labeling and magic (Nevin)

[From Bruce Abbott (951025.1420 EST)]

Bruce Nevin (951024.1200 EST) --

Another way to distinguish terms like "imprint" from terms like "control"
and "reference signal" is in the precision and specificity of their
definitions. So far as I know, no way of modelling imprinting has ever been
proposed, and no plausible neural mechanisms for it have been advanced,
beyond some vaguenesses about holonomic brain wave structures. Perhaps I am
wrong. If so, that is the proper way to defend "imprinting" as on a par
with the terms of control theory. The argument that both are metaphors is
all fog.

Yes, that was a stretch, I admit. I was trying to take the argument to
absurd lengths in order to make the point that most terms, even "control"
were derived by metaphor and analogy to provide a needed label where no
adequate label existed before. With usage the metaphor gets forgotten and
the word comes to represent the new concept for which it was invented. This
is especially true in scientific work where those working within the field
adopt their own specific definitions for terms that may have more general or
somewhat different meanings among the general public. Even theory-laden
terms like "response" eventually may come to lose their original
connotations during the evolution of scientific thought and usage.
"Response" originally referred to a reflex-like reaction to some sudden,
specific change in input (called the "stimulus") but after being co-opted
from physiology has gradually come to refer to any pattern of behavior
having well-defined observable characteristics.

The suitability of terms like imprinting or attachment in scientific
discourse is not an argument I am willing to go to the mat for; in arguing
for some leeway and tolerance in the naming of phenomena I am not arguing
against the basic thesis that terms should be precise and, if possible,
theoretically neutral. In discussing the phenomena to which these labels
have been attached (how's THAT for metaphor? Do you see any phenomena with
labels literally attached to them?), it is nice to have the labels as
substitutes for longwinded descriptions. Theory-free labels are difficult
to invent, for the reason that no observation is theory free, but in my view
the advantage of having labels, even theoretically biased ones, often
outweighs the potential disadvantage that someone might take term in its
literal rather than metaphorical meaning. As theoretical understanding
changes, the names remain, but their old theory-laden meanings are replaced
by different ones for those working in the research tradition, and no longer
evoke the spurious images to which Bill objects.

Regards,

Bruce

[From Bruce Abbott (951101.1615 EST)]

Bruce Nevin (951101.1122 EST) --

Beautiful post. I only wish I could think and write half as clearly and
elegantly as you are able to express your thoughts . . .

However (you knew there would be a "however," didn't you?), the difference
in precision between, say, such terms as "attachment" and "reference signal"
arises from the difference in the kinds of things to which the terms refer.
A reference signal is a functional part of a mechanical system, having the
same precision as terms like "carburetor" or "exhaust manifold." It is
something physical you can sense and directly measure. Attachment is not a
part or thing but an inferred process manifested by a number of
perceptual/behavioral relationships that presumably emerge as an outcome of
this process and that serve to indicate its successful operation. We look
for a familiar way to express this process/outcome, and "attachment"
provides a more-or-less adequate term to convey what we have in mind.
"Attachment" thus compares directly to "control," which is a process or
outcome of that process, and not to "reference signal," which is a directly
measurable thing. If the word "control" had not already existed as a label
for the process/outcome to which it refers, I wonder what term Bill P. would
have chosen.

Regards,

Bruce

From Bruce Nevin (951101.1122 EST)

Bruce Abbott (951025.1420 EST)
[Martin Taylor 951025 11:00]
[Hans Blom, 951026]

We've touched on the ineluctability of metaphor when we use language.

Metaphor reflects how we think by analogies. A metaphor is a word or phrase
that has a straightforward literal correspondence to objects and relations
in a familiar domain, but it is being used in another domain in which the
perceptions are less well understood. The metaphor invites us to suppose
that the objects and relations in the new domain are analogous to those
that we believe we know in the old.

PCT requires discipline of our analogies. This is what modelling is about.

In a constructed PCT model, the terms "reference level", "control", and so
on have a straightforward literal correspondence to well-understood
perceptions of computer science (or whatever the domain of model making may
be). We are explicitly invited to draw an analogy between the behavior of
the model and the behavior of the modelled.

It is this -- something like the notion of "operational definition" -- that
distinguishes the terminology of PCT from other usage sharply rather than
by continuous gradation as Hans has suggested by analogy to the heap
puzzle.

        Bruce