Its been a few days since the gentleman posted on this so I can't append
his message, but in general form the question :how to relate
heirchical control and the "learning curves" that have been observed,
spercifically ones with negatively decelarating reaction times.
First off, RT is a crap measure to understand what an organism might
be doing, and how it is organized to do such a thing.
With that preface (that I don't think anyone who cares about examining
the nature of purposive behavior should feel obliged to "explain" crap
data based on a faulty conception of behavior) I can think of three
reasonable hypotheses on why one might obtain such curves.
First, the curves don't even exist. Or, that in the search for a smmoth
transitions in "learning" and "effeciency" points not falling along
smooth functions were considered measurement error, hence discarded or
deemphasized through a variety of statistical techniques.
Second, that one minor final product of learning is such a RT curve,
but the changes that occur across the orders are distinct and abrupt so
that as this "trickles" down the heirarchy a smooth end change in RT is
Third, (suggested by Tom Bourbon) the slope is of the same form as error
traced over time for any proportional controller with a constant
disturbance, and this can also be seen in the adaptive controller.
So that possibly one of the variables being controlled is speed at which
a task might be completed which affects the K-factor in a subordinate loop.
If the experiments were conducted so that the controlled variables could
be specified then one of these hypotheses might be tested.
Unfortunately, that is not the case and psychology remains in an all-too
terribly ambiguous form.
Any comments appreciated.