[From Bill Powers (950926.1550 MDT)]
Brian D'Agostino (950926.0025) --
I am still waiting to hear corrections of errors I have actually
made, rather than errors B. F. Skinner or someone else made.
The errors I have found are not of the sort that readers of your paper
would notice, but they do create some jarring notes for one acquainted
In the control theory framework, all human behavior can be viewed
as the output of one or another control system, organized into a
hierarchy ... (p. 262).
"Behavior" is an ambiguous term, as it can refer either to overt actions
(applying a force to the environment) or to intended consequences of
those actions (closing a door). It is better to define control as a
process of _varying_ outputs in order to create and maintain _intended
perceptions._ Because the environment always contains unanticipated
disturbances, there is no way to predict exactly what action a person
will produce while controlling some perception.
In a hierarchical system, the output of all control systems but those of
the lowest level consists not of producing actions, but of manipulating
goals for lower systems.
At the apex of the individual's control hierarchy is the regulation
of self-image ... (p. 263)
This, too is ambiguous, as there are many system concepts that are not
concerned with self-image: mathematics, for example, or agronomy, or
economics, or religion. The hierarchy does not have one single apex but
many -- if the idea of system concepts is accepted as the highest
At every level of the individual's control hierarchy ... behavioral
output is jointly determined by a comparison of perceptual input --
or cognition -- with a reference perception (p. 263).
Perceptual signals in the hierarchy are functions of lower-level
perceptual signals and ultimately of external processes. Some perceptual
functions such as categorizing or recognizing logical states could be
called cognitive, but there must also be cognitive processes involved in
comparison and in the output process of selecting lower-level goals --
decision-making, for example. I would disagree with saying that the
perceptual functions are the sole locus of cognitive functions.
In this article, the behavior to be explained is support for or
opposition to militarist policies, which is viewed as output of the
individual's belief system.
A belief system selects lower-level goals to create perceptions, not
outputs. The output is the means of supporting the belief system, and
consists of processes that bring the experienced world into congruence
with the reference beliefs. If we say that beliefs are system-concept
perceptions, then the means of supporting them would be, according to
the proposed levels, the selection of particular _principles_ as goal-
states. Those in turn would be achieved by selecting particular
programs, rules, procedures, and so forth.
These are not devastating errors, and they are errors only if your
intent was to adhere to PCT rather than invent your own system. But a
PCT high priest would feel obligated at least to call attention to them.
A more serious problem exists in your Fig. 1. You have placed the goal
of femininity (with a reference level of zero) at the top level, and
have made the output of this system into a selection of (high) military
power as the means of reducing the perception of femininity. This is a
very specific proposal about the hierarchical relationship between the
perceptions of gender and of power. Alternative proposals are possible.
You could, for example, place the masculine/feminine reference signal
and the military power reference signal at the same level, so that these
become coexisting system concepts. You could also have placed the
concept of military power at the top level with the masculine/feminine
dimension, as a principle, at a subordinate level. In the latter case
you might have proposed that in order to maintain a system concept of a
militarily strong system, a "hawk" chooses, among other goals, the goal
of perceiving himself (and others in power) as showing a low level of
femininity, viewed as weakness and thus not desirable as a means to
maintaining military power.
I'm not arguing for or against any of these proposals. My point is that
the thesis you hope to support is built into Fig. 1, with no indication
that other arrangements are possible. Since your data show correlations,
not hierarchical dependencies, supportive correlations merely show that
these two factors tend to vary together (oppositely). The results do not
indicate which variable, self image or military power, occupies the
superordinate position. They would covary in either case. At best, you
have shown that they may be of different levels, since if they were at
the same level they might vary independently.
In order to demonstrate a hierarchical relationship, you have to show
that the higher perception is a function of the lower, and that if the
higher perception is disturbed, the lower reference signal will be
changed to defend it, if necessary. Since correlations are not
directional, they can't be used for this purpose.
Finally, the argument with cognitivists and others seems to rest on the
assumption that all people are the same. That is, either all people who
are against war are against it because it violates their self-images, or
they are all against it for cognitive anti-war reasons. Why can't it be
that some people are against war for some reasons and others are against
it for other reasons? Why can't some people be against war because
they're afraid of being exposed to danger, a simple emotional reason?
Why can't others be against it because it violates their religious
beliefs about killing?
The same sorts of questions apply, of course, to people who want to
maintain military power and are willing to use it. People can be in
support of military strength for all kinds of reasons, from a simplistic
macho sense of power to a timid and fearful desire to be protected
against danger -- just like the doves who want to avoid danger by being
against war. Many militarists quote from the same Bible that anti-war
types cite, showing that religious motivations can work either way.
It seems to me to be a mistake to try to settle the question of what
motivates "hawks" and "doves." All hawks are not alike and all doves are
not alike. "People" are not alike. The correlations in your tables are
ample evidence of this: on any dimension there are great overlaps
between hawks and doves or males and females. It's perfectly possible
that nothing you can say about the aggregate of 413 people applies to
any individual in that group.
A PCT approach to the same problem, as Rick Marken said, would not use
the survey approach at all. It would start with in-depth interviews of
individuals and an attempt to discover their unique structures of goals,
without pushing for any particular structure. This would be a much more
demanding and time-consuming approach and would obviously not be
suitable as thesis material. However, it would permit constructing
generalizations after the relevant data were known, rather than forming
the generalizations first and then trying to validate them from
population statistics. In my opinion that is a much more fruitful way to
investigate human nature, even if it is slower.