[From Bill Powers (970924.0905 MDT)]
Chris Cherpas (970923.1108 PT)--
Reorganization may vary
two properties of an existing control system
signal values (e.g., references) and mode connections
(i.e., observation, control, imagination, automatic).
I have a problem with reorganization acting on signal values. A neural
signal has to come from a cell body; there's no other source of signals
that I know of. I have thought of reorganization strictly in terms of
affecting synaptic weightings or thresholds -- properties of neurons which,
once set, will stay that way until changed again. It's possible, sort of,
for reorganization to create a signal _generator_ by increasing the weight
of a local positive feedback connection until the loop begins to fire
spontaneously. But I don't think that takes care of everything.
There's a special kind of reorganization that occurs early in the life of a
higher organism; neurotaxis. When neurons first start to grow in the brain,
their axons seem to grow toward specific targets, following chemical
signals. These signals depends on activity in neurons, so we find the
connections somehow being specific to what's going on. Once those
connections are made, however, the axon becomes more or less permanent
(although axons can atrophy again).
Anyway, I wonder where the information to switch
modes comes from and this is one possibility.
I still think of reorganization as basically a random process. Any
systematic effects on "modes" or connections would result from the
operation of a learned system. But this is all just guesswork -- learning
has never been studied with the idea of reorganization in mind, so there's
really no data.
I don't think it's worth spending too much time on a hypothetical process
until somebody actually does some studies.
More generally, though, how does reorganization "construct"
control systems? What are the "scratch" materials? Are
all the connections already there and its just a matter of
deciding what connections are "active" -- between control
systems in a hierarchy as well as within control control
systems (modes)? Or are little pieces moved around or
grown to connect. The latter seems a more relevant image
for epigenetic and up-to-system-level development, whereas
the former seems a more pleasing image for the mature organism.
I don't know. What's the evidence one way or the other?