A PCTer among the astrologers

[From Rick Marken (971118.0910)]

All psychological theories are, like PCT, theories of individual
behavior. These theories purport to describe processes in the
individual that give rise to visible behavior. Yet nearly all of
these theories are tested against group data. If the group averages
go (significantly) in the "right" direction (based on qualitative
"predictions" derived from the theory) then the theory is confirmed.
For example, if the average rating of liking for a task is higher
for a group that was paid little for doing a task than for a
group that was paid a lot, then a prediction of cognitive dissonance
theory is confirmed. Never mind that many individuals rated the high
pay task higher than the low pay task -- disconfirming the theory
at the individual level. It's the average results that matter.

I bring this up because I have been reading about research on
addiction and virtually all of this research tests theories of
individual behavior using group data. Thus, practitioners, who
deal with _individual_ addiction problems, have only group-data
based conclusions to guide their work. This stikes me as being
a rather unfortunate state if affairs.

An addiciton "theory" (such as it is) might say that a person's
"sense of self" determines whether or not the person stops drinking.
So a group of people are given a "sense of self" test to see whether
there is a relationship between scores on this test and some measure
of drinking cessation. If you find a relationship between test
scores and drinking behavior (a correlation of .5 or more will
do if you have a large enough sample) then you conclude that
the theory is correct: people have to have a high "sense of self"
in order to stop drinking. Never mind that many (possibly even
a majority) of the non-drinkers in the sample scored low and that
many of the drinkers scored high. The "theory" (high sense of self
needed to stop drinking) is now part of the "lore" of the field --
rather like "Pisces are sensative; Aries are tough".

I have read reviews of a PCT-based paper where the results of this
kind of research were cited as evidence _against_ control theory.
Control theory doesn't account for the "fact" that an individual's
"sense of self" (or "self-concept" or "sense of feedforward control"
or whatever) has an effect on their drinking behavior. Never mind
that this is not a real fact about individuals and that the "theory"
is not a real theory.

So we have the remarkable spectacle of a theory of individual behavior
(PCT) being rejected because people cannot derive verbalisms from it
that seem consistent with group findings that seem consistent with the
verbalisms derived from other theories of individual behavior.

Does anyone have any suggestions for how to deal with this problem?
How do I tell addiction researchers that their group data is
useless for testing theories of individual behavior and that their
theories of individuals are not real theories anyway? Or should I
just relax and kick back a few shots of Jack Daniels;-)

Best

Rick

ยทยทยท

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

1 Like

[From Bill Powers (971119.1028 MST)]

Rick Marken (971118.0910)--

All psychological theories are, like PCT, theories of individual
behavior. These theories purport to describe processes in the
individual that give rise to visible behavior. Yet nearly all of
these theories are tested against group data. If the group averages
go (significantly) in the "right" direction (based on qualitative
"predictions" derived from the theory) then the theory is confirmed.
For example, if the average rating of liking for a task is higher
for a group that was paid little for doing a task than for a
group that was paid a lot, then a prediction of cognitive dissonance
theory is confirmed. Never mind that many individuals rated the high
pay task higher than the low pay task -- disconfirming the theory
at the individual level. It's the average results that matter.

This post lays out one of the main reasons that PCT is not understood or
accepted by mainstream psychologists. It's not that PCT is one of the
experimentally-tested theories of how human beings work: it's the ONLY one.
All the other theories are based on group data, and amount to little more
than prejudices. A prejudice, to be specific, is a judgement we form about
individuals based on their sharing of superficial similarities with members
of some population: big noses, dark skins, menial occupations, social
position, clothing or uniforms, beliefs, age, sex, education, wealth, and
so on -- the list is endless, and is endlessly expanded by the "findings"
of psychology. The application of group data to individuals is clearly what
we mean by prejudice.

In addition, the use of statistical analysis without models leads to what
we know as superstition. Superstition is the belief that because one event
follows another in time, the first event caused the second event. One comes
to believe that if one causes the first event, the second will follow, or
if one prevents the first event, the second will be prevented (and all the
other combinations, depending on what is observed). When the relation is
clear, the main effect of superstition is to mislead us about mechanisms.
It seems that running out of gasoline keeps the power steering from
working, so the power steering must run on gasoline. When it is unclear, we
have to examine large numbers of events to see the "effect:" walking under
a ladder leads to bad luck. Even the attempt to deduce mechanisms is
frustrated, because real mechanisms do not work part of the time and fail
to work the rest of the time.

The method that can most reliably produce an unending output of
superstitions is known as Analysis of Variance.

Best,

Bill P.

[From Rick Marken (971119.1015)]

Me:

All psychological theories are, like PCT, theories of individual
behavior... Yet nearly all of these theories are tested against
group data.

Bill Powers (971119.1028 MST) --

This post lays out one of the main reasons that PCT is not
understood or accepted by mainstream psychologists. It's not that
PCT is one of the experimentally-tested theories of how human
beings work: it's the ONLY one. All the other theories are based
on group data, and amount to little more than prejudices.

So is the first round of Jack Daniels on you, Wild Bill? :wink:

Seriously, how do you tell addiction researchers what you said in
your post without sounding like, well.., me? Or, worse, Texas Tom
Bourbon. Or, still worse, Philly (The Kid) Runkel

Love

Deadeye Dick Marken

ยทยทยท

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

From Phil Runkel on 21 Nov 97.

individuals is clearly what we mean by prejudice." That's right, right,
right, right. Most (not all) of present-day psychological research is
built on prejudice, including, "Oh, well, he's the exception."

ยทยทยท

Bill Powers wrote on 10 Nov: "The application of group data to

[From Bruce Gregory (971126.1550 EST)]

Rick Marken (971126.1215)

This process could be very tedious and far more difficult than
I imagine (transcripts of verbal interactions are notoriously
incoherent). Does anyone have any other ideas? Suggestions?

This is probably a shocking suggestion, but why not try to teach
"subjects" to be their own observers? That is how most of us
convinced ourselves of the power of PCT. Your approach sounds
too much like the old-fashioned I experimenter, you subject
approach. Once an individual identified what seemed to be a
variable they were controlling, you could work with them to
develop a protocol to determine more precisely what was
going on.

Bruce

[From Rick Marken (971126.1215)]

Note: I am returning to the original title of this thread because
Bruce (the astrologer among the PCTers) Abbott no longer seems to
be contibuting to the thread (I presume that he was not convinced
of the folly -- if not the immorality -- of his ways and is back to
work on his statistics text, describing the "proper" way to use
group methods to study individual behavior).

In my original post in this thread I had asked:

How do I tell addiction researchers that their group data is
useless for testing theories of individual behavior and that their
theories of individuals are not real theories anyway?

I think it's clear now that it is impossible to convince anyone (who
isn't willing to believe it) that group methods are inappropriate
for the study of individual behavior. If Bill's ABS demo, my little
spreadsheet and, most especially, Richard Kennaway's mathematical
proofs won't do it, then clearly nothing will. Maybe, like the
students in Bruce Gregory's science learning group, the group
testing enthusiasts will eventually be able to figure it out for
themselves -- but I don't count on it.

As I said in an earlier post, I think one of the reasons group
research won't stop being used as a way to study individuals is
that it is so easy to do. And if you test enough subjects and/or
use a wide enough range of values of the IV you are virtually
guaranteed to get significant results (and find that the IV
"accounts for a large proportion of the variance in the DV"). So
it's a pretty seductive approach to doing "science". Heck, I'm
seduced by it myself (I rarely get a headache but when I do I
feel good about the opportunity to take an aspirin to reduce
my group-determined risk of a heart attack just a tad;-)

I think PCT research will be a lot harder to do than conventional
IV-DV research on groups. First, you have to think of possible
controlled variables. Then you have to think of various ways to
disturb those variables. Then you have to actually apply the
disturbances in a way that does not overwhelm the individual's
ability to keep the hypothetical controlled variable under control
(if it is, indeed, a controlled variable). Then you have to be
ready to revise your hypothesis about the controlled variable if
the evidence for control of the hypothesized controlled variable is
not clear. This all has to be done on one individual at a time and
it has be done using all the usual precautions of good experimentation
(trying to avoid confounding, measuring variables as accurately
as possible, etc). And, finally, the results you get may not seem
"interesting" to an audience that is used to seeing the kind of
results that come from conventional research. Instead of a
conventional result like "drug X injected into ventricle Y causes a
reduction in behavior Z", the PCT approach will probably produce
results like "All individuals tested were found to be controlling
for perceptions X, Y and Z, though all individuals differed in terms
of their reference level and gain for control of these perceptions
during the testing session".

But the only way to start a psychology of individuals is to start
doing it. We have made something of a start; we have many tracking
experiments that study the parameters of control one individual at
a time. There are also some studies of "more intersting" kinds
of individual control -- like the Robertson/Goldstein experiment
on "control of self-concept". I would like to develop methods for
studying "interesting" problems -- like "addiction" -- one
individual at a time.

My current inclination is to start with something like the
"ethnomethological" studies that thge late sociologist Harvey Sacks
used to do (and that I learned from him as an undergraduate at
UCLA). Basically, he would analyze transcripts of therapy
sessions and other human interactions (parent/child, etc) looking
for evidence of what we would call purposeful uses of language.
One idea I have is to go through transripts of therapy sessions
looking for evidence of controlled variables (in both the
therapist and client, assuming it's a two party interaction,
but I would focus on one person at a time). When I find what
seems like evidence of a response (say, on the part of the
client) to a disturbance (by the therapist) of a hypothetical
controlled variable (like "appearing to be able to control
alcohol intake") then I would look through the transcript for
other evidences of a response to what would be disturbances to
the same variable.

This process could be very tedious and far more difficult than
I imagine (transcripts of verbal interactions are notoriously
incoherent). Does anyone have any other ideas? Suggestions?

Thanks

Rick

ยทยทยท

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

[From Bill Powers (971126.1722 MST)]

Bruce Gregory (971126.1550 EST)--

This is probably a shocking suggestion, but why not try to teach
"subjects" to be their own observers? That is how most of us
convinced ourselves of the power of PCT. Your approach sounds
too much like the old-fashioned I experimenter, you subject
approach. Once an individual identified what seemed to be a
variable they were controlling, you could work with them to
develop a protocol to determine more precisely what was
going on.

Jeez, what do we have here, a flaming radical?

I love the idea, but maybe a little structure is needed. After all, we have
to do experiments we know how to do, including getting data. Maybe -- give
the participant a bunch of perceptions that can be affected by some action,
ask P. to play with it and pick something to make it do, then ask what P is
controlling? Of course for higher-level stuff I suppose you can just start
right out as in a therapy session: how are things going with you?
Everything working the way you want it to? Any difficulties with your world?

Best,

Bill P.

The situation suggested by Bill Powers sounds like biofeedback. A
person is given a visual and auditory display which represents some
aspect of a person's bodily function, for example, skin temperature or
skeletal muscle activity or brainwaves or sweating. A person is told to
take a playful attitude towards the display, and control it in some
way.

Rick's suggestion of using a therapy session is interesting. A few
years ago, I devised a coding system based on PCT to describe what was
happening.

Another suggestion, based on more public data, is to use tapes of a
talk radio program.

Bill Powers wrote:

ยทยทยท

From: David Goldstein
Subject: Re: A PCTer among the astrologers
Date: 11/27/97

[From Bill Powers (971126.1722 MST)]

Bruce Gregory (971126.1550 EST)--

>This is probably a shocking suggestion, but why not try to teach
>"subjects" to be their own observers? That is how most of us
>convinced ourselves of the power of PCT. Your approach sounds
>too much like the old-fashioned I experimenter, you subject
>approach. Once an individual identified what seemed to be a
>variable they were controlling, you could work with them to
>develop a protocol to determine more precisely what was
>going on.

Jeez, what do we have here, a flaming radical?

I love the idea, but maybe a little structure is needed. After all, we have
to do experiments we know how to do, including getting data. Maybe -- give
the participant a bunch of perceptions that can be affected by some action,
ask P. to play with it and pick something to make it do, then ask what P is
controlling? Of course for higher-level stuff I suppose you can just start
right out as in a therapy session: how are things going with you?
Everything working the way you want it to? Any difficulties with your world?

Best,

Bill P.

DW
I like your suggestion, Bruce. I'm reminded a bit of the European psychology tradition with introspection considered a mainstay of investigative techniques. I've also been pondering my own doctoral thesis in the area of psychophysics: 4 subjects, each listening to a tone that may or may not be getting very slowly louder or softer over a time period that went to several minutes. The issue: how rate of change influenced ability to detect change. I ended up with two remarkable subjects and two less remarkable. I'm also reminded of a fairly new field called collaborative inquiry, with Peter Reason at Bath Univ as its chief proponent. In a way, the approach aims to make people their own observers, working together in groups of 2 - 8.

I'm one of those who feels very much a PCTer but with some frustration at my own inability to think creatively about going beyond rubber bands and tracking ..... and mucking around in the messy vagaries of human behaviour where a multitude of controlled variables may be competing for attention at any moment in time. So, I'd like to urge you and others to explore your above suggestion.

David Wolsk
Victoria, BC Canada

P.S. I'm happy to hear that the next summer get-together is in Vancouver. Who will be organising it locally?

ยทยทยท

at 12:30 28/11/97 David Wolsk wrote:

at 10:00 28/11/97 At 15:48 26/11/97 EDT, Bruce Gregory wrote:

[From Bruce Gregory (971126.1550 EST)]>
Rick Marken (971126.1215)

This process could be very tedious and far more difficult than
I imagine (transcripts of verbal interactions are notoriously
incoherent). Does anyone have any other ideas? Suggestions?

This is probably a shocking suggestion, but why not try to teach
"subjects" to be their own observers? That is how most of us
convinced ourselves of the power of PCT. Your approach sounds
too much like the old-fashioned I experimenter, you subject
approach. Once an individual identified what seemed to be a
variable they were controlling, you could work with them to
develop a protocol to determine more precisely what was
going on.

[From Kenneth Kitzke (971129.20:00 EST)

Rick Marken wrote:

<I think it's clear now that it is impossible to convince anyone (who
isn't willing to believe it) that group methods are inappropriate
for the study of individual behavior.>

I agree they are inappropriate for the prediction of individual behavior.
Are you willing to concede that group methods are important fo determine
the behavior of groups of individuals?

Such studies provide knowledge on how individuals are likely to behave.
This can have value that may exceed knowing how any particular individual
will behave. Do you agree?

If not, I can see one reason why you and Bruce Abbott are on parallel
planes never to have much of a meeting of the minds. This may be good
though. That much concentrated power could cause a S-R event measurable on
the Richter scale and threaten life as we know it. :sunglasses:
Just wondering?

Peace,

Kenny