[From Bruce Abbott (980526.1155 EST)]
Bruce Nevin (980525.0839 EDT) --
If circumstances requiring learning are circumstances in which existing
control structures control with error, then this is a plausible mechanism
for reorganization. It answers the question, how are the new connections
localized in the brain. It is possible that the proliferation of redundant
connections occurs with some gross localization, just as in embryology the
growth cones of axons follow a chemical gradient to a region of the brain
The term "reorganization" encompasses at least two distinct types of change,
one involving the fine tuning of control parameters (especially gain), the
other involving the creation of new control systems. In both cases
presumably the process is driven by persistent error in the variable over
which the system is attempting to exert control, and in both cases the
current model assumes a process of unsystematic variation and selective
retention. However, there are other sorts of learning. As we experience
the world, some of that experience apparently is being continually stored in
some form, without regard to control errors. Would you be willing to
postulate that different mechanisms are involved in these two forms of
learning, or would you opt for parsimony and assume a single common
mechanim? If the latter, then you must either abandon the notion that the
neurological changes involved in learning are error driven, or assume, ad
hoc, that error must be involved in all learning. Something to think about.
Regards,
Bruce
[From Bruce Nevin (980527.0916 EDT)]
Bruce Abbott (980526.1155 EST)--
It seems likely that whatever mechanisms underly reorganization also
underly organization in embryology and in evolution, or at least that those
mechanisms have not gone away and still play a role even if something in
addition to them has developed specifically for learning. Cells controlling
their cellular inputs do things whose effects (probably many unintended by
them) constitute reorganization. For example, the axon grows along the
chemical gradient, pushing its growth bud around and through obstacles
until a synaptic connection is made in the appropriate region, doing this
for reasons that make sense in purely unicellular terms. Those conditions
and characteristics and supracellular effects of cell behavior have not
gone away. Beyond that admittedly rather vague premise, speculation is a
bit too askeletal.
As we experience
the world, some of that experience apparently is being continually stored in
some form, without regard to control errors.
I don't know anything about the mechanisms of memory either. I wouldn't bet
on it being without regard to error. Emotion-laden events are memorable.
My question was whether this finding of neurophysiology as described in
Gary's book had been discussed here. I'll take that as a no. Gary, are you
there?
Bruce Nevin
[From Bruce Abbott (980527.1025 EST)]
Bruce Nevin (980527.0916 EDT) --
Bruce Abbott (980526.1155 EST)
It seems likely that whatever mechanisms underly reorganization also
underly organization in embryology and in evolution, or at least that those
mechanisms have not gone away and still play a role even if something in
addition to them has developed specifically for learning. Cells controlling
their cellular inputs do things whose effects (probably many unintended by
them) constitute reorganization. For example, the axon grows along the
chemical gradient, pushing its growth bud around and through obstacles
until a synaptic connection is made in the appropriate region, doing this
for reasons that make sense in purely unicellular terms. Those conditions
and characteristics and supracellular effects of cell behavior have not
gone away. Beyond that admittedly rather vague premise, speculation is a
bit too askeletal.
I believe Bill P.'s view is that the development of the nervous system
unfolds as "reorganization" creates new levels of control atop those already
functioning. Thus the same physical principles observable in development
(such as those you describe) would also underly those later changes
attributed to learning. Although I think that Bill's view of neurological
development gives experience too great a role (because individual
differences in brain wiring are not as large as would be expected under the
hypothesis that brain structures result largely from the vagaries of
unsystematic reorganization), it is clear that such development is
influenced by experience, in that appropriate feedback is required in order
to maintain connections, once formed. The same or similar mechanisms are
probably at work to guide neural growth during initial development of the
nervous system and during subsequent fine tuning, including those changes we
call learning, although the conditions that initiate such growth are
probably different when we look at what triggers the appearance of the
relevant chemical gradients.
As we experience
the world, some of that experience apparently is being continually stored in
some form, without regard to control errors.
I don't know anything about the mechanisms of memory either. I wouldn't bet
on it being without regard to error. Emotion-laden events are memorable.
I didn't suggest that error has no effect on the creation of memories (it
clearly does), I suggested that memories probably can be created in the
absence of error.
My question was whether this finding of neurophysiology as described in
Gary's book had been discussed here. I'll take that as a no. Gary, are you
there?
Neural growth during learning has certainly been discussed, but I don't
recall whether the specific finding to which you refer has been.
Regards,
Bruce