[From Bill Powers (2000.11.13.0718 MST)]
...things can be
classified according to mechanism -- finding a classification scheme and
talking about underlying mechanisms are not necessarily the mutually
exclusive activities you imply they are.
I agree, but this implies that you know the mechanisms before you do the
classification. Then, of course, you can group phenomenona according to the
_known_ mechanism that produces them.
I _think_ what the authors are
trying to do is find the elements in each case that correspond to the three
components of Darwinnian evolution: variation, selection, and replication.
Each of these phenomena, however, can be produced by different mechanisms,
so not everything that is classed as "variation" (for example) has
something in common with all other phenomena put in the same category.
These are not the names of mechanisms, but of phenomena that are generated
by mechanisms. And the phenomena are not even the same across all instances.
This, it seems to me, is no different than, for example, trying to identify
what physical elements correspond to the comparator, error signal, and so on
in a specific biological control system.
I think it's quite different, even though something of the same problem
still exists. A comparator creates one clearly-defineable relationship
between two inputs and one output. It's true that this relationship can be
created by several different mechanisms (for example neurons, vacuum tubes,
and transistors, each operating according to its own laws), but the
underlying quantitative relationship can be defined in a way that is the
same for each known mechanism: e = r - p. There is no problem parallel to
that of defining "learning," where one term includes clearly different
With respect to the "evolution" of response force, one might note that in
the set of lever-presses recorded during an experimental session, there is
variation in response force. If the apparatus now is made to require that
the force fall within a particular "window" before it will deliver a food
pellet to the hungry rat, this constitutes a selection criterion for
response force. If, as a result of this selection, the population of
response forces shifts until most are within the window, this corresponds to
replication, and the situation being examined qualifies as an example of
Darwinnian evolution, broadly conceived.
This applies only in a disturbance-free environment, which gives a
distorted picture of all relationships. In a realistic environment, one
response force will not result in one unique environmental consequence. In
that kind of environment, it would become clear that it is control of the
consequence of a particular response force that is learned: a control
process is acquired, not a particular act. However, when the environment is
arranged so the same response force always has the same effect on other
variables, we have a degenerate case in which it is not possible to see
what is actually learned. This is true of all experiments in which a
behavioral action always has the same consequence. The only way to see what
is actually going on is to introduce disturbances and see what, if
anything, is maintained the same or nearly so.