actions/behaviour/output

[Martin_Taylor4]

Jeff Vancouver 980220.1442 EST to Bill Powers (980220.0354 MST)

Perhaps you can see now that action is also a relative term. The action of
one control system consists (at all but the first level) of specifying
reference levels for controlled perceptions at lower levels.

I am sorry, I thought actions were on the environment. Signals are passed
from the upper level to lower level units, not actions.

Think of an Elementary Control Unit (ECU). It consists _only_ of four
entities and their connections:

(1) a perceptual input function (PIF) that transforms the values from several
inputs into a single value called a "perceptual signal;" its inputs are
in most cased the perceptual signal values of lower-level ECUs.

(2) a reference input function (RIF) (seldom discussed) that transforms the
values from several inputs into a single value we call the "reference
signal;" its inputs are in most cases the output values of higher-level
ECUs.

(3) a comparator that takes the perceptual signal and the reference signal
as inputs and produces a single value called the "error signal."

(4) an output function (OF) that takes the error signal as input and produces
a single value we call the "output signal."

That's it for the _Elementary_ Control Unit. (More complex possibilities
include inputs that change the form or the parameters of any of the three
functions, and some models for adaptive perceptual control have used
such possibilities--e.g. Tom Bourbon's variable gain). Everything else
that happens when the ECU is active happens in its environment. That goes
equally for what happens elsewhere in the HPCT hierarchy both above and
below its own level, and for what happens in the world outside the
organism.

Outside the ECU, an external observer can see other things going on, other
signals in some ways of analyzing. But to the ECU, all that happens is
that the values input to the PIF and the RIF change, and as a consequence
the value of the output signal may change. If the ECU is actually controlling,
large changes in the value of the output signal can correspond to very small
changes in the value of the perceptual signal.

Why are there large changes in the output signal for small changes in the
perceptual signal (assuming the reference signal doesn't change)? One asks
the question this way only if one looks at the ECU in isolation from its
environment. The answer, of course, is "feedback." Somewhere outside the
ECU--in its environment--there is a connection that allows the output
signal to affect the inputs to the perceptual signal, and that connection
is made in such a way that it opposes whatever influence something else
(that we call a disturbing signal--the word "disturbance" has been
ambiguously used) has on the same or other inputs to the perceptual signal,
so that changes to the perceptual signal are minimized.

The output signal is a single time-varying value. Its variations have
consequences we call "actions," in various places. The "actions" are seen
by an outside observer. Where are they? Actions are the effects of the
output signal on some part of the environment of the ECU, not necessarily
of the organism. If the observer can look inside the hierarchy of ECUs,
as a simulation modeller can, then the immediate actions of the ECU are
its contributions to the reference input functions of several other ECUs.

But the observer can look elsewhere, specifically, outside the organism,
and can there also observe the effects of this output signal. When the
observer does that, the Test is available to check which controlled
perceptions are subject to the changed reference values that are the
"immediate actions" of the ECU of interest. All these different influences
on aspects of the outer world are "actions." They are actions that
represent the influences of controlled perceptions at many levels,
perceptions whose reference levels are influenced by the output of the
ECU we were considering. Those effects on reference levels are the
"immediate actions" of the ECU of interest, but all the different
observable effects on the world are its actions.

Rick's "umbrella in the rain" provides a good example. One uses an umbrella
as a way of satisfying a reference value of zero for the perception of
getting wet. If the weather was sunny and one was far from hoses and
sprinklers, one would probably not open the umbrella at all. But in using
the umbrella, as Rick pointed out, there are other controlled perceptions,
one of which was perception of torque at the wrist, with a reference value
of zero. By performing the Test and discovering that this is a controlled
perception, one also discovers that minimizing wrist torque is an action
of the control of the feeling wet perception with its zero reference.

A modeller can see "actions" at any level in a hierarchy. An external
observer can see influences on _his own_ perceptions (presumably uncontrolled
perceptions), and by analyzing these perceptions the external observer
can _infer_ the existence of actions internal to the person's hierarchy.

There was an article in "Science" a few years ago about this, by the way,
using as an example someone standing at a door. "What are you doing?"
I'm standing at the door; I'm pushing the button; I'm ringing the bell;
I'm visiting my aunt; I'm trying to get a good inheritance. (The detailed
examples are mine; the study had other examples, and was interested in the
effects of which level the subject was consciously focussed on).

Now "behaviour". This is a pretty generic term. "Behaviour is the control
of perception." This is the "purpose" view. It is what the actions are for,
and describes the whole system at work. "A control system with too high a
gain for its transport lag shows oscillatory behaviour" is a descriptive
use of "behaviour" as the observable effects of something on the environment.
"Behaviour is purposeful action" again refers to the observable influences,
but restricts them to those influences that form part of the environmental
feedback path, eliminating those that are only side-effects. One could
also use "behaviour" to refer to what I called "immediate actions."

There are other uses for the word "behaviour," but I think those are
the main ones used in technical discussions of PCT. They are similar enough
to cause confusion, and different enough that the confusions matter. I
suspect that it would be better to use "output" to refer to the specific
single value emitted by the output function of the ECU, and "actions" to
refer to the influences of the output on observable (or simulable) aspects
of the environment of the ECU, leaving "behaviour" out, except in colloquial
discussion.

···

--------------------
There are several other messages backed up awaiting comment, but I'm pretty
busy these days, and discussions that seem fruitless seem to get put off
more than cases like this, which seem to me not to be too controversial.

Martin

[R_Marken]

[From Rick Marken (980228.1020)]

Martin Taylor (980221) --

Nice post, Martin!

This part was particularly intriguing:

There was an article in "Science" a few years ago about this,
by the way, using as an example someone standing at a door...
the study had other examples, and was interested in the effects
of which level the subject was consciously focussed on).

This sure sounds like a VERY "PCT relevant" article. Could
you post the reference, please; I've love to see it. Thanks.

Bruce Gregory (980221.1250 EST)

My inclination is to say that OJ intended to kill his wife and
her friend and did not intend to leave his blood all over the
place.

Yes. I think Bill's [Bill Powers (980221.20200 MST)] recent
post on "responsibility" was (as usual) much better than
anything I said on the subject and it deals with something
I didn't touch on: the relationship between PCT and the most
common use of the word "responsible": "holding people responsible".

Best

Rick

···

--

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

[Jeff_Vancouver2]

[from Jeff Vancouver 980221.1440 EST]

Jeff Vancouver 980220.1442 EST to Bill Powers (980220.0354 MST)

Perhaps you can see now that action is also a relative term. The action of
one control system consists (at all but the first level) of specifying
reference levels for controlled perceptions at lower levels.

I am sorry, I thought actions were on the environment. Signals are passed
from the upper level to lower level units, not actions.

Martin, I am not sure what prompted your long response to this, but you
deserve a reply.

Think of an Elementary Control Unit (ECU). It consists _only_ of four
entities and their connections:

(1) a perceptual input function (PIF) that transforms the values from several
inputs into a single value called a "perceptual signal;" its inputs are
in most cased the perceptual signal values of lower-level ECUs.

(2) a reference input function (RIF) (seldom discussed) that transforms the
values from several inputs into a single value we call the "reference
signal;" its inputs are in most cases the output values of higher-level
ECUs.

(3) a comparator that takes the perceptual signal and the reference signal
as inputs and produces a single value called the "error signal."

(4) an output function (OF) that takes the error signal as input and produces
a single value we call the "output signal."

Nice, but I would nit-pick two points here. First, and this is purely for
the sake of parsimony, I conceive of the RIF as part of the comparator.
Now I know there is a danger in that the comparator is simply a subtraction
function, but if you look at Mary's diagram in the 1988 cybernetics paper,
she has the multiple outputs of the upper level systems feeding into the
comparators of the lower level systems. Thus, I conceive of the comparator
as r - p where r = f(higher-level o's). And you are correct, it is seldom
discussed. BTW, the reason I do not do that for p, is because p, which is
a function of the p's of lower-level systems, goes on up to a higher-level
system. Thus, if I up that function in the comparator, it would have to
have two outputs (p and e).

The second nit pick is that the output function presumably creates several
signals (although they may only have one value) as the output is passed to
several lower-level units. You, Martin, are _not_ necessarily _not_ saying
this, I just want to be explicit (See, Rick, how you can be respectful
while trying to complete a picture you think is half drawn).

The output signal is a single time-varying value. Its variations have
consequences we call "actions," in various places. The "actions" are seen
by an outside observer. Where are they? Actions are the effects of the
output signal on some part of the environment of the ECU, not necessarily
of the organism. If the observer can look inside the hierarchy of ECUs,
as a simulation modeller can, then the immediate actions of the ECU are
its contributions to the reference input functions of several other ECUs.

This is about carefully defining ours words, so let me give my 2-cents. I
think it is sloppy to think of the passing of a signal from one ECU to
another as an action (but I could be persuaded I am wrong about this). I
say this because I think it is a different thing to talk about (I could be
persuaded because the math is not that different) the changing of values
and parameters, and the changing of functions (i.e., actions are the
outputs of the intrinsic units on the functions of the perceptual units)
and physical environments (of which functions is one example and the
environment outside the system is another).

Now "behaviour". This is a pretty generic term. "Behaviour is the control
of perception." This is the "purpose" view. It is what the actions are for,
and describes the whole system at work. "A control system with too high a
gain for its transport lag shows oscillatory behaviour" is a descriptive
use of "behaviour" as the observable effects of something on the environment.
"Behaviour is purposeful action" again refers to the observable influences,
but restricts them to those influences that form part of the environmental
feedback path, eliminating those that are only side-effects. One could
also use "behaviour" to refer to what I called "immediate actions."

There are other uses for the word "behaviour," but I think those are
the main ones used in technical discussions of PCT. They are similar enough
to cause confusion, and different enough that the confusions matter. I
suspect that it would be better to use "output" to refer to the specific
single value emitted by the output function of the ECU, and "actions" to
refer to the influences of the output on observable (or simulable) aspects
of the environment of the ECU, leaving "behaviour" out, except in colloquial
discussion.

For the reasons stated above, I would agree with your use of the words
output and actions. Those uses exactly match the uses I was suggesting for
them (which I why I do not understand why you wrote this post). The
ambiguity of behavior as a term is also represented in my posts. Whether
we should drop it altogether I am not so sure. What I am sure of is that
before trouncing on someone for being a "behaviorist" or some other "evil"
entity (which I am not saying that you are doing to me, Martin), one should
reflect on all these posts this week.

Later,

Sincerely,

Jeff

[Martin_Taylor4]

[Martin Taylor 980222 02:00]

Jeff Vancouver 980221.1440 EST]

Jeff Vancouver 980220.1442 EST to Bill Powers (980220.0354 MST)

I am sorry, I thought actions were on the environment. Signals are passed
from the upper level to lower level units, not actions.

Martin, I am not sure what prompted your long response to this, but you
deserve a reply.

What prompted it was the statement requoted above--the claim that actions
are on the environment. I agreed with this, but not with what I thought
you meant by "the environment." I wanted to make it quite clear that
"the environment" of an ECU includes all the other ECUs in the control
hierarchy, as well as everything outside the organism.

Think of an Elementary Control Unit (ECU). It consists _only_ of four
entities and their connections:

(1) a perceptual input function (PIF) that transforms the values from several
inputs into a single value called a "perceptual signal;" its inputs are
in most cased the perceptual signal values of lower-level ECUs.

(2) a reference input function (RIF) (seldom discussed) that transforms the
values from several inputs into a single value we call the "reference
signal;" its inputs are in most cases the output values of higher-level
ECUs.

(3) a comparator that takes the perceptual signal and the reference signal
as inputs and produces a single value called the "error signal."

(4) an output function (OF) that takes the error signal as input and produces
a single value we call the "output signal."

Nice, but I would nit-pick two points here. First, and this is purely for
the sake of parsimony, I conceive of the RIF as part of the comparator.
... Thus, I conceive of the comparator
as r - p where r = f(higher-level o's).

There's no problem with combining the reference input function into the
comparator, if one wants, provided one recognizes that there has to be
such a function, and that the form and parameters of this function are
as susceptible to reorganization as are the form and parameters of the
other two functions (perceptual and output). Your point is hardly a nit
pick, more an expression of personal preference for a style of description.
I normally do as you do, except when I want to make the explicit point
that there must be a reference input function, because there is more
than one source contributing to the reference signal.

The second nit pick is that the output function presumably creates several
signals (although they may only have one value) as the output is passed to
several lower-level units.

No, I have to disagree here. The output function creates one signal value,
which is distributed _in the environment of the ECU_ to many different
places. Within the ECU, there is just one output signal. What happens in
its environment is just that--in its environment.

This is about carefully defining ours words, so let me give my 2-cents. I
think it is sloppy to think of the passing of a signal from one ECU to
another as an action (but I could be persuaded I am wrong about this).

I hope so, because if you can be persuaded you are wrong, the situation
becomes much cleaner. The labels become the same for every ECU. If you
don't call the effects of the output "actions" unless they impinge on
the world outside the ECU, then you have to conceive of the ECU as
"knowing" somehow what lies within and what lies without the organism.
What the ECU's output does within the organism (influence other reference
levels) is in some way different from what it does outside the organism.
I think that is an unnecessary complication, perhaps justifying the
word you used--"sloppy".

I
say this because I think it is a different thing to talk about (I could be
persuaded because the math is not that different) the changing of values
and parameters, and the changing of functions (i.e., actions are the
outputs of the intrinsic units on the functions of the perceptual units)
and physical environments (of which functions is one example and the
environment outside the system is another).

The output of an ECU never changes a function. That's reserved for the
outputs of the reorganizing system. (Unless that's what you mean by "the
outputs of the intrinsic units." If you mean that, then are you saying that
only reorganization should be assigned the word "action"?). I take "action"
to be the influencing of lower level reference values, and thereby the
influencing of things in the world outside the organism.

To belabour the point, an ECU is delimited by its two input functions and
its single output function. Where its signals come from and go to is its
environment, whether that environment is in the hierarchy or outside the
organism. What those signals do in its environment is of no consequence
to the ECU provided that they influence its perceptual signal in such
a way as to reduce the absolute magnitude of its error signal.

"Actions" are the observable influence of the output in the environment
of the ECU, inside or outside the organism, but they can be observed
only where the observer can see. That's why it is sometimes
convenient to think of the actions of an intact organism as happening
only outside the organism. It's the only place another intact organism
(the experimenter) can ordinarily see. But in a simulation, and by using
the Test, the experimenter can sometimes see--perhaps dimly--some of
those influences inside the organism. And I think it warranted to call
them "actions" there, too.

Martin

[Bill_Powers1]

[From Bill Powers (980221.0204 MST)]

Martin Taylor (980220 EST) (time-date stamp missing)

A nice job of organizing and explaining.

(1) a perceptual input function (PIF) that transforms the values from several
inputs into a single value called a "perceptual signal;" its inputs are
in most cased the perceptual signal values of lower-level ECUs.

(2) a reference input function (RIF) (seldom discussed) that transforms the
values from several inputs into a single value we call the "reference
signal;" its inputs are in most cases the output values of higher-level
ECUs.

This may be a good idea. The problem is where to put the transformation.
The output weights, which I usually consider to be only 1,0, and -1, could
be associated with the originating system or the receiving system, or with
a "connection matrix" between them. If there are other output weightings,
they really should go with the originating system, because they're relevant
to the higher-level control system and must be adjusted with knowledge of
the higher system's input weights. The resulting signals would simply sum,
positively or negatively, at the lower-level comparator. Neurologically, a
positive signal sums directly, and a negative signal sums via a Renshaw
cell that produces inhibitory effects on the comparator neuron.

The difficulty with associating the Reference Input Function with the lower
system is that the outcome has to be a single equivalent reference signal,
and the lower system has no basis for adjusting any weightings in the RIF.
The critical choice is to keep the feedback in the _higher_ loop negative
via all paths that go through lower systems. So whatever selects the
weighting of the higher system's output signal must be associated with the
higher system, not the lower one.

Another possibility is to treat this question as a problem for
reorganization. The output neurons that send axons to lower comparators
must be subject to neurotaxic effects; the neurons branch and grow toward
destination connections that have to depend on experience. Somehow. There
is a similar problem, perhaps with a similar solution, on the input side.
The lower perceptual signal pathways must branch and grow toward higher
input-function neurons. So between the lower and higher systems, on both
the input and output sides, we have a region in which connection matrices
develop as the organism matures and reorganizes. The connectivity in these
matrices establishes which lower systems are used by, and contribute to
perceptions in, which higher systems. But the weightings given to each
signal, either upgoing or downgoing, must be adjusted on the basis of how
the _higher_ system is performing.

There was an article in "Science" a few years ago about this, by the way,
using as an example someone standing at a door. "What are you doing?"
I'm standing at the door; I'm pushing the button; I'm ringing the bell;
I'm visiting my aunt; I'm trying to get a good inheritance. (The detailed
examples are mine; the study had other examples, and was interested in the
effects of which level the subject was consciously focussed on).

You may be thinking of an article by Robin Vallacher. He was fully aware of
my hierarchical model although he didn't cite my work that I know of.
Shortly after my book came out he invited me to IIT in Chicago to give a
long seminar on PCT, and I explained the hierarchical idea in the way you
describe. What Vallacher has done since then seems to be based directly on
my ideas. So you are really citing me via Vallacher.

Good post; I hope it accomplishes your (our) purpose.

Best,

Bill P.

[Martin_Taylor4]

[Martin Taylor 980222 17:10]

[Martin Taylor 980222 02:00 to Jeff Vancouver 980221.1440 EST

A bad typo:-( Probably not a good idea to write such things when one
should be going to bed:-)

I hope so, because if you can be persuaded you are wrong, the situation
becomes much cleaner. The labels become the same for every ECU. If you
don't call the effects of the output "actions" unless they impinge on
the world outside the ECU, ...

Obviously, should have been "the world outside the organism".

...then you have to conceive of the ECU as
"knowing" somehow what lies within and what lies without the organism.

Sorry for any confusion.

Martin

[Jeff_Vancouver2]

[from Jeff Vancouver 980223.1255 est]

[Martin Taylor 980222 02:00]

The second nit pick is that the output function presumably creates several
signals (although they may only have one value) as the output is passed to
several lower-level units.

No, I have to disagree here. The output function creates one signal value,
which is distributed _in the environment of the ECU_ to many different
places. Within the ECU, there is just one output signal. What happens in
its environment is just that--in its environment.

Signals are what travel down the pathways (the neurons). Given that they
connect to several comparators (or RIFs) means there must be more than one
of them. I respectfully maintain my opinion on this one.

This is about carefully defining ours words, so let me give my 2-cents. I
think it is sloppy to think of the passing of a signal from one ECU to
another as an action (but I could be persuaded I am wrong about this).

I hope so, because if you can be persuaded you are wrong, the situation
becomes much cleaner. The labels become the same for every ECU. If you
don't call the effects of the output "actions" unless they impinge on
the world outside the ECU, then you have to conceive of the ECU as
"knowing" somehow what lies within and what lies without the organism.
What the ECU's output does within the organism (influence other reference
levels) is in some way different from what it does outside the organism.
I think that is an unnecessary complication, perhaps justifying the
word you used--"sloppy".

Your way has the advantage of the "labels become the same." My way has the
advantage of highlighting the qualitative difference in the effects (of
crossing systems, whether the organism/physical environment or
reorganizing/perception system). I will try to remember to adopt your way,
but will probably forget.

The output of an ECU never changes a function. That's reserved for the
outputs of the reorganizing system. (Unless that's what you mean by "the
outputs of the intrinsic units." If you mean that, then are you saying that
only reorganization should be assigned the word "action"?). I take "action"
to be the influencing of lower level reference values, and thereby the
influencing of things in the world outside the organism.

I was using the meaning you have in parentheses. The reorganizing system
is a system of ECUs in my mind.

Sincerely,

Jeff

[Martin_Taylor4]

[Martin Taylor 980224 15:20]

from Jeff Vancouver 980223.1255 est to Martin Taylor 980222 02:00

The output function creates one signal value,
which is distributed _in the environment of the ECU_ to many different
places. Within the ECU, there is just one output signal. What happens in
its environment is just that--in its environment.

Signals are what travel down the pathways (the neurons). Given that they
connect to several comparators (or RIFs) means there must be more than one
of them. I respectfully maintain my opinion on this one.

Of course I agree that there are many pathways through which the (single)
output signal value is linked to the reference inputs of lower level ECUs.
The only question at issue is whether to consider that these pathways
exist within the ECU or outside it. It doesn't matter how you consider
them, inasmuch as the system works identically either way. What you
have to think about is which way is simpler to analyze and discuss. My
preference is to think of a set of ECUs at one level, a set of ECUs at
another level, and a set of wires connecting them (or, as Bill Powers
says, a connection matrix).

Bill P raised an interesting point, which is whether there is a distribution
function in the ECU associated with the weights that connect its output
to the lower level ECUs. He raised that point because it is plausible
that when an ECU is not controlling well, reorganization should be
associated with it, rather than with the lower-level ones to whose
reference inputs it contributes. They might all be controlling very well,
and there would be no reason for reorganization to affect them. I buy
that argument. But again, it is a question of convenience in analysis
rather than a difference in how the systems are connected and how they
work.

My way ["action" different from "influencing reference levels"] has the
advantage of highlighting the qualitative difference in the effects (of
crossing systems, whether the organism/physical environment or
reorganizing/perception system).

If you think of a whole "system" as a unit, such as "the reorganizing
system" or the "perceptual control hierarchy," then action _by the system_
is clearly on something outside the system. If you think of "system" as
a smaller entity, such as a group of ECUs connected in some way, then
I see "action" by the system as being any influence on something outside
that smaller system. Likewise, "action" by an ECU is an influence on
anything in the environment of the ECU. The most immediate influence is
on reference levels of the ECUs at the next lower level. I'd call that
its "immediate" action. The influence on its environment that matters to
its functioning is on the Controlled CEV to which its perceptual signal
corresponds. In the sense that "behaviour is the control of perception,"
its influence on the CCEV is also an "action" by the ECU. But then, so
also are all its side-effects.

The output of an ECU never changes a function. That's reserved for the
outputs of the reorganizing system. (Unless that's what you mean by "the
outputs of the intrinsic units." If you mean that, then are you saying that
only reorganization should be assigned the word "action"?). I take "action"
to be the influencing of lower level reference values, and thereby the
influencing of things in the world outside the organism.

I was using the meaning you have in parentheses. The reorganizing system
is a system of ECUs in my mind.

In mine, too. But I'm surprised that you exclude influences on the world
outside the organism--such as picking up a rock--from being called
"actions."

Martin

[Martin_Taylor4]

[Martin Taylor 980224 17:30]

Bill Powers (980221.0204 MST)]

The difficulty with associating the Reference Input Function with the lower
system is that the outcome has to be a single equivalent reference signal,
and the lower system has no basis for adjusting any weightings in the RIF.
The critical choice is to keep the feedback in the _higher_ loop negative
via all paths that go through lower systems. So whatever selects the
weighting of the higher system's output signal must be associated with the
higher system, not the lower one.

Another possibility is to treat this question as a problem for
reorganization.

I think this is the right answer, functionally. Whether the intrinsic
variables are better maintained near their reference level because a
lower-level perceptual ECU changes which master it serves or because a
higher-level one changes which slave it commands is of no concern to the
intrinsic system. Either might happen, and there could be both a
distribution weighting at the higher level and a Reference Input Function
(which need not be a summation) at the lower level. The result in
either case is what you called a "connection matrix."

There was an article in "Science" a few years ago about this, by the way,
using as an example someone standing at a door. "What are you doing?"
I'm standing at the door; I'm pushing the button; I'm ringing the bell;
I'm visiting my aunt; I'm trying to get a good inheritance. (The detailed
examples are mine; the study had other examples, and was interested in the
effects of which level the subject was consciously focussed on).

You may be thinking of an article by Robin Vallacher.

That name does sound familiar. It might be what I was thinking of. Do you
have the reference? I haven't found it yet. And I don't remember precisely
what the study tested, except that it had something to do with focus on
different levels of perception.

Martin

[Bill_Powers1]

[From Bill Powers (980224.2114 MNST)]

Martin Taylor 980224 17:30--

You may be thinking of an article by Robin Vallacher.

That name does sound familiar. It might be what I was thinking of. Do you
have the reference? I haven't found it yet. And I don't remember precisely
what the study tested, except that it had something to do with focus on
different levels of perception.

That's about all I remember now, too.

Best,

Bill P.

[Francisco_Arocha]

[from francisco arocha, 980225.9:14 EST]

[From Bill Powers (980224.2114 MNST)]

Martin Taylor 980224 17:30--

You may be thinking of an article by Robin Vallacher.

That name does sound familiar. It might be what I was thinking of. Do you
have the reference? I haven't found it yet. And I don't remember precisely
what the study tested, except that it had something to do with focus on
different levels of perception.

That's about all I remember now, too.

This is what you may be thinking of:

Vallacher RR, Wegner DM, What do people think they're doing? Action
identification and human behavior, Psychological Review 94 (1987) 3-15.

It is cited in Martin's AGARD paper.

francisco

[Jeff_Vancouver2]

[from Jeff Vancouver 980225.0855 EST]

[From Bill Powers (980224.2114 MNST)]

Martin Taylor 980224 17:30--

You may be thinking of an article by Robin Vallacher.

That name does sound familiar. It might be what I was thinking of. Do you
have the reference? I haven't found it yet. And I don't remember precisely
what the study tested, except that it had something to do with focus on
different levels of perception.

I am not familiar with the Science article to which Martin is refering, but
it certainly sounds like Vallacher (possibly along with Wegner). They have
a "theory" called action identification theory that states that people will
be conscious of what they are trying to do at the highest level possible
(cite is below). That is loose speak for the level that is reorganizing
and in thinking mode. They have some very clever research that I think
nicely communicates the levels of the hierarchy idea. It might be a good
example of what Rick is looking for (to trounce no doubt). Bill has
mentioned that Vallacher is using his ideas (and Vallacher cites Carver &
Scheier more than Bill). Indeed, the theorizing is not comparable to
Bill's. But the research is clever.

The example I use all the time is the dribble cup experiment. It was a
between-subjects design (I say this up-front so Rick can ignore it as he
reads instead of flushing it after he is done reading). Participants were
asked to drink a cup of coffee and asked what they were doing.
Participants in the regular (control) condition said they were trying to
get a caffine buzz, quench their thrust, etc. Participants in the dribble
cup condition said they were trying to get the liquid into their months,
not spill in themselves, etc. Vallacher & Wegner point out that the
participants in the control condition were trying to do the same thing as
the participants in the dribble cup condition, but the lower-level systems
in the control condition were operating (controlling) without need for
reorganization, while those systems in the the dribble cup participants did
need reorganizing because of the unique nature of the dribble cups.

        Vallacher, R.R., & Wegner, D.M. (1987). What do people think they're
doing? Action identification and human behavior. Psychological Review,
94, 3-15.

Okay, Rick, trounce away.

Sincerely,

Jeff

[Bruce_Gregory]

[From Bruce Gregory (980225.1150 EST)]

Jeff Vancouver 980225.0855 EST

The example I use all the time is the dribble cup experiment. It was a
between-subjects design (I say this up-front so Rick can ignore it as he
reads instead of flushing it after he is done reading). Participants were
asked to drink a cup of coffee and asked what they were doing.
Participants in the regular (control) condition said they were trying to
get a caffine buzz, quench their thrust, etc. Participants in the dribble
cup condition said they were trying to get the liquid into their months,
not spill in themselves, etc. Vallacher & Wegner point out that the
participants in the control condition were trying to do the same thing as
the participants in the dribble cup condition..

Ah, but were they? Not if you believe that behavior is the
control of perception.

Bruce

[Jeff_Vancouver2]

[from Jeff Vancouver 980225.1220 EST]

[Martin Taylor 980224 15:20]
Of course I agree that there are many pathways through which the (single)
output signal value is linked to the reference inputs of lower level ECUs.

Bill P raised an interesting point, which is whether there is a distribution
function in the ECU associated with the weights that connect its output
to the lower level ECUs.

I think we (you and I) are at a deeper level than categorizing what is the
environment of the ECU and what is part of it. I suspect the key issue is
the point-of-view of the ECU. No executive functions, no special knowledge
of the lags and states of the environment. Only the state of the CV and
assessed through the input function. I agree with all of this.

Bill P's point hits at a central question for me anyway. That is, how do
the layers of ECU's develop? I think the answer to that question will
require a careful consideration of the RIF. Another interesting topic for
another time.

My distinction between reorganizing and signal passing is also mute when
one gets to the level of what is happening neural-chemically (as opposed to
functionally). Releasing (or the frequency of releasing) neurotransmitters
versus growing new dentrites are both "actions," so lets not quibble.

In mine, too. But I'm surprised that you exclude influences on the world
outside the organism--such as picking up a rock--from being called
"actions."

I am surprised as well. I was calling influences on the world an action
from the start. What got me in trouble with you is that I was reserving
the word action for only that type of influence (then reorganizing systems
got involved, etc.).

Martin, you, me and Bruce A (and maybe others) need to get together
sometime and figure out how to convince ourselves or the others what the
causation thing is all about. Not now, of course, we are all too busy.

Later,

Jeff

Jeffrey B. Vancouver
Department of Psychology
Ohio University
Athens, OH 45701
phone: 614-593-1071
fax: 614-593-0579
vancouve@oak.cats.ohiou.edu

[Jeff_Vancouver2]

[from Jeff Vancouver 980225.1245 EST]

[From Bruce Gregory (980225.1150 EST)]

Jeff Vancouver 980225.0855 EST

The example I use all the time is the dribble cup experiment. It was a
between-subjects design (I say this up-front so Rick can ignore it as he
reads instead of flushing it after he is done reading). Participants were
asked to drink a cup of coffee and asked what they were doing.
Participants in the regular (control) condition said they were trying to
get a caffine buzz, quench their thrust, etc. Participants in the dribble
cup condition said they were trying to get the liquid into their months,
not spill in themselves, etc. Vallacher & Wegner point out that the
participants in the control condition were trying to do the same thing as
the participants in the dribble cup condition..

Ah, but were they? Not if you believe that behavior is the
control of perception.

What do you mean? They were controlling lower-level perceptions of
intension around the cup, of the cup to mouth, of liquid in the mouth, of
(whatever higher-level perception they were controlling). The difference
was that they (the control group) were controlling those perceptions. The
dribble cup group lost control at the "liquid in mouth" level. They had to
reorganize; that is, that ECU had to get into thinking mode. Which is
what, according to them, accounts for the self-report of what they were
doing. Perhaps I should have said "were trying _but accomplishing_ the
same things..."

Your conclusion "Not if you believe that behavior is the control of
perception" refers to a pretty complex perception of mine. How you glean
it from the paucity of information in these posts is amazing.

Sincerely,

Jeff

[Bruce_Gregory]

[From Bruce Gregory 9980225.1307 EST)]

Jeff Vancouver 980225.1245 EST]

Your conclusion "Not if you believe that behavior is the control of
perception" refers to a pretty complex perception of mine. How you glean
it from the paucity of information in these posts is amazing.

It's nothing really. Just a little skill I picked up from Rick.

Bruce