Memory in perceptual input (was Website updated)

[Martin Taylor 2005.08.28.19.45]

[From Bill Powers (2005.08.24.0301 MDT)]

Dag Forssell (2005.08.23.0815 PDT)--

Just one more peanut and I can go back to sleep.

You need memory to recognize each and every letter. Otherwise, "S" is just a couple of dots on your retina.

I think you're skipping over a few levels when you say you need memory to "recognize" something. By "recognize" do you mean saying the name of the letter? You don't need memory to generate a perceptual signal that is unique to the letter S and that can be used to stand for it at higher levels. All you need is a third-order input function. Input functions don't require memory to work (other than fractional-second short-term memory for generating such things as transition signals).

No, they don't "require" memory to work. They _embody_ memory. "S" would indeed be just a lot of dots if you had never encountered one before, or even if you had previously encountered such squiggles, but never in a context in which it mattered that the squiggles went "S-way" rather than "Z-way" or "N-way". That you see "S" rather than just one pattern among myriads in your visual field represents your memory of other "S's" and their contexts.

But I think Dag means something beyond that. There's lots of evidence suggesting that we more readily perceive things we are anticipating than things that are a complete surprise, and that we more readily perceive things of a kind of which we have much experience than things of a different class (think chess configurations as seem by a grandmaster or by a novice, or animal tracks in the jungle as seen by a city-dweller or a bushman). Between the two, we seem to see more readily things of a kind we have recently been working with, even though other things in the scene may be just as familiar.

I think it's a very dubious proposition to make the bald assertion that perceptual input functions _all_ don't use volatile memory.

Martin

[From Bill Powers (2005.08.28.1929 MDT)]

Martin Taylor 2005.08.28.19.45 --

No, they don't "require" memory to work. They _embody_ memory. "S" would indeed be just a lot of dots if you had never encountered one before, or even if you had previously encountered such squiggles, but never in a context in which it mattered that the squiggles went "S-way" rather than "Z-way" or "N-way". That you see "S" rather than just one pattern among myriads in your visual field represents your memory of other "S's" and their contexts.

I think you and I use the word memory differently. I say X is remembered if a signal is being replayed into a perceptual channel from storage, in the manner of replaying a tape recording. You say (if I understand you) that X is remembered if it is produced in any way that is dependent on past experience. I would not say that an S-recognizer embodies or employs memory just because it had to be trained, or resulted from reorganization. There is no need, in a general S-recognizer, for records of past instances of the sound or appearance of S to exist in the recognizer, so I would not say it requires OR embodies memory.

A thermometer can generate representations of temperature without either using or embodying memories of temperature -- under my definition. What would you say about a thermometer-type input function?

Best,

Bill P.

···

But I think Dag means something beyond that. There's lots of evidence suggesting that we more readily perceive things we are anticipating than things that are a complete surprise, and that we more readily perceive things of a kind of which we have much experience than things of a different class (think chess configurations as seem by a grandmaster or by a novice, or animal tracks in the jungle as seen by a city-dweller or a bushman). Between the two, we seem to see more readily things of a kind we have recently been working with, even though other things in the scene may be just as familiar.

I think it's a very dubious proposition to make the bald assertion that perceptual input functions _all_ don't use volatile memory.

Martin

[Martin Taylor 2005.08.28.22.31]

[From Bill Powers (2005.08.28.1929 MDT)]

Martin Taylor 2005.08.28.19.45 --

No, they don't "require" memory to work. They _embody_ memory. "S" would indeed be just a lot of dots if you had never encountered one before, or even if you had previously encountered such squiggles, but never in a context in which it mattered that the squiggles went "S-way" rather than "Z-way" or "N-way". That you see "S" rather than just one pattern among myriads in your visual field represents your memory of other "S's" and their contexts.

I think you and I use the word memory differently.

Only in the range over which the word is used. Psychologists distinguish several kinds of memory. "Episodic memory" is one, and that is the one that seems to fit what you call "memory". It is a memory os specific instances.

I don't know what the current thinking is on the various different kinds of memory, but in more colloquial terms, the memory of how to do something is quite separate from episodic memory. You may forget how to do something, just as you may forget what you had for dinner last Thursday, but when you do remember how to, say, ride a bicycle, you just do it, whereas when you remember what you had for dinner, you may re-perceive some aspects of the dinner.

Perceptual input functions that are not directly determined by genetics, and that develop differently consequent to different experiential history, would be similar in principle to the memory for how to do something. (Both, presumably, would be the result of reorganization, but that's a PCT-theoretic statement that isn't required for this discussion).

A thermometer can generate representations of temperature without either using or embodying memories of temperature -- under my definition. What would you say about a thermometer-type input function?

That depends on how it developed. If it developed because the organism, not being born with a predefined temperature sensor, had experienced different kinds of temperature in contexts where that made a difference to some controlled perception, then yes, the temperature sensor would embody memory. If the temperature sensor was constructed independently of the organism's experience, then it wouldn't.

Martin

PS. I'm only momentarily on-line again. I may contribute tomorrow, but probably not much more until next Sunday evening.

[From Bill Powers (2005.08.29.0757 MDT)]

Martin Taylor 2005.08.28.22.31 --

I think you and I use the word memory differently.

Only in the range over which the word is used.

I don't think that is the only difference.

Psychologists distinguish several kinds of memory. "Episodic memory" is one, and that is the one that seems to fit what you call "memory". It is a memory of specific instances.

Yes, I would agree to that. Specific instances of perceptions at most of the 11 levels (I'm not sure how far down memory goes). That includes sensations, configurations, transitions, events, relationships, categories, sequences, programs, principles, and systems.

I don't know what the current thinking is on the various different kinds of memory, but in more colloquial terms, the memory of how to do something is quite separate from episodic memory. You may forget how to do something, just as you may forget what you had for dinner last Thursday, but when you do remember how to, say, ride a bicycle, you just do it, whereas when you remember what you had for dinner, you may re-perceive some aspects of the dinner.

We don't really care what "current thinking" is, do we?

I don't see the difference you see. "How to do" something, in PCT, becomes simply a reference signal. Remember that people unacquainted with control theory think of "doing" something as an output, whereas we know it is really (at the appropriate level) a controlled perception of something happening. So memory of how to do something is memory of what to perceive. While it may not be a veridicial recall of one specific incident (what recall is?), it is probably an amalgam, a "typical" perception drawn from many recordings. Forgetting how to do something is no different from forgetting any other kind of recorded perception. If you forget the recorded phrases or images in which the rules of chess are, for you, couched, you have forgotten how to play chess.

Perceptual input functions that are not directly determined by genetics, and that develop differently consequent to different experiential history, would be similar in principle to the memory for how to do something.

"Similar in principle" is a weasel term. It means "they're different, but I still say they're the same." A perceptual input function generates an output (perceptual) signal that is a function of current values of input signals. What is learned is not any particular value of the perceptual signal, but the form of a function. The present signal values do not depend on any previous states of the world or of previous perceptions of the world. They are functions of present-time lower-order perceptions, which in turn are functions of still lower perceptions, until we reach the array of present-time incoming physical stimulations from outside the nervous system. So the principle of remembering and the principle of perceiving are totally different.

(Both, presumably, would be the result of reorganization, but that's a PCT-theoretic statement that isn't required for this discussion).

But that's the basis of your claim that there is a similarity between remembering and perceiving: they both come into being through reorganization during experience with the world. If you leave that out, no vestige of similarity remains. I don't agree that this is a real similarity, but you seem to be claiming that similarity of origin implies similarity in the product.

A thermometer can generate representations of temperature without either using or embodying memories of temperature -- under my definition. What would you say about a thermometer-type input function?

That depends on how it developed. If it developed because the organism, not being born with a predefined temperature sensor, had experienced different kinds of temperature in contexts where that made a difference to some controlled perception, then yes, the temperature sensor would embody memory.

I think this confuses the form of a function with a record of past values of signals generated by that function. A thermometer performs the function of converting its own temperature to the length of a column of fluid. That, not any particular temperature reading, and certainly not how the thermometer came into being, is what makes it a thermometer. Memories are replayed records of past values of signals generated by a perceptual input function. Perceptual input functions are not replayed recordings of anything; they are neural networks that perform certain computations, converting input signals to output signals. There is no similarity that makes a similarity between those concepts (to anti-paraphrase Bateson).

If the temperature sensor was constructed independently of the organism's experience, then it wouldn't.

The essence of memory is that it be recoverable. Can the form of a function be used to recover a memory of how it was formed? Absolutely not: there is an infinity of different ways in which any given physiological function might have come into being, and even in similar organisms, there is no reason to think that similar functions came about in the same way. And anyway, there is nothing about a function that can reveal its origins: it just does what it does. In y = 2x + 3z, there is no way to determine whether the 3z was written before or after the 2x, or the 2 before or after the x, or the plus sign first, last, or in between. How the function came to be written is irrelevant to its nature as a function. There is no way to deduce, knowing only the value of the function, y, what the values of x and z were.

I think that conceptions of memory in psychology, like conceptions of learning, have mixed together a lot of very different and incommensurate ideas.

Best,

Bill P.

[Martin Taylor 2005.08.29.16.39]

[From Bill Powers (2005.08.29.0757 MDT)]

Martin Taylor 2005.08.28.22.31 --

I think you and I use the word memory differently.

Only in the range over which the word is used.

I don't think that is the only difference.

I don't know what the current thinking is on the various different kinds of memory, but in more colloquial terms, the memory of how to do something is quite separate from episodic memory. You may forget how to do something, just as you may forget what you had for dinner last Thursday, but when you do remember how to, say, ride a bicycle, you just do it, whereas when you remember what you had for dinner, you may re-perceive some aspects of the dinner.

We don't really care what "current thinking" is, do we?

I do, when it is not incompatible with PCT. I find it a good guide. But that's sort of irrelevant in this context, since I explicitly said I wasn't using it!

I don't see the difference you see. "How to do" something, in PCT, becomes simply a reference signal.

Not at all. A reference signal is "what to achieve", which has no real connection with "how to achieve it (aka. "How to do it"). "How to do" it is embodied in the developed structure of the perceptual input functions, the output functions, and their interrelationships produced by reorganization.

Forgetting how to do something is no different from forgetting any other kind of recorded perception. If you forget the recorded phrases or images in which the rules of chess are, for you, couched, you have forgotten how to play chess.

I'd say that "forgetting how to do" something could be caused by subsequent reorganization, by changes in output functions, or by changes in perceptual input functions. And I'll add a caveat here on behalf of Dag (and myself, too, really), that perceptual memory of the episodic kind may indeed be a part of the input to perceptual input functions, even though that kind of linkage falls outside of canonical HPCT.

Perceptual input functions that are not directly determined by genetics, and that develop differently consequent to different experiential history, would be similar in principle to the memory for how to do something.

"Similar in principle" is a weasel term. It means "they're different, but I still say they're the same."

I don't think it's a weasel term at all, no matter how often you reiterate that it is. It's a shorthand way of saying that the underlying structures are functionally the same, although the surface manifestations may be different. It's the same as saying that a sandpile avalanche is similar in principle to the instability of the Bomb in the Hierarchy. No weasel there, either. You understand one, you understand the other.

A perceptual input function generates an output (perceptual) signal that is a function of current values of input signals. What is learned is not any particular value of the perceptual signal, but the form of a function.

Exactly!!!!!!!

So the principle of remembering and the principle of perceiving are totally different.

Of _episodic_ remebering.

(Both, presumably, would be the result of reorganization, but that's a PCT-theoretic statement that isn't required for this discussion).

But that's the basis of your claim that there is a similarity between remembering and perceiving: they both come into being through reorganization during experience with the world. If you leave that out, no vestige of similarity remains. I don't agree that this is a real similarity, but you seem to be claiming that similarity of origin implies similarity in the product.

Not in the slightest. What is a reasonable definition of "memory" but the present representation or effect of a past state? A "memory" that allows the re-evocation of a perceptual moment is one such representation. A memory that comes from the experience of many repetitions of some state is another. It's still memory, and I can't see any other way that perceptual input functions or output functions can develop in ways that permit perceptual control in a complex and dynamic environment. And then there's reorganization, too, which has a different kind of relation to the past experiences of the organism. It's the "winter leaf" effect -- if it ain't broke, don't fix it. Maybe that doesn't deserve the name of "memory", but it's an arguable point.

A thermometer can generate representations of temperature without either using or embodying memories of temperature -- under my definition. What would you say about a thermometer-type input function?

That depends on how it developed. If it developed because the organism, not being born with a predefined temperature sensor, had experienced different kinds of temperature in contexts where that made a difference to some controlled perception, then yes, the temperature sensor would embody memory.

I think this confuses the form of a function with a record of past values of signals generated by that function.

You may do the confusing, but I don't. I'm talking about the generation of the function.

A thermometer performs the function of converting its own temperature to the length of a column of fluid.

One kind of constructed thermometer does. How do the ones that allow you to perceive that you are touching a warm or cold surface do it?

That, not any particular temperature reading, and certainly not how the thermometer came into being, is what makes it a thermometer.

Of course, what makes it a thermometer is its ability to turn temperature into some other signal. I never intended to talk about the derivation of the word "thermometer", and you didn't ask me to. You asked whether a thermometer (not "a thermometer") embodied memory, and I answered with what I though was a tautology: if the thermometer developed because of the memory of past states, then "yes", but if its construction did not involve any of its past experience with temperatures, then "no".

Memories are replayed records of past values of signals generated by a perceptual input function.

Episodic memory again.

The essence of memory is that it be recoverable.

Really? But I guess that if you force that statement into a definition, then you can use it to make what argument you want about what is or is not memory. It's a bit like saying that the essence of civilization is that the people in it swear allegiance to the Stars and Stripes every morning, and then using that to argue that the Chinese aren't civilized.

I think that conceptions of memory in psychology, like conceptions of learning, have mixed together a lot of very different and incommensurate ideas.

That, precisely, is what several decades of conventional psychologists have been disentangling. I think that's one baby you shouldn't throw out with the bathwater. No matter what the background theory, it's hard to deny that there's a difference between remmebering how to ride a bike and remembering what you had for dinner last Thursday, and that both involve what most people would call "memory".

Martin

[From Rick Marken (2005.08.29.1510)]

Martin Taylor (2005.08.29.16.39)

Bill Powers (2005.08.29.0757 MDT)--

A perceptual input function generates an output (perceptual) signal
that is a function of current values of input signals. What is
learned is not any particular value of the perceptual signal, but
the form of a function.

Exactly!!!!!!!

So that should be that. This all started off with Martin saying:

Martin Taylor (2005.08.28.19.45) --

I think it's a very dubious proposition to make the bald assertion
that perceptual input functions _all_ don't use volatile memory.

Martin is supporting the idea, which is a big part of Dag's description of
the input (perceptual) function in his _Once Around the Loop_ paper
(http://www.livingcontrolsystems.com/intro_papers/once_around_loop.pdf),
that the input function is "a neural network that constructs interpretations
of them [the various signals] using both the current signals and signals
retrieved from memory". In fact there is no reason to believe that
perceptions are based, in part, on inputs from memory. In fact, there is
good reason to believe that they are not based on memory at all since this
would result in very poor control.

The evidence that Martin mentions for perceptions being based, in part, on
inputs from memory is the following:

There's lots of evidence
suggesting that we more readily perceive things we are anticipating
than things that are a complete surprise, and that we more readily
perceive things of a kind of which we have much experience than
things of a different class (think chess configurations as seem by a
grandmaster or by a novice, or animal tracks in the jungle as seen by
a city-dweller or a bushman). Between the two, we seem to see more
readily things of a kind we have recently been working with, even
though other things in the scene may be just as familiar.

I can't see how these examples suggest that perception might be based inputs
from memory. Assuming that "more rapidly perceive" means "identifying (with
a word) more quickly", I would say that identifying more quickly what we
anticipate sounds like a result of having a reference already set for the
"anticipated" perception. It's not a quicker perception but a quicker match
to a reference (the reference, not the perception, comes from memory). The
"familiarity" result sounds like a reference phenomenon as well; familiar
perceptions are one's which we have stored (as potential references) in
memory.

I would like to know why those of you who think so think that perception
depends in part on inputs from memory? My perceptions seem to remain
completely constant as I remember different things that seem pertinent to
those perceptions. For example, my perception of my computer screen stays
just the same as I reminisce about all the computer screens I used to know,
going way back to the little TV monitor I had attached, via RF modulator, to
my RCA Cosmac 1200. I remember all these things and yet they seem to have
absolutely no influence on what I am perceiving. Doesn't that seem like
rather strong evidence that perceptual functions have no memory inputs?

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

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

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[From Bill Powers (2005.08.30.0706 MDT)]

Martin Taylor 2005.08.29.16.39 --

I don't see the difference you see. "How to do" something, in PCT, becomes simply a reference signal.

Not at all. A reference signal is "what to achieve", which has no real connection with "how to achieve it (aka. "How to do it"). "How to do" it is embodied in the developed structure of the perceptual input functions, the output functions, and their interrelationships produced by reorganization.

To be more precise, "how to do it" is embodied as a negative feedback control system -- from the standpoint of an external observer. From the viewpoint of the system itself, however, "how to do it" is simply a remembered pattern of perceptions of lower order. "How to open a door" consists of perceptions of reaching, grasping, turning, and pulling. Those perceptions are created by producing certain configurations, and those are produced by producing certain sensations. The only physical outputs that are not simply reference signals are the tensions in muscles, which are similar no matter what we are "doing." The actual output processes involved are never experienced, and they do not resemble the perceptions they are used to control (not if disturbances are present, as they almost always are).

It's not surprising that psychologists have seen different kinds of memory, since they know nothing of reference signals or controlled perceptions, and generally believe that behavior consists of patterned outputs. "Episodic" memory (which refers to perceptions) seems different from "skill" memory (or whatever they call it) because they do not think in terms of levels of control or perception, and they believe that to produce a consistent result of action, it is necessary to calculate the actions in advance of carrying them out. It seems that there are things like "muscle memory," which of course we know is not the case.

However, we may be up against something similar to what occured in the old argument about what "control" really is. That word has been used for anything that is associated with negative feedback control processes, a case of synecdoche (as in describing a flotilla of ten ships as "ten sail"). So to "control" the path of a car, we "control" the "controls" -- meaning, we produce variable output actions to operate the wheel and pedals in whatever way is needed to make the perceived path of the car match a reference condition.

Regardless of the confusion around the use of the word control, we have to bow to custom and precedent, mostly to assure that what we say about control will not be misconstrued. It takes longer than saying "control", but it's clearer to say "acting on the world in whatever direction and amount is required to make it look and feel to us the way we intend for it to look and feel." If the listener is pretty sure to be unaware of PCT, that's how we have to talk to be understood.

The same is true of the word memory. What I intend to convey by that word is the process of storage and retrieval of perceptual experiences. Addressing memories allows a higher-order system to pick out reference signals having to do with lower-order perceptions that it is incapable of recognizing. To a relationship- controlling system events, transitions, and configurations mean nothing. My hypothesis (still tentative) is that it selects memories of them to use as reference signals by issuing an address for retrieval, so it does not need to know exactly what is stored at that address, or how it will appear to a lower-order system. At the lowest levels, of course, this "translation" becomes trivial and memory -- i.e., storage and retrieval of perceptul signals -- is probably not required.

The word memory is commonly used with many other meanings. You can buy a piece of "memory wire" which, on heating, returns to its original configuration no matter how it is bent while cold. As I noted in B:CP, large muscles can be said, metaphorically, to be memories of exercise. The present trace of any past occurrance can be said to be a memory of that occurance, even if there is no way to work backward from the present trace to its causes. None of these usages conveys the precise meaning I want to convey by the word memory, which is ONLY AND ALWAYS the storage and retrieval of perceptual signals.

So I suppose I will (if you insist) have to give up using the word memory, because as you describe it it is used to indicate many processes other than what I mean by the word, so whatever I say about it is likely to be misunderstood. I had to adopt a new term, "reorganization", to replace "learning" for the same reason; when I say "learning" other people will hear it in various ways I do not mean. I noted two in BCP: memorizing and abstract generalization.

I suppose this must be how jargon makes its way into new disciplines. New ideas simply require new words, or words lying around unused that can be stripped of their old meanings and retreaded. If one goes on using old words in the old ways, all that will be heard are old ideas.

Best,

Bill P.

[From Jeff Vancouver (2005.08.30.0940 EST)]

[From Rick Marken (2005.08.29.1510)]

>Martin Taylor (2005.08.29.16.39)
>
>> Bill Powers (2005.08.29.0757 MDT)--
>>
>> A perceptual input function generates an output (perceptual)

signal

>> that is a function of current values of input signals. What is
>> learned is not any particular value of the perceptual signal, but
>> the form of a function.
>
> Exactly!!!!!!!

So that should be that. This all started off with Martin saying:

> Martin Taylor (2005.08.28.19.45) --

> I think it's a very dubious proposition to make the bald assertion
> that perceptual input functions _all_ don't use volatile memory.

Martin is supporting the idea, which is a big part of Dag's

description of

the input (perceptual) function in his _Once Around the Loop_ paper

(http://www.livingcontrolsystems.com/intro_papers/once_around_loop.pdf),

that the input function is "a neural network that constructs
interpretations
of them [the various signals] using both the current signals and

signals

retrieved from memory". In fact there is no reason to believe that
perceptions are based, in part, on inputs from memory. In fact, there

is

good reason to believe that they are not based on memory at all since

this

would result in very poor control.

Rick, are you saying that input functions never receive signals from
memory? That seems contradictory to B:PC.

The evidence that Martin mentions for perceptions being based, in

part, on

inputs from memory is the following:

> There's lots of evidence
> suggesting that we more readily perceive things we are anticipating
> than things that are a complete surprise, and that we more readily
> perceive things of a kind of which we have much experience than
> things of a different class (think chess configurations as seem by a
> grandmaster or by a novice, or animal tracks in the jungle as seen

by

> a city-dweller or a bushman). Between the two, we seem to see more
> readily things of a kind we have recently been working with, even
> though other things in the scene may be just as familiar.

I would like to know why those of you who think so think that

perception

depends in part on inputs from memory? My perceptions seem to remain
completely constant as I remember different things that seem pertinent

to

those perceptions. For example, my perception of my computer screen

stays

just the same as I reminisce about all the computer screens I used to
know,
going way back to the little TV monitor I had attached, via RF

modulator,

to
my RCA Cosmac 1200. I remember all these things and yet they seem to

have

absolutely no influence on what I am perceiving. Doesn't that seem

like

rather strong evidence that perceptual functions have no memory

inputs?

How does an example of a perception not based on memory (let's assume
that is true) counter an argument that perceptions might sometimes, in
part or whole, be based on memory?

Also, going back to your poor control statement, this seems the crux
issue. The argument that control based on memory is not as good as
control based on real-time perception is, I think, unassailable (and
important). However, that argument does not mean that memory-based, or
supported, control is unreasonable. Otherwise, "modern" control theory
would not have developed. Yet, memory-based or supported control does
sometimes substantially improve control, hence it is programmed into
modern devices.

Does this mean we humans have this type of capacity? No, but it does
mean the poor control argument against memory is fallacious.

So what arguments support or refute memories sometimes use in input
functions? The argument refuting modern control theory models, as best
as I can tell, is that they are more sophisticated computationally then
are neuro-circuitry could develop. This can be counter in two ways.
First, unsophisticated neuro-circuitry can handle the job pretty well
(that is what I model). Second, our neuro-circuitry is pretty
sophisticated (the hierarchical notion is an argument along those
lines).

In terms of arguments supporting use of memory, it seems the strongest
is: what is it for? I also have a simple model of memory-based inputs to
an input function that nicely handles data from several studies. Add
that to Martin's arguments, and it seems the evidence is leaning to the
support side more than the refute side. But as I said earlier, this
might be a case of arguing all or none when both parties mean some.

Jeff Vancouver

[From Bill Powers (2005.08.30.0752 MDT)]

Rick Marken (2005.08.29.1510) --

The evidence that Martin mentions for perceptions being based, in part, on
inputs from memory is the following:
I can't see how these examples suggest that perception might be based inputs
from memory.

When you vacuum "a rug," what you are vacuuming is a large expanse of material with pieces missing under the sofas, chairs, tables, and even shoes left lying around. Right? Wrong. If you thought there were pieces missing just because you couldn't see them, you wouldn't be vacuuming the rug -- you'd be throwing it away. To keep from experiencing errors like that, you imagine that the rug continues under the furniture where it can't actually be seen. You are controlling something made up, in part, of imagined perceptions.

So I have to disagree with you here. Most of our perceptions contain "assumptions," which are actually imagined details that complete the perception but have to be filled in because they are not actually there. I don't go as high as Dag's 90%, but I'd agree that it's a substantial part.

Even the lower-order system behave this way. Think of the visual illusions in which blind spots in the visual field are filled in with the color of the surroundings. Think of what you experience when you reach in your pocket feeling for a quarter. A whole experience is often controlled just by controlling a few of its elements, with the rest simply changing passively or being imagined. When you park a car, you avoid scraping the right front fender against an adjacent car not by controlling the perception of the fender, which you can't see, but by controlling parts of the hood that you can see. Yet you imagine the fender just barely clearing the fender of the car next to you, and cringe if you imagine too small a space.

Since those imagined parts are details coming from lower-order perceptions, those details could be either imagined or real at the lower order; the higher system has no way of knowing. Only by attending to the lower order, and by experimentally wiggling things or looking for effects of disturbances, can we be consciously sure we are imagining, or not imagining.

Best,

Bill P.

[From Rick Marken (2005.08.30.0840)]

Jeff Vancouver (2005.08.30.0940 EST)

Rick, are you saying that input functions never receive signals from
memory? That seems contradictory to B:PC.

In B:CP, input functions receive signals only from lower level systems;
memory inputs go directly (via switch) into the perceptual signal path.
Whether input functions receive signals from memory is an empirically
testable question. So far, my informal tests (described below) suggest that
they do not.

For example, my perception of my computer screen stays
just the same as I reminisce about all the computer screens
I used to know...

How does an example of a perception not based on memory (let's assume
that is true) counter an argument that perceptions might sometimes, in
part or whole, be based on memory?

It's just one test. It doesn't prove that perception is never based on
memory inputs. I'd love to see a demonstration of how perceptual inputs
depend on memory inputs. Optical illusions may be an example.

...I also have a simple model of memory-based inputs to
an input function that nicely handles data from several studies.

I'd like to see that. That might be convincing.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

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

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Martin Taylor 2005.08.30.11.31]

[From Bill Powers (2005.08.30.0706 MDT)]

Martin Taylor 2005.08.29.16.39 --

I don't see the difference you see. "How to do" something, in PCT, becomes simply a reference signal.

Not at all. A reference signal is "what to achieve", which has no real connection with "how to achieve it (aka. "How to do it"). "How to do" it is embodied in the developed structure of the perceptual input functions, the output functions, and their interrelationships produced by reorganization.

To be more precise, "how to do it" is embodied as a negative feedback control system -- from the standpoint of an external observer. From the viewpoint of the system itself, however, "how to do it" is simply a remembered pattern of perceptions of lower order.

This is something quite new in my understanding of HPCT. I had thought that "how to do it" was embodied in the interconnections of the outputs from higher levels to the reference inputs of successively lower-level control units. And that boththe interconnections and the nature of the output functions were developed through experience (aka "reorganization").

I never considered that according to HPCT the higher-level outputs were passed through a device that evoked memories of specific past events in order to provide the signal value that went to the lower reference input. I had thought that the connections were as direct as are the connections from one level of perceptual input to the next.

"How to open a door" consists of perceptions of reaching, grasping, turning, and pulling.

Given that I have obviously long misunderstood the nature of the connections on the output side of the hierarchy, I'd like a little tutorial on how the value of the higher-level output selects which of the myriads of memories of opening doors to use when determining the reference levels for the lower units.

And while you are at it, since what the higher-level output evokes are remembered perceptions, why these perceptions are not perceivable -- as you say: "The actual output processes involved are never experienced, and they do not resemble the perceptions they are used to control (not if disturbances are present, as they almost always are)."

Also in connection with that last quote, I had been under the impression that the output processes produced signals that had some time-varying value, not perceptions that might or might not resemble other perceptions. I'm afraid my understanding of HPCT has been really out in left field for years and years.

Look, I'm not saying you are wrong in asserting that the reference values at each level are generated by evoking for each control unit the memory of the perception of a specific earlier occurrence. I'm just saying that I had never considered HPCT as incorporating a memory look-up function between the output at one level and the reference input at the next lower level, and that I don't understand how the signal value that is generated by the output function is used to evoke the correct memory from that look-up function out of the myriads of similar situations one would have encountered in a lifetime. And finally, I don't see why using the output signal to look up the perception of a past event is necessary; it seems an extra function that adds nothing to the system performance but an opportunity for things to malfunction.

It's not surprising that psychologists have seen different kinds of memory, since they know nothing of reference signals or controlled perceptions, and generally believe that behavior consists of patterned outputs.

Red herring alert!

Whether you are a "classical" psychologist or a student of PCT, the same evidence of difference applies: that some memories are of specific instances and can be re-played as perceptions of those instances (though perhaps in an impoverished way), whereas other memories are cumulative, built from many similar experiences. Regardless of your theoretical assumptions, it's hard to see how those two kinds of memory would be stored in the same way. (We are talking now about the perceptual input side, not the action side of the hierarchy).

"Episodic" memory (which refers to perceptions) seems different from "skill" memory (or whatever they call it) because they do not think in terms of levels of control or perception, and they believe that to produce a consistent result of action, it is necessary to calculate the actions in advance of carrying them out.

I can't see the logical sequitur in this sentence. If anything, the "because" part seems to suggest that they would be more, not less, likely to see episodic and skill memory as the same; they would be doing as you suggest is true of HPCT, using memories of past individuated events to compute what actions to perform. It's "classical HPCT" that denies this, at least as I previously understood it before reading the message to which this is a reply.

....
The same is true of the word memory. What I intend to convey by that word is the process of storage and retrieval of perceptual experiences.

....

The word memory is commonly used with many other meanings. You can buy a piece of "memory wire" which, on heating, returns to its original configuration no matter how it is bent while cold. As I noted in B:CP, large muscles can be said, metaphorically, to be memories of exercise. The present trace of any past occurrance can be said to be a memory of that occurance, even if there is no way to work backward from the present trace to its causes. None of these usages conveys the precise meaning I want to convey by the word memory, which is ONLY AND ALWAYS the storage and retrieval of perceptual signals.

So I suppose I will (if you insist) have to give up using the word memory, because as you describe it it is used to indicate many processes other than what I mean by the word, so whatever I say about it is likely to be misunderstood.

You could say "episodic memory", if you didn't object to using a technical term used also by "classical" psychologists. That would be easily understood by non-PCT people, and would have exactly the denotation and connotation you want. "Ep-mem" would be a suitable short form, which I haven't heard used elsewhere. However, in many contexts, you could simply use "memory", just as in an appropriate context "abstract" means a short resume of a document, rather than a non-concrete concept.

I suppose this must be how jargon makes its way into new disciplines. New ideas simply require new words, or words lying around unused that can be stripped of their old meanings and retreaded. If one goes on using old words in the old ways, all that will be heard are old ideas.

There's certainly some truth in that. The jargon words help make things precise for those who have the appropriate background, while making it hard for others to join the group in the know. The latter is a (possibly unfortunate) side-effect.

Martin

[From Jeff Vancouver (2005.08.30.1155 EST)]

[From Rick Marken (2005.08.30.0840)]

In B:CP, input functions receive signals only from lower level

systems;

memory inputs go directly (via switch) into the perceptual signal

path.

. . . which feeds into input functions in higher-order systems.

I'd like to see that. That might be convincing.

I am working on it for journal submission. It will likely take another
month at least given other priorities.

Jeff

[From Rick Marken (2005.08.29.0910)]

Bill Powers (2005.08.30.0752 MDT)]

Rick Marken (2005.08.29.1510) --

The evidence that Martin mentions for perceptions being based, in part, on
inputs from memory is the following:
I can't see how these examples suggest that perception might be based inputs
from memory.

To keep from
experiencing errors like that, you imagine that the rug continues
under the furniture where it can't actually be seen. You are
controlling something made up, in part, of imagined perceptions.

This does not seem to me to be an example of memory influencing perception,.
I think it's just a case of a perceptual function putting out a particular
output when there are sufficient inputs to generate that output. I would
still see the rug as a complete rug, for example, even if I remembered that
it had actually been cut in places to fit around the furniture. I still see
Kind Kong running around, beating up dinosaurs on Skull Island even though I
remember that these are just small puppets being move a little at a time.

So I have to disagree with you here.

And I have to disagree with your disagreement;-)

Most of our perceptions contain "assumptions," which are actually
imagined details that complete the perception but have to be filled
in because they are not actually there.

You'll have to prove that to me experimentally. Right now, I think I can
account for the same phenomenon with bringing in imagination. Subjectively,
imagination is a whole different thing. When I perceive the "continuous" rug
I am not actually seeing the missing parts; I see a rug with stuff on it,
obscuring parts of it. I can imagine what the whole rug looks like (without
the stuff on it) but when I look at the rug I perceive something different
than I imagine.

I don't go as high as Dag's 90%, but I'd agree that it's a
substantial part.

I'm still at 0%. You'll have to do a _lot_ better to convince me that
perception ever depends on memory inputs.

Even the lower-order system behave this way. Think of the visual
illusions in which blind spots in the visual field are filled in with
the color of the surroundings.

Again, that seems like that can be easily explained in terms of perceptual
input functions. A "line detector" input function can be designed to put a
perceptual signal indicating "line" when the input is

1 1 1 1 1 1 1 1

Or

1 1 1 1 1 1 1

The perception in both cases would be

···

______________

No imagination is needed to fill in for the blind spot given the perceptual
function model used by PCT.

But I would love to see a demonstration of the influence of memory on
perception. It seems to me that such a demonstration would showing that I
can change what I am perceiving by changing my memory. I can change what I
perceive by changing the lower level inputs to my perceptual functions. I
can do this really easily by changing my gaze from one location to another
(or by watching a movie, which changes the inputs for me). If perception
depends, to some extent, on memory inputs then I should see the same kind of
change in perception by changing what I am remembering. I haven't been able
to think of a situation where that works. Maybe you can. After all, you
think that nearly 90% of our perceptions are based on memory. Seeing a
change in what I perceive as a result of changing what I recall is an
experience I would really like to have.

Best

Rick
--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

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

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Martin Taylor 2005.08.30.11.58]

[From Rick Marken (2005.08.30.0840)]

Jeff Vancouver (2005.08.30.0940 EST)

Rick, are you saying that input functions never receive signals from
memory? That seems contradictory to B:PC.

In B:CP, input functions receive signals only from lower level systems;
memory inputs go directly (via switch) into the perceptual signal path.
Whether input functions receive signals from memory is an empirically
testable question. So far, my informal tests (described below) suggest that
they do not.

As you say, HPCT allows imagined (from memory) inputs to substitute for external-world inputs in any perceptual input function. You suggest that your informal tests argue that this doesn't happen. So be it. This message is just about HPCT as you cite it from B:CP.

The output from a perceptual input function is used as part of the input to a next higher-level function. The input to the lower-level function might come directly (through a switch) from memory. The rest of the input to the higher-level function comes from the outputs of other lower-level perceptual functions, which may be working on data from the external world. Therefore the input to that next higher-level function could be based partly on input from the external world and partly on input from memory.

On the surface, there are two questions at issue. One is whether the inputs to a perceptual input function could be partly based on memory and at the same time partly on the outer world. The other is whether the nature of a perceptual input function could be affected by input from memory.

If you look a little deeper, however, the two questions are really only one. Imagine a function such as x*(y^2) + (x^2)*y. Suppose x is fairly steady and y changes relatively quickly. While x is fairly small compared to y, the variation in the output of this function looks like y^2, whereas while x is large compared to y, it looks like y. Now, if the value of x comes from memory (through the imagination loop switch at the level below), the nature of the function can seem to change between linear and quadratic.

That's just a trivial example that can happen at a rather low level. At higher levels, where the functions get more complicated (especially when you get to logical levels), there are lots of possibilities for radical changes in the apparent behaviour of perceptual functions dependent on context, whether the context comes from memory or from the outer world.

Whether they do in practice is another question. I'm just saying that HPCT with the imagination loop allows for perceptual inputs to incorporate both memory and direct external data, and allows for context (whether memory or direct external data) to affect the overt form of the percetpual input function. Indeed, HPCT demands that both these effects occur.

Martin

[From Rick Marken (2005.08.30.1130)]

Martin Taylor 2005.08.30.11.58]

Rick Marken (2005.08.30.0840)

Jeff Vancouver (2005.08.30.0940 EST)

Rick, are you saying that input functions never receive signals from
memory? That seems contradictory to B:PC.

In B:CP, input functions receive signals only from lower level systems;
memory inputs go directly (via switch) into the perceptual signal path.
Whether input functions receive signals from memory is an empirically
testable question. So far, my informal tests (described below) suggest that
they do not.

I'm just saying that
HPCT with the imagination loop allows for perceptual inputs to
incorporate both memory and direct external data, and allows for
context (whether memory or direct external data) to affect the overt
form of the percetpual input function. Indeed, HPCT demands that both
these effects occur.

It just doesn't seem to me, based on my subjective experience, that my
perceptions ever include inputs from imagination (memory). When I imagine
something it's a purposeful process; I actively cull up images in memory.
But these imagined perceptions are clearly "in my head"; I can easily
distinguish them from real perceptions (so far;-)).

What I want is a demonstration where by voluntarily remembering (imagining)
something my perception changes. If it does, then I'll believe that
perceptions can be based, in part, on lower level inputs that have their
source in memory. However, I think the brain is actually pretty good about
keeping imaginations out of perception. Even when the imagination is carried
by the same perceptual signal that is an input to a higher level perception
I think that imagined input is kept out of the perception; it is clearly an
imagination. For example, I can imagine the "o" that completes the following
"w rd". So I cam imagine the "o" in "w rd" while I'm looking at "w rd" but I
don't perceive "word" until the "o" inputs are there (as they just were).

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

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

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Martin Taylor 2005.08.30.14.59]

[From Rick Marken (2005.08.30.1130)]

It just doesn't seem to me, based on my subjective experience, that my
perceptions ever include inputs from imagination (memory). When I imagine
something it's a purposeful process; I actively cull up images in memory.
But these imagined perceptions are clearly "in my head"; I can easily
distinguish them from real perceptions (so far;-)).

OK. Here's an example: If we had recently been talking about rivers and you hear the word "bank", do you perceive a financial establishment? If we had been recently talking about money matters and you hear the word "bank", do you perceive a slope leading down to a river?

. Even when the imagination is carried
by the same perceptual signal that is an input to a higher level perception
I think that imagined input is kept out of the perception; it is clearly an
imagination. For example, I can imagine the "o" that completes the following
"w rd". So I cam imagine the "o" in "w rd" while I'm looking at "w rd" but I
don't perceive "word" until the "o" inputs are there (as they just were).

If what you say happenned to be true, proofreading would be so trivially easy as not to be worth the name. You wouldn't perceive mis-spelled words as words at all, and the sentences that include them would make no sense whatever. Only when all the gaps are filled and the spellings corrected would everything leap into intelligibility as opposed to strings of letters.

However, proofreading is extremely difficult. It takes a LOT of training to keep imagination and memory out of the perception of what's on the page.

Martin

[From Rick Marken (2005.08.30.1330)]

Martin Taylor 2005.08.30.14.59]

Rick Marken (2005.08.30.1130)

It just doesn't seem to me, based on my subjective experience, that my
perceptions ever include inputs from imagination (memory). When I imagine
something it's a purposeful process; I actively cull up images in memory.
But these imagined perceptions are clearly "in my head"; I can easily
distinguish them from real perceptions (so far;-)).

OK. Here's an example: If we had recently been talking about rivers
and you hear the word "bank", do you perceive a financial
establishment?

If I'm looking at a bank (building) I'll see a financial institution. If I'm
standing next to a river I'll see a riverbank. You must mean "imagine" not
"perceive" when you ask "do you perceive a financial establishment"? I
imagine a riverbank whenn you are talking about rivers and a financial
institution when you are talking about banks of that sort. What I perceive
while you're talking depends on what's out there.

. Even when the imagination is carried
by the same perceptual signal that is an input to a higher level perception
I think that imagined input is kept out of the perception; it is clearly an
imagination. For example, I can imagine the "o" that completes the following
"w rd". So I cam imagine the "o" in "w rd" while I'm looking at "w rd" but I
don't perceive "word" until the "o" inputs are there (as they just were).

If what you say happenned to be true, proofreading would be so
trivially easy as not to be worth the name.

You clearly did not understand what I said. In fact, the model of perception
I described would predict that proof reading would be _very_ difficult, not
because of memory "filling in" but because incomplete lower level inputs can
still produce "correct" higher level perceptions.

My model of perception is basically this: If the lower level input to a word
recognition perceptual function is "word" or "w rd" or "wrd" the output of
will be a signal indicating that the word "word" is out there. This is what
causes the problem for the proofreader, who, in order to be successful, has
to attend to the lower level perceptions coming into the word recognition
function (that is, she has to attend at the configuration level perceptions
of the letters) to catch many (the not too far off) misspellings.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

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

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Martin Taylor 2005.08.30.16.56]

[From Rick Marken (2005.08.30.1330)]

Martin Taylor 2005.08.30.14.59]

Rick Marken (2005.08.30.1130)

It just doesn't seem to me, based on my subjective experience, that my
perceptions ever include inputs from imagination (memory). When I imagine
something it's a purposeful process; I actively cull up images in memory.
But these imagined perceptions are clearly "in my head"; I can easily
distinguish them from real perceptions (so far;-)).

OK. Here's an example: If we had recently been talking about rivers
and you hear the word "bank", do you perceive a financial
establishment?

If I'm looking at a bank (building) I'll see a financial institution. If I'm
standing next to a river I'll see a riverbank. You must mean "imagine" not
"perceive" when you ask "do you perceive a financial establishment"? I
imagine a riverbank whenn you are talking about rivers and a financial
institution when you are talking about banks of that sort. What I perceive
while you're talking depends on what's out there.

"You clearly did not understand what I said."

I'm not asking you to imagine looking at anything in particular. I'm asking what concept you perceive under the different conditions of what you remember the conversation recently to have been about.

. Even when the imagination is carried
by the same perceptual signal that is an input to a higher level perception
I think that imagined input is kept out of the perception; it is clearly an
imagination. For example, I can imagine the "o" that completes the following
"w rd". So I cam imagine the "o" in "w rd" while I'm looking at "w rd" but I
don't perceive "word" until the "o" inputs are there (as they just were).

If what you say happenned to be true, proofreading would be so
trivially easy as not to be worth the name.

You clearly did not understand what I said.

You seemed to say that you "don't perceive "word" until the "o" inputs are there". I understood that to mean that you wouldn't perceive "word" if the "o" inputs were missing and not supplied by imagination or memory.

I was commenting on that misunderstanding, not on a reasonable model of the functioning of perceptual input processing.

I do agree with you when you contradict yourself, as in:

My model of perception is basically this: If the lower level input to a word
recognition perceptual function is "word" or "w rd" or "wrd" the output of
will be a signal indicating that the word "word" is out there.

That's an example of the "how to" memory I've been talking about with Bill. Much experience with w o r d has allowed you to build a processor that recognizes the similarity of a pattern with what it "should" be.

But what would you perceive from "w rd" if you remembered that the recent conversation had been about legal guardianship of a minor ("ward"?). Why, other than from that memory, would you perceive "ward" rather than "word", as a good reader very probably would?

Martin

[From Rick Marken (2005.08.30.1445)]

Martin Taylor 2005.08.30.16.56]

  Rick Marken (2005.08.30.1330)--

Martin Taylor 2005.08.30.14.59]

OK. Here's an example: If we had recently been talking about rivers
and you hear the word "bank", do you perceive a financial
establishment?

If I'm looking at a bank (building) I'll see a financial institution...

I'm not asking you to imagine looking at anything in particular. I'm
asking what concept you perceive under the different conditions of
what you remember the conversation recently to have been about.

I would perceive you to be talking about a financial institution if you used
"bank to refer to a financial institution. This is not perception based on
memory input; this is perception based on lower level sensory/perceptual
inputs: your talk.

You clearly did not understand what I said.

You seemed to say that you "don't perceive "word" until the "o"
inputs are there".

I hope I cleared that up. I meant the opposite.

My model of perception is basically this: If the lower level input to a word
recognition perceptual function is "word" or "w rd" or "wrd" the output of
will be a signal indicating that the word "word" is out there.

That's an example of the "how to" memory I've been talking about with
Bill. Much experience with w o r d has allowed you to build a
processor that recognizes the similarity of a pattern with what it
"should" be.

But there are no memory inputs to the perceptual function involved in that
model of perception

But what would you perceive from "w rd" if you remembered that the
recent conversation had been about legal guardianship of a minor
("ward"?). Why, other than from that memory, would you perceive
"ward" rather than "word", as a good reader very probably would?

Again, I don't see that as an example of perception based on memory inputs.
The input string "w rd" could produce equal perceptual outputs from the
"word" and "ward" perceptual functions. The higher order perceptual
functions that are perceiving phrases or sentences would get both "word" and
"ward" as input and produce the perception of, say, "Robin was Batman's
ward" or "Robin heard Batman's word" depending on context.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

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

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Martin Taylor 2005.08.30.19.11]

This may well be my last CSGnet contribution until late Sunday. Maybe not, but it's likely.

[From Rick Marken (2005.08.30.1445)]

Martin Taylor 2005.08.30.16.56]

   Rick Marken (2005.08.30.1330)--

  Martin Taylor 2005.08.30.14.59]

  OK. Here's an example: If we had recently been talking about rivers
  and you hear the word "bank", do you perceive a financial
  establishment?

If I'm looking at a bank (building) I'll see a financial institution...

I'm not asking you to imagine looking at anything in particular. I'm
asking what concept you perceive under the different conditions of
what you remember the conversation recently to have been about.

I would perceive you to be talking about a financial institution if you used
"bank to refer to a financial institution. This is not perception based on
memory input; this is perception based on lower level sensory/perceptual
inputs: your talk.

You have a neat way of evading dealing with the situation I'm trying to get you to address. The "talk" in question was _recent_, not ongoing. As I tried to describe the situation, memory is likely to influence which meaning you perceive if you just hear "bank" (maybe from someone else entirely). Maybe the memory is only seconds or minutes, but it could be hours. Or you could have been influenced not by earlier conversation, but by having recently spent pleasant hours by a riverside, or unpleasant ones at a financial institution, either of which would involve memory.

>> My model of perception is basically this: If the lower level input to a word

recognition perceptual function is "word" or "w rd" or "wrd" the output of
will be a signal indicating that the word "word" is out there.

That's an example of the "how to" memory I've been talking about with
Bill. Much experience with w o r d has allowed you to build a
processor that recognizes the similarity of a pattern with what it
"should" be.

But there are no memory inputs to the perceptual function involved in that
model of perception

I didn't say there were. I pointed out that the perceptual fucntion embodied memory, not that it had memory inputs.

But what would you perceive from "w rd" if you remembered that the
recent conversation had been about legal guardianship of a minor
("ward"?). Why, other than from that memory, would you perceive
"ward" rather than "word", as a good reader very probably would?

Again, I don't see that as an example of perception based on memory inputs.
The input string "w rd" could produce equal perceptual outputs from the
"word" and "ward" perceptual functions. The higher order perceptual
functions that are perceiving phrases or sentences would get both "word" and
"ward" as input and produce the perception of, say, "Robin was Batman's
ward" or "Robin heard Batman's word" depending on context.

That's fair comment, when applicable. Maybe it's always applicable.

Martin