[From Rick Marken (951002.0800)]
I tried to send this yesterday, with no apparent success. Here we
go again; I hope it's not a repeat.
This is a post from Tom B. to Brian D. Tom's off CSG-L for a while,
while he sells his house. But he asked me to post this for him; I
will send him any replies or other posts relevant to this thread.
Best
Rick
···
---------------------
[From Tom Bourbon (950929.1340)]
Brian, I received your paper and read it some time ago. Thank you, for
sending it so promptly. I apologize for my delay in sending comments. We
are selling our house, an activity that seems to get in the way of every
other important thing in life.
I made some comments that pertain to issues that are broader than just
PCT, and others that are specific to PCT.
When you described your work at CSG, some of us tried to let you know we
were making that distinction in our comments to you. Perhaps I am wrong,
but I never believed you saw the distinction we were drawing. (By the way,
when Phil Runkel wants to talk, Phil Runkel talks. Don't let his genuine
sweetness and charm deceive you. 
Also, at CSG we tried to tell you that we knew you did not expect the kinds
of comments you encountered. We could readily see that you had used some
traditional experimental procedures and conventional statistical analyses,
of the kinds that, when they are used appropriately, might yield useful
descriptive data about people in naturally-defined groups. It was also
obvious that you had good knowledge, as defined by the highest criteria
used in traditional behavioral science, of your procedures, your analyses,
and your area of study. (That remark is intended as a compliment, not as a
criticism.) We did not fault you personally for knowing what you knew, or
for using the methods you used. We did try to tell you why those
traditional procedures, which _can be_, but are not always, appropriate for
studying groups of anonymous subjects, are of little interest or use _to
us_, in our studies of perceptual control by individuals taken one at a
time.
A. First, a few comments on matters not related exclusively to PCT:
1. Your experimental methods and statistical analyses are as mainstream as
they can get. I can't imagine how any traditional behavioral scientist
could object to them. I _can_ imagine some would object to the topic you
studied. I can also imagine some of them would object to the way you used
control theory as a different "perspective" on your data. On the other
hand, I know of several conventional behavioral scientists who identify
themselves as "control theorists" who probably would be comfortable with
your paper.
2. You used a Q-set procedure (also known as Q-sort). When it is used the
right way, it can be a source of ideas about some of the perceptions
controlled by an individual. However, you lumped the data from all of your
participants together in a few groups and did your analyses on the combined
data sets. Later, when you interpreted and discussed your data, you often
spoke of group statistics as though they could be applied to individuals in
the groups. You did say that some individuals did not match the
representative scores for your groups, but you went ahead and talked about
group statistics as though they might apply to all members of the groups.
For example, you wrote about the "gender specificity" of certain items, but
only certain proportions of males or females replied to certain items in a
particular way. The things you did are near-universal practices in
behavioral science, but they are not legitimate. (I am not saying you
initiated those practices.)
3. You ended up with too few females to satisfy the conditions that would
allow you to generalize from your sample to the larger "policy elites" and
"attentive publics," but you went ahead and talked about possible
generalizations. That is as mainstream as can be.
4. I wonder why you treated the data for females and males so differently.
Table I (page 272) contains data for females. You present six correlations
between "self variables" and their associated "self dimensions." You
indicate the level of statistical significance for each of the six
correlations. Table II (page 273) contains data for males. You present an
18X18 matrix of correlations. You do not indicate the level of statistical
significance for any of those correlations.
5. Concerning the correlations in your data, for females, all six of the
significant correlations are between -.39 and -.30. For males, 2
correlations (disregarding the sign) are between .50 and .53; 2 are between
.40 and .49, 6 are between .30 and .39, 30 are between .20 and .29, 57 are
between .10 and .19, and 63 are between 0.0 and .09. (Don't hold me to an
exact count, even if I did do it twice.) I know some of those correlations
must be statistically significant, but all of them are too low to support
any strong conclusions about people and their behavior.
For your largest correlation, +.53, the percentage of variance "explained"
is 28%. The coefficient of alienation, k, for that correlation is
SQRT(1 - r-squared), which is .85. That means the degree of lack of
relationship is .85, the degree of relationship, only .53. Another way to
describe this relationship is to say that, when r = .53, the Standard Error
of the estimate for the predicted variable is 85% as great as the margin of
error were you to make the prediction when you knew _nothing_ about the
degree of relationship between the variables. Those are weak data. It is
not your fault they are weak. It is not your fault that behavioral science
is built on a foundation of data that are similarly weak, albeit
statistically significant.
We can perform a similar assessment of alienation for your multiple
regression analysis. Your Multiple R was .66296 (rounded to .66). For
Multiple R = .66, K = .75. The interpretation of this relationship would
follow, exactly, the style of the one above. Whether or not PCT exists,
these correlations are weak and any science built on them must be flawed,
especially if the scientists purport to explain the behavior of individual
people. It is not your fault that most behavioral science is built on
"facts" that are fatally flawed.
6. You rely heavily on a mainstream strategy in which writers portray
certain behaviors they do not like as "psychopathological," and people who
encourage those "behaviors" are said to use "propaganda." Hence,
"militarism = psychopathology," there is a "problem of militarism," and
there is "militaristic propaganda."
A male who controls for a perception of "not feminine" evinces "male gender
insecurity." Psychopathy is rampant!
B. Now a few comments on matters related directly to PCT.
1. On your first page, page 259, you say control theory is about
"self-regulating, negative-feedback systems." Maybe yes, maybe no. This
statement is often part of a frequent misconception about PCT -- that it is
about self-control, or self-regulation, in the sense of "being in control
of yourself and your actions." I don't think that was what you meant. I
usually try to avoid that misconception altogether, by not saying PCT is
about self-regulation.
2. On page 269, in your description of the Q-set, you say the procedure is
one in which each item is a potential disturbance to a controlled
self-perception. Fine. As a suggestion about how to use the Q-set in a
study that is more clearly PCT-ish than yours, you could work with
participants, one at a time, and follow up on their original sorting of the
items. You could tell them the sorting was "wrong", or that you think some
of the items they said were important should be in another place that
indicates unimportance, or you could present an alternative sorting and ask
the person to "make it right," and so on. Procedures like those would let
you test more directly for controlled perceptions.
3. On page 278, you describe some thought experiments in which you
discuss the different "effects of" a given perception on three individuals,
with three different reference levels for the perception. This is a good
topic, and I can imagine your treatment of it might be novel in the
literature of political science. I would describe the situation a little
differently than you, given my quirky aversion to speaking about
perceptions as things that affect people -- that's a little too
Cause-Effect-like for my radical tastes.
4. You introduce control theory on page 262. You use a thermostat as an
example of, "the basic structure of all control systems, including living
organisms." You continue, "In the case of the thermostat, the system
controls room temperature by continuously comparing its perception of this
variable with a "reference perception," namely, the thermostat setting.
When perception deviates significantly from the reference perception, the
resulting error signal activates a heater or air conditioner until room
temperature is returned to the level established by the thermostat
setting." Not too bad, but a few fine points are not very clear.
For one thing, the distinction between the roles of the thermostat and the
output devices is not clear, hence, it isn't totally clear which element(s)
you have in mind when you say, "the system controls room temperature."
Also, as part of the entire system, the thermostat controls its own
"perceptual signal," independent of the actual state of room temperature.
A person can use the thermostat as a way to affect a system that affects
the person's perceptions of the temperature of the room. Nowhere in the
assemblage of person-thermostat-output_device is there an element that
controls the objective temperature of the room.
5. On page 263, you write about "the behavior to be explained" (support
for or opposition to militaristic policies) as "the output of the
individual's belief system." In Footnote 2, you amplify on that
characterization: "The output side (of a control system) consists of
political preferences that guide behavior." These statements leave the
real issue a little foggy for a reader not familiar with PCT. The
"behavior" you studied is a means by which people control certain
perceptions, which may be (and I would bet, often are) quite different from
any you discuss in your study. I am also leery of the idea that
preferences" (reference perceptions?) "guide behavior." That idea is
common in some of the more egregious bastardizations of control systems
theory. I am not sure you mean your footnote to be read in exactly that
way. In the role of references perceptions, preferences would guide
(specify) perceptions, not behaviors, which would necessarily vary in order
to maintain perception at the reference value.
6. On pages 280, 281, you write about a case of "gender insecurity" (you
refer to the PCT model illustrated in Figure 1, page 264) in which a male
has a reference level of "zero" for being "feminine," but in your example
he, "succeeds only imperfectly because of continued identification with the
mother." "Such identification can be understood as a chronic intrapsychic
'disturbance' in the control theory sense . . .." Brian, I think your
example might be represented better by a system with _two_ high-level
reference signals, and two high-level control systems, one for "don't be
feminine," and the other for whatever you mean by, "be identified with
mother." That way, you would have pure internal conflict. Eliminating
error for one system would necessarily create error for the other system,
assuming the two reference perceptions are mutually exclusive.
This is long enough, for now.
Later,
Tom