Egg-retrieval

CSGnet Archive CSG1999
1

[From Bruce Abbott (990401.1045 EST)]

Bill Powers (990401.0435 MST)

I'm having some doubts about the description of Tinbergen's experiment. As
I recall it, a darning egg was substituted for the real egg, and the goose
pulled the darning egg in and sat on it. That would be much simpler to
explain, of course. But maybe both experiments were done. Mary's off trying
to find the original description in our limited library. We may have to
wait until we can get to a real library, unless someone out there can find
the original source of the story.

Here is Tinbergen's (1951) description of his observations of egg-retrieving
in the grey lag goose, based on a 1938 report:

···

-------------------------------------------------------------------------
The grey lag goose reacts to eggs that have rolled out of the nest by
stretching the neck towards it, bringing the bill behind the egg and with
careful balancing movements rolling it back into the nest (Fig. 69). The
innate releasing mechanism of this response reacts to relatively few
stimuli; objects of very different shape and size, provided they have a
rounded contour, release it. In spite of the balancing movements the bird
sometimes loses control of the egg and then the egg slips sideways. In this
case the egg-rolling movement does not always break off, but it may be
completed, very much as if it were a vacuum activity. If this happens, the
sideways balancing movements are absent. This indicates that the balancing
movements are dependent on continuous stimulation from the egg, probably of
a tactile nature, while the other component, a movement in the median plane,
is not dependent on continuous stimulation but, once released, runs its full
course. At the same time, the spatial orientation of this latter movement
is, from the very beginning, not influenced by external stimuli. It is an
internally fixed pattern. The stereotyped nature of this component was
especially obvious when the bird was offered an egg that was much too large.
It could not adapt the movement to the abnormal size and got stuck half-way
when the egg was pressed between the bill and the breast (Fig. 70). The
balancing movements could further be eliminated by presenting the goose with
a cylinder. This model, when offered on an entirely flat and smooth
surface, made no sideways movements. The result was a straight egg-rolling
movement without any sideward deviations of the head (Lorenz and Tinbergen,
1938).
  From Tinbergen, N. (1951/1969). _The study of instinct_. New York: Oxford.
--------------------------------------------------------------------------
Clearly the goose is controlling a number of variables as it retrieves the
egg -- even when the egg has rolled away the movement of the head and bill
toward the goose's body is doubtless produced by a control system that
maintains the bill a certain distance (plus or minus) above ground despite
disturbances produced by muscle contractions that bring the head and bill
toward the body and despite changes in the height of the terrain along the
way. But the sequence of acts consisting of pulling the head and bill
toward the body, standing, and tucking the bill under the body appear to be
capable of continuing absent tactile and perhaps visual feedback from the
egg, suggesting that the relevant reference signals are being varied by an
internal pattern generator in open-loop fashion. Clearly the sideways
movements of the bill are control movements that tend to compensate for
deviations of the egg's path toward the goose's body: they are absent when
the egg is absent and also when the "egg" (cylinder) takes a straight path
requiring no correction. I think it would be quite interesting to conduct a
more thorough investigation of this act, designed to identify the relevant
controlled variables and to determine to what extent the pattern can unfold
in the absence of the usual sensory feedback at various points. Anyone
happen to have a goose handy?

Bruce

2

[From Rick Marken (990401.0940)]

Bruce Abbott (990401.1045 EST)

But the sequence of acts consisting of pulling the head and
bill toward the body, standing, and tucking the bill under
the body appear to be capable of continuing absent tactile and
perhaps visual feedback from the egg

Are you saying that taking the egg away takes away tactile and
perhaps visual feedback (as suggested by your phrase "absent
tactile and perhaps visual feedback )? Come on, Bruce. You can
imitate PCT-speak better than that. Egg removal doesn't take
away feedback; it just disturbs tactile and visual variables that
are affected by the goose's actions (variables that may or may
not be under control).

I think it would be quite interesting to conduct a more thorough
investigation of this act, designed to identify the relevant
controlled variables

An excellent suggestion!

Anyone happen to have a goose handy?

Whoever does this is going to need more than a goose. They are
going to need an understanding of how to test for controlled
variables. They could have gotten this understanding from
your research methods textbook if you had bothered to rewrite
it based on what you (should have) learned after over four
years on CSGNet. But now they'll just have to learn about it
in the "underground" PCT press (Runkel's "Casting Nets...",
many articles by Powers, my "Dancer..." paper, discussions
on CSGNet, etc).

Bruce Gregory (990401) --

Try http://www.pigeon.psy.tufts.edu/psych26/fap.htm

Excellent find!

I looked at the Quicktime. This is no fixed action pattern!
What it looks like to me is control of pressure against the
inside of the bill. When the egg is removed (bringing the
perception of pressure instantly to zero) the goose moves
the bill (in what appears to be an accelerated rate) in the
direction that _would_ increase the pressure on the inside of
the bill if the egg were still there. Neck movement seems to be
the output variable of a pressure control system. This system
is organized so that movements of the neck towards the body
are used to _increase_ pressure perception and movements away
are used to _decrease_ pressure perception. This would be the
right way to have the sign of the error (difference between actual
and reference pressure) affect the output variable (direction of
next movement) to have negative feedback (control of pressure).

The apparent "exponential runaway" when the egg is removed
is exactly what you would expect from a system with an integrating
output controlling for a non-zero perception of pressure by
pulling back on an object that is no longer there. The system's
(integrating) output increases exponentially, driven by a constant,
non-zero error. Of course, more tests are needed to prove that
pressure is controlled. But it's pretty clear from the Quicktime
that this "egg-retrieval" behavior is not necessarily (indeed, not
likely to be) a "fixed action pattern".

I think Tinbergen saw a "fixed action pattern" instead of a
pressure control system here for the same reason that
behaviorists see S-R instead of control when studying "reflexes".
Like the behaviorists, Tinbergen used disturbances that were
large and abrupt. The goose's apparently stereotyped egg-retrieval
actions are no more a "fixed action pattern" than is the leg kick
that occurs when a hammer tap is applied to the patellar tendon.

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 (990401.1059 MST)]

Bruce Abbott (990401.1045 EST)--

Thanks for the 1951 citation from Tinbergen. It would be nice to see the
original 1938 report. Mary will look for it. Among other things, it would
be interesting to see how much of that theory-loaded language was present
in the original report. I thought ethologists were trained to give
scientific descriptions of behavior without slipping in theory-based
interpretations. It makes one wonder just how much influence theory had on
the original perceptions.

You refer to the bill tucking under the breast as part of the "open-loop
sequence." I didn't see any reference to that in the citation. All that is
said on this matter is "In this case the egg-rolling movement does not
always break off, but it may be completed, very much as if it were a vacuum
activity." That does not tell us how complete the movements without the egg
were. The movement is not, of course, open-loop, except with regard to
moving the egg. Considerable control is required to extend the bill, lower
it to a point just above the ground, and pull it back toward the goose.
That act produces large quantities of both kinesthetic and visual feedback
(remembering that the eyes of the goose necessarily partake of the
movements of the bill). Without at least the kinesthetic feedback, such
refined movements, I would claim, are impossible to carry out.

Otherwise I agree with your analysis, including your inference of a
pattern-generator. However, it is not an output pattern generator, I would
propose, but a pattern of reference signals for visual and kinesthetic
perceptions. This could be tested.

Best,

Bill P.

4

[From Bruce Gregory (990401.1325 EST)]

Rick Marken (990401.0940)

I think Tinbergen saw a "fixed action pattern" instead of a
pressure control system here for the same reason that
behaviorists see S-R instead of control when studying "reflexes".
Like the behaviorists, Tinbergen used disturbances that were
large and abrupt. The goose's apparently stereotyped egg-retrieval
actions are no more a "fixed action pattern" than is the leg kick
that occurs when a hammer tap is applied to the patellar tendon.

Nice argument. By the way, I want to reveal my "secret". For anyone who
has not yet tried it, www.google.com is a relatively new and very
powerful search engine. I use it all the time and am constantly amazed
at how good it is in uncovering relevant sites.

Bruce Gregory

5

[From Bruce Abbott (990401.2015 EST)]

Bill Powers (990401.1059 MST) --

The movement is not, of course, open-loop, except with regard to
moving the egg. Considerable control is required to extend the bill, lower
it to a point just above the ground, and pull it back toward the goose.
That act produces large quantities of both kinesthetic and visual feedback
(remembering that the eyes of the goose necessarily partake of the
movements of the bill). Without at least the kinesthetic feedback, such
refined movements, I would claim, are impossible to carry out.

I believe I said as much. It's only open loop with respect to tactile
feedback from contact with the egg. If one views the "goal" of this
activity as rolling the egg back into the nest, then the apparent fact that
the patterned activity may continue to completion in the absence of the egg
might be rather surprising, from a control-theoretic point of view. Loss of
contact with the egg "ought" to be a rather serious disturbance to such a
control system, and yet the system's actions continue unaffected.

Otherwise I agree with your analysis, including your inference of a
pattern-generator. However, it is not an output pattern generator, I would
propose, but a pattern of reference signals for visual and kinesthetic
perceptions. This could be tested.

If you will reread my description, you will see that the pattern-generator I
suggested manipulates reference signals, not outputs. I think we're in
agreement here as well.

Bruce

6

from [ Marc Abrams (990401.2104) ]

[From Bruce Abbott (990401.2015 EST)]

... If one views the "goal" of this
activity as rolling the egg back into the nest, then the apparent fact...

These are perceptions and reference levels _you_ ( or some researchers ) are
attributing to the goose. This is not a statement of what the _goose_ might
be controlling for. _Everything_ else that follows, comes from these
assertions.

It's certainly not a crime to feel this way, :slight_smile: and you may even be right.
How would you know?

Marc

Marc

7

[From Bill Powers (990401.1955 MST)]

Bruce Abbott (990401.2015 EST)--

If you will reread my description, you will see that the pattern-generator I
suggested manipulates reference signals, not outputs. I think we're in
agreement here as well.

Yes, although both of us are guessing.

Thanks to Mary's efforts we now have four old books by Tinbergen.

There is an interesting discussion of egg-retrieval in

Tinbergen, N. (1960);p _The Herring-Gull's World.,New York: Basic Books (p.
140 ff_.

A number of amusing instances of conflict are described. In one experiment,
while the gull is away, the three eggs are removed and placed in plain
sight about a foot outside the nest. "... the returning bird usually goes
to the empty nest, and, often after some hestitation, sits down in it."

Then later, after some other discussion: "These tests do not always give
exactly the same results. Most gulls do not choose rigidly either the
nest-site or the eggs, but they hesitate, looking from one to the other,and
even sitting down on each alternately. Also, one bird may be more attracted
by the nest-site, while another may be more inclined to choose the eggs."
This somewhat contradicts the impression given by the the "usually" above,
but it does show that even birds can be of two minds, however tiny.

Somewhat in the way that Tinbergen has an evident conflict about explaining
behavior as stimulus-driven or "spontaneous."

Best,

Bill P.

8

[From Bill Powers (990402.0849 MST)]

Bruce Abbott (990401.2015 EST)--

It's only open loop with respect to tactile
feedback from contact with the egg.

Having now seen the movie that Bruce G. steered us to, I'm not sure how
much tactile feedback there is. I think rolling the egg must be much like
our trying to do the same thing using a tongue depressor -- the beak is
hard and can't be very sensitive to touch. Pressure could be sensed, of
course, where the beak joins the head. I was trying to see whether the bird
could possibly see the egg, and I think it could, at least the ends at the
two sides. When the egg is removed, there is an immediate change in the
actions, in one case the head swinging to the bird's right side and in the
other the bird's head pulling straight back (the raking motion ceases
immediately) and then going clear under the body to the front left, which
it never did when there was still an egg there. In both cases the lateral
movement is in the direction toward which the egg disappeared. In both
cases I could easily imagine a little balloon over the bird's head saying
"WHERE'D IT GO?" I think visual feedback predominates. I agree with Mary,
that nothing about this behavior looks really open-loop, and we both agree
with Rick that removal of the egg is primarily a disturbance. After the egg
is removed, the pattern of behavior changes as if to try to re-establish
contact with it.

What we see is heavily influenced by our interpretation of what's going on.
To see whether the pattern really changed when the egg was removed, as I
think it did, it would be necessary to measure the beak position and
orientation (6 numbers) in each frame, and fit dynamic curves to the
motions. I think you are quite mistaken when you say "Loss of contact with
the egg "ought" to be a rather serious disturbance to such a control
system, and yet the system's actions continue unaffected." They do not seem
to me to "continue unaffected" at all. As Tinbergen himself pointed out,
the raking sequences (up, forward, down, back) cease as soon as the egg
disappears, as do the sideward "balancing" movements. The movement looks to
me much more the way Rick described it, as if the system is trying to
re-establish contact by a continuing motion in the direction that would
have increased pressure on underside of the bill when the egg was still
there. At the end of the second sequence, the bill did not stop where it
did when the egg was present, but instead continued clear under the body to
the front left, as if still pursuing the vanished egg.

Otherwise I agree with your analysis, including your inference of a
pattern-generator. However, it is not an output pattern generator, I would
propose, but a pattern of reference signals for visual and kinesthetic
perceptions. This could be tested.

If you will reread my description, you will see that the pattern-generator I
suggested manipulates reference signals, not outputs. I think we're in
agreement here as well.

Yes, you're right.

I found Tinbergen's discussion on p. 140-141 of "The Herring Gull's World"
very interesting (aside from his quaint behavioristic "descriptions"). Here
he emphasized how slow, uncertain, and clumsy all this behavior is. The
bird might sit on its two eggs, eyeing the egg that's out of the nest, for
as long as an hour before getting up to roll the stray egg back into the
nest. The problem can't be in the bird's (low level) ability to control its
body, for in other respects is is very skillful. What's wrong is the
rudimentary state of its higher level control systems (as I interpret
Tinbergen's words). Correcting the relationship of the egg to the nest is
very nearly beyond the bird's mental capacities. Of these abilities,
Tinbergen says

"The conclusion to be drawn from these, and hundreds and thousands of
similar facts, is that the animal could do much more with its sense-organs
and executive organs if it had only the central nervous mechanmisms to use
them in more ways. ... In other words, behavior is dependent on something
like definite mechanisms or systems in the central nervous system, and [in
these birds] the number of these mechanisms is limited" (p. 141).

Best,

Bill P.

9

from [ Marc Abrams (990402.1331) ]

[From Bill Powers (990402.0849 MST)]

What we see is heavily influenced by our interpretation of what's going on.

So far so good.

To see whether the pattern really changed when the egg was removed, as I
think it did,

Still with you :slight_smile:

it would be necessary to measure the beak position and
orientation (6 numbers) in each frame, and fit dynamic curves to the
motions.

Still there.

I think you are quite mistaken when you say "Loss of contact with
the egg "ought" to be a rather serious disturbance to such a control
system, and yet the system's actions continue unaffected."

When and how did this become a control process? Could you be seeing
inadvertent side effects of something you have no clue about?

Marc

10

[From Bill Powers (990402.1316 MST)]

Marc Abrams (990402.1331) --

I think you are quite mistaken when you say "Loss of contact with
the egg "ought" to be a rather serious disturbance to such a control
system, and yet the system's actions continue unaffected."

When and how did this become a control process? Could you be seeing
inadvertent side effects of something you have no clue about?

There are definitely some clues. First, on all 4 trials in the movie, the
beak extended outward and then downward until its underside touched the
egg, even though the head started from a different position and had to
reach out by a different amount each time. So that end result is likely to
be a controlled variable (not proven, but at least worth going on with).
Second, the path of the egg, rolling through tallish grass, was disturbed
by irregularities in the ground and by the grass, and the beak moved from
side to side in just the way needed to counteract those disturbances and
keep the egg on a path toward the nest. A model that would reproduce this
behavior would automatically behave as the bird did when the egg was
removed: speed up as if to restore the contact of the beak with the egg. If
a PCTer had done Tinbergen's experiment, other informative disturbances
could have been applied. For example, instead of just removing the egg, the
experimenter could have moved it closer to the bird or to one side or the
other, or gently pushed on the egg to make it resist the bird's push a
little, to see if the bird would increase and decrease its pulling efforts
as the disturbance was varied. All these variations would help us define
more closely what the bird was controlling. Of course Tinbergen, not
knowing enough to search for controlled variables, did't think of applying
more finely-adjustable disturbances. He did look at effects of egg size,
shape, markings, and color, but that was pretty random, and not designed to
refine a definition of a controlled variable already partly known.

Best,

Bill P.

11

from [ Marc Abrams (990402.1316) ]

[From Bill Powers (990402.1316 MST)]

There are definitely some clues. First, on all 4 trials in the movie,

...

Thanks Bill. I have no problem with anything in this post. I guess I am
bundling a few concerns together. Let me try to be a bit more specific.

For research of this type ( "behavioral research with non human animals ),
what would the purpose be? To establish the fact that control exists in
living biological systems? Are we literally trying to approximate what the
Goose, Rat, etc are controlling for? And if so, why? To what end?

Marc

12

[From Bill Powers (990402.1511 MST) --

Marc Abrams (990402.1316)--

For research of this type ( "behavioral research with non human animals ),
what would the purpose be? To establish the fact that control exists in
living biological systems?

Yes.

Are we literally trying to approximate what the
Goose, Rat, etc are controlling for?

Yes.

And if so, why? To what end?

To understand them and what they have in common with us and all living
systems.

Best,

Bill P.

13

[From Dick Robertson,990403.0704CST]

Bill Powers wrote:

[From Bill Powers (990402.0849 MST)]

What we see is heavily influenced by our interpretation of what's going on.
To see whether the pattern really changed when the egg was removed, as I
think it did,...

I found Tinbergen's discussion on p. 140-141 of "The Herring Gull's World"
very interesting (aside from his quaint behavioristic "descriptions"). ..
Bill P.

So, it would be very interesting to see how an ethologist who _did_view events
with a PCT perspective
would describe similar behavior. In the meantime, it seems to me, Franz Plooij
has gone a long way down
that path, with both fasciniating and practical results (as far as child-rearing
advice).

Best, Dick R.

14

from [ Marc Abrams (990403.1029) ]

[From Bill Powers (990402.1511 MST) --

To understand them and what they have in common with us and all living
systems.

Bill, I understand this line of reasoning and I guess, would have an easier
time accepting it ( I.e. feeling better about putting _tons_ of research
money into it ) if in fact the study of behavior in animals was somehow
"easier", "faster", or more "humane" then with human subjects. But, Because
we cannot currently communicate with animals ( not to the degree I believe
necessary to study control ), I think the research becomes more difficult
and open to _lots_ more interpretation(s) regardless of the data collected.

Marc

Marc

15

[From Rick Marken (990403.0930)]

Marc Abrams (990403.1029) --

Because we cannot currently communicate with animals ( not to
the degree I believe necessary to study control ), I think the
research becomes more difficult and open to _lots_ more
interpretation(s) regardless of the data collected.

The fact that you cannot communicate with an animal (you can't ask
the goose "Are you controlling the pressure you feel when moving the
egg into the nest?", for example) does make things more difficult.
But I think this is good; it forces you to do the Test appropriately;
no "one shot" Testing, for example. You must prove to your audience
(the scientific community) that a particular variable (or variables)
is actually under control. This can only be done by exhaustive Test
and a willingness to show that any variable that you believe to be
under control is, indeed, protected from every suggested disturbance
that should have an effect on that variable (and _not_ protected
from every suggested disturbance that should _not_ have an effect
on the variable).

Also, when you Test animals you can't validate the results of your
Test by asking your subject "Were you really controlling X?". But
this is the same as the situation in most sciences: you can't ask
water whether it is really composed of hydrogen and oxygen, for
example. Also, communication can confuse as well as enlighten. Look
at Bill's description of the Test ("The Coin Game") in B:CP where
he talks about how subject and experimenter can disagree about the
_name_ of the controlled variable (you -- the Tester -- conclude
that the subject is controllling for an "N" pattern and the subject
disagrees and says it's a "Z"). As Bill notes, the Test makes such
verbal disagreements "beside the point". But people don't always
like to be told what they are controlling so such verbal disagreements
are probably inevitable when your subject is human (even though
you can prove to the subject that you know which influences on a
variable are disturbances that s/he will resist and which are not;
that is, you know what variable is being controlled, regardless of
what the subject wants to call it). Happily, this problem doesn't
arise with animals.

Best

Rick

···

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

16

[From Bill Powers (990403.1051 MST)]

Marc Abrams (990403.1029)--

Bill, I understand this line of reasoning and I guess, would have an easier
time accepting it ( I.e. feeling better about putting _tons_ of research
money into it ) if in fact the study of behavior in animals was somehow
"easier", "faster", or more "humane" then with human subjects. But, Because
we cannot currently communicate with animals ( not to the degree I believe
necessary to study control ), I think the research becomes more difficult
and open to _lots_ more interpretation(s) regardless of the data collected.

It's not as much of an advantage to be able to talk with people as you may
think. Human beings are not necessarily able to describe their own
controlled variables accurately (or at all), and in fact if they believe in
S-R theory they are likely to describe their own behavior that way. When
someone tells me what he is controlling for, I always like to do a little
test just to make sure. The test may not tell me what he IS controlling
for, but it can sure tell me if he's not controlling for what he says he
is. [I never thought that sentence would work out]

Consider the Coin Game. This game is carried out without any verbal
communication other than "no error". Even that could be omitted, although
the experimenter might take longer to realize that there isn't going to be
any correction. It's perfectly possible to apply the test nonverbally,
which means it can be used with animals -- and even with artificial control
systems.

It just occurred to me that I _have_ used the Test with artificial control
systems. Every time a control system I built didn't seem to be controlling
in a reliable way, I would have to find out why. After eliminating the
comparator and actuator, I would end up looking at the sensor, and find
(for example) that instead of sensing light intensity, it was sensing some
combination of light intensity and temperature. So while it was controlling
perfectly well, the variable it was controlling was not the one I had in
mind (light intensity alone). Back then, of course, I would have said that
the photocell was being thrown out of calibration because it was
temperature-sensitive.

Best,

Bill P.

17

from [ Marc Abrams (990403.1416) ]

[From Bill Powers (990403.1051 MST)]
[From Rick Marken (990403.0930)]

Thank you both for two excellent posts. I am going to use Bill's Post to
further the conversation rather then possibly starting two seperate threads
Bill:

It's not as much of an advantage to be able to talk with people as you may
think. Human beings are not necessarily able to describe their own
controlled variables accurately (or at all), and in fact if they believe in
S-R theory they are likely to describe their own behavior that way.

Yes, this is an interesting phenomenon ( i.e. a side effect of control )
that is accounted for with PCT.

Bill:

When someone tells me what he is controlling for, I always like to do a

little

test just to make sure. The test may not tell me what he IS controlling
for, but it can sure tell me if he's not controlling for what he says he
is. [I never thought that sentence would work out]

Bill, I assume you are using the concept "controlling" interchangeably with
"intention" here. If so, I agree. If not, could you clarify?

Bill:

Consider the Coin Game. This game is carried out without any verbal
communication other than "no error". Even that could be omitted, although
the experimenter might take longer to realize that there isn't going to be
any correction. It's perfectly possible to apply the test nonverbally,
which means it can be used with animals -- and even with artificial control
systems.

The questions I have do not concern the "viability" of control. I am firmly
convinced that the control process ( i.e. PCT control ) exists and is
important to any number of existing disciplines.

Both you and Rick feel that communication might be more of a hinderence ( in
some situations ) then a help in studying PCT. I can see your point, and I
understand, as scientists, your quest for knowledge and understanding to
add to the general knowledge pool is paramount. I guess my practical side
says lets worry about understanding things that are of concern to us
meaning humans ) first, then worry about adding to the general knowldege
pool later. I just think it's highly unlikely that we will "learn" anything
about human "behavior" ( i.e. the PCT defined kind :slight_smile: ) by studying
"behavior" in animals and I think we would have a much higher upside by
sticking to humans, at least for a while.

Marc

18

[From Bill Powers (990403.1457 MST)]

Marc Abrams (990403.1416)--

I guess my practical side
says lets worry about understanding things that are of concern to us
meaning humans ) first, then worry about adding to the general knowldege
pool later. I just think it's highly unlikely that we will "learn" anything
about human "behavior" ( i.e. the PCT defined kind :slight_smile: ) by studying
"behavior" in animals and I think we would have a much higher upside by
sticking to humans, at least for a while.

Then by all means you should stick to studying humans, shouldn't you? I
presume you weren't laying out what you think _I_ should be studying.

Best,

Bill P.