Clark's levels: what's the principle?

[From Bill Powers (930923.0700 MDT)]

Bob Clark (930918.1715 EDT) --

I'm having a lot of difficulty understanding the growing list of
"levels" that you're presenting. It seems to me that instead of
describing classes of perceptions that people control in a
hierarchical-related way, you're developing a systematic but
arbitrary taxonomony of behaviors without much regard to either
the perceptions involved (from the behaving person's standpoint)
or the hierarchical relations among them.

It's hard for me to see, for example, how "skill" can be a
category of controlled perception except in a very narrow sense.
Are you saying that people perceive the skilfulness of their
environments and themselves, and act so as to bring examples of
skilfulness to some desired level, low to high? I can see how a
person might perceive and control skilfulness under certain
circumstances. For example, a clown might deliberately act in an
unskillful way to amuse others, or an athlete trying out for a
team might try to act in a way that he or she perceives as
exceptionally skillful, in the hopes that the judges will also
perceive a high degree of skill. So skillfulness could
conceivably be a controlled variable with reference levels
ranging from low to high on the scale of skillfulness. But as a
dimension of control, it seems to me that skillfulness is too
specific a perception to merit a level of its own.

How does one recognize skill? In my scheme, I would place this
perception at the "principle" level, because it is a
characterization, a generalization, about the way in which
programs or sequences are carried out. When you look at an
athlete or a mathematician or a manager, or your own attempts at
acting in such roles, you can perceive whether the execution of
the sequences and programs is clumsy, merely competent, skillful,
or masterful. There is nothing in the details of any particular
program or sequence that can be pinned down as the criterion of
skillfulness; skillfulness is perceived over a range of control
behaviors, as a generalization about them that itself is neither
a program nor a sequence. So in my scheme skillfulness is a
principle-level perception, of a higher level than the programs
and sequences that are being perceived as skillful to some
degree, but only one example among myriads of perceptions of the
same level.

I recognize that my attempts to define levels of perception and
control are probably both deficient and excessive in various
regards, but behind them there is a rationale that imposes some
discipline on the procedure. Each type of perception is intended
to be defined so that if you try to decompose any example of that
type into its elements, the elements prove to be perceptions of a
lower-level type already defined. The new level adds some aspect
or consideration of a new type not exemplified in any one of the
lower-level perceptions, yet it depends for its existence on at
least several perceptions of the lower types. To bring the
higher-level perception to a particular state requires altering
specific lower-level perceptions.

So, for example, a perception of the type "relationship," when
analyzed into components, proves to be a function of events,
transitions, configurations, sensations, and intensities, but is
not to be found in any one of those components. In order to alter
a relationship, it is necessary to alter some of the lower-level
perceptions. A specific relationship cannot exist unless at least
two perceptions of those lower types also exist. "Separation"
requires perceptions of at least two objects (independent
configurations). These criteria seem to be met by any perception
we would class as a relationship, regardless of sensory modality,
context, or circumstances -- whether the elements are associated
with people or things, objects or movements, living or nonliving

I have tried, with incomplete success, to define levels that have
all these characteristics in relation to other levels:
decomposition into and dependence upon lower-level perceptions,
as well as the requirement that control requires alteration of
lower-level perceptions. This is what the system concept of
hierarchical control means to me: it is the appearance of these
principles in all of behavior, regardless of the kind or

In looking at your definitions of levels, I fail to see the
system concept that generates them, or the specific principles
that constrain the arrangement of levels. It also seems to me
that your categories are too narrow to cover the full range of
human experience -- where, for example, do you put the perception
and control of physical theories? The writing of computer
programs? Logical reasoning? Generalizations? It would help my
understanding of your system greatly if you were to discuss how
you decide that one level is higher or lower than another, and
give examples showing how your criteria apply.



Bill P.