Sequence control and 'amnesia'

The following is quoted from pp. 58-60 of
“The problem of amnesia in waking and hypnotic states”, in Rossi, E. L. (ed.) 1980, The collected papers of Milton H. Erickson on hypnosis, Vol. III: Hypnotic investigation of psychodynamic processes, NY: Irvington, pp. 58-70.

Perhaps the best examples of spontaneous amnesias of the waking state are those that occur frequently in the experience of most people. One example is the experience of being introduced to someone, shaking hands and acknowledging the introduction by repeating the person’s name, and a moment or so later wondering desperately what the person’s name is. Another common experience is asking directions when intent on reaching a certain destination, repeating the directions that are given, only to wonder a few minutes later at which intersection to turn and whether to turn right or left.

Then there is that common occurrence where a professor carefully specifies to an attentive class the day, the hour, and the room in which the final examination of the course is to be given, only to find members of the class a few minutes later in the hallway wrongly debating the hour, the day, and the room specified. This rapidly teaches instructors to say, as a preliminary remark, “Now write this down carefully,” before giving such vital information. Even so, students highly interested in the course and fully expectant of a high grade will discover their need later, perhaps by the time they reach the dormitory, to consult their notes in order to know what instructions they had been given.

One may conclude that only the second of these—forgetting a step in the directions one has been given—has to do with sequence control. Consider now the author’s comment on these:

In all three examples the primary element is something quite different from lack of attentiveness. In the first instance the social gathering is of primary importance—not some stranger’s name, no matter how attentively received. In the second instance the reaching of the destination, not the precise noting of distances, is of primary importance. Similarly, in relationship to the predicament of the students the final completion of the course is important, not the hour of the exam, which is usually different from the class hour; not the day, which has no bearing upon the course; not the specified room, which so often is not the regular classroom. We tend to spontaneously forget the parts or details of a situation when we are fixated or motivated by the total Gestalt or major goal of that situation.

Still another type of peculiar and spontaneous waking amnesia is that encountered when asking directions from a stranger for a specific goal well known to that stranger. For example, the author was riding in a car with some colleagues when we arrived at a small town. our general information was that the meeting we were to attend was being held in a hall across the street from St. Mary’s Church. Noting a woman who had just concluded an animated conversation with a gas station attendant, our driver stopped at the curb and, as the woman reached the sidewalk, asked her if she knew where St. Mary’s Church was. She answered promptly, “I certainly do. I go there for Mass every Sunday. Just drive straight ahead for exactly one mile, and at the top of the hili there, turn right. It’s right near the corner.” our driver thanked her, whereupon she asked curiously why we were going there at that time in the evening. The driver explained that we were attending a medical meeting in a building opposite the church. She replied, "Oh, yes, I know that medical building, but it’s nor straight across from the church. It’s about two houses further down.’ She was thanked, and our driver drove on two blocks, then stopped the car and declared, “Something is wrong. She was too explicit, too precise, so completely certain in each statement. I’m going back to that gas station and get correct directions.”

The gas station attendant was told precisely the directions the woman had given. The man looked puzzled, then his face cleared and he laughed and said, “My wife gave you exactly the right directions, only she must have forgotten she was downtown and not at home. you’re headed in the right direction now, but at that stoplight one block down, turn right, go about three blocks, and at the top of the hill you’ll see it-the only church there.” His directions were correct.

What brings about such an occurrence? There is a dominance of a well-established item of memory that takes precedence over the incidental, immediate realities of the situation which are spontaneously forgotten so that a response is made accurately in terms of the well-established memory belonging to a totally different situation. More than once the author has experimentally inquired for directions to the location of a landmark familiar to him. Upon receiving explicit instructions, the author has carefully repeated them with an introduction such as, “Now let’s see, we are at the corner of 7th and Wall Street and. . . .” Not infrequently the informant has remarked in astonishment, “Oh, no! I told you the way I usually think. I just forgot for the moment that I was here,” and then corrects the previous instructions.

In some of these examples it is not obvious that sequence control is involved, and concurrent control of more than one sequence in parallel, but I believe that case is not difficult to demonstrate.

The PCT model suggests that in sequence control the reference for Step n+1 ‘appears’ from memory by association with the perceptions that result from control of Step n. The relation of multitasking to working memory supports the hypothesis that the ‘amnesia’ described here is due to the need to devote working memory to the successive steps of a sequence and limitations of working memory in supporting concurrent sequence control.

Discuss.

Hi Bruce

This jogged a lot of interesting thoughts – interesting to me, anyway. Thoughts like:

  1. How do we know that we have forgotten something?
  2. How do we know that we do remember something but can’t bring it to awareness (the “Tip of the Tongue” phenomenon famously studied by Roger Brown?
  3. Where does memory fit into the PCT model?
  4. How do we know (or do we ever know) that our memories are accurate?

BN: Right now I am confident only about my answer to 3): memory fits into the PCT model for control systems at all levels of the hierarchy that control perceptions that occur over time. Very short term memory is involved in the perception of transitions; longer term memory is involved in the perception of temporal relationships (before, after), events/sequences (words); and very long term memory is involved in the perception of programs, principles and system concept perceptions.

BN: One may conclude that only the second of these—forgetting a step in the directions one has been given—has to do with sequence control.

RM: I think all the examples involve control in imagination, not actual control. Controlling the directions would involve carrying them out; producing the components of the directions in sequence. The directions would also have to be remembered as a sequence, unless contingencies are part of the directions (such as “if there is construction going on on 6th St go north on Elm Ave and turn right on 8th”), in which case what has to be remembered is a program.

BN: In some of these examples it is not obvious that sequence control is involved, and concurrent control of more than one sequence in parallel, but I believe that case is not difficult to demonstrate.

BN: The PCT model suggests that in sequence control the reference for Step n+1 ‘appears’ from memory by association with the perceptions that result from control of Step n. The relation of multitasking to working memory supports the hypothesis that the ‘amnesia’ described here is due to the need to devote working memory to the successive steps of a sequence and limitations of working memory in supporting concurrent sequence control.

BN: Discuss.

RM: I think this could be the basis for some experiments to test this idea. But the first step should be to turn this verbal model into a model that produces behavior (in the form of success and failure at recall. Maybe you could start with a flow diagram of your concept of control of sequence in memory.

Best

Rick

When what we have forgotten is part of a more complex perception, other parts of which we currently perceive with awareness. Very often, I think, the more complex perception is a sequence, but it could be e.g. a configuration (sketching a picture of someone’s house and can’t remember which side the chimney is on).

“Tip of the tongue” is a failure of association between words (“the tongue”) and visual memory. Like the jazz standard with the line “I’ll never forget … what’s her name”, I can reconstruct a visual perception of her face from memory but the associated language doesn’t connect. Leaning on authority of e.g. Elizabeth Loftus and Jill Bolte Taylor I jokingly refer to this as having to drill through the corpus callosum.

In recent years I’ve been controlling the process of making and strengthening name-person associations. When I see the person again, I’ll venture the name that comes to mind. If it’s wrong I might say “Gee, that was way off, Alicia.” If bears some similarity, I might say “Well, I was on the right track, César.” Either way, they really like it that I care to know their name, and when I get it right, well that’s a pleasure to me. Turns out to be possible to strengthen that ability.

At every synapse, according to neuroscience, so I should think that should also be true of the PCT model.

The same way that we know that our perceptions are accurate. We don’t, in any ultimate sense, but so long as control involving them is adequate they’re good enough.

Let’s look, case by case.

In this case, we may presume that unless he remembers the name he will be unable to control some other perception, probably a sequence, or a sequence as part of a program. The reference for the desired outcome of that sequence is already controlled. It is inherent in sequence control that actual control of the final perception is deferred while earlier perceptions leading to it are successively brought under control. It is inherent in sequence control that perception n+1 is brought under control only after perception n is controlled.

In passing, note that when perception n can be controlled by diverse alternative means, and disturbances to controlling one such means and/or circumstances facilitating control of another such means can have the appearance of if/then program logic to an analytical observer. There is much that is opportunistic about everyday sequence control.

At the time of the lapse of memory, the person seeks to control the next step of the sequence actually, not in imagination.

Language is much used as an aide memoire, which is of course why the experienced professor counsels them to “write this down carefully”. Language specifications are much more memorable when coupled with remembered or imagined (‘visualized’) non-language perceptions.

The “goal of the situation” is the perception to be controlled by successively controlling the perceptions that are linked together in the sequence.

From subjective observation of planning processes I believe that a novel sequence is constructed by working backward from the goal state to imagine (and create a reference for controlling) some antecedent perceptual state that is more immediately attainable (iterating this), and alternatingly/concurrently to imagine (and create a reference for controlling) some perceptual state that is immediately attainable from the presently perceived situation (again, iterating this) and which seems to provide some conditions necessary for controlling imagined later states in the developing chain of CVs serving as preconditions for control.

“Getting directions” from someone requires us to remember the words, the recipe. It’s easiest to remember when we know the territory and can recall the landmarks from memory. When we’re new to the area we are reduced imagining perceptions which are very unlikely to be a detailed match to perceptual inputs as we drive along.

She didn’t forget. She was accurately recalling from memory the landmarks that she uses, each perception n which she must control as a precondition for controlling perception n+1. She just failed to assume the point of view of the person who she was trying to help to control “arriving at the medical building across from the church”.

Another aside: Like the TCV, helping requires us to perceive things not only through control-theory glasses, but also to perceive them (as best we can) from the helped person’s point of view. So helping presupposes the TCV, and mishaps in attempting to help are perhaps best understood as tests that might further disclose what the person is controlling.

There’s an initial sketch on p. 6 of what I presented in Manchester in 2019 and posted here in Discourse a few weeks later.

All of this is in pursuit of Bill’s program of identifying what perceptions people (organisms) control and how they control them.

Hi Bruce

RM: How do we know that we have forgotten something?

BN: When what we have forgotten is part of a more complex perception, other parts of which we currently perceive with awareness.

RM: My idea was quite similar. Though I would be more explicit in my hypothesis about the basis on which we know we have forgotten something. I would say we know we have forgotten something when the higher level system “requesting” the memory (by sending reference “address” signals to lower level systems) doesn’t get the stored perception it wants resulting in an error signal that tells the higher level system that something has been forgotten.

RM: There are still problems with this model since it implies that the reference input to the higher level system knows what complex perception it wants to perceive (in imagination). And if that were true why doesn’t the higher level system just replay that reference into the perceptual input of the the still higher level systems that presumably sent it? This convinces me that building a testable PCT-based memory model is not going to be easy. But I think if we could make a plausible model of remembering and forgetting it should be possible to develop tests of the model.

RM: How do we know that we do remember something but can’t bring it to awareness (the “Tip of the Tongue” phenomenon famously studied by Roger Brown?

BN: “Tip of the tongue” is a failure of association between words (“the tongue”) and visual memory. Like the jazz standard with the line “I’ll never forget … what’s her name”, I can reconstruct a visual perception of her face from memory but the associated language doesn’t connect.

RM: This describes the phenomenon (sort of; the Tip of the Tongue phenomenon is a failure to be able to retrieve a memory that you know is there). But this description doesn’t explain why it happens. A good model of memory should be able to account for both knowing when something is forgotten and knowing (or thinking, possibly incorrectly) that it is not forgotten but difficult to retrieve.

RM: Where does memory fit into the PCT model?

BN: At every synapse, according to neuroscience, so I should think that should also be true of the PCT model.

RM: In some sense that may be true but in terms of the PCT model I think one of the main places where memory occurs - - in terms of storing variable values over time – is in perceptual function that construct of perceptual variables that represent events that occur over time. And memory is probably also involved in storage of references (addresses) sent to the used to teh lower level system that control the components of these perceptions that occur over time.

RM: I think this could be the basis for some experiments to test this idea. But the first step should be to turn this verbal model into a model that produces behavior (in the form of success and failure at recall. Maybe you could start with a flow diagram of your concept of control of sequence in memory.

BN: There’s an initial sketch on p. 6 of what I presented in Manchester in 2019 and posted here in Discourse a few weeks later.

RM: Yes, that’s a good one for sequence production:
image
Though there are still some details missing that would make it work properly. For example, there has to be some mechanism that sequentially shifts the reference from the top level system to be the reference for each lower level system. Also, to make this a memory model you would have to add the storage unit and a reference input that addresses the contents of that unit and show where these components (storage and reference) are located. I would put the storage unit above the reference input to the top level system and a reference from an even higher level system going into it with the existing perceptual (p) signal going up to that system as the remembered sequence. Several more components would have to be added as well but once you have a working model of a system that can stored and retrieve specified sequences you will be in a position to develop some tests of the model. And we would be able to compare this PCT-based model of memory to others developed by conventional cognitive psychologists.

BN: All of this is in pursuit of Bill’s program of identifying what perceptions people (organisms) control and how they control them.

RM: I think it’s a good start at trying to develop and test a PCT-based model of recall of a type of sequence perceptions, which would be a start at developing a PCT model of one of the most important phenomena studied by conventional cognitive psychologists: memory. I didn’t discuss possibilities of PCT-based research in my book because none has been done yet. But I think I should have at least discussed the PCT model of memory and imagination, some of the evidence for it (such as that discussed by Bill in B:CP) and perhaps discuss ways to proceed from Bill’s sketch in B:CP. And done it in a chapter called Cognitive Control. But, alas, I didn’t. Nevertheless, any ideas on this topic are welcome in discussions of my book.

Best

Rick

The figure you quoted is the figure for sequence recognition. It’s no more than Bill’s diagram with the several ECS inputs sketched in the places where Bill put J, OO, and S to represent phonemes of the word “juice”.

Bill’s proposed ‘recirculation loops’ are synaptic structures which do not appear in a control loop. They bear some resemblance to input functions. When the first ECS controls its input, a branch of that perceptual signal kicks off the first the first sequence-step loop. Not visible in the diagram, but assumed, other copies of that perceptual signal are distributed elsewhere concurrently. (It’s unlikely that e.g. the perceptual signal for the taste of lemonade is constructed by different perceptual input functions for different sequence perceptions and for other purposes which are not sequences, with the possibility of two perceptual input functions for the ‘same’ perception disagreeing!) The inhibitory loopback from the end is what makes it short-term memory at this level, and the perceptual signal going up from the last element is what makes a different longer-term memory at the level calling for that sequence.

Yes, the reference signals are problematic. That’s the reason one branch of the reference signal r ends at a question mark. In Manchester I confessed how unclear it was how this might work.

In the modification below, a reference signal from the calling system kicks off the first element, and then if each successive perceptual input happens to be present the sequence completes and the calling system gets the perception of completion for the sequence.

Arguably, that’s one form of sequence control (and it’s cool to note that there can be more than one form), but we certainly also want the form in which each step of the sequence gets a reference signal to control it.

This approach might work:

The idea is that as soon as perception n is controlled a signal goes to its ‘recirculation loop’, triggering input to the next ‘recirculation loop’, and in parallel at the same time a branch of that signal goes to the reference input for controlling perception n+1.

Of course it is immaterial whether or not perception n+1 is already present, but this conception captures that. Control starts with input of r, even if r=p.

In Bill’s original diagram, the input perception is assumed to be binary, fully present or not—that is, categorial. The resolution of ambiguous input among alternative candidate categories is a multi-level affair. I gave some simple language examples in my chapter of the Handbook, LCS IV.

Hi Bruce

RM (referring to the diagram below): Yes, that’s a good one for sequence production

image

BN: The figure you quoted is the figure for sequence recognition. It’s no more than Bill’s diagram with the several ECS inputs sketched in the places where Bill put J, OO, and S to represent phonemes of the word “juice”.

RM: Yes, the diagram is great. I said it was a good one for sequence production because you have the output of the highest level system (the arrow coming out of the function O) becoming the reference input to the three lower level “phoneme” control systems. If it were a recognition system the highest level would be in “passive observation mode” which means it’s output would be disconnected from all other signals; it’s just listening.

BN: Bill’s proposed ‘recirculation loops’ are synaptic structures which do not appear in a control loop. They bear some resemblance to input functions.

RM: Actually, these loops are components of the input function of the sequence control system; this input (perceptual) function is made out of the three reverberating circuits. You have correctly shown the output of this perceptual function – the line labeled p – to be the perceptual variable controlled by the sequence control system at the top of the diagram.

BN: The inhibitory loopback from the end is what makes it short-term memory at this level, and the perceptual signal going up from the last element is what makes a different longer-term memory at the level calling for that sequence.

RM: I don’t see it that way. There is, indeed, short term memory in this diagram but it is in the perceptual function itself: it is the reverberation of each circuit until the perceptual signal is completed and sent to the sequence control system. While the perceptual signal is being sent to the sequence control system the sequence perceiver is being reset to 0 by the inhibitory loopback. There is no long term memory in this model. So I don’t see how this model can account for the ‘sequence amnesia’ in the title. But it’s a nice start at a model of sequence production.

Best

Rick