[From Bill Powers (921006.0945)]
Bruce Nevin (921006.0835) --
I am troubled by the noun "percept" that you have introduced,
Martin, and have been using for some time. Maybe you can clear it
up for me.
I agree with Martin's definition. In my 1960 paper with Clark and
McFarland (A general feedback theory of human behavior, Part I;
Perceptual and Motor Skills _11_,71-88) I defined it this way:
"A _percept_ is the basic unit of experience. It is that "bit" of
perception which is self-evident to us, like the intensity of a light,
or the taste of salt.
"A _variable_ is always a combination of two classes of percept. One
class contains percepts which _do not vary_; by these percepts we keep
track of the "identity" of the variable. The other class contains
percepts which _do_ change; these percepts carry information about the
"magnitude" of the variable. "Magnitude" is used here in its most
general sense, including the meanings of "intensity," "size", or any
other word for the general class of variable attributes." (p. 3-4 in
LCS).
All this fancy definition-making did NOT lead to Principia
Gubernatoriae, although I had, then, ambitions in that direction. But
it's clear that Clark, McFarland, and I started right out defining the
world from the standpoint of the perceiver, with the percept being the
basis for everything. What I had in mind, I think, was to define
control theory without any reference at all to an objective outside
world. I obviously gave up on that as being impractical. As Martin
Taylor (921006.1045) puts it so neatly:
···
(1) I control the CEV, (2) I control the percept, (3) I control the
CEV, realizing it is really the percept.
-------------------------------------------------------------------
Rick Marken (921003.1100) --
I think that Martin has been saying lately that statistics is used
improperly in psychology (predicting individual behavior from mass
measures, etc.), but that it still has uses in understanding signa-to-
noise ratio, observational error, and the like. I think this dead
horse will stay dead now, at least locally.
I take a somewhat different tack, which is that the world of
experience doesn't seem statistically uncertain. Whatever the actual
noise level of neural signals (and it isn't zero), it is low enough on
the time scale of perception and action that it plays no important
part in HPCT. The world looks smooth, continuous, and sharp down to
the limit of resolution -- which is all we care about. This tells us
that most neural signals involved in ordinary behavior have magnitudes
large enough to make noise unimportant.
Uncertainty that crops up in behavior is, I think, to be attributed to
conflict, not noise. Only when we're poised on the knife-edge between
conflicting wishes, or when unusual circumstances require us to see in
the dark, smell something at the threshold of detection, or hear
something on the bring of inaudibility, does the underlying noise
level make any difference. What people call "choices" I would call
"conflicts." When there is no conflict, you simply act to produce and
maintain the world you want, choice being unnecessary.
-------------------------------------------------------------------
Best to all,
Bill P.