syllogism

[Bruce_Abbott]

[From Bruce Abbott (951019.1355 EST)]

Bill Powers (951018.1530 MDT) --

There's only one problem for the defense. How are you going to decide whether
a given action is relevant to the behaving organism until you have determined
what that organism is controlling?

By applying the Test, naturally. Where in anything I have said have I
stated or implied that behavioral phenomena identified through conventional
research methods cannot, should not, or need not be investigated using PCT
methodology? This has not been my argument. Rather, my argument has been
that such observations provide strong hints about how this complex
hierarchical control system may be organized, hints that cry out for a
PCT-style investigation. Whether someone else has misinterpreted these
phenomena is no reason to reject the observations and deny their relevance
to a science of behavior.

So, allow me to turn the question around. How are you going to determine
what the organism is controlling if you don't know what to look for and when
to look for it? Where do your hypotheses come from? Consider the male
stickleback. At a particular time of the year it scoops out a shallow
depression in the bottom of the stream, constructs a nest, and then attempts
to drive away any other male sticklebacks that venture within a yard or so
of the nest. A gravid female, in contrast, will be greeted with a pecular
pattern of movement, to which she often responds with movements of her own;
if the series of interacting movements goes to completion the result is that
the female will lay her eggs in the male's nest.

Now in this series of observations I see the operation of some very
interesting control systems. Without these observations I have no idea what
perceptions a male (or female) stickleback may or may not control, and I
very much doubt that you or Rick or anyone else would ever come to a
complete understanding of the stickleback's innately-given control-system
hierarchy unless you knew first what patterns of behavior occur naturally in
this fish, under what conditions, and why. You would not know what
controlled perceptions to test for, nor would you have any idea under what
conditions such control might be active or inactive.

The phenomena uncovered in "conventional" research may or may not turn out
be "irrelevant side-effects of control." Much current behavioral research
cannot even adequately address the questions the researcher who designed the
study set out to answer. Much of it provides little or no information about
the organism's control systems and is therefore of no relevance for PCT. I
am not defending such research, nor am I arguing that PCT researchers must
attend to all these findings. I agree with you that what is required is a
systematic study of the organism's control-system structure from the ground
up, using methods approprite to that study. Instead, I am arguing that
certain observations are highly relevant and can be of great help by
providing specific examples of control systems at work, which can guide
efforts to understand what these systems are, what perceptions are under
control, under what circumstances, and the functions of these systems in the
lives of their possessors.

This is a very different argument from the one you seem to think I have been
making.

Regards,

Bruce

[Richard_Marken]

[From Rick Marken (951019.1500)]

Bruce Abbott (951019.1355 EST) --

my argument has been that such observations provide strong hints about how
this complex hierarchical control system may be organized, hints that cry
out for a PCT-style investigation.

Then you agree that conventional research methods are useless except as a
source of hints about what organisms control. I would suggest that even then
such methods are a waste since we can get better hints with less wasted
effort by just looking around at behavior (though control theory glasses, of
course).

Conventional reseach methods add nothing to the process of observation --
except, perhaps, obfuscation, since they encourage us to look at behavior
through cause-effect glasses. I've gotten more ideas for studies of what
organisms control by looking at people going about their business in
everyday life than I have ever gotten from reading the hundreds of bizarre
conventional research studies published in journals like JEP or JEAB.

This is a very different argument from the one you seem to think I have been
making.

It certainly is. Your argument now is that what has been done using
conventional research methods can be a source of hints about where to start
the real research on control systems. I thought you were arguing that
conventional research results could actually help the PCT scientist
_understand_ behavior. This impression was reinforced by your insistance that
PCT account for the results of conventional research. It was futher
reinforced by your reluctance to agree that the behavioral sciences should
abandon conventional research methods since they clearly cannot reveal
anything about the variables organisns control.

Now in this series of observations I see the operation of some very
interesting control systems.

Not quite. In these observations you see _hints_ about variables that
sticklebacks _might_ be controlling; hypothetical controlled variables. These
observations should be the _start_ of research aimed at testing these
hypotheses (which are almost certainly wrong in detail and will have to be
refined based on the results of your tests). This research cannot be done
with the conventional methodology; it can only be done using _the test for
controlled variables_.

"Hints" about what organisms are controlling can come from the results of
conventional research (though I maintain that this is a very arcane source of
such hints) or from observation of everyday behavior. But regardless of where
they come from, once you "take the hint" and start testing to see if a
variable is actually under control, you are doing PCT research; conventional
research methods are no longer necessary. In fact, behavioral scientists
would be better of just not using conventional methods any more since these
methods tell us nothing about control. Conventional data can be a source of
hints about control, but that's all; there is no way to use this data as the
basis for determining controlled variables.

So, if I understand your position correctly, you are advocating the use of
conventional research data as a source of hypotheses about controlled
variables, hypotheses that should be tested using PCT methodology. If
this is your position, then I agree with it; I would just add that
observation of everyday behavior is probably a more fruitful place to look
for hints about controlled variables. And I would also suggest that once
we've started looking for controlled variables it is no longer necessary to
do conventional research; the search for controlled variables will simply
become an ongoing process; research that identifies variable X as controlled
will, itself, provide hints that, perhaps, variables Y and Z are under
control too; and then we test for control of Y and Z; and so on.

Thanks to PCT, conventional research methods in psychology can go the way of
alchemical research methods in chemistry, and the study of living systems
will be better off when they do.

Rick

[Bruce_Abbott]

[From Bruce Abbott (951019.1855 EST)]

Rick Marken (951019.1500) --

Bruce Abbott (951019.1355 EST)

my argument has been that such observations provide strong hints about how
this complex hierarchical control system may be organized, hints that cry
out for a PCT-style investigation.

Then you agree that conventional research methods are useless except as a
source of hints about what organisms control. I would suggest that even then
such methods are a waste since we can get better hints with less wasted
effort by just looking around at behavior (though control theory glasses, of
course).

No, I don't believe I said that. Conventional research methods are not
categorically useless, although they may be the wrong tools for addressing
some questions, particularly the kinds of most interest within PCT. Useless
for what purpose? And which methods? They range from between-groups
studies involving random assignment of participants to groups to
single-subject designs to naturalistic observation to correlational studies
to survey methodology. Whether a particular method (conventional or
otherwise) is "useless" or not all depends on what question you are trying
to answer. I prefer to keep my toolbox well stocked. That way, I'll always
have the right tool for the right job.

Conventional reseach methods add nothing to the process of observation --
except, perhaps, obfuscation, since they encourage us to look at behavior
through cause-effect glasses. I've gotten more ideas for studies of what
organisms control by looking at people going about their business in
everyday life than I have ever gotten from reading the hundreds of bizarre
conventional research studies published in journals like JEP or JEAB.

Again, you will have to do a better job of identifying which specific
conventional methods you have in mind, and for what purposes they are
inappropriate. Most of the examples of observations from conventional
research I have given in this debate have come from ethology. Ethological
explanations have mostly been of the S-R type (perhaps even more so than
those from psychology), but I am concerned here with the observations, not
with their interpretation or the sometimes theory-laden labels given them.
Are you suggesting that such careful studies of phenomena such as imprinting
or courting in birds, or termite mound-building, and so on are no better
sources of research hypotheses than your own casual observations?

Now in this series of observations I see the operation of some very
interesting control systems.

Not quite. In these observations you see _hints_ about variables that
sticklebacks _might_ be controlling; hypothetical controlled variables. These
observations should be the _start_ of research aimed at testing these
hypotheses (which are almost certainly wrong in detail and will have to be
refined based on the results of your tests). This research cannot be done
with the conventional methodology; it can only be done using _the test for
controlled variables_.

When I said that I see control systems at work, I meant that I could see
evidence of perceptions under control, not that I could positively identify
all the systems at work and their controlled perceptions. And again when
you state that the research cannot be done with conventional methodology, I
want to know which methodologies you include. I have seen plenty of what
you would call conventional research that has identified controlled (usually
misidentified as controlling) variables. Tinbergen's research on the male
stickleback is a case in point: he was able to demonstrate conclusively that
the intruding male is identified by the defending male by the intruder's red
belly-patch and black eyespot, and not by other features such as its
specific shape, smell, or patterns of movement. But then, you might be
willing to place Tinbergen's study under the heading of PCT methodology, I
don't know. Tinbergen's study was aimed at identifying the "releasing
stimulus" for territorial defense against the intruding male--clearly not a
PCT framework for the research.

So, if I understand your position correctly, you are advocating the use of
conventional research data as a source of hypotheses about controlled
variables, hypotheses that should be tested using PCT methodology. If
this is your position, then I agree with it; I would just add that
observation of everyday behavior is probably a more fruitful place to look
for hints about controlled variables. And I would also suggest that once
we've started looking for controlled variables it is no longer necessary to
do conventional research; the search for controlled variables will simply
become an ongoing process; research that identifies variable X as controlled
will, itself, provide hints that, perhaps, variables Y and Z are under
control too; and then we test for control of Y and Z; and so on.

I don't want to repeat what I just said in another post today, so I'll just
remind you of my argument that a simple search for controlled variables will
never disclose the whole system unless that search is guided by knowledge of
the ordinary behaviors of the critters under study and the conditions under
which these behaviors occur. Hypotheses about the functional significance
of the system under study can also help to guide such research, as these
suggest what particular variables may need to be controlled in order to
successfully carry out that function. There already exists a large body of
knowledge and hypothesis concerning many interesting functional systems of
behavior which can provide such insights. I would prefer to reserve the
right to judge the relevance of a particular finding for myself rather than
have its relevance rejected out of court by pronouncements about the general
irrelevance of data provided by conventional research.

Thanks to PCT, conventional research methods in psychology can go the way of
alchemical research methods in chemistry, and the study of living systems
will be better off when they do.

Researchers have often used the wrong methods or have pursued the wrong
questions with them, but this does not mean that the methods themselves are
useless for all purposes and all questions. Thus, there will probably
always be a use for what you are calling conventional research methods in
psychology, although there will come a time when their inappropriateness for
answering certain questions will come to be generally recognized. The
alchemists suffered a different problem. The problem with alchemy was not
so much in its methods as in its incorrect theoretical framework. When you
start with the assumption that earth, air, fire, and water are the four
basic elements, not too many of your experimental results make sense.

My guess is that most of the difficulties you attribute to use of the wrong
methods have their origin in the fact that the questions being asked have
their origins in the wrong theory of behavior.

Regards,

Bruce

[William_T_Powers4]

[From Bill Powers (951019.1630 MDT)]

Off to Boulder tomorrow (Friday) morning; back late Monday.

Bruce Abbott (951019.1355 EST) --

There's only one problem for the defense. How are you going to decide
whether a given action is relevant to the behaving organism until you
have determined what that organism is controlling?

     By applying the Test, naturally. Where in anything I have said
     have I stated or implied that behavioral phenomena identified
     through conventional research methods cannot, should not, or need
     not be investigated using PCT methodology? This has not been my
     argument. Rather, my argument has been that such observations
     provide strong hints about how this complex hierarchical control
     system may be organized, hints that cry out for a PCT-style
     investigation. Whether someone else has misinterpreted these
     phenomena is no reason to reject the observations and deny their
     relevance to a science of behavior.

I've specifically said that I'll take the observations; it's the
interpretations I would rather ignore. I agree with you that any
observations at all serve to give us ideas about what variables might be
under control, so we can do the Test to pick out controlled variables.

     So, allow me to turn the question around. How are you going to
     determine what the organism is controlling if you don't know what
     to look for and when to look for it?

If I were embarking on a career at this late date, what I would do would
be first to replicate the experiments that others seem to have found
interesting, and see for myself what is going on. Knowing only what has
filtered through someone else's point of view and theoretical overlays
-- and overly-focused attention -- is very unsatisfactory. The only way
to _understand_ something is to start from scratch and observe it for
yourself.

     Where do your hypotheses come from?

From guesses about how things might work. And before you ask, it's

guesses all the way down.

     Consider the male stickleback. At a particular time of the year it
     scoops out a shallow depression in the bottom of the stream,
     constructs a nest, and then attempts to drive away any other male
     sticklebacks that venture within a yard or so of the nest. A
     gravid female, in contrast, will be greeted with a pecular pattern
     of movement, to which she often responds with movements of her own;
     if the series of interacting movements goes to completion the
     result is that the female will lay her eggs in the male's nest.
     Now in this series of observations I see the operation of some very
     interesting control systems.

That's a very sketchy description and not particularly useful for PCT
purposes. Saying that the female responds to the male's movements with
movements of her own not only asserts an S-R connection as if it were
data, but it is so qualitative that nobody could determine what was
being controlled on the basis of the description. Before saying what
controlled variables were involved, I would want a record of the 3-D
positions and orientations of the fish with a time resolution
commensurate with the time-scale on which the fish can correct errors.
Finding that time-scale would involve much preliminary experimentation,
and design of suitable equipment.

What I find lacking in reports of experiments like those with the
stickleback is that they don't really take the problem seriously. It's
as though the experimenter thinks he can just leap into the middle of a
study of a very complex system and merely on the strength of what his
eyeballs can tell him, arrive at a fundamental understanding of the
phenomenon. To study the control systems involved we would have to do
the experiments all over again, with the necessary instrumentation.

     The phenomena uncovered in "conventional" research may or may not
     turn out be "irrelevant side-effects of control." Much current
     behavioral research cannot even adequately address the questions
     the researcher who designed the study set out to answer. Much of
     it provides little or no information about the organism's control
     systems and is therefore of no relevance for PCT.

It's pretty certain that if a researcher doesn't report on the variables
that an organism was controlling through its actions, the data will
concern only the actions. This means that even if we found the
experiment to be interesting and relevant, we would still have to do it
over ourselves in order to find out what was being controlled. If we're
going to have to repeat all the experiments that seem possibly useful,
why not just create our own, and design them to get the necessary
information from the start?

     I am not defending such research, nor am I arguing that PCT
     researchers must attend to all these findings. Instead, I am
     arguing that certain observations are highly relevant and can be of
     great help by providing specific examples of control systems at
     work, which can guide efforts to understand what these systems are,
     what perceptions are under control, under what circumstances, and
     the functions of these systems in the lives of their possessors.

OK, fine, we agree that observations are required. But I have had very
little success in taking other people's observations and making sense of
them; if the observer wasn't observing with PCT in mind, it's very
unlikely that the data will contain answers to my questions. Not
impossible, but unlikely. When I did that analysis of Verhave's rat
experiments, I had to contact Verhave and get his raw data; there wasn't
enough information in the published paper to do the analysis I had in
mind. And even then, I had to give up on several ideas because Verhave
didn't take the data that would be needed.

     Avery's comment has, I think, been misinterpreted. Naturalists who
     have studied termite mounds have done an excellent job of showing
     what functions the mound serves in the life of the species. Once
     you know what purposes the mound serves, you are in an excellent
     position to identify which among the myriad possible controlled
     variables are actually involved in the termite's control processes.

Statements of function are not statements of purposes. As I tried to
point out, we can read functions into any situation which are irrelevant
to the actual control processes. A termite mound would protects termites
from small meteors; is that the function of the mound? Behavior has many
products, many consequences, but the only way to see which of those were
intended is to disturb them and see if the disturbance is resisted. Such
tests have been done occasionally, but there is no systematic method
based on that approach.

     Until we determine that the child crys when its mother leaves the
     room, why would we even ask whether crying is being used as a means
     to control its distance from its mother?

I understand what you mean and agree with your point. We have to observe
before we can look for controlled variables. But I want more than a
count of how many babies cry when mothers leave the room. I want to see
what actually happens, both when the baby cries and when it doesn't. I
want a look at the baby and the mother, to see what else is going on
between them. I want a chance to notice everything that the person who
did the experiment ignored or at least failed to record. I want to make
up my own mind whether we're talking about a real phenomenon.

Obviously, I'm never going to repeat all these experiments about which I
have questions. The only way I have of reaching any conclusion about the
reports is to ask whether the person who made the observations knew
anything about PCT, which is equivalent to asking whether that person
would have noticed something relevant to a PCT analysis. If I know that
the person understood PCT and would have notice the right things, I can
be reasonably comfortable with the fact that I wasn't there watching.
But if it's obvious that the experimenter knew nothing about PCT, I
can't really trust the results that are reported. Is that so
unreasonable? Would an archeologist trust a gardener to dig up a site
and report on any interesting discoveries?

···

----------------------------------------------------------------------
Best,

Bill P.