Attention and Control

[From Rick Marken (990311.0900)]

Bruce Gregory (990310.0940 EST)--

_If_ I were perceiving myself drinking coffee, the reference
signal would command the perception, but since I am _not_
drinking coffee, the reference signal is not playing an active
role.

The reference signal is still specifying (commanding) a value
of a perception of sweetness/bitterness. It's probably commanding
a value of zero sweetness/bitterness while you drive -- and it's
getting that perception as long as there is no disturbance to this
perception. If, while you are driving, a piece of organic matter
flies into your mouth, disturbing the perception of sweetness/
bitterness, you will probably salivate and (possibly) pull over
for a sip of water in order to bring the sweetness/ bitterness
perception back to its reference.

We control hundreds of perceptual variables simultaneously,
without any awareness (attention) that we are doing so.

Suppose we performed the Test. The car in front of me brakes
rapidly while I am tuning the radio. I collide with the car in
front. As an observer, you would conclude that I was _not_
controlling for distance, would you not?

That's true. As I said, while you are tuning the radio you
are not perceiving the distance variable, so you can't control
it at that time. The crash indicates that you were not
controlling visual distance at the time of the crash; but it
doesn't indicate that you had no reference for that perception.
Indeed, even if you are not injured in the crash I think you
would be upset by the crash because it represents an undesired
state of a controlled perception: "distance from car in front".
If you were really _not_ controlling for "distance from car in
front" the crash would not upset you at all (no error).

I'm willing to wager that your ability to track a target depends
critically on your paying attention to the task.

Why not do the research (or suggest an experiment) to test this?
But before you put a lot of money down on this wager I think you
should consider how many perceptual variables you are controlling
successfully at this very moment without paying any attention to
the fact that you are doing this (hint: your posture, the wetness
of your clothes, what you are looking at, where you are located,
etc).

So there are at least two types of tasks, those that require
attention and those that do not.

I think you are basing this on your own imagination. How
do you know that there are tasks that require attention and
tasks that don't? According to the PCT model all tasks (all
controlling -- from controlling intensities to controlling
system concepts) can be done without attention. What evidence
do you have that _any_ control task requires attention?

Best

Rick

···

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

[From Bruce Gregory (990311.1245 EST)]

Rick Marken (990311.0900)

I think you are basing this on your own imagination. How
do you know that there are tasks that require attention and
tasks that don't? According to the PCT model all tasks (all
controlling -- from controlling intensities to controlling
system concepts) can be done without attention. What evidence
do you have that _any_ control task requires attention?

Since PCT does not explicitly model attention, I shouldn't be surprised
at your question. I must admit, however, that I find it bizarre.

Bruce Gregory

[From Bruce Gregory (990311.1255 EST)]

Rick Marken (990311.0900)

> I think you are basing this on your own imagination. How
> do you know that there are tasks that require attention and
> tasks that don't? According to the PCT model all tasks (all
> controlling -- from controlling intensities to controlling
> system concepts) can be done without attention. What evidence
> do you have that _any_ control task requires attention?

Since PCT does not explicitly model attention, I shouldn't be
surprised
at your question. I must admit, however, that I find it bizarre.

p.s. What sort of evidence would you find persuasive?

Bruce Gregory

[From Bill Powers (990311.1111 MST)]
Bruce Gregory (990311.1245 EST)--

replying to Rick Marken (990311.0900)

Since PCT does not explicitly model attention, I shouldn't be surprised
at your question. I must admit, however, that I find it bizarre.

Good, that means that you have found something in your thinking
incompatible with Rick's PCT-based statement. The next step is to figure
out what your misunderstanding was.

Best,

Bill P.

[From Bruce Gregory (990311.1350 EST)]

Bill Powers (990311.1111 MST)

Bruce Gregory (990311.1245 EST)--

replying to Rick Marken (990311.0900)
>
>Since PCT does not explicitly model attention, I shouldn't
be surprised
>at your question. I must admit, however, that I find it bizarre.

Good, that means that you have found something in your thinking
incompatible with Rick's PCT-based statement. The next step
is to figure
out what your misunderstanding was.

If Rick gives me an example of the evidence he would find convincing,
perhaps I can make some progress. I'd welcome an example from you as
well. I had no trouble understanding your post (990311.1053 MST), by
the way.

Bruce Gregory

[From Rick Marken (990311.1130)]

Me:

What evidence do you have that _any_ control task requires
attention?

Bruce Gregory (990311.1255 EST) --

What sort of evidence would you find persuasive?

A good start would be data that cannot be explained by the
existing PCT model (which includes no attention mechanism).
What evidence persuaded you that there are tasks that require
attention and tasks that don't?

Best

Rick

···

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

[From Hank Folson 990311.2100)

Bruce Gregory (990311.1245 EST)

According to the PCT model all tasks (all
controlling -- from controlling intensities to controlling
system concepts) can be done without attention. What evidence
do you have that _any_ control task requires attention?

Since PCT does not explicitly model attention, I shouldn't be surprised
at your question. I must admit, however, that I find it bizarre.

There are many words like "behavior" and "attention" that we have heard
so often in our lives that we assume these words describe something that
is true, rather than an illusion. The problem is that these words were
developed without an understanding of how organisms really work (assuming
PCT is right and everything else is wrong).

So most believe that what we do is produce our "behaviors" in passive
response to environmental forces. "Attention" assumes that something is
turned off and on in our heads. In pre-PCT psychologies, I suppose this
means the environment is throwing the switch?

Now look at PCT: Rick Marken & Bill Powers have said (and probably
modeled) that there is no such thing as "attention". "Attention" is
another phlogiston in PCT. For "attention" (as classically defined) to
change from one thing to another, an on-off switch is required. I have
never seen such a switch proposed in any control system diagram.

Imagine that such switches do exist. We control thousands of variables.
Each would require an on-off "attention" switch. Under PCT, each switch
would require a control system above it to turn the switch on and off.
Would each higher system have an on-off switch, too? If so, there would
be an exponential number of systems and switches needed.

The only way to imagine a practical "attention" system would be to have
only one level that has on-off switches. But the level above would not,
so all those control system loops would always be on. If so, there is no
advantage to having the layer of systems with on-off switches. So I don't
think classic "attention" exists in PCT.

So how does PCT get along without "attention"? Does a phenomenon exist
that is mistakenly labeled as "attention"? Let's look at your adjusting
radio while driving example. Yes, your focus is on the radio, not the
road. But your control systems involved in your driving do not shut down.
My understanding of PCT is that a comparator(s) at a higher level is
comparing the parallel error signals related to traffic conditions and
those related to good music. If traffic perceptions are well under
control, the comparator will let you temporarily concentrate the vision
inputs at the radio dial. Hopefully, your systems are organized well
enough that only the output of one hand will be used to turn the knob.
Hopefully, your comparator will allow only a limited time of focusing on
the dial before returning the use of vision to the driving system inputs,
even if the perfect radio station has not been found to reduce that error
signal. Hopefully your systems are well organized so that other
perceptions of driving disturbances like brake lights in your peripheral
vision or the sound of brakes screeching will return the vision system to
driving related controlling.

No special "attention" system is needed. The sound of a horn will create
a big error signal in the driving system, and the higher comparator will
override the radio error signals, allowing you to use the visual input
for driving again. If there is no input perceived by the driving system
as you overtake the car ahead, you will crash. The PCT solution to this
problem is this: Limit distractions from driving to prudent small periods
during which the lack of control of the driving systems is unlikely to
lead to disaster. Turning the radio off is a simple and fast PCT correct
solution.

I find that the parallel-hierarchical arrangement PCT proposes is elegant.

···

------------

An unintended consequence: Verbal examples of how control systems work
show the limits, or at least dangers, of just talking about theory (and
using metaphors). While I don't get excited about making models myself, I
do think that they force us to be more thorough in our efforts to learn
PCT. (My verbal ramblings above can be very flawed.)

Sincerely,
Hank Folson

704 ELVIRA AVE. REDONDO BEACH CA 90277
Phone: 310-540-1552 Fax: 310-361-8202 Web Site: www.henryjames.com

[From Bruce Gregory (990312.0838 EST)]

Hank Folson 990311.2100

So how does PCT get along without "attention"? Does a phenomenon exist
that is mistakenly labeled as "attention"? Let's look at your adjusting
radio while driving example. Yes, your focus is on the radio, not the
road. But your control systems involved in your driving do not shut down.
My understanding of PCT is that a comparator(s) at a higher level is
comparing the parallel error signals related to traffic conditions and
those related to good music.

I was unaware of such a mechanism in PCT. Can you refer me to a description
in B:CP or elsewhere?

If traffic perceptions are well under
control, the comparator will let you temporarily concentrate the vision
inputs at the radio dial. Hopefully, your systems are organized well
enough that only the output of one hand will be used to turn the knob.
Hopefully, your comparator will allow only a limited time of focusing on
the dial before returning the use of vision to the driving system inputs,
even if the perfect radio station has not been found to reduce that error
signal. Hopefully your systems are well organized so that other
perceptions of driving disturbances like brake lights in your peripheral
vision or the sound of brakes screeching will return the vision system to
driving related controlling.

I do sense a lot of hope, but I do not understand the model you seem to have
in mind.

No special "attention" system is needed.

Perhaps you can explain the following. I am sitting at the keyboard looking
out at the snow falling. My "attention" is on what I will type next. What is
the proper PCT model for this situation?

Bruce Gregory

[From Rick Marken (990312.0900)]

Hank Folson (990311.2100) --

There are many words like "behavior" and "attention" that we
have heard so often in our lives that we assume these words
describe something that is true, rather than an illusion...

So most believe that what we do is produce our "behaviors" in
passive response to environmental forces. "Attention" assumes
that something is turned off and on in our heads

Now look at PCT: Rick Marken & Bill Powers have said (and probably
modeled) that there is no such thing as "attention". "Attention"
is another phlogiston in PCT.

I think you correctly note that the problem with "attention" is
the same as the problem with "behavior". But I think the problem
is not so much that these words refer to things that don't exist;
the problem is that it is not clear _to what_ these words do refer.
Also, the problem is that people assume that they _do_ know to
what these words refer.

A careful analysis shows that the word "behavior", for example,
is quite ambiguous. Once you take a close look at the various
phenomena to which the word "behavior" refers you see that this
word has been used to refer to what (in PCT) we see as several
different performance variables: output variables, controlled
variables and "side-effect" variables. So when we talk about
"behavior" in this group we are talking informally, knowing
that we are referring to all the observable variables involved
in the process of control.

We can still talk about "behavior" in this group (and in a post
PCT world). We just know that we are using this word to point to a
process: control of perception. When we want to be precise in our
discussion of control we know that we have to talk about output
(or action) variables, controlled (or perceptual) variables and
side-effect variables.

It's really the PCT model that has given precision to the
term "behavior". The same kind of precision has not been
given to the term "attention" because we have not tried to model
it in any detail. And we have not modeled it (I think) because
we don't yet know how to study whatever phenomena are being
pointed to by the word "attention". (But I think the clinical work
done on the "method of levels" counts as a start at research on
attention).

I think "attention" does refer to real behavioral phenomena
that are not yet captured by the current PCT model. I think
the word "attention" is pointing to aspects of "consciousness",
particularly "awareness" and the fact that awareness can be
"directed" at one thing and then another. Bill Powers did take
a cut at putting awareness (and volition) into the PCT model;
these efforts are described in B:CP (in the learning chapter,
I think). But so far no "hard" research has been done to test
this aspect of the model.

I'm sure that the term "attention" refers to important behavioral
phenomena. But so far, whatever contribution attention makes
to the process of control of perception has not shown up as
"unexplained variance" in our studies of control. The existing
control model explains 99% of the variance in the behavior we
have observed in our little tests of the PCT model. But I'm sure
one could design experiments where there would be unexplained
variance that could _only_ be accounted for by the addition of
some kind of attentional mechanism to the PCT model. Designing
an experiment that reveals the need to an attentional mechanism
is the "fertile ground for research" to which Bill Powers referred
in an earlier post.

Bruce Gregory (990312.0838 EST) --

Perhaps you can explain the following. I am sitting at the
keyboard looking out at the snow falling. My "attention" is
on what I will type next. What is the proper PCT model for
this situation?

There is only one PCT model; it is the one described in _Behavior:
The Control of Perception_. I think it would be a good exercise
for you try to figure out, on your own how, the PCT would explain
the situation you describe.

Best

Rick

···

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

[From Hank Folson (990312.0900)]

Bruce Gregory (990312.0838 EST)

I was unaware of such a mechanism in PCT. Can you refer me to a description
in B:CP or elsewhere?

I do sense a lot of hope, but I do not understand the model you seem to have
in mind.

No special "attention" system is needed.

Perhaps you can explain the following....

Sorry, Bruce. I do not know what variables you are trying to control
here. I do not have the time to run the Test For The Controlled Variable
on you to find out. It makes no PCT sense to respond directly to your
questions (outputs) which you (or any other living organism) have no real
interest in, as you are really trying to reduce internal error signals.
It would be a waste of your time and mine, to say nothing of the rest of
CSGnet, to address your variable outputs instead of your inner error
signals and reference levels. Of course, this is what is usually done on
CSGnet :slight_smile:

I will say this:
1. I may have assumed/imagined how we prioritize the controlling of
variables which are competing for limited sensor access. But it sounds
good to me. :slight_smile:

2. PCT is based on the concept of a parallel and series hierarchy of
control systems.

3. PCT requires the presence of Comparators.

4. Try making a model, or at least go through the precursor process of
exploring whether what I talked about fits or doesn't fit PCT.

Sincerely,
Hank Folson

704 ELVIRA AVE. REDONDO BEACH CA 90277
Phone: 310-540-1552 Fax: 310-361-8202 Web Site: www.henryjames.com

[From Bruce Gregory (990312.1236 EST)]

Rick Marken (990312.0900)

Bruce Gregory (990312.0838 EST) --

> Perhaps you can explain the following. I am sitting at the
> keyboard looking out at the snow falling. My "attention" is
> on what I will type next. What is the proper PCT model for
> this situation?

There is only one PCT model; it is the one described in _Behavior:
The Control of Perception_. I think it would be a good exercise
for you try to figure out, on your own how, the PCT would explain
the situation you describe.

Semantic issue here. When I ask for a PCT model I am asking for the
particular deployment of the principles laid out in B:CP to explain a given
situation. This requires one to identify the perceptual variable being
controlled, the reference level and its source, and the output that affects
the perceptual variable. Looking out the window is not obviously an example
of anything being controlled. Of course one might reason that since I am not
overtly acting, there is no error signal being generated. The PCT model
would apply to a perception related to what I will type. Presumably this is
a controlled perception. The only reason I would begin to type would be an
error associated with the perception "I am typing". Presumable a higher
level control system would have to issue a command "perceive yourself
typing" in order for this error signal to arise. This higher level control
system would only issue such a command because a yet higher level system
issued a command to it. And so on up the hierarchy until the command was
issued to decrease error in an intrinsic variable.

Bruce Gregory

[From Bruce Gregory (990312.1243 EST)]

Hank Folson (990312.0900)

4. Try making a model, or at least go through the precursor process of
exploring whether what I talked about fits or doesn't fit PCT.

It makes no PCT sense to respond directly to your
questions (outputs) which you (or any other living organism) have no real
interest in, as you are really trying to reduce internal error signals.

Bruce Gregory

[From Hank Folson (990312.1000)]

Rick Marken (990312.0900)

A careful analysis shows that the word "behavior", for example,
is quite ambiguous. Once you take a close look at the various
phenomena to which the word "behavior" refers you see that this
word has been used to refer to what (in PCT) we see as several
different performance variables: output variables, controlled
variables and "side-effect" variables. So when we talk about
"behavior" in this group we are talking informally, knowing
that we are referring to all the observable variables involved
in the process of control.

A modest Test For The Controlled Variable:

Rick, is your position that the following statement is consistent with
PCT: A stranger can simply state a diametrically opposed definition for
an important word in daily use for decades of an individual's life, and
the individual will have no trouble segregating the two uses of the same
word, and communications relating to the subject will be effective,
unconfused and unambiguous?

Sincerely,
Hank Folson

704 ELVIRA AVE. REDONDO BEACH CA 90277
Phone: 310-540-1552 Fax: 310-361-8202 Web Site: www.henryjames.com

[From Rick Marken (990312.1030)]

Bruce Gregory (990312.1236 EST)--

Looking out the window is not obviously an example of anything
being controlled.

I can think of several variables, some or all of which might
be controlled when "looking out the window" (I list different
possible values of each variable in parentheses to show that
these words refer to _variables_ -- aspects of experience that
can be in different states at different times):

1. Direction of gaze (at window vs at wall vs...)
2. What is seen (snow vs wall vs...)
3. Posture (standing vs sitting vs...)
4. Comfort (cozy vs freezing vs...)
5. Typing (not typing vs random pecks vs...)

The only reason I would begin to type would be an error associated
with the perception "I am typing". Presumable a higher level
control system would have to issue a command "perceive yourself
typing" in order for this error signal to arise. This higher
level control system would only issue such a command because a
yet higher level system issued a command to it. And so on up
the hierarchy until the command was issued to decrease error
in an intrinsic variable.

I think this is basically OK. The only little nit I have is
with the phrase "issue a command"; I think it's better to
think of reference signals as _always there_ and, thus,
always commanding _some_ perception. The higher level system
that uses "typing" as the means of controlling its perception
(the perception of a completed book, say) doesn't "issue a
command" and then "unissue" it. The "book control system" is
continuously _varying_ the reference (command) for typing
as the means of producing the book perception it wants;
sometimes its commanding a lot of typing, sometimes less
typing and sometimes no typing at all (so that undesired
sentences are not placed in the book).

Best

Rick

···

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

[From Rick Marken (990312.1040)]

Hank Folson (990312.1000)--

Rick, is your position that the following statement is consistent
with PCT: A stranger can simply state a diametrically opposed
definition for an important word in daily use for decades of an
individual's life, and the individual will have no trouble
segregating the two uses of the same word, and communications
relating to the subject will be effective, unconfused and
unambiguous?

I don't know whether the statement is consistent with PCT or
not. But if the "stranger" is a PCTer using the word "behavior"
and the "individual" is someone listening to the PCTer talk
about "behavior" then the individual will probably have some
considerable trouble "segregating the two uses of the same word"
(his own and the PCTer's use of "behavior") until he (the
individual) has really mastered control theory.

Best

Rick

···

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

[From Bruce Gregory (990312,1508)]

Rick Marken (990312.0900)

I'm sure that the term "attention" refers to important behavioral
phenomena. But so far, whatever contribution attention makes
to the process of control of perception has not shown up as
"unexplained variance" in our studies of control. The existing
control model explains 99% of the variance in the behavior we
have observed in our little tests of the PCT model. But I'm sure
one could design experiments where there would be unexplained
variance that could _only_ be accounted for by the addition of
some kind of attentional mechanism to the PCT model. Designing
an experiment that reveals the need to an attentional mechanism
is the "fertile ground for research" to which Bill Powers referred
in an earlier post.

Suppose I am tracking a target. My behavior is accurately modeled by PCT.
After a while I grow "bored" with the task. My "attention wanders" and my
control deteriorates. Can the PCT model predict how control will
deteriorate? This, I suggest, is your "unexplained variance". (Or the reason
why you never find this variance--you discard the data.)

Bruce Gregory

[From Rick Marken (990312.1400)]

Bruce Gregory (990312,1508)--

Suppose I am tracking a target. My behavior is accurately
modeled by PCT. After a while I grow "bored" with the task. My
"attention wanders" and my control deteriorates. Can the PCT
model predict how control will deteriorate? This, I suggest,
is your "unexplained variance". (Or the reason why you never
find this variance--you discard the data.)

These are good questions. People do sometimes stop doing
control tasks. I have had only a couple experiences with
this and in the one case I recall the person stopped, not
because he was bored or because his attention was wandering
but because the results seemed to be disturbing some higher
level perception (I was a conventional psychologist at the
time and the person being tested was a fellow conventional
psychologist).

I think the PCT model can, in principle, "predict how control
will deteriorate" in situations where a person does spontaneously
stop doing a control task. Again, my experience is that when
this does happen, control doesn't really "deteriorate"; it
stops (the RMS measure of control, for example, goes from some
low value, like 3 pixels, to some maximum value, like 100 pixels);
the person just let's go of the mouse, say.

There are several possible way to explain what might be going on
in this situation , at least one of which requires no change in
the model at all. One "no model change" explanation has the person
controlling two high level variables -- like "helping the
experimenter" and "going out to meet chicks". The person is
doing the control task to help out the experimenter but if
this goes on too long it's taking up time to be used meeting
chicks. So the longer the person does the control task, the
bigger the error in the meet chicks system; eventually, the
error in the meet chicks system forces the reference for
"do the control task" to zero; the person stops doing the
task and goes out to meet chicks. If he still wants to help
the experimenter he might cut his cavorting short and return
to the lab.

Of course, the person might be stopping the control task
because he's bored and his attention is wavering. But then
you've got to figure out how to put "boredom" and "attention"
into the PCT model and figure out experiments to show that
this explanation is better than the one based on conflict.

Best

Rick

···

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

[Martin Taylor 990312 17:08]

[From Rick Marken (990312.1400)]

... One "no model change" explanation has the person
controlling two high level variables -- like "helping the
experimenter" and "going out to meet chicks". The person is
doing the control task to help out the experimenter but if
this goes on too long it's taking up time to be used meeting
chicks. So the longer the person does the control task, the
bigger the error in the meet chicks system; eventually, the
error in the meet chicks system forces the reference for
"do the control task" to zero; the person stops doing the
task and goes out to meet chicks. If he still wants to help
the experimenter he might cut his cavorting short and return
to the lab.

How is this a "no model change" explanation? I've never heard of high
error in one of two conflicted control systems forcing the reference
to another to go to zero. "Do the control task" is not in the hierarchy
of lower-level perceptions that support "meet chicks" is it? How does
the output of "meet chicks" affect its reference value, then, especially
when there is a "please the experimenter" reference for a simultaneously
controlled perception of experimenter pleasure? "Do the control task"
is a reference value that may be the _only_ way to achieve "pleased
experimenter".

I think what you need is a way within the model to resolve conflicts
other than by reorganization. It can be done, but not within the standard
quasi-linear organization of the hierarchy we ordinarily assume, I think.

Martin

[From Rick Marken (990312.1500)]

Martin Taylor (990312 17:08) --

How is this a "no model change" explanation?

No new mechanisms are introduced.

I've never heard of high error in one of two conflicted control
systems forcing the reference to another to go to zero.

I think you could design a system that works this way. For
example, I think it would work this way if one of the
conflicting control systems were of much higher gain than
the other and subtracted from the reference signal value
determined by the outputs of the two.

"Do the control task" is not in the hierarchy of lower-level
perceptions that support "meet chicks" is it?

That's a good point. I suppose that the time spent in the
experiment would be a disturbance to this variable. So the
conflict situation I was describing is more like the one in
my "Cost of Conflict" demo at my my demo site.

Best

Rick

···

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

[From Bruce Gregory (990312.2110 EST)]

Rick Marken (990312.1400)]

I think the PCT model can, in principle, "predict how control
will deteriorate" in situations where a person does spontaneously
stop doing a control task. Again, my experience is that when
this does happen, control doesn't really "deteriorate"; it
stops (the RMS measure of control, for example, goes from some
low value, like 3 pixels, to some maximum value, like 100 pixels);
the person just let's go of the mouse, say.

But this does not describe what happens when I move my attention from the
road to tuning the radio. I do not let go of the steering wheel and "stop"
controlling. Do I?

There are several possible way to explain what might be going on
in this situation , at least one of which requires no change in
the model at all. One "no model change" explanation has the person
controlling two high level variables -- like "helping the
experimenter" and "going out to meet chicks". The person is
doing the control task to help out the experimenter but if
this goes on too long it's taking up time to be used meeting
chicks. So the longer the person does the control task, the
bigger the error in the meet chicks system; eventually, the
error in the meet chicks system forces the reference for
"do the control task" to zero; the person stops doing the
task and goes out to meet chicks. If he still wants to help
the experimenter he might cut his cavorting short and return
to the lab.

Would you apply the same interpretation to driving and tuning the radio?
Does an error in the tuning the radio control loop force the reference for
the driving control loop to zero? Sounds very unlikely to me. How is it that
I glance up even before I've found the station I'm looking for? Does the
reference for the driving control loop somehow recover and force the tuning
control loop reference to zero? What mechanism is responsible for this
"reversal of fortune"?

Of course, the person might be stopping the control task
because he's bored and his attention is wavering. But then
you've got to figure out how to put "boredom" and "attention"
into the PCT model and figure out experiments to show that
this explanation is better than the one based on conflict.

It seems rather limited to be require all shifts from one control system to
another to be the result of conflict, but if that's the PCT model, that's
the PCT model.

Bruce Gregory