contrast process, discrete words

[Marken_Richard_S1]

[From Rick Marken (931026.0900)]

Martin Taylor (931025 15:05) --

Several participants in the discussion on contrast and the discussion on
category that followed it have suggested that category is something that
happens at every perceptual level, rather than representing in itself a level
of perception.

This was not my suggestion. My point (and Bill's too, I think) was that
"categorization" is what perceptual functions do, in the sense that
the output of ANY perceptual function could be the result of many
possible inputs. Give this fact about the nature of perception, the
problem is to figure out what might be unique about the "categorization"
done by what we have been calling the "category" level of perception; in
other words, what TYPE of perception is a "category" perception. Maybe it's
not a separate type -- but, experientially, it seems like it is.

Bill P. has proposed that categorization is something done in
the perceptual input functions of logic level ECSs.

Again, this is not quite correct. "Categorization", as a many-to-one
mapping, is carried out by ALL perceptual functions (as noted above).
What I think Bill was saying is that what you, at one point, were calling
a "category" perception (ie. something that was in one class but NOT
another) was more like a "logical" type of perception.

I would like to propose an alternate view.

Category is defined not by a place in the hierarchy, but by a process.
That process is "contrast," and the contrast process can occur between
any two or more perceptions at any single level of the hierarchy.

Both association and contrast are seen as processes rather than perceptions.
The process of contrast is the suppression of perceptual signal PB as a
consequence of inputs to PIF-A that make PA go high; the process of
association is the enhancement of PB when PA goes high.

So a Mach band is a category perception?

The revisionist element I am proposing in this posting is that the PIFs in a
contrastive relationship, apart from their cross-linked inputs, can be of any
kind in the whole hierarchy, rather than being of one specific level.

But that doesn't seem to address the question of whether there are
"category" type perceptions and what they might be.

Bruce Nevin (Tue 931026 09:35:58 EDT) --

The phonemic contrasts of English partition the articulatory/acoustic
space available to all humans in a way that is different for English than
it is for French or Lakota.

How can contrasts (whatever those are) do anything -- let alone partition
a space? Are contrasts now control systems that perceive a particular
partitioning of articulatory/acoustic space, compare it to the intended
partitioning (the one for English, say, rather than French or Lakota)
and act to keep the perceived partitioning in the reference state? These
contrasts sure are amazing things -- I can see why linguists think that
they are so important.

I said:

I presume you mean that I hear many different acoustical variants
of the same word as being the same and each different word as being
discretely different because of "the way phonemic contrasts
exhaustively partition the available articulatory space". In other
words, these phonemic contrasts are what break a continuum of acoustical
signals into the discrete events that we call words. Is this right?

You say:

Yes.

Then are "photemic" contrasts responsible for the fact that I
see different visual variants of the same object as being the same
and each different object as being discretly different? For example,
pick up a copy of "Mind Readings" or some lesser work; note that it has
a rectangular cover. Now rotate the book (slowly) in front of you. Note
that the book still has a rectangular cover. Yet the visual (retinal)
representation of the book is changing considerably as you rotate it.
A continuously varying visual object produces the same, discrete
perceptual result: "rectangle". The "rectangle" perception (resulting
from all those variants) can be easily contrasted with (seen as different
from) a "trapazoid" perception (if you had a trapazoidal book lying
around - or just cut out a trapazoid to see how different it looks from
the book --even when the book is casting what you know is a trapazoidal
image on your retina). This is a particularly interesting visual "contrast"
because, in many cases, the retinal shape of the rectangular object is
trapazoidal while the retinal shape of the trapazoidal object is often
rectangular.

As I've said before, there is nothing "special" about the perception of
words (speech) -- although I will apparently never be able to convince
a linguist of that fact. Linguists are familiar with many of the
perceptual phenomena involving language (speech). But I don't think
we're going to see much progress in this discussion of word perception
("contrasts") until we start dealing with actual models of speech
perception.

Best

Rick

[bnhpct]

[From: Bruce Nevin (Tue 931026 15:28:56 EDT)]

Rick Marken (931026.0900) --

You ridiculed this:

The phonemic contrasts of English partition the articulatory/acoustic
space available to all humans in a way that is different for English than
it is for French or Lakota.

The phonemic contrasts that a speaker and hearer of English perceives
partition the space. That is, the speaker/hearer partitions the space.
Rick, how could I have possibly intended any meaning but that? Or do you
seriously believe that I am proposing that contrasts are socially
instituted control systems that control the speakers of languages? I
thought my bona fides here were better established than that.

If I were to visit you and say "your home is very welcoming--thank you!" I
would not expect you to reply "How can a home (whatever that is) do
anything, let alone welcome somebody? Are homes now control systems that
perceive people and welcome them?" I would be surprised. I would
perceive such a response as a peculiarly literal interpretation, arriving
at one meaning that the words could have, to be sure, but a meaning so
obviously not the intended one that I could take it only as a joke--or if
you pressed on, I would wonder what perception of yours I had disturbed
such that you wanted me to leave -- I would assume that's why you were being
offensively obtuse and insulting.

Do you want me to go away and stop disturbing you, Rick? I can offer you
a real easy solution to that problem.

In what follows you confound the discreteness of words with their
repeatability or perceived "sameness" despite phonetic variation. In
visual perception, edge perceptions are important for the kinds of
phenomena you describe. But in the perception of words in the speech
stream there are no edges. There are no reliable invariants in the
acoustic continuum of speech by which you perceive the edges of the words
in it. You project the edges of words onto the perceived acoustic
continuum out of your memory and imagination of how you control the words
that you perceive in it. This is more like an optical illusion--
probably you can recall one in which you perceive an edge that is not
supported by the sensory data coming in through the eye, but is rather
projected on the basis of suggestion or expectation.

There is no way to continuously vary a word, as it occurs in the
continuum of speech, in the way that rotating your book (I'll soon have a
copy, Rick) varies the visual image of your book. This is because the
word occurs in the dimension of time, and variation in the word occurs
between instances of its occurring.

A better analogy to visual perception might involve the gestures of, say,
classical Indian dance, or mudra. A particular movement of left hand and
arm may be repeated (by a fluent dancer). There could be variations,
some of them depending upon differences in immediately preceding
positions of the body in the course of dancing, tempo, and also
expressive nuance, and so on. An outsider could imitate this gesture.
The dancer seeing the imitation might say it was OK, or might say, no,
that's an instance of this other gesture, and perform a different
gesture. I can't cite specific examples, not knowing these movements,
but I do know enough about them to feel sure the analogy would go through.

I am not attached to using the word contrast. What I am defending is the
fact that phonemes are controlled relative to one another and not
absolutely, the way a perception of "rectangle" is controlled. That is,
the reference values of acoustic or kinesthetic perceptions for phonemes
are controlled, at a higher level, for their separation from one another.

If you don't want to call that contrast, fine, I'll give up the word.
Furthermore, I have no doubt that the same kind of control turns up other
than in speech. I have no interest in defending your idea (it is your
idea, not mine) that speech is somehow "special" and different from all
other perceptual control. It might turn up in classical Indian dance.
Any system of elements established by social convention is a likely
candidate. But where perceptual control does not involve
conventionalized norms for social cooperation, as in your example of
recognizing the same book as you turn it around, it is not analogous to
perception of phonemes and words. So yes, something "special" is indeed
going on when control systems control an appearance of conventionality
for the sake of cooperation with others. But it's not unique to
language.

    Bruce
    bn@bbn.com

[Tom_BOURBON3]

From Tom Bourbon [931026.1251]

[From Rick Marken (931026.0900)]

Martin Taylor (931025 15:05) --

Several participants in the discussion on contrast and the discussion on
category that followed it have suggested that category is something that
happens at every perceptual level, rather than representing in itself a level
of perception.

Rick:
This was not my suggestion. My point (and Bill's too, I think) was that
"categorization" is what perceptual functions do, in the sense that
the output of ANY perceptual function could be the result of many
possible inputs. Give this fact about the nature of perception, the
problem is to figure out what might be unique about the "categorization"
done by what we have been calling the "category" level of perception; in
other words, what TYPE of perception is a "category" perception. Maybe it's
not a separate type -- but, experientially, it seems like it is.

Martin:

Bill P. has proposed that categorization is something done in
the perceptual input functions of logic level ECSs.

Rick:
Again, this is not quite correct. "Categorization", as a many-to-one
mapping, is carried out by ALL perceptual functions (as noted above).
What I think Bill was saying is that what you, at one point, were calling
a "category" perception (ie. something that was in one class but NOT
another) was more like a "logical" type of perception.

Much of the discussion on "categorization" seems to cluster around two
"contrasting" ideas: perception of categories as a special class of
perception, and all perception as category-like in certain ways. Martin, in
the post from which I quoted above, has suggested (at least for the sake of
discussion) a second discrete or logical hierarhy, everywhere parallel to
the original analog hierarchy in PCT. After reading through his post twice,
I still see Martin's parallel discrete hierarchy as a complex way of
accomplishing what is done by traditional perceptual input functions.
(Perhaps I should read the post several more times, or wait until Martin is
back on the net to elaborate on his ideas.)

Much of the recent discussion has revolved around Rick's remark that, in many
ways, *all* perceptions, at all levels in the hierarchy, are
"category-like." Throughout the discussions, I believe Rick and Bill
have maintained that idea, that all perceptions are category-like in the
sense that all perceptual functions perform many-to-one mappings. Much of
the discussion has been directed at higher levels of percpetion, but
category-like many-to-one-mappings characterize perception at all levels.
At the intensity level, perceived brightness, loudness, and tension are
simply *different*, not by virtue of any process of categorization, but by
virtue of different input functions each producing their own signals as
analogues of different "parts" of Boss Reality. Or perhaps they are creating
their own signals as analogues of the *same* parts of the Boss. The fact
is, the direct, unnamed experiences are different.

So too at the second level, where figure-ground perception, pitch, and
effort occur. The experiences are different, but not categorically so.
There is no experience of category at that level. Each perception
simply *is*. Any contrasts or distinctions between them arise from the
activity of different input functions, large numbers of which perform
many-to-one mappings from the level of perceived intensity, one step down in
the hierarchy. (Some 2nd-level perceptions of sensation arise from signals
that have "skipped" a level. That also happens at the 3rd level,
configurations, and the 4th, transitions, for certain signals that originate
in receptors at the joints of our skeletons.)

The names we have used to identify "kinds" of experience reveal the
workings of logical categorization, something removed from the direct
experiences. Those are forever unnamed. The Tao that can be named is not
Tao. The categories of "lower-level" perceptions that can be named are not
categorical lower-level perceptions.

Until later,

Tom

[_Martin_Taylor]

[Martin Taylor 931029 22:00]
(Rick Marken 930126 0900)

In an interchange with Bill Powers, I said and he responded as follows:

I would prefer to say that an outside observer can categorize
the inputs to the function, using his/her category-level
perception. The perceptual function itself doesn't categorize
unless it IS a category-level function.

Good. This expresses in another way the general problem that I
bring up now and then: the use of a human level of perception to
characterize the functions of a level in a model or in someone
else that is actually lower (like trying to characterize the
stretch reflex as a program, as in the TOTE unit). If any level
of perception is seen as a categorizing level, then we can be
sure of one thing at least: the person doing the seeing is using
a category level of perception. Whether this is justified is
another question.

Rick says:

My point (and Bill's too, I think) was that
"categorization" is what perceptual functions do, in the sense that
the output of ANY perceptual function could be the result of many
possible inputs.

If you can read this into Bill's agreement with my position contradicting
your statement, your reading is more liberal than mine.

Categorization is what people looking at the output of a many-to-one
function can do if they want. It is not what the function does, unless
it is a categorizing function. Not all subspaces are categories of the
space within which they reside. At a minimum, it would seem necessary
that the subspace occupied by a category should have a measure greater
than zero. In other words, specification of a measure by a point on the
real number line is insufficient to define a category in itself.

For a category to be valuable as a category perception, there must be
a probability greater than zero that another member of the category will
be perceived during the (finite) lifetime of the perceiver.

So a Mach band is a category perception?

Mach bands are not, so far as I am aware, generated by the kind of
recursive inhibition that I discussed. They are, I think, generated
by the Bekesy kind of forward lateral inhibition that generally is
involved in edge sharpening. Mach bands may not be category perceptions,
but "Mach bands" certainly are. (A label for a kind of phenomenon denotes
that the phenomenon is a category).

The revisionist element I am proposing in this posting is that the PIFs in a
contrastive relationship, apart from their cross-linked inputs, can be of any
kind in the whole hierarchy, rather than being of one specific level.

But that doesn't seem to address the question of whether there are
"category" type perceptions and what they might be.

The posting you purport to be responding to was an attempt to do exactly
that--to show what "category" type perceptions might be, and moreover,
to show how they might be created. At the same time, it was intended to
resolve the issue of contrasts as perceptions, by proposing that they are
processes rather than perceptions. Specifically, they are processes
that define categories. So

How can contrasts (whatever those are) do anything -- let alone partition
a space?

which you wrote sarcastically to Bruce Nevin, is answered in my posting.
What contrasts DO, as processes whose function it is to DO things, is to
partition spaces defined by continuum perceptual signals.

Martin