[From Bill Powers (920610.0800)]
Chuck Tucker: $90 should cover it, including membership, banquet for both,
no other meals for you or guest. You can purchase meals with the rest of
us, as desired (Breakfast $4.50, Lunch $5.50, Dinner $6.25).
Andy Papanicolaou & Tom Bourbon (920606) --
RE: habits and speech
I guess I misunderstood the thrust of your questions about habit formation.
But I still think the problem lies at least PARTLY in focusing too much on
outputs. I don't mean to dismiss the fact that output learning (learning of
lower-level control systems) must take place, only to point out that many
output processes are as they are only because the physical extra-neural
world is as it is.
So, we believe that it is generally true (and, hopefully, the linguists
will concur or persuade us otherwise) that every time a /ba/ or a /ga/ >or
a /tu/ is heard the corresponding patterns of articulatory gestures
contain a set of invariant features.
OK, this is true, even if it is also true that there are many sounds that
can be made by producing variable articulator patterns. Consider, however,
the role of the non-neural environment in the way I open the front door of
my house when entering from outside. It is my habit to grasp the knob with
my left hand, turn the knob clockwise, lean slightly toward the door while
turning the knob and pushing, and extend my arm so that the door swings
open, inward. There are slight variations in this habit, but most of the
time that is how it's done.
The question is, how does this habit come into being, and why this one
instead of some other? We're all agreed, I think, that to answer "Because
the opening of the door reinforces the movements that lead to its opening"
would leave us no closer to an explanation. What we would like to know is
how I come to do just those acts that will open the door, and why I do
them.
The most general answer is "Because I want to get inside the house." The
reference signal that motivates the opening of the door is a picture of
myself stepping through an opening that does not initially exist. If I
began this process in total ignorance, I might try pushing the door out of
the way, pulling it out of the way, sliding it out of the way, waving my
hands and crying out "Open, Sesame!", or walking back and forth in a
figure-eight pattern. None of these actions, however, would be effective --
not because they are foolish or superstitious, but simply because they
don't have the effects on the door that suffice to open it. When one has no
knowledge of the properties of the door, there is nothing to say that one
act would work any better than another. Trying one action that one knows
how to produce is no more foolish than trying another.
If I'm solving this problem by reorganization -- by running randomly
through different uses of skills I already possess -- I will simply keep
trying until I hit on the combination that makes the actual perception (a
closed door) change to match the desired one (an open door).
If I'm solving this problem by reason and insight (in the way a Djinn would
who is used to doors that open with "Open Sesame"), I will see that there
is a knob, and devices holding the door on its left side that look as if
they could pivot and would prevent sliding. By examining the door frame, I
will see that it is prevented by a strip of wood from opening outward, and
deduce that it must be pushed inward. I will realize that the knob, being
rotationally symmetrical, is probably intended for grasping and turning. I
will see that I should push the knob with my left hand so my arm will be
out of the way if or when I move through the door. I will then end up doing
exactly the same actions I would have done at the end of a random search
for an effective means.
The fact is that the door is so constructed that only a combination of
knob-turning and pushing will make it swing open and create the wanted
perception of an opening sufficiently large to walk through. That fact is
quite independent of the organization of my nervous system. Doing this with
the left hand will leave me free to move through the door while keeping
control of it. It is the physical door and its properties that determines
what lower-level variables I MUST control in order to gain control of the
higher-level variable that is my reason for wanting to open the door. While
there are slight variations in the way different people might go through
this process, everyone who opens that door will end up doing it in
essentially the same way -- not because there is any propensity of people
to develop similar habits of acting, but because there is a limited number
of actions that will open the door. Given the purpose of perceiving an open
door, and the initial condition that the door is closed and latched,
essentially everyone will perform the same acts in opening it. This tells
us nothing about the nervous system, but much about the door.
Now transfer this parable to the pronunciation of certain sounds. If, in
fact, there is only one manipulation of the articulators that will produce
a given sound, /ba/, and given that a person wishes to experience the sound
of /ba/ (perhaps simply in imitation of a remembered sound), then however
one succeeds in doing this, by random trial and error or through visual
cues or by clever reasoning, the final result will be to close the lips,
start the sound, and release the lips. This is not because of any
propensity to develop that habit, but because that is the set of physical
processes that produces the wanted sound. If different people end up with
the same articulator habits in making the sound /ba/, this is not because
their nervous systems have similar inclinations to develop that habit, but
because the physical construction of different people is, in the relevant
regards, very similar. That is where the invariant features of
"articulatory gestures" come from.
Suppose that there is more than one way to open the door: suppose it can
swing open either inward or outward, also also slide sideways into its
frame, after the knob is turned either clockwise or counterclockwise.
Starting in ignorance, I might try just pushing and pulling, or walking in
figure eights. Eventually, I will hit on a combination of acts that creates
the perception of an open door that I want. This combination then becomes a
means of opening the door. I turn the knob counterclockwise and slide the
door open. Why, then, do I use this act again the next time I open this
door? Why does this act become a "habit" even though other acts would work
just as well?
The answer is simple: if the higher order goal is simply to get through the
door, and the output of this higher order control system has become
connected to reference inputs of lower systems so as to create the desired
perception, there will be no error to motivate another random search (or
another period of clever reasoning). The alternate solutions will go
undiscovered because there is no need for them. The "habit" of
accomplishing the goal by one specific means remains by default. Only if
this means entails awkwardness or pain or some other error-creating
consequence would there be any motivation to extend the search.
Now consider the Japanese who has trouble producing an 'l' sound, but is
quite capable of producing an 'r'. The goal, in a Japanese environment,
would initially have been to hear the same 'r' sound heard in others'
speech. So the reference signal would have been this remembered sound. The
Japanese baby, by experimentation, finds a way to manipulate the
articulators so they create the desired sound. After much learning, the
feeling of making the sound is added to the auditory experience, and
remembered, so the effective reference signal includes kinesthetic as well
as auditory components of lower order.
Now the sound 'l' is heard, and for some reason the (now-older) Japanese
wishes to reproduce it. Initially, only the 'r'-input-function responds to
the heard sound of 'l'. So the Japanese understands that the goal is to
produce the 'r' sound, and does so. "No, no," the Western teacher says,
"You must say it this way: 'l'." And the Japanese, hearing 'r', says 'r'.
This could go on forever: it seems that the Japanese pupil simply has too
strong a habit of making the articulations that result in the sound of 'r'.
In fact, there is no "habit." The Japanese is simply repeating back the
sound that is heard.
The Western teacher, who has different perceptual functions that respond to
'l' and 'r', concludes that the Japanese is lacking the ability to SAY the
sound 'l'. In fact, what is lacking is the ability to perceive the sound
'l' as something different from 'r': both are heard as 'r', although the
'l' might not generate quite an optimal sense of 'r'. The puzzled Japanese
hears himself or herself reproducing the same sound that the teacher is
emitting, and doesn't understand what the problem is. From the Japanese
viewpoint, neither the Japanese nor the English language contains the sound
'l'.
If, by some means, the Japanese could be persuaded to search for perceptual
differences between the two sounds, and ultimately to be able to indicate
which sound is being heard, the problem would still remain of creating
those sounds -- just as a person who perceives that a door can swing open
as well as slide open must still learn to use different actions to create
this new perceptual consequence. Now kinesthetic learning must take place.
The new perceptual function (and comparator) must produce an error signal,
and that error signal must be connected to produce new reference signals
for the way the mouth and tongue feel while uttering a sound. If the
Japanese understood that this is now the problem, he or she could begin
experimenting until the sound was successfully produced. Then, of course, a
period of practice would be needed so that the initial sound-only
perception became a sound-plus-feel perception, the way the 'r'-sound
controller has developed.
It occurs to me (thanks to Martin Taylor and his unrelated namesake J. G.
Taylor) that the perceptual learning might be facilitated if the person
were given control of the sound -- not through the articulators, but
through some simple means the person can already use, such as a control
handle. On hearing the teacher say "road" and "load" over and over, the
person would hear an artificially-produced word "road" being repeated over
and over, with the initial consonant being adjustable smoothly between the
sound of 'r' and the sound of 'l'. The task would be to move the handle
until the artificially-generated word matched the sound of the word as
pronounced by the teacher. Once control of the auditory 'r'-'l' difference
is attained by this means (assuming it ever is), perhaps the task of
finding a different means -- using normal methods of articulation -- would
then be mastered more easily. Martin Taylor has pointed out repeatedly that
perceptual learning works best in a control context -- perhaps ANY control
context is better than none.
Marvin Brown says that once a wrong pattern of articulation is learned, it
is too late to learn native speech. Perhaps the stage of using an
alternative means of producing the right sounds might overcome this
problem.
Marvin received your message, Gary, and will probably reply by snail mail
-- I have to conclude now to have a final conversation with him and wave
goodbye as he heads back to Salt Lake City.
ยทยทยท
--------------------------------------------------------------------
Best,
Bill P.