[From Bruce Abbott (991109.2015 EST)]
Bill Powers (991105.2340 MDT) --
Bruce Abbott (991105.1215 EST)
Bruce Abbott (991104.1100 EST)
I very much agree. The diagram I pictured (and hoped you would provide in
this reply) is one ugly monster. As you say, it raises some important
questions that thus far have gone unanswered.
That is equally the case for the cued-chained-behaviors model. ...
This is not a problem for the behavioral chaining model, any more than for
the sequence-control model.
Somehow that seems to me to add up to "equally the case."
It adds up to "equally NOT the case." (You and I were referring to a
specific problem, not 'some important questions' as the context supplied
above unfortunately suggests.)
Although I verbally described "turn left," we
all know what this means to real drivers. In the real-world case, the
driver is controlling a number of variables simultaneously, just as the
circles do in the crowd demo. These include such things as staying in the
proper lane, slowing sufficiently prior to the turn, and many others. You
are attempting to make the control system we are discussing do the work of
all these other systems. To simplify the discussion (where we are focusing
on the issue of generating the sequence) I have left them out, and for
current purposes that is appropriate.
Is it? The systems you mention are subordinate to any system that sets
their reference levels; they must be told where on the road to stay, at
what point to start the turn, and so forth. I know what you have in mind:
the higher system, upon being cued, sets the reference signals for all the
lower systems required to execute a left turn. So whereas the lower systems
autonomously negotiate the turn, the system that tells them to do so is
simply triggered off by the cue.
O.K.
I think you overlook the possibility that these other systems are in fact
sequence control systems, and that the system you imagine to be in charge
of them is not needed. And you also overlook the need of these analog
systems to be given _quantitative_ reference signals, not just on-off
reference signals. As you approach the street corner, at what point does
the lettering on the street sign suddenly become a "cue" and tell the other
systems to start the left turn? If the lettering becomes a cue too early,
you will turn before reaching the corner; if too late, you will miss the
corner. And for someone who is not going where you want to go, it should
not serve as a cue at all.
If the "cue" indicates that one should turn at the next intersection, I
don't see any problem. Also, one can be maintaining some variables within a
particular range of values independently of the system orchestrating when
one should turn or go straight. It's a massively parallel system. So I can
have a system that keeps the car centered in its lane, and this system will
continue to do its job whether some other system orders a left turn or not.
One can have a system that controls the centripital force experienced by
adjusting the speed of the auto according to the sharpness of the turning
angle, and so on. I don't see that these systems must be receiving their
references from the system that is setting references for turning here and
going straight there.
A system that is keeping the car in its proper lane during a turn is
obviously controlling the relationship of the cross-street to the car; in
other words, it is employing the same perceptions that you see as "cues."
Your way of conceiving this process is reflected in many other concepts of
behaviorism. For example, when a reinforcer occurs, it is said to increase
the probability that the behavior which led to the reinforcement will occur
again under the same circumstances.
Another way of stating this is that the system will begin to reorganize so
as to control for the perception of the reinforcer. The organism _wants_
the reinforcer and reorganizes so as to be able to produce it on demand.
But that is only one of two choices of
how to interpret that situation. We might, alternatively, see the
reinforcer as being caused by the behavior, and see any increase in the
frequency of reinforcement as a consequence of an increase in the frequency
of the behavior that causes it. That is the interpretation that would be
preferred by anyone who sees the behavior as the organism's means of
controlling the delivery of reinforcements.
If what I stated above is correct, then your dichotomy is false. Control
gets established because the reinforcer is desired, and reorganization
occurs until that desire is fulfilled (if that is possible under the
circumstances). The pellet appears because the behavior produces it (the
behavior is the cause of the reinforcer), but the control system exists
because the rat wanted the pellet, and reorganized until its behavior
produced it.
In a parallel way, we have a choice of seeing the cued behavior
differently. We can see a cue simply as a perceptual consequence of the
behavior that preceded it, rather than as a cause of the behavior that
follows it. Instead of seeing the chain as being grouped cue-behavior,
cue-behavior, we can see it as behavior-consequence, behavior-consequence,
with the consequences (which you call cues) being perceived. The organism,
under this view, selects the next consequence to be produced,
Yes, but this leaves unanswered why the organism makes any particular
selection. I would say that it does so because it has learned that a
particular selection makes further action possible, which eventually leads
to a desired perceptual state. If following those selections reliably leads
to the desired end-state, then with practice the "decisions" will become
automatic and stimulus-bound.
and this
results in the actions that produce it. As soon as each selected
consequence is achieved, the next one in the sequence is selected. Thus the
sequence is directed by the organism's sequential selection of
reference-states, not by the environment. If a behavior is emitted that
results in the wrong next perception, the behavior will be changed to
maintain the same sequence, rather than the "wrong" cue simply leading to a
correspondingly "wrong" next behavior.
Yes, I'm not disputing that. I'm suggesting that under certain conditions
the organism may default to the simpler process of using the cues as they
arise consequent on the control processes that produce them as the souces of
the next reference values.
That, incidentally, is the kind of
expeirmental test we need.
Yes, clearly so. I think I've mentioned this myself in an earlier post on
this topic.
The behavior-chaining interpretation puts control of the sequence into the
environment. It depicts each behavior as being triggered off by the
preceding cue. The experimenter can determine the sequence by determining
what cue will be produced at the end of each behavior. This means that
there is no actual sequence control in the organism; if the experimenter
changes the cue that a given behavior will produce, and that cue has
previously become linked to another behavior, that other behavior will
occur. As far as the organism is concerned, each step in the sequence is a
simple independent S-R pair, unlinked to any other S-R pair. The only
sequence control is what happens in the experimenter as the series of cues
is set up.
That is correct.
As for the HPCT sequence-controller having a mechanism to report the state
of the sequence, it seems to me that the signal which this mechanism
generates is too ambiguous for the purpose.
If you want to propose a mechanism that is too ambiguous, you can. Why not
propose one that is adequate, once you know what is needed?
Pardon me? The mechanism I described is _your_ proposal, not mine. So, why
don't you ask yourself that question? (:->
Mechanisms are
a dime a dozen; what matters, as Rick keeps reminding us, is to define what
the model has to do. Among other things, it has to provide for a person
being able to say, "Wait just a minute, I'm almost through with [this
sequence]." In order to say that, the person must be able to perceive the
progress of the sequence.
Of course other information is needed, too, to allow the person to know
when one step has been finished and the next one can be initiated.
Like, for instance, the appearance of what I've been calling a "cue?"
And the
person must be able to detect when a wrong sequence step has occurred, so
as to be able to correct the error. Assuming, of course, that people do
make mistakes and that they do correct them.
Yes, I noted this somewhat later in the post. (See below.)
It begins with zero value and
increments by some amount as each perception in the sequence appears. A
moderate level of this signal at the end of the sequence may remain below
reference, but this by itself cannot inform the output function what part of
the sequence was incorrect.
Then that is not what this kind of perception is for.
It is what you proposed for sequence control in B:CP and it is the variable
Rick so confidently pointed to when asserting that sequence control is based
on a scalar variable whose intensity indicates the degree of error in the
sequence.
But some such
perception is needed, because people can judge, I claim, whether they are
near the beginning, middle, or end of a sequence in progress.
This only tells me that people can perceive the sequence in memory and
locate where they currently are in the remembered sequence.
A moderate level of sequence-perception signal
could mean that elements A, B, and D appeared in proper sequence but that C
was missing, or that the proper elements occurred but in the wrong order, or
that four elements appeared, in proper order, except that C was replaced by
E. The output function cannot know which element needs attention.
No, but the control system can be designed to behave appropriately (i.e.,
as real subjects do) under these various sequence errors. If a person
restarts the sequence for any error whatsoever, we can design a system to
do that. If the person will tolerate one or more errors and press on to the
next element, we can design a system to do that, too. The design is not a
problem; the problem is to find out what people really do. Maybe different
people do things differently; in that case we would have to model them
differently.
Well, I can certainly agree with that! I see no trouble designing a system
that will behave appropriately, once we know what that is. What I was
having problems with is the specific proposal made in B:CP, which seems
incapable of doing the job as I see it.
As long as the sequence is
occurring correctly, all that is required is a perception that the current
element has occurred, to keep the sequence going. In fact, the PCT model
would behave in a way that looks just like your cued-behavior model. The
difference would be seen only when disturbances occur that cause a problem
with production of the specified sequence. This is usually the case in
comparing SR with PCT models. If there are no disturbances, they behave the
same. It's only when there is a disturbance, and the system acts to resist
it, that you can tell it's not an open-loop system.
Exactly so. I have suggested that this open-loop mode might develop under
conditions where such disturbances are unlikely.
That is not sufficient. If a sequence has been progressing without error,
how do you know that the first error is not about to occur?
Why would I have to? Does your sequence-controller know when an error is
"about to occur"? No. So what's the point?
And anyway, the
idea of a sequence error doesn't even apply in the cued-behavior model --
the only error that can occur is responding in the wrong way to whatever
cue occurs.
Right!
The design issues we're talking about can't be settled in an ad-hoc way;
each involves considerable complexities so that a neural substrate suitable
for one kind of system would not be suitable for the other. Sequence
control versus cued chaining is a design difference that would result from
evolving along very different lines. The concept of an open-loop response
gets into prediction and modeling of the lower systems and the environment,
and all the complexities implied by that inverse-dynamics,
inverse-kinematics, world-modeling approach.
You keep slipping this in. It's not relevant. Only the initiation of the
next element in the sequence is carried out open loop in my proposal. We're
talking stimulus-reference, not stimulus-response, and the problems of
inverse-dynamics etc. do not enter in.
A control system organization
accomplishes the same overall effects, but in a far simpler and
structurally very different way.
If sequences _can_ be produced as my proposal suggests (which is not the
same as saying that they _are generally_ produced that way), it is actually
simpler than sequence-control, requiring less cognitive overhead. If it
does occur, the reduced cognitive load it permits would be the reason why
such a method evolved.
I just don't think it's a tenable idea to
say that the organism adopts whichever strategy works for it. Once you've
committed to one organization, I think you've ruled out the other.
Well, that's one place where we strongly disagree.
It would then work just as
well as sequence control but without the cognitive overhead. Because the
end-states that trigger shift to the next element in the sequence are the
references of controlled variables, they will occur unless the currently
active control system meets with an overwhelming disturbance.
There is nothing to say that those cues are the reference-states of
controlled variables. In fact, they are not. The cue that triggers off the
next behavior is not the reference state to be achieved by the next
behavior, but the final state of the world generated by the previous
behavior. If it's the reference state of anything, it's the reference state
of the variable controlled by the previous behavior. If a rat has to press
a bar 10 times to turn on a light, under your proposal the state of the
light is at best the controlled variable of the system that does the
pressing. It is not the controlled variable of the next system to come into
play, the one that presses a different lever when the light is on to get
food. The cue is simply a signal for the next process to start, and is
unaffected by the next process; it's generally gone, or irrelevant, by the
time the next process has started.
Perhaps I did not make this sufficiently clear. In my proposal, the "cue"
serves as a _retrieval_ cue for associative memory; it retrieves the
parameters required to set up the next control process in the sequence
(e.g., the reference value).
Is
this how you determined that there was a reliable sequence produced without
any sequence control?
I haven't determined any such thing.
I was misled by your speaking as if you knew that reliable
sequence-production could occur open-loop. I now understand that you were
only conjecturing that it might be possible.
Yes.
I began this exercise by showing how a
reliable sequence might be produced in the absence of sequence control. I
said that the proposal is testable. You obviously agree, as you are now
suggesting tests that would discriminate the two modes of operation (open
and closed loop).
Yes.
.. showing the reaction time involved in perceiving a change in sequence and
doing something about it, and also showing that the elements of the
sequence are being generated independently.
Independently of what?
Of the environment. When the person repeats part of the sequence after the
target movement has stopped, this shows that there is a sequence-generator
inside the person.
If the cues arise as a result of completion of each behavioral act, then the
sequence would continue under the open-loop proposal as well.
We also know that from the fact that prior to the
target's stopping, the person is matching movements to the target movements
at a speed too great for simple tracking to work.
Please explain.
It seems to me what what is going on in sequence control is that the
individual is recalling [the] what needs to be produced one element at a time
and then engaging in the actions required to bring about the element.
That's what I mean by independent generation of the sequence. Storage of
the sequence in memory is one basis for producing a sequence that is not
being driven by present-time target movements.
Good, we're thinking along the same lines here. Keep in mind, however, that
at this point I'm speculating about how sequence _control_ might work, not
cued behavioral chains.
If a
mismatch occurs between the element that appears and the one remembered, an
error is generated. Depending on circumstances, the individual may simply
attempt to continue (ignore the error) or may halt and attempt to correct
the element.
Right.
This description has a surface simplicity that obscures a considerable
complexity of organization required to carry it out. Important and
unanswered questions include the following: Just what is remembered? Do
these memories become reference perceptions or do they act as comparison
perceptions where the reference comparison is for identity (zero
difference), or are both organizations possible, for use under different
circumstances?
Sorry, I have no idea what that means. How is a reference perception
different from a comparison perception?
If a sequence of perceptions is being read out of memory, this may _supply_
the references for the lower-order systems. Alternatively, the sequence of
memory-perceptions may be compared in real time to the arriving sequence of
perceptions; the output of the perceptual function would be a series of
same-different judgements. The reference of this system might be for
"same," in which case a "different" perception would generate an immediate
error when it occurred.
I haven't really even begun to puzzle through these possibilities, so I'm
not going to offer a detailed proposal for critique at this point. I offer
these rather vague sketches here to underscore the fact that there are many
alternative constructions that could be developed and tested.
Right, but we're not just pursuing them at random. The control architecture
is very different from an open-loop architecture for doing the same thing,
so the alternatives are mutually exclusive. What we are looking for are
experiments that will rule out one in favor of the other.
Again, I don't agree that the architecture has to be either/or. I
understand that this violates one of the HPCT design principles and that it
should not be accepted without solid empirical evidence. But it is a
testable proposition, and seems consistent with some of my own inner experience.
Regards,
Bruce