[From Bill Powers (970224.1935 MST)]
Bruce Abbott (970224.1620 EST)--
Bill, this is a different case from the one we've been examining -- one in
which one gathers enough information to make a reasonable guess about
where some visually-acquired moving object will be in the near future and
then acts on that projection, without further input, until the final stage
when the image is reacquired. One can hardly keep the object at a
constant bearing under these conditions, and I was specifically talking
about a situation in which I was not in motion at the time the observation
was made.
Good, you didn't lose your temper at me, but I'll bet you were tempted.
What I'm trying to get you to see is that the customary way of explaining
these phenomena makes assumptions for which there is no evidence. And this
includes what we think, to ourselves, about how we do these things. A lot of
what we think is based on imagination, not observation.
And no, I'm not "echoing" explanations I've been "fed." This is my own
interpretation, based on my own first-hand experience. If it's wrong, I
take full credit. (:->
Do you mean that nobody in your psychology courses or papers you read
offered the same explanation? It seems strange that I've heard this
explanation so many times, while you, a professional psychologist, never
encountered it before you took up baseball.
If you and the approaching ship are on a collision course, the bearing
will be constant regardless of the angle between the courses. Your guess
that the angle has to be 90 degrees is wrong.
I _didn't_ guess that the angle has to be 90 degrees, so I can hardly have
been wrong about that. What I said (evidently not clearly enough) was
that the situation is one where the angle is difficult (if not impossible)
to judge. (The angle of regard is too close to the surface, given the
distance involved, to produce a reasonably accurate projection.)
I see that I misunderstood you. You're saying that when your point of view
is sufficiently above the surface, you can extend the two courses, yours and
that of the other object, to see where they intersect. Then, I presume you
mark off distance intervals along these two paths in proportion to the
velocities of yourself and the other object, and if the number of intervals
is not the same at the point of intersection, you adjust your speed and/or
course until this is the case. Having done this, you no longer need to
observe the other object, because you know how fast and in what direction to
travel.
Is this more or less what you experience when you turn your back on the ball
and start running at a calibrated speed? If not, how DO you make these
estimates? What makes you sure that these estimates are even being made?
That means that
this example is one where the sort of projection I am talking about is
likely to fail -- badly. Therefore it is no surprise to me that one would
have to rely on other optical relationships that require repeated sampling
to detect (e.g., constant bearing).
The constant-bearing approach does not require continuous sampling, except
at the beginning. Once you have established a speed and direction of
movement, and for any reason you can't observe any more, what is the most
effective thing to do? Keep traveling the same way at the same speed. I
hinted at this in speaking of trying to collide with another airplane in
Flight Simulator. This method accomplishes exactly what the "projection"
method accomplishes, and requires the same initial observations, but does it
without actually making any projections.
Who was "exercised" by this explanation? I don't recall any great furor
arising from this article . . .
After the article about the experiment using video cameras (something like
20 years after the initial article on catching a baseball) there were a
number of letters proposing other methods, mostly of the type you suggest,
and strong rejection of the article's interpretations. I know I saw these
commentaries, but can't remember where.
And note that the article dealt with a
slightly different situation -- the fielder was able to repeatedly glance
up and make adjustments to his path of travel. Because this involves
repeated sampling, there can be essentially closed-loop control
throughout. I have absolutely no problem with the proposed explanation for
this performance, under this condition.
Good. Now, starting with that method, what is the least change necessary to
account for a similar performance when the ball is not in sight at all
during the middle portion of its trajectory? Is it necessary to go all the
way to "calculating the landing point?" I've already indicated how I think
it's done.
Now, how does he intercept a grounder? Different method?
Same method, with modifications. Generally, infielders are taught to "get
their bodies behind the ball," which in perceptual terms means moving
yourself (if possible) until the ball is not moving either left or right.
When it's not moving left or right, it's coming right at you (same as
landing an airplane in a crosswind). Of course it's still going to hit a
bump and jump over or under your glove. If you can't get your body behind
the ball, you try to get your glove on a collision course with the ball.
The constant-bearing hypothesis (or some constant-perception idea) is
right and the "projection" hypothesis is wrong, Bruce.
I'm sure it is -- under circumstances where it can work. I'm asking about
what you do when it can't.
You ask what other _real present-time_ perceptions can be controlled that
will give almost as good a result. You assume that anything involving very
complex calculations will fail miserably, and will be used only as a last
resort, without much hope of success.
Do you know what the
giveaway is? When you imagine projecting your own position to some future
position of the object you want to intercept, can you see yourself in the
picture? If the answer is yes, then you're imagining something you can't
perceive. When you actually do it, _you_ aren't in the picture.
What I perceive is the point I estimate I will have to reach if I am to
have any hope of catching the ball, and some idea of the amount of time
available in which to get there. These become references for where I will
run and how fast.
How do you perceive that point? By what method do you carry out this
estimation? How do you perceive an "amount of time?"
If, as you say, you can make these estimates while standing still, you ought
to be able to specify the landing location immediately. So a test ought to
be fairly simple. Lay out a numbered string grid in the possible landing
area, and have someone throw balls, with you specifying the grid location of
landing immediately after the ball is thrown (or batted, of course). Then
record the guess and the actual landing position.
Actually, if you pause after the ball is hit to make these estimates, before
you start running, you are never going to make it to the majors. A typical
major-league second baseman, going after a blooper to center, turns
immediately and starts running, twisting back to watch the ball as he gets
up to speed, and then turns and runs for a while at the same speed before
turning around to adjust position for the final part of the trajectory. When
a deep drive starts off, a center-fielder playing in close may well start
running while watching the ball, and then just turn and go hell for leather
to the wall, because he's unable to go fast enough to establish the right
perceptual condition (constant trajectory, constant slow rise, whatever) and
just runs as fast as he can hoping to catch a carom off the wall, or make a
showboat leap.
I don't know how this second-baseman or center-fielder would describe how he
catches a ball. He might say "I just run to where it's going to come down."
That may well be how it seems -- but if he catches the ball often enough to
earn his bloated salary, that's not likely to be how he actually does it.
Being good at something doesn't mean you understand how you really do it.
It's easy to understand this when you hear a batting instructor explaining
to an interviewer that batted balls have more "carry" when the air is humid
(I've heard commentators say the same thing about golf balls, so gee, maybe
impetus theory really does work).
By "computations" I don't mean taking sines and cosines of angles and
all that mathematical nonsense. This is an analog system, remember.
You don't perceive the analog computations, either.
Oh, for goodness sake. The perceptual machinery does the analog
"computations." _I_ just perceive their result.
Don't you "Oh for goodness sake" ME, pal. You perceive what you perceive.
You're GUESSING that it's the result of a computation. You think it is
reasonable that it be the outcome of a computation. But you don't know WHAT
computation, and you don't know what the computation is based on. It might
be based on perceptual variables which you aren't attending to consciously,
and the computation might be completely different from the one you imagine.
You keep going back to this non-applicable example. I have already
granted that if I have the ball in view, I may very well use this type of
perceptual control, but the example I proposed is one in which this
strategy is impossible. Therefore we must either deny that I am able to
arrive at some location anywhere near the ball's path under my scenerio,
or develop an alternative explanation for my ability to do so.
This is getting repetitious, but I'm glad you tacked on that final clause. I
think it is relatively easy to find an alternative explanation that doesn't
call for unnamed and unsensed calculations of very complex nature that go on
where you can't see them. Maybe such calculations do go on, but before I
would be driven to refer to them, I would exhaust all the simpler
explanations based on control of perceptions that are clearly available and
don't need to be imagined.
What's so difficult about accepting the idea that under some conditions we
can and do rely on estimates or forecasts to improve our ability to >control?
It's not difficult unless you're trying to account for skilled behavior. It
doesn't take much skill to predict that the sun is going to rise tomorrow,
and plan accordingly, or to look at thunderheads out your window and reach
for an umbrella. This is the sort of simple stuff we can do with cognitive
planning, because it's elementary. But as soon as the forecasts become at
all complex, we have to rely on slow artificial methods; adding up the
checkbook to see if we have any money left, programming a computer to
predict what tomorrow's weather will be; solving equations with pen and
paper to see the best number of newspapers to stock at the newsstand. If you
have to estimate anything involving a quadratic or cubic equation, you can
count on being off by a large amount - unless you actually do the algebra.
Even the running ball-player who can only occasionally glance back at the
ball must run for a time without the continuous input that would allow him
to keep the ball on a constant visual bearing, and then correct any error
that has built up between sightings.
An expert pool player can visualize where each ball will go after being
struck, including any effect of putting "english" on the cueball. The
only direct control possible is of the motion imparted to the cue -- after
that the whole scene goes ballistic, i.e., open loop. From one
perspective the player is only controlling the perception of the cue
stick's position and motion, but where did the references for those CVs
come from? My guess is that they come from the player's ability to
project the paths of each ball while mentally varying the cue stick's
tip-position and angle relative to the cueball, and the pattern of force
and motion imparted to the stick. This requires a sophisticated intuitive
model of the physics involved, one that requires considerable experience
to develop. The task is made somewhat simpler, however, by the relatively
consistent characteristics of the table, cue stick, and balls.
Right, no "twisted cue, and a cloth untrue, and elliptical billiard balls."
Part of my misspent high-school youth was misspent while managing a
community-house rec-room with a real billiard table. I learned the fine art
of three-cushion billiards from an underemployed house painter named,
honestly, Rembrandt Noble. I didn't get good enough to beat Rem, but he
taught me enough that nobody else would play with me. He taught me all those
things you mention, about stick speed, spin, and angles. Rem was obviously
not a physicist; emphatically not a physicist. He was an empiricist. He
taught me how to chalk a cuetip around the edges but not the center, to
control spin. He taught me how to hold the cue and stroke it so you could
feel the living echoes of the impact running up and down the cue, the sign
of a perfect contact. He taught me how to pick my spot on the cueball to
impart spins of varying degrees (about three degrees is all, but three
standardized spins is all you need). He taught me how to use the diamonds on
the rails to estimate the bounce angles, and showed me how contact with the
rail could impart spin, and how an initial spin would affect the angle of
reflection off the rail. He showed me how a ball behaves in the corner with
and without spin, close to the corner and farther and farther from it, for
shallow and steep approaches. He showed me how to pick angles in corners
that were forgiving of errors. He showed me how to stroke a curve ball.
All this was obviously planning and forecasting of the highest order. And it
was all completely cognitive. It boiled down to learning what perceptions to
control at what reference levels in order to create, most of the time, a
desire result. It was about what to imagine, too, but mostly it was about
what to see and feel. Of quantitative estimation there was very little; that
was mostly about learning the angles at which balls would bounce apart when
contacting with the line of centers at various angles.
This is why I don't utterly reject the idea of SOME kind of model-based
control. But it's also why I associated it almost totally with the higher
levels of perception, and with reasoning and conscious imagination.
The same thing goes for golf, as you mention, and other sports like bowling
and baseball. The whole trick, in all these sports, is learning what
perceptions to control and in what reference conditions. The reason is that
this is ALL you can control while "in the act." The less estimating and
predicting you have to do, the better you will be at the sport. You have to
learn to remember what you did and what happened, so you can SLOWLY vary
what you do, time after time, to control the average state of the result.
But above all, you have to see through all the false explanations and
theories and misdirections to the variables that can be controlled directly,
simply, and easily.
I rather pounded on you in my post; got carried away (I visited Rick only a
week ago). I was reacting to what, to me, sounded like a traditional banal
explanation that everyone knows is true and nobody has really tested.
Obviously, we do project and predict and forecast and even simulate things.
But I contend that we do this mainly when accuracy doesn't matter or when
lower-order systems will take care of the details. Planning, I contend, is
greatly overrated; that's probably why large organizations spend so much
money on it. It's vulnerable to inaccurate or plain wrong assumptions, and
unexpected contingencies, and inability to carry out exactly the prescribed
actions. And it usually produces about the result you would expect from
knowing those deficiencies. In spite of this, there are people who just love
planning, and are willing to excuse just about any failure by blaming it on
something other than the plan -- which is, of course, at the same time a
legitimate excuse and the basic reason that plans don't work. Something
other than the planned future is always happening. Everything would have
worked perfectly if my secretary hadn't got sick on that particular day.
So you see, you're the victim of a pet peeve.
Best,
Bill P.