[From Bill Powers (2006.07.11.1325 MDT)]
Marc Abrams (2006.07.11.1508) –
Bill, what does the number
coming out of the input function supposed to represent? That is, what
characteristic of a perception? If I increase that number, what am I
increasing? The intensity? If so what does that mean?
It means the amount of whatever perception the input function is
organized to detect. Each kind of perception requires its own input
function in this model, and yes, that means a very large number of input
functions (and control systems). However, because of the hierarchical
nature of this perceptual model, it could be that each function can be
simpler than it is in other kinds of models. And because each control
system controls only one one-dimensional variable, the actual control
systems themselves are very simple. That’s the tradeoff I
chose.
Other perceptual models have the signal coming out of the input function
be a code that indicates which thing is being perceived. The problem with
that kind of model is that only one perception at a time can be
recognized. And the input function still has to have the ability to
discriminate among different perceivable objects, which is usually done
by feature recognition or some sort of holographic encoding. In fact, all
the same functions are needed as in my model – they’re just packed into
a single gigantic function.
Still others (which I doubt are much believed today) assert that the
signal inside the brain is simply a copy of the external thing being
perceived. Others still, such as Gibson’s (implied) model, says that
perception is simply “transparent” – the brain looks through
input functions at the world itself as it really is. Good trick.
The “coding” concept of perception is not very useful for
control systems. You get codes identifying the perception: that’s an
“A”, and now it’s a “B” or a taste of
“Chocolate,” but the idea of controlling the A to make it be
tall or fat or lean left or right is impossible to handle. There are no
continuous variables in the coding model: it’s either this or it’s
that. In HPCT, that sort of perception is confined to a single
level, the category level. And even that level has some continuous
features – better and worse examples of a given category.
In the PCT model, perceptual signals are like variable numbers; all that
determines their meaning is the function or functions that generated
them. Reference signals also are just variable numbers; what they specify
is determined by the control system they enter and the meaning of the
perceptual signal in that system. This means they can be compared with
the perceptual signals; they have the same units.
It’s tempting to speak of neural signals as if they share some
characteristic of the thing they represent. But any attempt to do that
brings in infinite regress, because you have to ask how the brain
recognizes that characteristic of the signal. Another perceptual function
of the same kind is implied. In that sort of concept, you need an
infinite hierarchy of perceptions. And anyway we petty well know that
neural signals are all alike; you don’t have one kind of signal to
represent chocolate and a different kind of signal to indicate vanilla.
or salty, or bright. People have imagined that they see temporal codes in
neural signals, but whether they are seeing a real code or just rapid
variations in rates of firing has never been established.
A basic problem with the PCT model of perception is that the perceptual
signals don’t look like the things being perceived, and that seems to
contradict experience. But if you examine any perceptual experience in
isolation, a color or a sound or even a complex pattern, the
distinctiveness of the perception weakens or disappears. It seems that
each perception looks “natural” only when it’s
experienced in a network of other perceptions. This would seem to imply
something about the observer of the perceptual signals rather than the
signals themselves. I don’t know what that might be. It’s a problem that
remains unsolved, as far as I’m concerned.
Sorry about the drink from the firehose.
Best,
Bill P.