[From Rick Marken (961015.0850)]
Bill Benzon (961014) --
How does PCT model the process by which one chooses to pursue one or
another of those behaviours? Given that one is already pursuing one of
them, that is, seeking perceptions which satisfy the need, PCT says alot
about how that pursuit is conducted. But what does it say about choosing,
e.g. to eat, hunt, urinate, or groom?
The behaviors McCulloch describes are just the visible side-effects of
perceptual control. When we see an animal eating, we are seeing the visible
side effects of the animal's control of perceptions such as the taste of
food, the forces involved in chewing and swallowing, the chemical results of
disgestion, etc. Same is true when we see hunting, urinating, grooming, etc.
What we see is typically _not_ what the animal is controlling. So what looks
to an observer like a switch from eating to hunting "modes" is just an
observer's perceptual classification of side-effects of control.
There are no behavioral "modes" in PCT. Apparent changes in "modes" are the
observer's interpretation of what is seen when the organism changes
references for certain perceptions in order to control higher level
perceptions. The lion, for example, changes its reference for the perception
of moving toward its prey (one of the perceptions controlled during what we
see as "hunting") once the prey is no longer moving; it also changes its
reference for the perception of tasting the flesh of the prey (one of the
perceptions controlled during what we see as "hunting") once the prey is
immobile. The observer sees the side effects of these changing references as
a change from "hunting" to "eating" mode.
There is also no "choosing" to pursue one behavior rather than another in the
PCT model of behavior. The whole idea of "choosing behavior" is, I think,
based on a misconception of what behavior is. You have to "choose" behavior
if behavior is a "product" -- something produced by the organism for the
sake of the observer. Behavior, viewed as "product", seems to roll off a
behavior assembly line one item at a time -- hunting then eating then
urinating then grooming and so on. PCT shows that this view of behavior is
wrong.
PCT shows that behavior (from the point of view of the behaving system) is
really controlled perceptions -- like tastes, forces, movements, speeds, etc.
PCT assumes that controlled perceptions are hierarchically structured and
controlled; lower level perceptions are controlled relative to varying
reference specificatons in order to maintain control of higher order
perceptions. This hierarchical control process can look like "choosing
behaviors" to an external observer; the animal that was chasing the deer
is now seen as eating it. But these changes are really just visible side-
effects of the hierarchical control of perceptions.
If you want to see how hierarchical control works and you are lucky enough to
have Excel 5.0 running on a Mac then I suggest that you go to the CSG web
site:
and select "Computer Programs", "Macintosh", "Marken Programs", "Hierarchical
Control" and finally click "Hierarchical Control.sea" which will quickly put
my 3 level hierarchical control spreadsheet model on your desktop. Once you
load it into Excel (be sure to set "Iteration" to "on") and run it, you will
see three levels of control systems (six systems at each level) gracefully
changing the references of lower level perceptions (or, in the case of the
lowest level systems, changing the output effects on a physical variable) in
order to maintain their perceptions at the reference levels specified by
higher level systems (see the "Read Me" text that comes with the program for
a more detaild description).
I think it [PCT] needs to add something like McCulloch's mechanism to its
account of the relationship between the perceptual stack and the
reorganizing system
I think that's highly unlikely, considering the fact that McCulloch clearly
didn't know what behavior is: behavior, of course, is the control of
perception.
John Anderson (961015.0430) --
My problem is that I dont understand what C2 is doing, whether you call
r2 "offset", "criterion", or "bias". What is it comparing?
You are correct; I had r2 entering a comparitor when there was really no loop
involved. I think it would be better to think of r2 as a specification for a
perception of the _relationship_ between levels of perceived restriction and
saying "restricted" or "not restricted". Then changes in r2 would be requests
for changes in this perceived relationship. For example, moving r2 up might
specify changing the perceived relationship so that only very high levels
of perceived restriction are called "restricted"; moving r2 down would
specify changing the perceived relationship so that even very low levels of
perceived restriction will be called "restricted". So now changing r2 changes
the "bias" of the perceived relationship between perception of restriction
and saying "restricted" or "unrestricted". This also makes r2 a specification
for the level of a controlled variable (the relationship between what is
perceived and what is said) which is what it should be if it's called a
"reference".
Does this help?
Best
Rick