Let's get perceptual

[From Rick Marken (930828.1600)]

Michael Fehling (27 Aug 1993 18:01:36) --

In PCT, the word "perception" refers to the presense of a signal of some
amount in an afferent neuron. It is hypothesized that ALL experience is
perception. This includes what are called sensations, perceptions and
beliefs but it also includes desires, intentions and actions (to the
extent that these are experienced -- we can only experience desires,
intentions and actions as perceptions). Everthing we can talk about is
perception because all we experience is afferent neural signals. Talking
itself is a perception; the meaning of the talk is perception; PCT is
perception; physics is perception, feelings are perceptions, it's ALL
perception (at least according to PCT). Some perceptions are caused by
events in the environment outside the nervous system (including the
"internal" environment of the body -- muscles and organs). These are
the "common" perceptions -- what you experience as waking reality.
Other perceptions are based on events originating deeper inside the
nervous system itself. These perceptions are called "imaginations".
Imaginations differ from perceptions ONLY in terms of what causes them
(external events in the case of perceptions, internal nervous system
events in the case of imaginations); both are physically EXACTLY
the same -- they are signals in afferent neurons. Imaginations include
much of what you call belief (like the belief that my dead grandmother
isn't really standing here or the belief that there are perceptions
and beliefs); imaginations are also called thoughts (which can include
thoughts about how to model human behavior and about why one might even
want to do it anyway).

I think this is basically what Bill P. meant when he said "Beliefs,
as I see the term used, are _primarily_ perceptions". The "primarily"
part has to do with the fact that, when people talk about their beliefs
(imaginations) they imply preferred states for them (perceptions and
imaginations are variables). If you see your late grandmother before
you (p) and you believe (imagine) that she is not there you are preferring
to imagine "p is not true" rather than "p is true"; there are two
possible values of this logical belief and you prefer to have this belief
at one (rather than the other) of those values. So PCT has no trouble handling
the fact (I imagine it's a fact -- I've never experienced it; well, now
that I think of it, I have experienced it -- at magic shows; I perceive
one steel ring passing through another (p) but I believe that that's not
what is actually happening) that people can believe that they are not
perceiving what they are percieving; it also explains why they might
want to have this belief at one level (grandma's not really there)
rather than another (grandma is there); the reference level for this
belief (perception) would be set in order to satisfy higher order
references for perceptions of one's rationality or sanity, for example.

Your test of PCT as "designed to disprove [PCT's] claim..." sounds like the
Popperian view of science. I'll pass on this, but Kuhn, Feyerabend, Lakatos,
and others have challenged this approach to science in interesting and
influential ways.

"The test" that Bill described is NOT a test of PCT; it is a test to
determine whether you are actually dealing with the phenomenon of
control. Many phenomena resemble control -- things like the behavior
of a mass on a spring, a raindrop moving to the sea, a marble coming to
rest at the bottom of a bowl, an organization remaining headquartered
in the same town, etc. I think it is important to know the nature of
the phenomenon you are dealing with before you start trying to model it.
It would be a big mistake (I think) to start applying a control model
to, say, planetary motion before you have good reason to believe that
the planets are control systems (we already have a lot of evidence
that they are not).

Finally, I guess we may just have to agree to disagree about a theory/model
as a representation of "reality" versus a model as a view of reality.

This might turn out to be a very important disagreement. PCT (like
Newton's laws) purports to be a model of the unseen variables and the
functional relations between them that result in what we do see (in
the case of PCT, what we see is purposeful behavior -- control). If
a model is just a way of "viewing" reality then how do we test which
"view" is better? I prefer to evaluate behavioral models by comparing
their performance to that of real organisms. If you agree that that's
an acceptable way to evaluate models then we have no problems; you can
call a model a "view" of reality; we'll call it a representation of
the reality that underlies our "view".



In re Rick Marken 930828.1600 --


Judging from your remarks about my earlier post (930827.0745), I may need to
clarify. I think I'm clear on your remarks. However, some of your points
seem not to be responsive to the issues I attempted to raise. So, I'll try to
be more precise.

  Let I'll restate how one demonstrates that sensations, beliefs, desires,
intentions, and actions are distinct functional categories. For example, one
can say 'I perceive p' and yet, freely say either (a) 'But I don't believe p'
or (b) 'And I do believe p.' This demonstration simply refutes the claim that
_belief_about_p_ and _perception_of_p are the same. I am employing here a
very basic method of rational analysis that your remarks seem to pass over--if
one can find a feature present in one case and absent in another, then the two
cases are distinct (up to that feature). One may use the same method to
distinguish sensation from perception. This time, I will borrow your own
distinctions to do the work. In an earlier post you discussed how an input,
i, is transformed into a percept, p, by a function, f--i.e., p = f(i). The
sensation is the input to f. It better be "in" the organism, else f couldn't
operate on it. And, unless f(i) = i, it must be that i is different from p.
So, sensations are distinct from perceptions.

  This type of analysis clarifies what it means to call something a belief, a
perception, a desire, etc. Philosophers often call these categories of
"propositional attitudes" to reflect that each distinct type depicts an
_attitude_ or disposition toward a _proposition_. For example, belief-in-p is
an attitude of some degree of subjective commitment to the factual status of
p. In contrast, perception-of-p conveys a different attitude. The
proposition, p, toward which the attitude is directed can be common to to more
than one propositional attitude. In my example, p is the proposition 'Grandma
is there.' However, the propositional attitudes with common propositional
content are, nevertheless, distinct. (In my earlier example p is the
proposition 'Grandma is standing before me' toward both perception and belief
are expressed.)

  Unfortunately, your discussion of the grandmother assertions glosses over
this important distinction. Once again, I hear what you proclaim--it's all
perception for the agent. But, the preceding analysis suggests that
proclamation may not be logically sound. Actually, you appear to be using the
word "perception" as a common term for all propositional attitudes since you
hypothesize "that ALL experience is perception...[including]...sensations,
perceptions and beliefs...[as well as]...desires, intentions and actions."

  To clarify my point I'll accept for now this expansive construal of
perception. I'll stipulate that each propositional attitude is merely a form
of perception (sic). (For terminological clarity let's rename the attitude I
have been calling a "perception" and call it an "interpretation.") The point
propositional attitudes define perceptual categories whose distinct properties
and functions are evidently critical in fully explaining the structure and
function of perceptual control systems. (This is one of the central theses
that I am exploring in my own modeling of organizations and individuals as
control systems.) With this restatement in hand, I hope you will deal with
the this analysis of propositional attitudes on its own terms. It would be a
great service to many theorists like my if PCT can show how/why these
distinctions are useless, or wrong.

  On the other hand, I propose that these distinctions can provide a basis for
giving a constructive account of "information" in PCT's control systems. This
is why I am raising these issues to Bill, and the rest of this discusion
group. In a number of previous posts various people have pointed out that the
role of information in PCT has not been significantly addressed and/or is very
unclear. Perhaps this is one place where work like my own might make some
small contribution to the work on PTC and HPCT.

  Now, let me react to some of your other comments. First, as for my remarks
about "The Test," you are right. I put my point very imprecisely. Indeed, I
understand that this test is part of the PCT methodology to determine if a
system qualifies as a control system. It is not a test of PCT itself. I
meant to say "Your PCT-Test as 'designed to disprove the claim that some
system is a perceptual control system' sounds to me like the Popperian notion
of scientific method as the examination of falsifiable claims. I only wanted
to mention the existence of challenges to this Poperian view by Feyerabend and
others as an antidote to the assumption that falsifiability as a cornerstone
methodological principle stands unchallenged.

  Now, I hope you will clarify another point. I would never have thought to
state, as you did, that a system such as a mass on a spring, "resembles
control." It is just a dynamic system. A dynamic system, as perceived, is
the _object_ of control. The same would be said for a "raindrop moving to the
sea," or "a marble coming to rest in the bottom of a bowl." I.e., these
dynamic systems are just "plants," i.e., candidate controlled systems.
Control is defined in terms of how controllers excercising (perceptual)
control laws are (or, more properly, perceive themselves to be) coupled to
these plants. What were you driving at?

  Toward the end of your message you react to my stand for models as views by
saying, "This might turn out to be a very important disagreement...If a model
is just a way of 'viewing' reality then how do we test which 'view' is
better?" Well, it's easy. Just compare models in terms of how well thay
account for the observations (of phenomena) in which you are interested.
Don't confuse the winning model with the phenomena for which it (partially)
accounts. Keep in mind that even the terms you use to identify those
phenomena (e.g., your "real organisms") are conceptual constructs that you use
to make the phenomena meaningful for you (e.g, your "real organisms"). And,
since models of complex phenomena may themselves be complex, there may not be
a model that wins uniformly. One model may be superior in some respects while
another may be superior in others. So, the winning model may be partly
determined by your purposes in using that model. Once again, these are all
non-Popperian issues talked about by such writers as Lakatos, Feyerabend,
Kuhn, etc. What puzzles me is that these non-Popperian remarks would raise
the hackles of a devotee of PCT. After all, these points about models or
theories as views just go to say that a scientific proposition conveys a
_perception_ of reality. The reality accounted for may really be "out there,"
but it's beyond direct access by the scientist. That's what is meant by a
view. A model presents a structured perception of scientists' inputs. I
would have thought that PCT'ers like you and Bill be the first to say that
that's all science can do.

  Finally, please believe me: Due to the effort of Bill Powers, you, and
others, I think I've got it -- PCT is about behavior as the control of
perception. I find it to be a very clear, simple, and deep insight. It seems
to directly address some serious problems in conventional psychology and
social science that have concerned me since I became a student of these
subjects. But, PCT's insights also seem quite robust. They stand quite
nicely without dismissing other, refining distinctions such as the
epistemological distinctions proposed above and previously. I proposed them
because I think they suggest a way to analyze information in PCT terms. In
fact, I suspect that these added distinctions might even serve to amplify the
clarity and importance of PCT's insights. As Einstein supposedly said of a
theory, "Things should be as simple as possible, but no simpler."

- michael -


from my last message can now be restated as follows: These various types of

[Martin Taylor 930829 17:15]
(Michael Fehling 930829 13:00 or thereabouts)

[A point of order--It is customary, in CSG-L, to identify a posting by the
poster's name and the date and time of the posting, to ease back reference,
and to list the back reference, if any, either at the start of the posting,
or if there are several distinct ones, at the start of the section referring
to each.]

Michael is concerned with the differences among the denotations of "perception"
"belief" intention" "desire" "sensation" and perhaps a few others.

Whenever one uses a word, one assumes a background world within which the
word has connotations. It means something only within a world model that
is assumed to be largely shared by the reader/listener. The word "perception"
has taken on a meaning within PCT that is different from its meaning in
everyday speech. This is acknowledged, and it is also acknowledged that
the change of denotation is confusing. The re-use of the word "perception"
for a different concept has been excused by Bill P. on the grounds that
it is the closest word available for a construct that does not occur in
everyday language.

The PCT model, and in particular the variety of hierarchic PCT (HPCT) that
forms the foundation for CSG-L, takes "perception" to refer to any neural
signal that is derived eventually from sensory input by means of any number
of transformations, or that occurs within the neural system in a place
that sometimes conforms to that definition (thus covering memory, imagination,
and similar constructs). An immediate consequence of that definition is
that a perception is (a) unidimensional, and (b) definable in terms of
its intensity. A further consequence is that it is NOT what we normally
call a perception, which is conscious and of very large dimensionality.
So when a PCT writer says "It's all perception" or the like, the statement
hinges on a redefinition of the word. In PCT terms, anything that could
be part of a conscious (everyday) perception is itself a perception, but
lots more besides also is perception, if it is based on sensory input.

When Rick says that a belief is a perception, that is true within the PCT
definition of perception, if only loosely. A belief must be a constellation
of PCT-unidimensional perceptions, including what Michael points out as
necessary, an intensity of the sense that this constellation refers to
the real world. That unidimensional value (a PCT perception) is part of
the entire constellation. The same constellation except that it has a
zero value for "sense of reference to the real world" might be an imagination
or perhaps a hallucination (though hallucinations probably have an
intermediate value for SRTRW).

Again, PCT has at least a dual usage for "sensation". Every Elementary
Control System (ECS) controls one single-valued perceptual signal, but
that signal is a function of many inputs. Those inputs are often called
the "sensory" inputs to the ECS, or its "sensations." Unless the ECS
is connected to the peripheral sensory system directly, its sensory
inputs are the perceptual signals of other ECSs. In the strictly
hierarchic form of PCT, these inputs are from lower-level ECSs, but one
could imagine other forms of PCT in which cyclic connections of perceptual-
to-sensory-to-perceptual-to-sensory signals might occur. The other use
for "sensation" is just the intensity of a sensory signal, especially
one coming from a peripheral sensor.

Now consider the other, output, side of the control hierarchy. Since PCT
does not explicitly deal with emotion, and "desire" has an emotional
connotation, it is not a specific construct within PCT. Nevertheless,
it seems reasonable to relate "desire" to "reference signal" in much
the same way that the everyday use of "perception" is related to the
PCT use of "perception." One might say that a "desire" is a constellation
of consciously developed reference signals. It is a wish that a belief
be obtained (by actions that affect the real world).

There is a complication, based on what is known as the "degrees of freedom
argument." The number of degrees of freedom a person has for action are
several orders of magnitude fewer than the number of degrees of freedom
for sensation (or perception). Hence, at any moment, almost all perceptions
are not under active control, even though they may not be at their reference
values. Hence, even though one may "desire" a constellation of perceptions
that include a sense that the constellation refers to the real world, yet
it is possible that one does not act so as to bring that belief into being.
Desires can be left unsatisfied.

"Intention" clearly refers to active control. The "intention" is the
bringing under active control of those perceptions involved in a desire.
A small subset of desires lead to intentions, which is to say to actions
in the real world intended to lead to the fulfilment of the desires.

One final point of wording: "Action" and "behaviour" are distinct. Action
is the PCT word for the output of an ECS. It sets reference levels for
lower-level ECSs, in the strictly hierarchic version of PCT. Actions
(and this is most important) are UNCONTROLLED. They form part of a control
loop, and therefore are modified as the controlled perception deviates
from or approaches its reference level, but in themselves, they are just
blind output. "Behaviour" is the term for the complex of control, in which
the actions are seen as affecting the controlled perception. There are
lots of side-effects to actions, altering parts of the world unrelated
to the controlled perceptual signal. Behaviour has no side effects: it
is the outer-world counterpart of the intention. It is what is done to
control the perception.

I'm sure that there will be disagreement with some of this from the regular
contributors, but I hope it isn't too far from canonical orthodoxy.


[Michael Fehling 930829:3:06PM PDT]

In re Martin Taylor 930829 17:15 --


You given just what I had hoped for. A reconstruction in PCT terms of
cognitive/connative/affective constructs that that form a part of my own work.
As per my last message, I got the idea from Rick that PCT has its own special
meaning for "perception," but you've clarified this even more with your
mapping of propositional attitudes into PCT terms. Thanks.

  Incidentally, the propositions to which I referred can be arbitrarily simple
or complex. Thus, the unidimensional signals also fall within the scope of my
analysis, unless one insists on a strict consciousness criterion (which I,
among others, do not).

  One clarification. Desire as many use the term essentially refers to a
value function. It states relatively the status of the content of the desire
as an purpose. So, desire may well correspond to PCT's reference perceptions,
but no connotation of emotion is presumed.

  Now a question. Has any of the hard work been done to formalize and fill in
the details of the mapping we're now discussing?

- michael -

[From Rick Marken (930829.2030)]

Martin Taylor (930829 17:15) --

I join Michael in thanking you for an excellent job of explaining
and expanding on what I tried to say about the nature of "perception"
in the PCT model.

I'm sure that there will be disagreement with some of this from the regular
contributors, but I hope it isn't too far from canonical orthodoxy.

I tried, but I couldn't find a thing with which to disagree. Nice job.

Michael Fehling (930829:3:06PM PDT) --

Now a question. Has any of the hard work been done to formalize and fill in
the details of the mapping we're now discussing?

I'm not sure I know what you mean by "formalize and fill in the details" but
I'm pretty sure that, whatever you mean, the answer is VERY LITTLE. There
are VERY FEW people who are actively interested in PCT and far fewer who
are doing the research and modelling needed to flesh out the theory. Most
of the work that has been done has involved control of VERY SIMPLE
perceptual variables. Bill P. has just recently published a paper on the
control of what would surely be called "cognitive" variables (but are
just called "higher level perceptions" in PCT. I don't know the reference
for the article. Tom Bourbon has done some nice research on social control -
but still with "simple" variables. I have started doing some work on
control of "higher order variables" but this work is not published. I
recommend my collection of papers called "Mind Readings" ($18.00 from
CSG Press) for a description of some of the "spadework" that has been
done to test the basics of the model. But the fact of the matter is
that there has been very little done to test the mapping of the kinds
of variables you mention ( beliefs, values, etc) to the PCT model.
Much of that kind of research will necessarily be of an almost clinical
nature -- and in this sense there are clinicians (for both individuals --
like Ed Ford -- and for organizations -- like Dag Forssell) out there
"testing" the PCT mapping of these concepts to what actually happens in
their practice. But the bulk of the quantitative research and modelling
is being done by only a handful of people, most of whom (like me) do
it "on the side" while making a living in some other profession. That's
why it would be great to have someone with the resources (graduate
students and TIME) do the work that needs to be done to "flesh out" the
rest of the model.

What we have tested of PCT works awfully well -- I would be surprised if the
basic organization of the model had to change drastically once we started
to move into the study of higher level variables. But we do need to start
studying some of the kinds of control processes you described in
your earlier post, for example. Why, for example, do people vary their
beliefs about the truth of propositions or the probability of the correctness
of implications. Studies of these kinds of things have been done -- but
usually in the "wrong" way, from a PCT perspective (eg. results based
on averages over subjects, individual results too noisy to be useful, etc).

So much of the PCT model is the way it is for the sake of 1) simplicity
2) internal consistency and 3) the fact that it works. The PCT model
is unquestionably weakest (in terms of research results) in precisely the
areas where you are working. So I would imagine that you would be
able to help the PCT model a lot more than the PCT model could help you.
So I hope we can keep up the dialog.



[Hugh Petrie 930830 09:00]
(Martin Taylor 930829 17:15)

As a former philosopher who used to be quite concerned with some of the
philosophical aspects of action which also are exercising Michael Fehling, I
just wanted to emphasize one of Martin's points about the distinction between
action and behavior. In the philosophical literature of some ten to fifteen
years ago, anyway, a gross reading of the accepted analysis would be
        action = behavior + intention
Although there is not a perfect transformation into PCT terms, roughly that
implies that PCT uses the words, action and behavior, in almost the opposite
way in which they tended to be used in philosophical theories of action a
decade or so ago. So beware!

Hugh G. Petrie
367 Baldy Hall
Graduate School of Education
University at Buffalo
Buffalo, NY 14260


[Martin Taylor 930830 10:50]
(Hugh Petrie 930830 09:00)

I just wanted to emphasize one of Martin's points about the distinction between
action and behavior. In the philosophical literature of some ten to fifteen
years ago, anyway, a gross reading of the accepted analysis would be
       action = behavior + intention
Although there is not a perfect transformation into PCT terms, roughly that
implies that PCT uses the words, action and behavior, in almost the opposite
way in which they tended to be used in philosophical theories of action a
decade or so ago. So beware!

Interesting. Should PCT switch its nomenclature, or would that be even more
confusing? Probably it would be more confusing to PCT-ers, but how much of
the outer world that eventually must be totally converted is aware of the
philosophers' distinction? Who switches? Or do we use new terms such as
"PCT-action", "PCT-behaviour". Since PCT-action is simply output, perhaps
we should drop the term entirely. At least I'm confused, now.

A point I forgot in yesterday's glossary of terms. An output of an ECS
(PCT-action) provides references for lower ECSs, and is therefore a potential
generator of PCT-behaviour at the lower level. I say "potential" for three
reasons: (1) because the reference level for any ECS is usually derived
from the outputs of several higher ECSs, rather than from a single one,
(2) because the new reference level just possibly might be such as to reduce
the existing error in the lower ECS (this happens a lot in my Layered Protocol
theory of communication), and (3) because the lower ECS may be operating
in imagination mode so that no externally visible acts occur.

So, though the distinction between PCT-action and PCT-behaviour is clear
when one is looking at a single ECS, it blurs when one looks at the hierarchy.
An uncontrolled "action" at one level can be seen as well controlled
"behaviour" (i.e. actual control of perception) at many lower levels.

Rick, Thanks for the vote of confidence. I hope this addendum also fits
your view.


[From Chris Malcolm]

[Martin Taylor 930830 10:50]
(Hugh Petrie 930830 09:00)

In the philosophical literature of some ten to fifteen
years ago, anyway, a gross reading of the accepted analysis would be
       action = behavior + intention
Although there is not a perfect transformation into PCT terms, roughly that
implies that PCT uses the words, action and behavior, in almost the opposite
way in which they tended to be used in philosophical theories of action a
decade or so ago. So beware!

Interesting. Should PCT switch its nomenclature, or would that be even more
confusing? Probably it would be more confusing to PCT-ers, but how much of
the outer world that eventually must be totally converted is aware of the
philosophers' distinction?

In at least one thread of modern philosophy of mind which is concerned with
cognitive science and ethology "behaviour" is used in a sense which includes
purpose; indeed, the most compact description of behaviour is in terms of its

I suspect the earlier definition of "behaviour" in terms empty of purpose was a
temporary US attempt to be politically and scientifically correct within the
aegis of Skinnerian behaviourism.

Chris Malcolm