Control of sequences and programs

[Abbott]

[From Bruce Abbott (2018.02.12.1100 EST)]

Although I have no doubt that one can control sequences and programs, I doubt that we do very much of either.

To control a sequence implies (a) knowing what the sequence is, (b) recognizing deviations from the sequence, and © having a means of correcting the deviations. At the very least, the sequence reference specifies what to perceive at any given step in the sequence; actions are initiated to carry out the step, resulting in a perception that either matches or does not match the reference for control of the perception associated with that step. The sequence-reference sets the reference for the next step, and so on until all steps in the sequence have been carried out.

For example, typing the word “cat” entails having a reference for the sequence c – a – t. To type “cat,” the sequence controller sets a reference for a lower-level controller to perceive typing the “c,” then “a,” then “t.” Typing the “c” involves several levels of control that may themselves involve sequence-control, i.e., placing the fingers on the keyboard, moving a finger over the “c” key, then pressing and releasing the key, and then (for a touch typist) moving that finger back to the “home” position on the keyboard. These hand and finger movements are carried out by even lower level controllers.

The sequence controller may not wait for tactile or visual feedback to confirm that the “c” has been typed before initiating the typing of the “a.” Once a mistake has been made, by the time the sequence controller receives the feedback it is too late for on-the-fly correction. I’m a fairly good touch typist but when I make a mistake I’ve usually typed a number of additional letters before becoming aware of it (the tactile perception is wrong). Error correction in sequence control usually involves interrupting the sequence to make the correction before continuing, and sometimes involves starting the sequence over again.

In the example just given, the sequence to be produced is known in advance and one can compare the perception of that sequence to its reference sequence. As in typing or playing music, there is a known pattern to be reproduced. (The pattern may include more than just the sequence of letters or notes, but also patterned variations in the speeds of execution, intensity of the perceptions, etc.) We can learn to control sequences. But I believe that most examples in which a sequence of activities is carried out do not involve pre-specified sequence references and therefore are not examples of sequence control. The same may be said of program control.

In program control, there is a prespecified “program” to be carried out – a program reference that specifies what is to be done at each step, including alternatives whose execution depends on current perceptions: If this perception, then do that, else do this other thing. Certainly we can learn what should be perceived at each step, what alternatives if any are available at that step, and whether what should be perceived is being perceived. But I’m doubtful that we do much of this. Instead, we have a goal to achieve and string together activities on the fly that (we hope) will move us closer to achieving that goal, based on prior experience and what options appear to be available to us at the time. After we have achieved the goal and look back on how it was accomplished, it may appear that we were executing a pre-planned program, but most of the time we just did what occurred to us at the time, step by step.

Most of what we do is improvisation, not programmed execution.

Comments?

Bruce A.

[_Erling_Jorgensen]

[From Erling Jorgensen (2018.02.12 1150 EST)]

Bruce Abbott (2018.02.12.1100 EST)
BA: Although I have no doubt that one can control sequences and programs, I doubt that we do very much of either.
BA: ...[example of typing]...
BA: We can learn to control sequences. But I believe that most examples in which a sequence of activities is carried out do not involve pre-specified sequence references and therefore are not examples of sequence control.

EJ: Well, ... a word count of your post suggests you engaged in sequence control over 600 times. Your post seems more like evidence that we DO control sequences. It would help to hear examples of where it _seems_ like sequence control but where you have your doubts.
EJ: I'm not sure what you're emphasizing by singling out "pre-specified sequence references." Whenever there is corrective action, isn't that a sign of a pre-existing or previously-specified reference? Are you questioning the so-called steps in a sequence? Are you thinking that the reference for the second step arises from completion of the first step? I view that only as a gating function.
EJ: It seems to me that a reference for the whole enactment (see discussion below about that piece) is put in place, as well as references for each step of the process. The gating of when to move from one step to the next could be handled via gain modulation, although those variations would need to occur at a very rapid rate for some sequences. As a first approximation of a model, the gain for step 2 would be low until there is a signal of completion from step 1. This doesn't fully attend to the situation you describe of mistakenly typing a word, where "by the time the sequence controller receives the feedback it is too late for on-the-fly correction." Perhaps we should think in terms of a rolling gain function, like a sine wave, where each step ideally takes place as its corresponding gain reaches the crest of the wave. This would still allow for glitches if a given step got interrupted while the sine wave gain of a later step overtook the interrupted one.
EJ: There is still room for initiating a given step before there is time for confirming feedback from a previous step. Bill Powers wrote -- I think it was in B:CP -- about a pianist playing a passage faster than the perceptions for that level could occur, where the only way to deal with an error would be to ignore it or use the method you describe of interrupting the whole sequence and playing it again.
EJ: Let me go back to the notion above of "the whole enactment." It seems to me that, if there be sequence control, a reference for the entire sequence is present, even if the steps are being gated. So there are really two things going on, the whole and its parts. I notice that Bill P. had a word for the wholes we are considering here: He called them "Events." That is the word he utilized for the 5th level of his perceptual hierarchy, when he relocated "Sequences" to a higher spot closer to "Programs," when the original 9 levels of the hierarchy expanded to 11 levels.
EJ: I am of the mind that Sequences and Events are more intimately tied, and that a whole enacted event is comprised of its sequentially ordered parts. In my Part 1 contribution to the forthcoming LCS-IV, "How the Brain Gets a Roaring Campfire: 1. Structuring for Perceptual Results," I argue that the Sequence level should be relocated back down just below the Event level.
EJ: I claim that the essence of a Sequence is the connecting link "then". By contrast, a properly enacted Event seems to require sequences timed-in-sequence. There is neurophysiological evidence that I cite in my chapter, that non-specific regions of the thalamus may provide the kind of timing signal that the neocortex needs to turn a sequence into a timed-sequence, i.e., an event.
EJ: In any event, contra to the doubts that you allude to, it seems to me that sequence control is pretty ubiquitous in how we operate. Perhaps that's just because I am currently thinking from that layer, and so 'everything is a sequence'. I would like to hear more evidence as to why you think otherwise.
All the best,
Erling

Confidentiality: This message is intended only for the addressee, and may contain information that is privileged and confidential under HIPAA, 42CFR Part 2, and/or other applicable State and Federal laws. If you are not the addressee, or the employer or agent responsible for delivering the message to the addressee, any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this in error, please notify the sender immediately and delete the material from your computer. Thank you for your cooperation.

Please also note: Under 42 CFR part 2 you are prohibited from making any further disclosure of information that identifies an individual as having or having had a substance use disorder unless it is expressly permitted by the written consent of the individual whose information is being disclosed or as otherwise permitted by 42 CFR Part 2.

[MartinT]

[From Bruce Abbott (2018.02.12.1100 EST)]

      Although I have no doubt that one can

control sequences and programs, I doubt that we do very much
of either.

      To control a sequence implies (a) knowing

what the sequence is, (b) recognizing deviations from the
sequence, and © having a means of correcting the
deviations. At the very least, the sequence reference
specifies what to perceive at any given step in the sequence;
actions are initiated to carry out the step, resulting in a
perception that either matches or does not match the reference
for control of the perception associated with that step. The
sequence-reference sets the reference for the next step, and
so on until all steps in the sequence have been carried out.

      For example, typing the word “cat” entails

having a reference for the sequence c – a – t. To type “cat,”
the sequence controller sets a reference for a lower-level
controller to perceive typing the “c,” then “a,” then “t.”
Typing the “c” involves several levels of control that may
themselves involve sequence-control, i.e., placing the fingers
on the keyboard, moving a finger over the “c” key, then
pressing and releasing the key, and then (for a touch typist)
moving that finger back to the “home” position on the
keyboard. These hand and finger movements are carried out by
even lower level controllers.

      The sequence controller may not wait for

tactile or visual feedback to confirm that the “c” has been
typed before initiating the typing of the “a.” Once a mistake
has been made, by the time the sequence controller receives
the feedback it is too late for on-the-fly correction. I’m a
fairly good touch typist but when I make a mistake I’ve
usually typed a number of additional letters before becoming
aware of it (the tactile perception is wrong). Error
correction in sequence control usually involves interrupting
the sequence to make the correction before continuing, and
sometimes involves starting the sequence over again.

      In the example just given, the sequence to

be produced is known in advance and one can compare the
perception of that sequence to its reference sequence. As in
typing or playing music, there is a known pattern to be
reproduced. (The pattern may include more than just the
sequence of letters or notes, but also patterned variations in
the speeds of execution, intensity of the perceptions, etc.)
We can learn to control sequences. But I believe that most
examples in which a sequence of activities is carried out do
not involve pre-specified sequence references and therefore
are not examples of sequence control. The same may be said of
program control.

      In program control, there is a prespecified

“program” to be carried out – a program reference that
specifies what is to be done at each step, including
alternatives whose execution depends on current perceptions:
If this perception, then do that, else do this other thing.
Certainly we can learn what should be perceived at each step,
what alternatives if any are available at that step, and
whether what should be perceived is being perceived. But I’m
doubtful that we do much of this. Instead, we have a goal to
achieve and string together activities on the fly that (we
hope) will move us closer to achieving that goal, based on
prior experience and what options appear to be available to us
at the time. After we have achieved the goal and look back on
how it was accomplished, it may appear that we were executing
a pre-planned program, but most of the time we just did what
occurred to us at the time, step by step.

      Most of what we do is improvisation, not

programmed execution.

Comments?

Bruce A.

[ts1]

Found the sequencer (?)

http://www.cell.com/current-biology/fulltext/S0960-9822(18)30033-2

control independent of specific movements

Bob Eichler

···

From: Bruce Abbott

Sent: Monday, February 12, 2018 8:02 AM

To: CSGnet

Subject: Control of sequences and programs

[From Bruce Abbott (2018.02.12.1100 EST)]

Although I have no doubt that one can control sequences and programs, I doubt that we do very much of either.

To control a sequence implies (a) knowing what the sequence is, (b) recognizing deviations from the sequence, and © having a means of correcting the deviations. At the very least, the sequence reference specifies what to perceive at any given step in the sequence; actions are initiated to carry out the step, resulting in a perception that either matches or does not match the reference for control of the perception associated with that step. The sequence-reference sets the reference for the next step, and so on until all steps in the sequence have been carried out.

For example, typing the word “catâ€? entails having a reference for the sequence c – a – t. To type “cat,ât,â€? the sequence controller sets a reference for a lower-level controller to perceive typing the “c,â€? then “a,â€? then “t.â€? Typing the “câ€? involves several levels of control that may themselves involve sequence-control, i.e., placing the fingers on the keyboard, moving a finger over the “câ€? key, then pressing and releasing the key, and then (for a touch typist) moving that finger back to the “homeâ€? position on the keyboard. These hand and finger movements are carried out by even lower level controllers.

The sequence controller may not wait for tactile or visual feedback to confirm that the “c� has been typed before initiating the typing of the “a.� Once a mistake has been made, by the time the sequence controller receives the feedback it is too late for on-the-fly correction. I’m a fairly good touch typist but when I make a mistake I’ve usually typed a number of additional letters before becoming aware of it (the tactile perception is wrong). Error correction in sequence control usually involves interrupting the sequence to make the correction before continuing, and sometimes involves starting the sequence over again.

In the example just given, the sequence to be produced is known in advance and one can compare the perception of that sequence to its reference sequence. As in typing or playing music, there is a known pattern to be reproduced. (The pattern may include more than just the sequence of letters or notes, but also patterned variations in the speeds of execution, intensity of the perceptions, etc.) We can learn to control sequences. But I believe that most examples in which a sequence of activities is carried out do not involve pre-specified sequence references and therefore are not examples of sequence control. The same may be said of program control.

In program control, there is a prespecified “programâ€? to be carried out – a program reference that specifies what is to be donee at each step, including alternatives whose execution depends on current perceptions: If this perception, then do that, else do this other thing. Certainly we can learn what should be perceived at each step, what alternatives if any are available at that step, and whether what should be perceived is being perceived. But I’m doubtful that we do much of this. Instead, we have a goal to achieve and string together activities on the fly that (we hope) will move us closer to achieving that goal, based on prior experience and what options appear to be available to us at the time. After we have achieved the goal and look back on how it was accomplished, it may appear that we were executing a pre-planned program, but most of the time we just did what occurred to us at the time, step by step.

Most of what we do is improvisation, not programmed execution.

Comments?

Bruce A.

[rsmarken]

[From Rick Marken (2018.02.12.2130 PDT)]

Bruce Abbott (2018.02.12.1100 EST)--

Â

BA: Although I have no doubt that one can control sequences and programs, I doubt that we do very much of either.Â

 RM: The demo is designed only to show what it means to control a sequence and a program perception. I think the question of how much we actually control such variables can be answered only by research. But I see that Erling made a pretty good observation about how much you were controlling for sequences in your post. I suspect people control sequence perceptions far more than you think!

BA: To control a sequence implies (a) knowing what the sequence is, (b) recognizing deviations from the sequence, and (c) having a means of correcting the deviations.

RM: Exactly. You have to be able to perceive that a sequence is in progress; you have to have a reference for the desired state of the sequence; the reference, of course, defines what constitutes a deviation from the sequence. And you must be able to act to keep the sequence perception in the desired state, protected from disturbances which would "push" the sequence away from its reference state. In the demo, the means (actions) that keep the sequence in the desired state is very simple; a press of the space bar.
Â

 BA: At the very least, the sequence reference specifies what to perceive at any given step in the sequence;

RM: This is not how it's conceived of in PCT. I recommend (to everyone) that they read the section of B:CP on control of sequences (pp. 139-146 in the 2nd edition). On p. 139 Bill notes that, according to the model, "the perceptual signal emitted by a fifth order [sequence perception] input function will not be a sequence, but will stand for the presence of a particular sequence" and perceptual signal "...must be a one-dimensional signal if unambiguous control is to be achieved...". This means that, as in all control systems in the PCT model, the reference signal for a sequence perception is a one-dimensional signal that that specifies the magnitude of the one-dimensional perceptual signal whose magnitude indicates the degree to which a particular sequence is occurring.Â
Â

BA: actions are initiated to carry out the step, resulting in a perception that either matches or does not match the reference for control of the perception associated with that step. The sequence-reference sets the reference for the next step, and so on until all steps in the sequence have been carried out.

RM: This is not the way a sequence control system works in PCT. But that's irrelevant to my demo, which demonstrates the fact of control of a sequence and the fact that what is controlled when you control a sequence is not a sequences of actions that produce the sequence but a perception of the sequence that is a consequence of those actions (which are not necessarily a sequence, as in the demo)  combined (as usual) with effects produced by the environment (disturbances); in the demo these disturbances are the program statements that produce the display of the shapes of different sizes and colors. >

Â

BA: For example, typing the word “catâ€? entails having a reference for the sequence c – a – t. To type †“cat,â€? the sequence controller sets a reference for a lower-level controller to perceive typing the “c,â€? then “a,â€? then “t.â€? Typing the “câ€? involves several levels of control that may themselves involve sequence-control, i.e., placing the fingers on the keyboard, moving a finger over the “câ€? key, then pressing and releasing the key, and then (for a touch typist) moving that finger back to the “homeâ€? position on the keyboard. These hand and finger movements are carried out by even lower level controllers.

Â

BA: The sequence controller may not wait for tactile or visual feedback to confirm that the “c� has been typed before initiating the typing of the “a.�  Once a mistake has been made, by the time the sequence controller receives the feedback it is too late for on-the-fly correction. I’m a fairly good touch typist but when I make a mistake I’ve usually typed a number of additional letters before becoming aware of it (the tactile perception is wrong). Error correction in sequence control usually involves interrupting the sequence to make the correction before continuing, and sometimes involves starting the sequence over again.

RM: This is all correct. I think it would be great if you could create a system that controls a perception of sequence in this way. Â >

Â

BA: In the example just given, the sequence to be produced is known in advance and one can compare the perception of that sequence to its reference sequence. As in typing or playing music, there is a known pattern to be reproduced. (The pattern may include more than just the sequence of letters or notes, but also patterned variations in the speeds of execution, intensity of the perceptions, etc.) We can learn to control sequences. But I believe that most examples in which a sequence of activities is carried out do not involve pre-specified sequence references and therefore are not examples of sequence control. The same may be said of program control.

RM: That may be. It's certainly easy to find examples of behaviors that look like intended (controlled) sequences or programs that are not. For example, a person will move the mouse in a back and forth sequence of motions to protect the cursor from a sine wave disturbance in a compensatory tracking task; so the person is producing a sequence of movements but it is not produced intentionally (controlled). And compensatory tracking itself can look like a program of actions of the form " if cursor moving left, move mouse to right; if cursor moving right move mouse left; if cursor is stationary, don't move mouse". These are more examples of behavioral illusions, where the side effects of control are seen as emitted (caused) outputs; that is, they are seen as what the person is "doing", when they are not.>

Â

BA: In program control, there is a prespecified “programâ€? to be carried out – a program rreference that specifies what is to be done at each step,

RM: No, the reference for a program specifies what is to be perceived; behavior is the control of perception, remember. My demo shows what it means to control a program perception, without having to do anything other than to perceive the fact that the program is "happening"; when the program perception stops happening all you have to do is press the bar to get it to keep happening again. The maintenance of programs in real life typically requires a great deal more than just pressing a bar; but that's why, in real life, it often looks like a program is a process of emitting a contingent sequence of actions-- another behavioral illusion. My demo pares things down so that it's easier to see that carrying out a program is the control of input (perception) not the emission of output.

BA: including alternatives whose execution depends on current perceptions: If this perception, then do that, else do this other thing. Certainly we can learn what should be perceived at each step, what alternatives if any are available at that step, and whether what should be perceived is being perceived. But I’m doubtful that we do much of this.Â

RM:Â I don't know why you think this. We even have a phase we use when we see that a program is not being controlled correctly: "get with the program". I think the control of program perceptions is very common in people and other creatures; but, again, how prevalent it is is a matter for research. I'm just trying to make sure that the PCT concept of sequential and programmatic behavior as the control of sequence and program perceptions, respectively, is understood.

BA: Instead, we have a goal to achieve and string together activities on the fly that (we hope) will move us closer to achieving that goal, based on prior experience and what options appear to be available to us at the time. After we have achieved the goal and look back on how it was accomplished, it may appear that we were executing a pre-planned program, but most of the time we just did what occurred to us at the time, step by step.Â

Â

BA: Most of what we do is improvisation, not programmed execution.

RM: All of what we do is control! Some of what we do is building control systems through trial and error, a process called reorganization (which can look like "improvisation"). None of what people do is programed execution; it's control of perceptions, some of which are program perceptions.
 >

 BA: Comments?

RM: You got 'em. Â
Best

···

Â

Bruce A.Â

--
Richard S. MarkenÂ
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

[Abbott]

[From Bruce Abbott (2018.02.13.0930 EST)]

Thanks for this, Erling. I had hoped to stimulate some discussion, for the benefit of my own thinking if for no other reason.

Erling Jorgensen (2018.02.12 1150 EST)]

Bruce Abbott (2018.02.12.1100 EST)

BA: Although I have no doubt that one can control sequences and programs, I doubt that we do very much of either.

BA: …[example of typing]…

BA: We can learn to control sequences. But I believe that most examples in which a sequence of activities is carried out do not involve pre-specified sequence references and therefore are not examples of sequence control.

EJ: Well, … a word count of your post suggests you engaged in sequence control over 600 times. Your post seems more like evidence that we DO control sequences. It would help to hear examples of where it seems like sequence control but where you have your doubts.

I used typing a word as an example of sequence control so of course my post is full of such examples. But when I pay attention to my own activities, I don’t perceive myself to be executing such prespecified sequences all that often. Even when typing, I am composing my sentences on the fly and thus the specific words I type are not part of a pre-specified sequence (although I do impose grammatical constraints on their ordering and other features).

EJ: I’m not sure what you’re emphasizing by singling out “pre-specified sequence references.” Whenever there is corrective action, isn’t that a sign of a pre-existing or previously-specified reference? Are you questioning the so-called steps in a sequence? Are you thinking that the reference for the second step arises from completion of the first step? I view that only as a gating function.

According to PCT, sequence control is control that produces a particular sequence of perceptions. There is a reference that specifies what sequence is to be perceived. When typing, if I just hit keys at random, I produce a sequence of letters, but there was no control system organized to produce that specific sequence of letters.

EJ: It seems to me that a reference for the whole enactment (see discussion below about that piece) is put in place, as well as references for each step of the process. The gating of when to move from one step to the next could be handled via gain modulation, although those variations would need to occur at a very rapid rate for some sequences. As a first approximation of a model, the gain for step 2 would be low until there is a signal of completion from step 1. This doesn’t fully attend to the situation you describe of mistakenly typing a word, where “by the time the sequence controller receives the feedback it is too late for on-the-fly correction.” Perhaps we should think in terms of a rolling gain function, like a sine wave, where each step ideally takes place as its corresponding gain reaches the crest of the wave. This would still allow for glitches if a given step got interrupted while the sine wave gain of a later step overtook the interrupted one.

I like the fact that you are considering possible models of sequence control; up to now the only proposal on the table is the one outlined by Bill Powers in B:CP, which Bill admitted was only provided to show how sequence control might be accomplished using simple neural elements. The actual mechanism behind sequence control probably is far more complex.

EJ: There is still room for initiating a given step before there is time for confirming feedback from a previous step. Bill Powers wrote – I think it was in B:CP – about a pianist playing a passage faster than the perceptions for that level could occur, where the only way to deal with an error would be to ignore it or use the method you describe of interrupting the whole sequence and playing it again.

EJ: Let me go back to the notion above of “the whole enactment.” It seems to me that, if there be sequence control, a reference for the entire sequence is present, even if the steps are being gated. So there are really two things going on, the whole and its parts. I notice that Bill P. had a word for the wholes we are considering here: He called them “Events.” That is the word he utilized for the 5th level of his perceptual hierarchy, when he relocated “Sequences” to a higher spot closer to “Programs,” when the original 9 levels of the hierarchy expanded to 11 levels.

EJ: I am of the mind that Sequences and Events are more intimately tied, and that a whole enacted event is comprised of its sequentially ordered parts. In my Part 1 contribution to the forthcoming LCS-IV, “How the Brain Gets a Roaring Campfire: 1. Structuring for Perceptual Results,” I argue that the Sequence level should be relocated back down just below the Event level.

EJ: I claim that the essence of a Sequence is the connecting link “then”. By contrast, a properly enacted Event seems to require sequences timed-in-sequence. There is neurophysiological evidence that I cite in my chapter, that non-specific regions of the thalamus may provide the kind of timing signal that the neocortex needs to turn a sequence into a timed-sequence, i.e., an event.

EJ: In any event, contra to the doubts that you allude to, it seems to me that sequence control is pretty ubiquitous in how we operate. Perhaps that’s just because I am currently thinking from that layer, and so ‘everything is a sequence’. I would like to hear more evidence as to why you think otherwise.

Again, although there are many activities involving actual control over the sequence to be perceived, it is my impression that, most of the time, we are not engaged in producing such sequences. Behavior is inherently sequential – do this, then do that – but sequence control exists only when there is a sequence controller in place to produce a given sequence of perceptions.

Do you intend to implement your proposed model in a simulation?

Bruce

[Abbott]

[From Bruce Abbott (2018.02.13.1050 EST)]

Rick Marken (2018.02.12.2130 PDT)]

Bruce Abbott (2018.02.12.1100 EST)–

BA: Although I have no doubt that one can control sequences and programs, I doubt that we do very much of either.

RM: The demo is designed only to show what it means to control a sequence and a program perception. I think the question of how much we actually control such variables can be answered only by research. But I see that Erling made a pretty good observation about how much you were controlling for sequences in your post. I suspect people control sequence perceptions far more than you think!

As I pointed out to Erling, I used the typing of words as an example of sequence control, so it is no surprise to me that I did a lot of sequence control when typing that post.  But most of my activities do not seem to involve controlling specific sequences, so far as I can tell. Even the words I am typing now are not part of a pre-specified sequence of words, as would be the case if I were typing the words to a song that I had previously memorized.

BA: To control a sequence implies (a) knowing what the sequence is, (b) recognizing deviations from the sequence, and (c) having a means of correcting the deviations.

RM: Exactly. You have to be able to perceive that a sequence is in progress; you have to have a reference for the desired state of the sequence; the reference, of course, defines what constitutes a deviation from the sequence. And you must be able to act to keep the sequence perception in the desired state, protected from disturbances which would “push” the sequence away from its reference state. In the demo, the means (actions) that keep the sequence in the desired state is very simple; a press of the space bar.

BA: At the very least, the sequence reference specifies what to perceive at any given step in the sequence;

RM: This is not how it’s conceived of in PCT. I recommend (to everyone) that they read the section of B:CP on control of sequences (pp. 139-146 in the 2nd edition). On p. 139 Bill notes that, according to the model, “the perceptual signal emitted by a fifth order [sequence perception] input function will not be a sequence, but will stand for the presence of a particular sequence” and perceptual signal “…must be a one-dimensional signal if unambiguous control is to be achieved…”. This means that, as in all control systems in the PCT model, the reference signal for a sequence perception is a one-dimensional signal that that specifies the magnitude of the one-dimensional perceptual signal whose magnitude indicates the degree to which a particular sequence is occurring.

In Bill’s model, there is a separate control system for every controlled sequence. The reference signal for each is a scalar signal whose intensity indicates – what? What does it mean if the reference is zero? Some other value? The reference signal itself carries no information as to what is to be perceived. It is the organization of the currently active sequence control system that specifies what is to be perceived and in what order. The perceptual signal of the sequence control system is likewise a scalar.  The intensity of this signal indicates the degree to which the sequence has been completed, so perhaps it increases with each completed step. If so, this suggests that the reference signal’s intensity would be equal to the sum of the intensities of each successfully completed step. When the two intensities match, the sequence is complete.Â

But what if there is an error? Say I’m trying to type c-a-t and I type c-v. I assume that the sequence controller produces each individual letter-perception by setting a reference to a typed-letter control system. If this system produces “v� instead of “a,� that’s an error, but the “v� is already typed. The failure to successfully produce “a� means that the intensity of the sequence perception does not increase, but at this point the situation is ambiguous. Is this lack of increase because the letter has not yet been typed or because it was typed incorrectly? Is the execution of the sequence halted in some way by the error that has occurred in the “type ‘a’ control system? How, if at all, does the timing of the sequence enter in? It seems that there are a lot of unresolved issues here. Bill himself said in B:CP that his suggestion was only intended as a check on feasibility and not as a serious model of sequence control.

Well-practiced sequences can be executed without conscious attention, but in many cases we are consciously aware of what is to be perceived at each step in the sequence. I know that my intention is to type “cat.â€? It seems to me that I have a reference “imageâ€? that I compare to the stream of perceptions actually being produced as I execute the sequence. Perhaps the reference is merely a pointer to a stored sequence-image, and it is the imagined sequence of perceptions that is being compared to the sequence of perceptions being produced. (The “imageâ€? is not necessarily visual; in typing it could be tactile as well – what iif feels like to type “cat.â€?) Errors are mismatches between the reference image and the developing sequence perception.

Bruce

[_Erling_Jorgensen]

[From Erling Jorgensen (2018.02.13 1305 EST)]

Bruce Abbott (2018.02.13.0930 EST)
BJ: Do you intend to implement your proposed model in a simulation?

EJ: Do I know how to do such a thing?
EJ: I appreciate your encouragement. However --
a) "there are a lot of unresolved issues here" (as you note in your later reply to Rick), and
b) there are a lot of unresolved abilities (as I note in this reply!) so
c) it still seems like too many moving pieces to me.
All the best,
Erling

Confidentiality: This message is intended only for the addressee, and may contain information that is privileged and confidential under HIPAA, 42CFR Part 2, and/or other applicable State and Federal laws. If you are not the addressee, or the employer or agent responsible for delivering the message to the addressee, any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this in error, please notify the sender immediately and delete the material from your computer. Thank you for your cooperation.

Please also note: Under 42 CFR part 2 you are prohibited from making any further disclosure of information that identifies an individual as having or having had a substance use disorder unless it is expressly permitted by the written consent of the individual whose information is being disclosed or as otherwise permitted by 42 CFR Part 2.

[rsmarken]

[Rick Marken 2018-02-15_09:36:28]

Bruce Abbott (2018.02.13.1050 EST)--

 BA: At the very least, the sequence reference specifies what to perceive at any given step in the sequence;

RM: This is not how it's conceived of in PCT. I recommend (to everyone) that they read the section of B:CP on control of sequences (pp. 139-146 in the 2nd edition). On p. 139 Bill notes that, according to the model, "the perceptual signal emitted by a fifth order [sequence perception] input function will not be a sequence, but will stand for the presence of a particular sequence" and perceptual signal "...must be a one-dimensional signal if unambiguous control is to be achieved...". This means that, as in all control systems in the PCT model, the reference signal for a sequence perception is a one-dimensional signal that that specifies the magnitude of the one-dimensional perceptual signal whose magnitude indicates the degree to which a particular sequence is occurring.Â

Â

BA: In Bill’s model, there is a separate control system for every controlled sequence. The reference signal for each is a scalar signal whose intensity indicates – what? What does it mean if the reference is zerro? Some other value?Â

RM: Zero reference means "no sequence perception signal" so if the sequence is not happening, the magnitude of the perceptual signal is also zero and there is no error. What some other reference means is that some approximation to the sequence is desired. But I suspect that when one wants to produce a sequence the sequence reference signal magnitude is set to maximum, demanding that the sequence be happening.
Â

BA: The reference signal itself carries no information as to what is to be perceived.Â

RM: The reference signal specifies the desired magnitude of the perceptual signal. This is true for perceptual signals at all levels of the control hierarchy, whether the magnitude of the perceptual signal is an analog of a muscle tension, a sequence or a system concept. So in PCT the reference signal "informs" the control system of the magnitude that the perceptual signal should be brought to and maintained at.
Â

BA: It is the organization of the currently active sequence control system that specifies what is to be perceived and in what order.Â

RM: You may be right but this is not the PCT model of sequence control. Again I suggest you read the section of B:CP on sequence control. In PCT the sequence to be perceived is defined by the perceptual function that transforms whatever sequence is (or isn't) happening into the magnitude of a scalar perceptual signal.
Â

BA: The perceptual signal of the sequence control system is likewise a scalar. The intensity of this signal indicates the degree to which the sequence has been completed,

RM: The way Bill describes it in the sequence control section of B:CP I would say that the perceptual signal represents the degree to which the sequence is occurring; when a perceptual signal is present in a sequence control system it indicates "...that a particular sequence is in progress". He goes on to note that "...since this must be a one-dimensional signal if unambiguous control is to be achieved, we must ask about the dimensions of the space in which sequences exist; that is, the ways in which what we call a sequence can change". That is, to make a perceptual function that produces a perceptual signal indicating the degree to which a particular sequence is happening we have to understand the ways in which a sequence can vary.
Â

BA: so perhaps it increases with each completed step. If so, this suggests that the reference signal’s intensity would be equal to the sum of the intensities of each successfully completed step. When the two intensities match, the sequence is complete.Â

RM: My guess is that the sequence perception function produces maximal perceptual signal magnitude as long as the sequence is in progress. The magnitude of the perceptual signal will go down in proportion to the degree to which the sequence that is occurring deviates from the one that the perceptual function is designed to detect; it goes to zero if that sequence isn't happening at all. I think when a sequence is to be produced the reference for that sequence perceptual signal is at maximum value so that there is an error to the extent that there is a "break" in the sequence. If the error is large enough it will drive lower level control systems to restart the sequence; if the error is small then the output function will continue to try to produce the sequence perception; if successful, the sequence will be produce with an "error" that will have gone undetected. The "hard part" of designing a sequence control system, it seems to me, is figuring out how to turn error into outputs that restart or restore the sequence. I think the paper Bob Eichler posted:
Â
<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.cell.com_current-2Dbiology_fulltext_S0960-2D9822-2818-2930033-2D2&d=DwMFaQ&c=OCIEmEwdEq_aNlsP4fF3gFqSN-E3mlr2t9JcDdfOZag&r=G2rjwc9SjlT6Blyc8su_Md8P_xOsOTRMJ5teQVBC2qU&m=pO6o6IDkvJGRNe7FK2FmWgQ7u5XlZMcZ5-bCYaDzeCA&s=T8eXH8NyI_Rn5p7Ygsxbi-Paq9d-poMEXpgcGQPR_0w&e=&gt;http://www.cell.com/current-biology/fulltext/S0960-9822(18)30033-2
Â
suggests a possible answer; error leads to the production of a sequence of references for lower level systems that produce the sequence. These are the sequence of outputs (references to lower level systems) that are producing the perceived sequence. If the error is relatively small the sequence of outputs continues as is; if the error becomes large enough the sequence of output is reset to start over again.Â

BA: But what if there is an error? Say I’m trying to type c-a-t and I type c-v. I assume that the sequence controller produces each individual letter-perception by setting a reference to a typed-letter control system. If this system produces “v� instead of “a,� that’s an error, but the “v� is already typed. The failure to successfully produce “a� means that the intensity of the sequence perception does not increase, but at this point the situation is ambiguous. Is this lack of increase because the letter has not yet been typed or because it was typed incorrectly? Is the execution of the sequence halted in some way by the error that has occurred in the “type ‘a’ control system? How, if at all, does the timing of the sequence enter in? It seems that there are a lot of unresolved issues here. Bill himself said in B:CP that his suggestion was only intended as a check on feasibility and not as a serious model of sequence control.

RM: I agree.  >

Â

BA: Well-practiced sequences can be executed without conscious attention, but in many cases we are consciously aware of what is to be perceived at each step in the sequence.Â

RM: I agree. But I think it's best to try to figure out how sequences are controlled before getting into the question of how consciousness plays into it. Me sequence and program control demos were developed to show that we can study how people control sequence and program perceptions in the same way as we study how people control the perception of the position of a cursor on a computer screen. We don't have to worry about the involvement of consciousness in the control of sequence and program perceptions any more than we need to worry about it in the control of cursor position. Once we have a good model of control of sequence and program perception then we can start developing tests to see how consciousness plays into control of these perceptions.
Best
Rick
Best
Rick
Â
Â

I know that my intention is to type “cat.â€? It seems to me that I have a reference “imageâ€? that I compare to the stream of perceptions actually being produced as I execute the sequence. Perhaps the reference is merely a pointer to a stored sequence-image, and it is the imagined sequence of perceptions that is being compared to the sequence of perceptions being produced. (The “imageâ€? is not necessarily visual; in typing it could be tactile as well – what if feels like to type “cat.â€?) Errorrs are mismatches between the reference image and the developing sequence perception.

 >

···

Â

Bruce

--
Richard S. MarkenÂ
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery