coin game analogy -- dedux

[Marken_Richard_S1]

[From Rick Marken (931008.0930)]

Ok, Bruce. I revisited your "coin game analogy" post
(Wed 93106 11:49:21 EDT) and here are my detailed comments.

Let's look at the analogy [of the pair test] to the coin game.

To make this a fair analogy to a linguist applying the Test to a language
whose phonemes he does not yet know, we have to take away some of this
shared and known-to-be-shared knowledge.

The teste (T) has to be able to come up with hypotheses about what the
subject is controlling. Apparently this can be done by T in both the
coin game and in the pair test becuase the T's in both cases are
DOING the test. The quality or correctness of their hypotheses about
what is controlled depends on the cleverness of T, in both cases.

You are apparently trying to call into question the analogy between
the pair test the coin game. Why? You claimed that the pair test
is a version of "the test"; we (Bill and I) saw that the pair test
IS a version of "the test", being carried out almost exactly like the
coin game described in BCP. Now you are apparently trying to argue that
the pair test is NOT like the coin game: WHY? It is very peculiar for
me to have conceded a point to you (as I did, by conceding that
the pair test is a version of "the test for the controlled variable")
only to have you start to argue as though I were wrong to say you
were right. But, whatever.

Suppose the subject S is controlling for an isosceles triangle. This
imposes a categorization on the possible arrangements of three coins:
isosceles vs. not isosceles.

That's a way of looking at it; but all the S is controlling for is
the isosceles configuration of coins. By doing so, he is also
"imposing" this categorization, which seems like a fairly odd way
to look at it since "not isosceles" includes EVERYTHING that is not
an isosceles triangle -- ie. 747s, Marilyn Monroe, e-mail, etc.

All that S can (without training) bring to awareness is a perception
of a triangle that is "balanced" (two sides equal) or not. The coins
and the details of their relationships are not accessible to
consciousness without careful attention, training, and practice.

This is all completely irrelevant to "the test", which should be able
to reveal what variable is being controlled whether S is aware of
it or not. "The test" is a test of purpose, not consciousness (a
separate phenomenon, best investigated with hallucinogenic drugs --
wheeeee).

(If T is not
trained, then T comes in knowing how to control for a different
categorization, say, scalene vs. non-scalene, which is normal for the
"language" spoken where T comes from. T can only interpret S's responses
in terms of T's categories, and vice versa, and they generally get pretty
muddled.)

If, in fact, T doesn't know about isosceles triangles then T will
not be able to hypothesize that S is controlling for isosceles tri-
angularity. If a non-linguist (me) doesn't know about the possibility
that some aspect of the p sound in spin might NOT be controlled,
then the non-linguist would never think of testing this by giving
sbin as a comparison. You can't test for controlled variables unless
you have some idea what those variables might be. I could not
have thought of testing to see whether some aspect of p in spin is
controlled because I know diddly squat about phonetics. But Harris did
guess that some aspect of p MIGHT controlled in spin (the aspect of p
that is shared with the b in sbin), tested it and found that, indeed,
that aspect is NOT controlled (because the subject does not resist a
change to that aspect of the sound).

Indeed, it is only by virtue of this
training and experience that T perceives the game as taking place over a
field of possible arrangements of coins; S can perceive neither coins nor
arrangements of coins nor field of possibilities, S can perceive only
isosceles vs. non-isosceles ("balanced-unbalanced").

What matters (in order to carry out the test) is what T knows -- NOT
what S knows. T is the one who must come up with the hypotheses. I
would assume that a linguist, using the pair test, could figure out
what variables I am controlling when I say certain words; I guarantee
you that I have NO idea what variables I'm controlling -- I'm just
controlling them.

[Aside: Do linguists really believe that you have to be a linguist
in order to be able to talk (ie. control the articulations, sounds,
phonemes, morphemes, grammars, semantics)??]

T wants to show S that what is being controlled is a categorization of
isosceles vs. scalene. But this recognition is attainable only if other
categorizations are possible. For S, no other categorizations are
possible, and, indeed, only the category perceptions exist.

This is where you confused me last time. I don't see what all this
has to do with "the test". I don't see the point of this paragraph.
If the S is controlling a "categorization of isosceles vs. scalene
triangles" then this will be revealed by the test. If S cannot
perceive the "categorization of isosceles vs. scalene" then
S CANNOT be controlling it. If S is controlling the "categorization
of isosceles vs. scalene" then S MUST be perceiving it (unless you
have a model that shows how unperceived variables can be controlled).
It is certainly possible that S can be controlling a perception that
he has great difficulty becoming aware of; this is precisely what goes
on in speech. I think I speak fairly well but I have only the fuzziest
notion of what variables I am controlling when I speak; that's what
linguists are for, no? To figure out what people are controlling when
they speak.

It is not quite right to say that S cannot perceive coins and
relationships.

It is not quite right to say that S cannot perceive coins and
relationships. When T talks to S about coins and relationships, S says
"yes, I see that", but in fact S sees it only in terms of the
categorization into balanced vs. unbalanced triangles. Here, the analogy
is difficult to use because it is simplistic. What I am referring to
here is the way Rick and Bill identify the phoneme p as a particular
sound, and then (against perceptual input that is accessible to them)
assert that this is the sound that occurs in "spin", and that it is the
substitution of b in "sbin" that is anomalous. That is why I introduced
the experiment with the strip of paper. The affect of egressive air on
the paper reinforces with visual perceptual input the abovementioned
perceptual input that is available but is ordinarily not accessed. It
may even be "blanked out" to avoid accessing it. There certainly seems
to be resistance to becoming aware of sensory input that contradicts the
categorization of the [p] of spin as identical to the [p`] of pin and in
contrast to the b of bin. After s, there is no phonemic contrast that
may be represented by voiced sounds vs voiceless aspirated sounds.

This seems to be a crucial paragraph. Hopefully, the first part of
my discussion shows why I think it is irrelevant whether the p
in spin is the same as the p in pin (as far as the pair test being
analogous to the coin test is concerned). All I can say is that
what S consciously knows (or says to T) about the details of the
perception being controlled is irrelevant to "the test". I know
that, as the subject in the coin test, I would know a LOT about
possible geometric arrangements of coins that I might control.
I also concede that, as a subject in the pair test, I would have
no idea what the possible sounds are that I might control -- p
or p' or b or whatever. But this is all irrelevant to the main
point: the pair test is perfectly analogous to the coin game --
but the coin test can be done by a dumb slug like me who can at least
be expected to be able to come up with some hypoetheses about possible
arrangements of coins that might be controlled. The pair test, on
the other hand, can only be done by a skilled linguist (I couldn't
do it) who can come up with hypotheses about the kinds of variables
controlled in speech (like the p vs b sound) and who can figure
out ways to test these hypotheses (the linguist can figure out what
sounds would not match the reference if the variable were controlled).

Does that help?

Best

Rick

[bnhpct]

[From: Bruce Nevin (Fri 93108 14:54:04 EDT)]

( Rick Marken (931008.0930) ) --

I was wrong to say my (Wed 931006 11:49:21 EDT) was primarily an analogy
to our conversational process about phonemes. As it reads, it was
primarily about an analogy between the coin game and the pair test, as
you say. My additional motivation, one of drawing an analogy to our
conversational process, was not at all clearly expressed.

The passage you singled out for ridicule in your (931006.1100) "Telling
others what they want, aspirate" was a part of the analogy to what I had
been trying to accomplish here on the net, rather than a part of the
analogy to the pair test. I said (931006 11:49:21):

T wants to show S that what is being controlled is a categorization of
isosceles vs. scalene.

You ridiculed this as follows (931006.1100):

Spoken like a true linguist. Unlike a linguist, a PCT researcher (T)
doesn't try to convince S of what S is controlling; S convinces T (by
resisting disturbances to hypothesized controlled variables) of what
S is controlling.

By the way, if you actually have the goal of getting people to
agree with your interpretation of what they are _really_
up to (controlling for) then expect disappointment (and a very
small social circle).

In response to this, I asked how you would have gone about directing your
and Bill's attention to subphonemic perceptions. I don't wonder that you
haven't answered this question, since it makes sense only if you
recognized the analogy between the test and our communication process and
only if you were aware of what I was trying to do with that analogy in
that post.

So I take responsibility for the confusion here. The second analogy
amounts to a hidden agenda because I did not state it clearly. Because I
did not distinguish the two, the inexplicit one steps on the toes of the
explicit one. With the help of your latest today, (931008.0930), I hope
I can separate these two analogies-- one between the coin game and the
pair test, and the other extending beyond the test (be it coin game or
pair test) to go up a level and take a snapshot of our discussion from a
point of view outside it.

The coin game and the pair test are both instances of the test. As you
say, in both cases "The quality or correctness of their hypotheses
about what is controlled depends on the cleverness of T, in both cases."
I pointed to a difference in what T knows, ahead of time, about the
perceptions that S is controlling. In the coin game, S and T agree ahead
of time that they are concerned only with geometrical arrangements of
coins in a plane delimited by a certain region of the table between them
(or whatever the details of the environment happen to be). They each
know this and know that the other knows this, and they each know and know
that the other knows what coins are, etc. T does not know what
perceptions S is controlling in the agreed upon "envelope" of possible
arrangments of coins on the plane surface before them.

In the language situation, S is typically naive about phonetics and
linguistics and T is presumably trained in these matters. This does not
mean that the pair test is not an instance of the test. You go too far
in saying "You are apparently trying to call into question the analogy
between the pair test the coin game." This confusion is an instance of
the inexplicit analogy stepping on the toes of the explicit one.
But I see you agree with the above two paragraphs at the close of your
post, as follows:

I know
that, as the subject in the coin test, I would know a LOT about
possible geometric arrangements of coins that I might control.
I also concede that, as a subject in the pair test, I would have
no idea what the possible sounds are that I might control -- p
or p' or b or whatever.

I have tried repeatedly over a couple of years to work out the relation
between phonemes and the actual sounds that people pronounce. This
relation is not simple and direct. It cannot be accounted for by saying
that the appearance of phonemic contrast is a byproduct of controlling a
perception of a particular phonetic output called (for example) the
phoneme p or the phoneme b. But in order to make this case, I have to
talk about speech sounds as perceived below the level of phonemic
category perceptions. The difficulty is that, without phonetic training,
one is not aware of those perceptions, one is aware only of the category.
The consonant after the s in "spin" is a p phoneme of course, identical
to the p in pin and different from the b in bin.

So I emphasized the differences between the coin game and the pair test.
T does not need to make S aware of these perceptions below the level of
phonemic category perceptions in order to do the pair test. The aim of
the pair test is to determine what S's category perceptions are--which
phonetic productions (actually, fractions of the phonetic productions of
words) are the same, and which are different phonemic categories. But my
aim in our conversations on this net has been different from T's aim in
the pair test. My aim here has been to figure out a PCT account of
language, and specifically of phonology. To talk about this, I needed to
be able to talk about phonemic category perceptions (or, equivalently,
perceptions of phonemic contrast). But many of my readers identified
their category perceptions so closely with the sounds they produced that
they thought there were no category perceptions--no contrast
perceptions--at all. They thought there was only the control of speech
sounds. Before there could be awareness that there are category
perceptions that are not simply perceptions of sounds there had to be
awareness of that categorized as something different from the categories.
I needed to get my readers to pay attention to phonetic perceptions below
the phonemic category level. Before that, I had to get my readers to
acknowledge that such perceptions existed, and that they could be at
variance from what their category perceptions would lead them to believe.

What happens with such exploration is that it becomes conceivable to the
subject, S, that a given sub-category perception might be assignable to a
different category among those that S knows, or even that a different set
of categories might be possible. (But talking about foreign languages
with their different phonemic categories has not at all been effective
for communicating on this net.)

This is why I said

T wants to show S that what is being controlled is a categorization of
isosceles vs. scalene. But this recognition is attainable only if other
categorizations are possible. For S, no other categorizations are
possible, and, indeed, only the category perceptions exist.

Again, you are quite right, this has nothing to do with the test. It has
to do with my other agenda of trying to go up a level and look at our
miscommunication from outside of it.

You single this out as a crucial paragraph:

It is not quite right to say that S cannot perceive coins and
relationships. When T talks to S about coins and relationships, S says
"yes, I see that", but in fact S sees it only in terms of the
categorization into balanced vs. unbalanced triangles. Here, the analogy
is difficult to use because it is simplistic. What I am referring to
here is the way Rick and Bill identify the phoneme p as a particular
sound, and then (against perceptual input that is accessible to them)
assert that this is the sound that occurs in "spin", and that it is the
substitution of b in "sbin" that is anomalous. That is why I introduced
the experiment with the strip of paper. The affect of egressive air on
the paper reinforces with visual perceptual input the abovementioned
perceptual input that is available but is ordinarily not accessed. It
may even be "blanked out" to avoid accessing it. There certainly seems
to be resistance to becoming aware of sensory input that contradicts the
categorization of the [p] of spin as identical to the [p`] of pin and in
contrast to the b of bin. After s, there is no phonemic contrast that
may be represented by voiced sounds vs voiceless aspirated sounds.

Again, you are right,

it is irrelevant whether the p
in spin is the same as the p in pin (as far as the pair test being
analogous to the coin test is concerned). [...]
what S consciously knows (or says to T) about the details of the
perception being controlled is irrelevant to "the test".

These concerns become relevant when you are interested in developing a
PCT account of what people do when they talk to each other and understand
what one another is saying. You and Bill have made pronouncements about
what such an account ought to be like without knowing these facts about
what perceptions are being controlled. This has been a source of some
considerable frustration to me. Hopefully (I say) things will be better
now that we have got past this hump.

    Bruce
    bn@bbn.com