not believe. I was
sorely tempted to do the same, but chose instead to stay at it.
Now, with a growing body of literature on PCT, on feedback
processes and on experimental design and statistics from the PCT
perspective, it is easier to teach the material. But that is a
mixed blessing: the kids in my classes (most of them ARE kids,
compared to me!) make up their own minds about what they hear, read
and do in my classes, compared to what they get elsewhere, and
increasing numbers of them are interested in, or downright excited
about, PCT and its implications.

But these people are getting B.A.s, or at most M.A.s, and there is
no place for them in the world. Several of my colleagues are
somewhat tolerant of me and of students who turn on to PCT, but
some of the others are not so open or supportive. The person who
asked seniors in a statistics course to present a talk on some
controversial topic concerning uses of statistics in psychology was
not prepared to have one student give a reasoned discussion of the
"coefficient of failure," discussed on CSG-L a while back. Nor was
he ready for another student who, by all accounts, gave an elegant
review of Phil Runkel's critique of abuses of the method of
relative frequencies.

Students are told by some people that they don't care what kind of
evidence he (me) might present, PCT isn't right and it isn't
psychology (I believe that!). During a discussion with several
students who invoked PCT as part of a challenge to his pet
theories, another faculty member blurted out, "What does he (me) do
to you people, brainwash you?"

Hugh Petrie is talking of attending CSG in Durango. If he does, he
wants people who teach about PCT to share ideas and experiences.
Bill Williams (an economist who is not on CSG-L) wants to talk
about sample curricula for teaching people about PCT. This is all
well and good -- I share their interests, but for several of us,
the issue of what happens to the students we expose to PCT looms
larger with every new development in the model and in its
applications.

Something for several of us to discuss at Durango. This has turned
into free association, and on the net, no less. I guess I was
wishing some of you could have been there this afternoon, when Wade
deserved a better audience.

Best wishes, Tom Bourbon

Continued in part 2