Definitions

[Bruce_Abbott]

[From Bruce Abbott (980309.1700 EST)]

Below is my stab at some PCT definitions. Comments welcome, as always.

disturbance: A variable external to a control system that exerts an
             influence on that system's controlled variable. Symbolized
             as d.

disturbance The influence of a disturbance on the controlled variable.
influence Basically, what effect the disturbance would have on the CV
             if the feedback loop were to be cut. The disturbance
             influence = F(d), where F is some function. No symbol.

perception The scalar output variable of a perceptual input function.
             Also called a perceptual signal. Symbolized as p.

controlled The scalar output variable of the perceptual input function
perception of a control system. Symbolized as p.

input An input variable to a perceptual input function. In a
quantity control system, this variable represents some aspect of an
             environmental variable, such as its amount or its rate of
             change. Symbolized as qi.

error The difference, in a control system, between the controlled
signal perception and the reference signal. Symbolized as e.

reference A scalar variable, external to a control system, whose value
signal defines the level at which the controlled perception of the
             system will tend to be maintained against the influence of
             disturbances. Symbolized as r.

output In a control system, the output of the output function. This
             output will be some function of the error signal. In general,
             the output function acts as a power amplifier and the output
             does work in the environment external to the control system.
             In a control system, the output is feed back via the output
             function to the controlled variable. Symbolized as qo.

x (Tentative) The output of the environmental feedback function.

controlled That aspect of the environment external to the control system
variable which is acted upon by the output (via the environmental
             feedback function) and disturbance (via a disturbance function)
             and which is represented to the control system as the input
             quantity. Symbolized as cv.

input Mechanism that transduces the input quantity (environment) to
function a perceptual signal (internal representation).

output Mechanism that transduces the error signal (internal to system) to
function an output (action in the environment).

comparator Mechanism that outputs the error signal as some function of the
             difference between the reference signal and perceptual signal.

environment Relationship between a system's output and the influence of that
function output on some environmental variable. In a control system, the
             feedback function through which output influences the cv.

                                     r
                                     >
                            p v e
                [inp f]----------->[comp]------------>[outp f]
                   ^ |
                qi | | qo
                   +------------[cv]<------[env f]<------+
                             f(d) ^ x
                                  >
                                  d

system That part of a control system receiving inputs through
             the perceptual input function and producing outputs through
             the perceptual output function. This includes the input
             function, p, r, the comparator, e, and the output function.

environment That part of a control system receiving inputs through
             the perceptual output function and inputing to the perceptual
             output function. This includes the environment function,
             x, cv, d, f(d), and qi.

behavior What a control system attempts to accomplish: the control of
             perception.

action Output of a control system by means of which the influences of
             disturbances to the controlled perception are countered.

behavioral The illusion that by manipulating d and recording qo, one has
illusion identified a relationship between d and qo that describes
             something about the internal organization of the control
             system, when in fact one has only found the inverse of the
             environmental feedback function.

Marken The illusion that all experiments in psychology involve
behavioral manipulating a disturbance to a control system, recording
illusion the resulting variation in output of the same system, and
             from this believing that one has deduced the organism
             function.

difference Proof that the person holding the differing opinion (a) does
of opinion not understand control theory, and/or (b) is trying to impose
             a stimulus-response analysis on a closed-loop system in order
             to preserve his or her investment in "conventional" psychology.

information A fictitous quantity of no use in understanding control and its
             limits.

words Vehicles for misunderstanding.

Bruce A smug pain in the ass. The first person to be put against the
wall
Abbott when the revolution comes.

Regards,

Bruce

[Bruce_Gregory]

[From Bruce Gregory (980309.1727 EST)]

Bruce Abbott (980309.1700 EST)

Bruce A smug pain in the ass. The first person to be put against the
Abbott wall when the revolution comes.

Damn. Do we have to wait that long?

Bruce

[R_Marken]

[From Rick Marken (980309.1600)]

Bruce Abbott (980309.1700 EST) --

Below is my stab at some PCT definitions. Comments welcome, as
always.

An excellent job, Bruce!

Here are some suggested changes:

disturbance The influence of a disturbance on the controlled variable.
influence Basically, what effect the disturbance would have on the
            CV if the feedback loop were to be cut.

I don't think so. It's just the effect of the disturbance on the CV;
the output has a simultaneous effect on the value of the controlled
variable so you can't actually observe the influence of the
disturbance but you can calculate it if you know the effect of the
output on the cv.

input An input variable to a perceptual input function. In a
quantity control system, this variable represents some aspect of
             an environmental variable, such as its amount or its

               rate of change. Symbolized as qi.

The input quantity is not the _input_ to a perceptual function. It
is itself a function of a set of environmental variables (v's) that
corresponds to the perceptual fucntion that produces the perceptual
signal that is controlled by the control system.

             In a control system, the output is feed back via the
             output function to the controlled variable.

No. The output is fed back via the feedback function. The output
function converts the error signal into an output variable.

controlled That aspect of the environment external to the control
variable system which is acted upon by the output (via
              the environmental feedback function) and disturbance
             (via a disturbance function) and which is represented
             to the control system as the input quantity.

Not quite. It's better to give the cv the same definition (mine) as
the one for the input quantity. What is "acted upon" by the output
may be some environmental component of the cv that is completely
independent of the component influenced by d. I think your definition
could lead people to think that a controlled variable is always like
the knot in the rubber band demo; it's (k)not;-)

input Mechanism that transduces the input quantity
function (environment) to a perceptual signal (internal
             representation).

This definition gives the impression that perceptual functions
are transducers, converting some variable in the environment (qi)
into a percpetual signal (p). According to PCT, it doesn't work
that way. Input functions construct or define perceptions as a
function of sets of environmental variables (v's). There is no "input
quantity" that is the input to an input funtion; the input quantity
is _defined_ by the nature of the input (perceptual) function.

words Vehicles for misunderstanding.

And understanding.

Bruce A smug pain in the ass. The first person to be put
Abbott against the wall when the revolution comes.

Smug, yes. But not much of a pain. Rather fun, I think.

According to Abbott, PCT isn't revolutionary; conventional psychology
knows all about control and controlled variables already. So the
only reason I could see for putting Abbott up against a wall is
if it's a wall papered with all the psychology articles containing
the term "controlled variable". If the wall is covered with such
articles than Bruce will know that the revolution has, indeed,
happened. Otherwise I think it will slip right past him.

Best

Rick

···

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

[bnhpct]

[From Bruce Nevin (980309.2239)]

Bruce Abbott (980309.1700 EST)]

Great job! I appreciate your perseverance and fortitude.

Here's what I would do with the diagram, elaborating a bit on Bill Powers
(980308.1620 MST):

                              > r
                              >
                     ------->COMP ---------> e
                    > >
                    > p = v1 + v2 - v3 | [CONTROL SYSTEM]
                    > (for example) |
  - - - - - - Fi - - - - - - - - Fo - - - - - -
                  / | \ |
                 / | \ |
                > > > >
          Fd1 | | | Fe1 |
--> d1-------> v1<-|-- | -------------- |

        Fd2 | | | \ |
   d2-------- |-->v2 | -o- [ENVIRONMENT]
        Fd3 | | | Fe3 /
   d3 ------- | --|-> v3 <-------------
              > > >
              > > >
               \ | /
                \ | /

Fo'- - - - - -Fi' - - - - - - - - - - - -

                  >
                  > qi' = v1 + v2 - v3 [OBSERVER]
                  > (for example)
                  >

e' <----- COMP' ----
           >
           > r' This or qi' is the CV, CEV, CCEV, etc.

Note: this observer is not taking into account d2 and d3.

Question: Is it possible for the net effect of v1, v2, and v3 to be
controlled without an explicit Fe2 to v2? For example, v1 is target, v3 is
cursor, v2 is a sleeve that snags on something. I think no. But only Fe1
and Fe3 are metered. They too are perceptions controlled by the observer,
no? But because d2 is unidentified and unmetered it is not in the diagram
above. If we're measuring mouse movements instead of muscle effort it's OK
to leave it out. This demonstrates that the set of identified d, v, Fe() is
a construct in the observer's perceptual world, perhaps in the
participant's as well (e.g. mouse and target cursor). It's virtually
certain that more disturbance and control goes than is noticed.

I think some changes in your definitions would follow from the changed
diagram, for example

input An input variable to a perceptual input function. In a
quantity control system, this variable represents some aspect of an
            environmental variable, such as its amount or its rate of
            change. Symbolized as qi.

How's this for a quick try:
The observer's perception of what the net input of environment variables to
a perceptual input function must be. Corresponds to p in the observed
control system. Represented as qi (input quantity) in a model. May be
thought of as a controlled variable (CV) or controlled complex variable
(CCV) in the environment (CEV, CCEV), but this is a projection of the
observer's p (or qi) into the environment.

  Bruce Nevin

[Bruce_Abbott]

[From Bruce Abbott (980309.1150 EST)]

Rick Marken (980309.1600) --

Bruce Abbott (980309.1700 EST)

Here are some suggested changes:

disturbance The influence of a disturbance on the controlled variable.
influence Basically, what effect the disturbance would have on the
            CV if the feedback loop were to be cut.

I don't think so. It's just the effect of the disturbance on the CV;
the output has a simultaneous effect on the value of the controlled
variable so you can't actually observe the influence of the
disturbance but you can calculate it if you know the effect of the
output on the cv.

If the control system is doing its job, there is little effect of the
disturbance on the cv, so this can't be it. It seems to me that F(d) gives
the independent influence of d on the cv, just as the environmental feedback
function gives the independent influence of the output on the cv. These are
the changes in the cv that would be produced by one of the two variables (d
or o) if the other influence were absent. Imagine trying to hold your
finger over a spot on a table while I push against the side of your hand.
If you are successful, your finger remains over the spot, but meanwhile the
side of your hand is being compressed by the force we are exerting against
one another as I push and you push back. We could measure that force and
determine what effect it would have on your finger position if you were not
resisting.

input An input variable to a perceptual input function. In a
quantity control system, this variable represents some aspect of
             an environmental variable, such as its amount or its

              rate of change. Symbolized as qi.

The input quantity is not the _input_ to a perceptual function. It
is itself a function of a set of environmental variables (v's) that
corresponds to the perceptual fucntion that produces the perceptual
signal that is controlled by the control system.

I didn't say anything about the input quantity being a scalar. It could be
a vector.

             In a control system, the output is feed back via the
             output function to the controlled variable.

No. The output is fed back via the feedback function. The output
function converts the error signal into an output variable.

Of course. I find myself making more these errors of inattention more as I
get older. I meant to type environment function, in line with the diagram I
provided.

controlled That aspect of the environment external to the control
variable system which is acted upon by the output (via
              the environmental feedback function) and disturbance
             (via a disturbance function) and which is represented
             to the control system as the input quantity.

Not quite. It's better to give the cv the same definition (mine) as
the one for the input quantity. What is "acted upon" by the output
may be some environmental component of the cv that is completely
independent of the component influenced by d. I think your definition
could lead people to think that a controlled variable is always like
the knot in the rubber band demo; it's (k)not;-)

(Note that I got the bit about the environmental feedback function right
this time.) There may be some use in distinguishing them. The height of
water in a tank is not the same as an electrical reading giving the height
of the water.

input Mechanism that transduces the input quantity
function (environment) to a perceptual signal (internal
             representation).

This definition gives the impression that perceptual functions
are transducers, converting some variable in the environment (qi)
into a percpetual signal (p). According to PCT, it doesn't work
that way. Input functions construct or define perceptions as a
function of sets of environmental variables (v's). There is no "input
quantity" that is the input to an input funtion; the input quantity
is _defined_ by the nature of the input (perceptual) function.

Your point is well taken, but a transducer simply changes variation in one
variable (or set of variables) into variation in another. It seems to me
that that's what the input function does, even in the typical case, in which
the perception corresponds to no particular variable in the environment. As
I noted earlier, the input quantity is not necessarily a scalar.

According to Abbott, PCT isn't revolutionary; conventional psychology
knows all about control and controlled variables already. So the
only reason I could see for putting Abbott up against a wall is
if it's a wall papered with all the psychology articles containing
the term "controlled variable". If the wall is covered with such
articles than Bruce will know that the revolution has, indeed,
happened. Otherwise I think it will slip right past him.

I don't recall ever making the claim that conventional psychology knows all
about control and controlled variables already. It is, I suppose, a
perception of yours which you have constructed out of inferences based on
things I actually have said, but it doesn't necessarily correspond to
anything actually out there in the environment.

Regards,

Bruce

[Bruce_Abbott]

[From Bruce Abbott (980310.0025 EST)]

Bruce Nevin (980309.2239) --

I think some changes in your definitions would follow from the changed
diagram, for example

input An input variable to a perceptual input function. In a
quantity control system, this variable represents some aspect of an
            environmental variable, such as its amount or its rate of
            change. Symbolized as qi.

How's this for a quick try:
The observer's perception of what the net input of environment variables to
a perceptual input function must be. Corresponds to p in the observed
control system. Represented as qi (input quantity) in a model. May be
thought of as a controlled variable (CV) or controlled complex variable
(CCV) in the environment (CEV, CCEV), but this is a projection of the
observer's p (or qi) into the environment.

Your definition makes good sense when the idea is to call attention to what
the analyst of the control system perceives as the input quantity. It makes
less sense when the idea is to label the input to the PIF of a control
system. qi is the input to the control system's PIF, not the observer's
idea of what the input is. The control system doesn't care what the
observer thinks its inputs are; it will work the same regardless.

What do you mean by "corresponds to p" in your definition? If i is a sound
wave varying in frequency, and p is a varying pattern of activity in the
nervous system, they are obviously not the same thing.

By the way, in your diagram and in a previous post you place the reference
signal r "inside" the control system. It may be inside the organism, but
that is not the same as being inside the system. In HPCT the output of a
control system at one level (located in the environment of that system) is
also the reference signal of the control system beneath it. Like qi, r is
an input to the system and should properly be located in its environment.

Regards,

Bruce

[R_Marken]

[From Rick Marken (980309.2220)]

Bruce Abbott (980310.0025 EST) to Bruce Nevin (980309.2239) --

qi is the input to the control system's PIF, not the observer's
idea of what the input is.

Actually, qi is _precisely_ the observer's idea of what the
controlled perceptual input is; it is _not_ the input to the
perceptual function.

The control system doesn't care what the observer thinks its
inputs are; it will work the same regardless.

That's not the point. The point is that the input quantity is
a perception in the observer that corresponds to the perception
controlled by the controller. Suppose that the controller is
controlling for a perception of "squareness" (trying to keep that
perception at "square"). This perception is proportional to the
difference between the width (x) and height (y) of a polygon:
p = f (x-y) (this perception is at the reference when x-y = 0).
The input quantity is the same function of x and y as computed
by the observer_; using my suggested notation, p' = f (x-y),
where p' is the input quantity.

So when an observer is able to perceive (or compute) the same
function of input variables (like x and y) that is perceived and
controlled by the controller, then the observer has identified
the _controlled perception_.

···

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

[Bruce_Gregory]

[From Bruce Gregory (980310.1037 EST)]

Bruce Abbott (980309.1150 EST)]

>Rick Marken (980309.1600) --

>According to Abbott, PCT isn't revolutionary; conventional psychology
>knows all about control and controlled variables already. So the
>only reason I could see for putting Abbott up against a wall is
>if it's a wall papered with all the psychology articles containing
>the term "controlled variable". If the wall is covered with such
>articles than Bruce will know that the revolution has, indeed,
>happened. Otherwise I think it will slip right past him.

I don't recall ever making the claim that conventional psychology knows all
about control and controlled variables already. It is, I suppose, a
perception of yours which you have constructed out of inferences based on
things I actually have said, but it doesn't necessarily correspond to
anything actually out there in the environment.

I think we might all be better off if we gave up thinking that
anything is "actually out there in the environment",at least
until God let's us know exactly what She is perceiving. You show
up in Rick's world the way he describes you. (You know how you
show up in mine, but I've already apologized for that...)
Arguing with Rick is not going to change his world as I'm sure
you have already noticed. But this exchange shows that all is
not lost...

Bruce

[Bill_Powers1]

[From Bill Powers (980310.0825 MST)]

Bruce Abbott (980310.0025 EST)--

Bruce Nevin (980309.2239) --

Your definition makes good sense when the idea is to call attention to what
the analyst of the control system perceives as the input quantity. It makes
less sense when the idea is to label the input to the PIF of a control
system. qi is the input to the control system's PIF, not the observer's
idea of what the input is. The control system doesn't care what the
observer thinks its inputs are; it will work the same regardless.

The point here is that the environmental input to the PIF is _in general_
multiple, and _in general_ not linearly related to the perceptual signal.
So there is not in general a single input quantity that corresponds
linearly to the perceptual signal, as your definition requires.

The example I used for the new diagram shows a fairly general case. In the
environment we have two positions, target and cursor, which enter the PIF
to generate the "distance" perception. I assumed a linear PIF, so that
question didn't arise, although it could. Clearly there is no "qi" in the
environment of which p is a direct representation. If the observer looks at
the target and cursor and see the same "distance" perception that the other
system sees, that is because the observer is applying the same input
function to the two environmental variables, t and c. And this is in fact
how we use the idea of an input quantity in our models: we say that p is
the perception of the "real" distance which we call qi. This is very
convenient for setting up equations. But it is a fiction, because in the
environment there are only the two positions, target and cursor. There is
no "distance" in the environment.

There are cases in which a single environmental variable corresponds to a
single perception. But a general definition has to apply in all cases, not
just one special case.

What do you mean by "corresponds to p" in your definition? If i is a sound
wave varying in frequency, and p is a varying pattern of activity in the
nervous system, they are obviously not the same thing.

The fact that you refer to the sound wave as varying "in frequency" shows
that a perceptual interpretation has already been applied. The sound-wave,
in physical terms, consists of air pressures that exist at a given moment.
But the perception of pitch or frequency is something that remains constant
while the pressures vary. So when we speak of the sound waves in terms of a
stationary measure like frequency, we are projecting a perceptual
transformation onto the external world. If the "frequency" of the
sound-waves is constant, this does not mean that there is a constant
pressure at the sensory interface; on the contrary, the pressure must be
continually fluctuating. It is the perceptual input function that converts
these continual fluctuations into a steady perceptual signal. When we
define qi as the frequency of the sound-waves, we're defining it in terms
of a perceptual signal, not the objective sound-wave.

···

-----------------------------------------------------------
Before anyone gets in too deep here, I must remind us all that exactly the
same problem exists when we look too closely at those environmental
variables at the input of the PIF. I used target position and cursor
position as the environmental variables in my example. But what is
"position" but another kind of perception? The perceived "position" of the
cursor, for example, depends on other perceptions like the positions of
background objects, the way shadings and edges appear, shapes and
curvatures. Exactly what is the "position" of a cursor that is 1 cm long
vertically and 0.2 cm wide? It depends on how our visual systems perceive
it, not on any God-given definition of position. As usual, we're looking
down the hierarchy of perception, and nowhere is the real environment to be
seen.

What it comes down to is that whatever level we're looking at, or through,
when we try to define an input quantity we're talking about our own
perceptions. Because it's ALL perception -- no time-outs or waivers for
especially objective people like scientists or bookies.

Best,

Bill P.

[Bruce_Abbott]

[From Bruce Abbott (980310.1125 EST)]

Rick Marken (980309.2220) --

Bruce Abbott (980310.0025 EST)

qi is the input to the control system's PIF, not the observer's
idea of what the input is.

Actually, qi is _precisely_ the observer's idea of what the
controlled perceptual input is; it is _not_ the input to the
perceptual function.

O.K., I'll make a note of that. In PCT the input quantity to a control
system is whatever the observer thinks it is. It is _not_ the input to the
perceptual function.
                                         r
                                         >
                                 p V e
                        [PIF]-------->[COMP]---------->[OPF]
                          ^ |
                        z | | qo
                          +--------[CV]<----[EFF]--------+
                                     ^
                                     >
                                     d
As it is _not_ the input to the perceptual input function, I'll need another
symbol to indicate that input; for now I've just called it "z." I'm not
sure where qi goes. Somewhere in the brain of the observer, I'd guess, but
that doesn't appear in my diagram.

Regards,

Bruce

[Bruce_Gregory]

[From Bruce Gregory (980310.1216 EST)]

Bruce Abbott (980310.1125 EST)

>Rick Marken (980309.2220) --

>>Bruce Abbott (980310.0025 EST)

>> qi is the input to the control system's PIF, not the observer's
>> idea of what the input is.

>Actually, qi is _precisely_ the observer's idea of what the
>controlled perceptual input is; it is _not_ the input to the
>perceptual function.

O.K., I'll make a note of that. In PCT the input quantity to a control
system is whatever the observer thinks it is. It is _not_ the input to the
perceptual function.
                                         r
                                         >
                                 p V e
                        [PIF]-------->[COMP]---------->[OPF]
                          ^ |
                        z | | qo
                          +--------[CV]<----[EFF]--------+
                                     ^
                                     >
                                     d
As it is _not_ the input to the perceptual input function, I'll need another
symbol to indicate that input; for now I've just called it "z." I'm not
sure where qi goes. Somewhere in the brain of the observer, I'd guess, but
that doesn't appear in my diagram.

Your accommodating nature is always appreciated. In this case,
however, it is unnecessary. The important point is that the
diagram is drawn by the observer. The elements it contains all
exist in his/her perceptual world.

Bruce

[Bruce_Abbott]

[From Bruce Abbott (980310.1220 EST)]

Bill Powers (980310.0825 MST) --

Bruce Abbott (980310.0025 EST)

The point here is that the environmental input to the PIF is _in general_
multiple, and _in general_ not linearly related to the perceptual signal.
So there is not in general a single input quantity that corresponds
linearly to the perceptual signal, as your definition requires.

Here's the definition I offered:

input An input variable to a perceptual input function. In a
quantity control system, this variable represents some aspect of an
            environmental variable, such as its amount or its rate of
            change. Symbolized as qi.

As I noted to Bruce Nevin, my definition does not state that qi is a scalar
quantity. Vectors are welcome. (Alternatively, there could be many scalar
qi's, each arising from a different aspect of the environment, i.e.,
multiple inputs to the PIF. But this is just another way of saying that qi
is a vector.) Secondly, I see nowhere in the definition any statement
suggesting that qi be _linearly_ related to the perceptual signal. Your
assertions that my definition requires a single input quantity and a linear
relation to the perceptual signal thus are seen to be incorrect.

Clearly there is no "qi" in the
environment of which p is a direct representation.

I agree with your position. I have not defined qi as a single environmental
variable of which p is a direct representation. However, if you think the
definition unclear in this respect, I'd be happy to accept a clearer substitute.

The fact that you refer to the sound wave as varying "in frequency" shows
that a perceptual interpretation has already been applied. The sound-wave,
in physical terms, consists of air pressures that exist at a given moment.

I'm using accepted terminology from physics. I recognize that our
understanding of these things is all based on perception (we can't know
physical reality), but I don't see that this buys us anything in terms of
understanding how control systems, or perceptual input functions, work. In
the last analysis, everything we think we know about the world external to
our direct perceptual experience is only inference. I am not arguing
against this view, but am questioning its utility in describing the elements
of control systems as entities.

Regards,

Bruce

[R_Marken]

[From Rick Marken (980310.0930)]

Bruce Abbott (980310.1125 EST) --

In PCT the input quantity to a control system is whatever the
observer thinks it is.

Not quite. The input quantity is a perception in the observer
that corresponds to the perception controlled by the controller.
The input quantity is not "whatever the observer thinks it is".
If the observer thinks that his perception of the length of a
side (x) of a quadrilateral is the controlled perception when
the controller is actually controlling a perception of x-y,
then this perception in the observer _is not_ the input quantity
because it does not correspond to the perception controlled by
the controller. The input quantity is not what you disparagingly
refer to as "whatever the observer thinks it is"; it is an
analog of the controlled perception. An observer can only
perceive the input quantity (controlled perceeption) after he
has successully performed The Test for the Controlled Variable.

The input quantity is not something "out there" in the world
waiting to be noticed by an observer; it is a perceptual aspect of
whatever is "out there", and the only way for an observer to find
that perception (the input quantity) is by Testing to determine
which of his own perceptions corresponds to the perception controlled
by the system under study. This is why conventional psychologists
have never seen an "input quantity"; they have never Tested to
determine which of their own perceptions corresponds to the
perception controlled by the organism under study. This is why we
are harping on this apparently trivial point about the input quantity.

We draw the input quantity in our control diagrams as though it
were an objective aspect of the world. This is a convenient
_fiction_ that allows us to talk about controlled variables as
the "correlates" of controlled perceptions. But it is a fiction
that might also lead the unsuspecting (or those with a particular
agenda) to imagine that controlled variables can be "found" by
simply looking in the controller's environment. In fact, controlled
variables (or input quantities) can _only_ be "found" by systematically
Testing to determine which of one's own perceptions correspond
to the perception controlled by the controller.

This Testing cannot be done by simply observing disturbance-output
relationships because any particular disturbance-output relationship
could result from control of any one of many _different_ input
quantities. This is why I keep saying that the results of
conventional psychological research provide, at _best_, the
starting hypothesis for appropriately conducted PCT research.
This apparently esoteric discussion of the nature of the input
quantity in PCT really gets to the heart of why conventional
approaches to psychological research are useless; it's because
qi is not a variable in the environment; it's a variable in the
observer's brain; it's a perception that can only be discovered
by doing the Test.

···

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

[Bruce_Abbott]

[From Bruce Abbott (980310.1300 EST)]

Bruce Gregory (980310.1216 EST) --

Your accommodating nature is always appreciated. In this case,
however, it is unnecessary. The important point is that the
diagram is drawn by the observer. The elements it contains all
exist in his/her perceptual world.

Of course it is drawn by the observer. And subject to revision given new
evidence.

Implications, please. Does this change the way we do science?

Regards,

Bruce

[Bruce_Gregory]

[From Bruce Gregory (980310.1307 EST)]

Bruce Abbott (980310.1300 EST)]

>Bruce Gregory (980310.1216 EST) --

>Your accommodating nature is always appreciated. In this case,
>however, it is unnecessary. The important point is that the
>diagram is drawn by the observer. The elements it contains all
>exist in his/her perceptual world.

Of course it is drawn by the observer. And subject to revision given new
evidence.

Implications, please. Does this change the way we do science?

Indeed it does. See Rick's post (980310.0930)

Bruce

[bnhpct]

[From Bruce Nevin (980310.1330)]

Bruce Abbott (980310.0025 EST)--

in your diagram and in a previous post you place the reference
signal r "inside" the control system. It may be inside the organism, but
that is not the same as being inside the system. In HPCT the output of a
control system at one level (located in the environment of that system) is
also the reference signal of the control system beneath it. Like qi, r is
an input to the system and should properly be located in its environment.

Depends. If you're talking about a hierarchical control system the diagram
is truncated at the top. If you're talking about an elementary single-loop
control system (ECS) I'm not sure you should call the origin of r an
environment of the ECS; and if you do, it's not the same environment. If it
were the same environment, then the higher level of control for the
observed control system could have a direct connection (through the
environment) to the higher level of control for the observer. This may be
our fond illusion (as in the projection of qi into the environment), but it
ain't so.

Looking just at one ECS, r is inside the ECS even if its origin is not.
There's presumably a function that typically unites multiple signals into a
single reference signal r. As with Fi() and Fo(), Fr() would be at the
boundary with the next-higher level of the hierarchy.

                          e1 e2 e3
                           \ | / [NEXT LEVEL OF CONTROL]
                            \ | /
- - - - - - - - - - Fr - - - - - - - - - - -
                              >
                              > r
                              >
                     ------->COMP ---------> e
                    > >
                    > p = v1 + v2 - v3 | [CONTROL SYSTEM]
                    > (for example) |
  - - - - - - Fi - - - - - - - - Fo - - - - - -
                  / | \ |
                 / | \ |
                > > > >
          Fd1 | | | Fe1 |
--> d1-------> v1<-|-- |---------------- |

        Fd2 | | | \ |
   d2-------- |-->v2 | -o- [ENVIRONMENT]
        Fd3 | | | Fe3 /
   d3 ------- | --|-> v3 <-------------
              > > >
              > > >
               \ | /
                \ | /

Fo'- - - - - -Fi' - - - - - - - - - - - - - -

                  >
                  > qi' = v1 + v2 - v3 [OBSERVER]
                  > (for example)
                  >

e' <----- COMP' ----
           >
           > r' This or qi' is the CV, CEV, CCEV, etc.
           >
  - - - Fr' - - - - - - - - - - - - - - - - -
         / | \
        / | \ [NEXT LEVEL OF CONTROL]

The loop function Fr has multiple error signals e on one side and one
reference signal r on the other, and the function itself handles the
branching. In the interface with the environment through Fi and Fo, it is
physics that accounts for the branching. (Can there be multiple sensors
connecting to one lowest-level Fi?)

What if the comparator COMP is not at the lowest level of the hierarchy?
Other than adding more levels explicitly, I can see three ways to
reinterpret this diagram. Fi could be taken as representing one or more
lower levels of hierarchical control connected eventually to the immediate
input function of the level depicted. Or the environment could be taken as
including the lower levels. Or the environment could be considered as
though it consisted of only the immediately connected lower levels of
perception.

Would it be possible to model higher levels of control in this way?
Simulate an environment consisting of perceptual signals from lower levels?
The effect would be as though the lowest levels, those not directly
connected to Fi but indirectly connected through other ECSs, were in
imagination mode.

  Bruce Nevin

[Bruce_Abbott]

[From Bruce Abbott (980310.1350 EST)]

Rick Marken (980310.0930) --

Bruce Abbott (980310.1125 EST), paraphrasing Rick:

In PCT the input quantity to a control system is whatever the
observer thinks it is.

Not quite. The input quantity is a perception in the observer
that corresponds to the perception controlled by the controller.

Not quite. It is a perception in the observer of a correspondence between
the observer's perception of the input quantity and the observer's
perception of the perception controlled by the controller as the observer
perceives the controller.

Just where is this getting us, except mired ever more deeply in confusion?

This apparently esoteric discussion of the nature of the input
quantity in PCT really gets to the heart of why conventional
approaches to psychological research are useless; it's because
qi is not a variable in the environment; it's a variable in the
observer's brain; it's a perception that can only be discovered
by doing the Test.

I'm talking about what qi represents in the control system diagram, and that
certainly isn't something that exists only in the observer's brain.

Regards,

Bruce

[Bill_Powers1]

[From Bill Powers (980310.1213 MST)]

Bruce Abbott (980310.1125 EST)--

O.K., I'll make a note of that. In PCT the input quantity to a control
system is whatever the observer thinks it is. It is _not_ the input to the
perceptual function.
                                        r
                                        >
                                p V e
                       [PIF]-------->[COMP]---------->[OPF]
                         ^ |
                       z | | qo
                         +--------[CV]<----[EFF]--------+
                                    ^
                                    >
                                    d
As it is _not_ the input to the perceptual input function, I'll need another
symbol to indicate that input; for now I've just called it "z." I'm not
sure where qi goes. Somewhere in the brain of the observer, I'd guess, but
that doesn't appear in my diagram.

The input to the perceptual function is the set of signals that enters the
perceptual function -- in most cases, multiple signals. What the observer
sees is the collection of signals, interpreted as the observer habitually
interprets such things. If the observer sees a target and a cursor, the
observer is likely also to see something called the "distance between
target and cursor." If the positions are changing, the observer is likely
to see "motion."

These perceptions are not constructed as a whim or simply for the purposes
of some experiment. They are the way reality looks to the observer, and the
observer will quite likely defend their reality very energetically,
especially if not committed to solving the implied problems. The observer
may look at the target and the cursor, and say "You'll never convince me
that there's no distance between them!" Of course the person doing the
tracking task would be likely to agree, and point out that since they can
_both_ see it, it must be real. Heck, they could throw a party and ask all
their buddies in, and show that _everybody_ is convinced that there's
really a distance there between the target and the cursor.

But it's still a perception. We all see it because we perceive similarly.
And nobody can walk up to the monitor and put a finger on the distance
between target and cursor, because it's not there. It's in the perceptions
of the observers.

Here one way I could redraw your diagram:

                                         r
                                         >
                                 p V e
                        [PIF]-------->[COMP]---------->[OPF]
                          >>> >
                  ^ / |\ z <------EFF----> /|\
                  > z | z z z z | z
                  > \/ \ z---< z z \--z-----
                  qi z/ z \z \ z z
                  > z/ z\ z [Fd] <----qo--->
                  v z\ z z
                                  \z z \z
                                 <---d----->

Now all the z's are the individual variables in the environment of the
control system. The observer can draw fuzzy circles around subsets of the
z's next to the labels qi,d, Fd, EFF, and qo, and construct his own
higher-level perception of the environment, with nice simple variables in
it related by nice simple functions. The problem, of course, is to
construct this environment-model so the observer sees it the same way the
control system does. That is what the Test is for -- to see if the
observer's version of the environment is reasonably close to that of the
control system. When all that has been done, we end up with the observer's
model:

                                         r
                                         >
                                 p V e
                        [PIF]-------->[COMP]---------->[OPF]
                          > >
                          qi <-------- EFF <-------------qo
                          >
                         Fd
                          >
                          d

Actually, qi, as I indicated in my own diagram of yesterday, is not in the
environment at all; it's in the observer's head, as are all the other
functions and variables. But the observer projects his perceptions into his
model of the environment and places it all outside himself, so the above
diagram is how s/he sees it. The "reality" is more like the upper of the
last two diagrams -- it is what would be seen if one examined such things
as d, qi, and qo more closely, or from a lower-level point of view.

Which of these diagrams one uses is partly a matter how much information is
available about the nature of the environment, and how complex we want the
definition of the PIF to be. In the last diagram the PIF is just a
multiplier of 1. In my diagram of the tracking experiment, we can handle a
bit more complexity, so we can show both the cursor and the target as
environmental z's, or v's. I think it's unlikely that we would ever want to
go into the kind of detail in the middle diagram, although 30 years from
now, who knows what we will be prepared to do?

Best,

Bill P.

[Bill_Powers1]

[From Bill Power (980310.1301 MST)]

Bruce Abbott (980310.1220 EST)--

As I noted to Bruce Nevin, my definition does not state that qi is a scalar
quantity. Vectors are welcome. (Alternatively, there could be many scalar
qi's, each arising from a different aspect of the environment, i.e.,
multiple inputs to the PIF. But this is just another way of saying that qi
is a vector.) Secondly, I see nowhere in the definition any statement
suggesting that qi be _linearly_ related to the perceptual signal. Your
assertions that my definition requires a single input quantity and a linear
relation to the perceptual signal thus are seen to be incorrect.

OK, I read too much into your diagram. I'm somewhat reluctant to admit that
qi, as a vector, is the same thing as my collection of v's. For one thing,
a vector is not just any random collection of variables; as I understanding
it, there's some ordering scheme in the list. Anyway, my point is that the
v's, or z's, are the underlying observations, while grouping them into a
set that is treated as a single entity is a perceptual artifact.

Clearly there is no "qi" in the
environment of which p is a direct representation.

I agree with your position. I have not defined qi as a single environmental
variable of which p is a direct representation. However, if you think the
definition unclear in this respect, I'd be happy to accept a clearer
substitute.

Nothing to offer just now.

The fact that you refer to the sound wave as varying "in frequency" shows
that a perceptual interpretation has already been applied. The sound-wave,
in physical terms, consists of air pressures that exist at a given moment.

I'm using accepted terminology from physics. I recognize that our
understanding of these things is all based on perception (we can't know
physical reality), but I don't see that this buys us anything in terms of
understanding how control systems, or perceptual input functions, work. In
the last analysis, everything we think we know about the world external to
our direct perceptual experience is only inference. I am not arguing
against this view, but am questioning its utility in describing the elements
of control systems as entities.

Its utility depends entirely on what you're trying to do. If you just want
to model a single control behavior in the simplest way, use the simplest
diagram. But if you want to start thinking about the hierarchy, I think you
always have to question the units into which we unthinkingly divide our
experience.

Best,

Bill P.

[R_Marken]

[From Rick Marken (80310.1215)]

Me:

The input quantity is a perception in the observer that corresponds
to the perception controlled by the controller.

Bruce Abbott (980310.1350 EST)--

Not quite. It is a perception in the observer of a correspondence
between the observer's perception of the input quantity and the
observer's perception of the perception controlled by the controller
as the observer perceives the controller.

Now you are just trying to redefine qi the way you need it to be
defined. You're getting colder!

Just where is this getting us, except mired ever more deeply in
confusion?

I'm afraid it is getting you mired more deeply in confusion. But
this is only happing because you are trying to defend the notion
that qi as an environmental input to the perceptual function. Since
I am not trying to defend this notion, I am finding this discussion
most enlightening and it's helping me clarify my understanding of
controlled variables (input quantities) and the value of the Test
in contrast to conventional behavioral methods (about which, not
surprisingly, you have thus far said nothing).

I'm talking about what qi represents in the control system
diagram, and that certainly isn't something that exists only
in the observer's brain.

Yes it is, although the computations that derive this perception
may not exist in the brain. For example, our perception of the
input quantity in a tracking task is a number that represents the
distance from target to cursor at various time instants. This
number is computed by processes external to the brain; but the
resulting perception exists only in the brain. The input quantity
(distance from target to cursor) doesn't exist in the environment;
it exists only as the output of a computation (a surrogate for the
observer's perceptual function) that takes two arguments (cursor
and target position) that are themselves perceptual representations
of what is "really" in the environment.

···

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