Avery Andrews sent this note. My response follows.
Gavan,
Since you appear to know a fair amount about the Gibsonian view
of things, I have a question. The lead-up is a bit long, the
the query itself is short.
One of the central ideas of the `neo-Gibsonian' approaches to movement
is that of a `coordinative structure', a constraint relating a
collection of `indices over muscle function, such as, for example
joint angles (Fowler & Turvey 1980:395). For example, if you want
to hold a coffee-cup vertically with a 3 jointed arm (wrist, elbow,
shoulder), all joints in the vertical plane, it can be achieved
by obeying the constraint:
0 = wrist + elbow + shoulder
(assuming appropriate measurements for the three angles)
Specific angles of tilt for the cup can furthermore be produced
by parameterizing (`tuning') the constraint:
T = wrist + elbow + shoulder (eq. 1)
So far, so good.
My problem is that I can't figure out what the implementation theory for
coordinative structures is supposed to be. Surely not Saltzman & Kelso
`task dynamics' since it seems most implausible that orgnanisms actually
perform the kinds of calculations proposed there (or is there something
simple I'm missing, that organisms could plausibly do?).
On the other hand, the 2nd order feedback systems of Powers seem to
be a plausible implementation theory for coordinative structures.
I'm not sure that the original 1973 versions would actually work very
well, because they don't have `slowers' (leaky integrators) in the
output functions, but the more recent versions described by Rick Marken
in various papers, and implemented in his `spreadsheet model', certainly
do work, and the approach raises what seems to me to be potentially
empirical issues. For example, eq. 1 can be enforced in various ways,
such as (a) setting the reference level for wrist to be
T - shoulder - elbow (a rather rigid, and I suspect, bad, approach),
(b) taking a more general 2nd order approach, whereby a too-large
value for T will tend to decrease the references levels for
all three joint angles (though some of them may be subject to
other constraints that would normally to the wrist producing
most of the effect anyway).
So, as far as I can make out, the Powers-Marken 2nd order systems are
the only specific proposal I have run into as to how coordinative
structures might be implemented in organisms. The question is then:
are there any other substantive ones in existence?
Thanks if you have time for this,
My response
I suspect (although I do not know for sure) that some ideas forwarded by
Fowler and Turvey are no longer highly regarded. I know I had some trouble
with the paper, and I no longer see it cited in recent work.
The notion of Coordinative Structures remains central. There are a number of
models, although I have not kept up with all of them. The Kelso lab has done mo
st
of the work in this area -- Kelso in collaboration with Schoner, Scholz, and
others, and Turvey has done some with his graduate students. There is a
recent paper by Turvey in American Psychologist that gives a good account of
the current no thinking. My copy has it "in press". I have had it about a
year. I could track down the current data if it is needed.
I do not think the claim ever has been that actors calculate differential
equations in constructing an action. These are, I think, presumed to be
models of direct physical, learned, intentional ... processes. As an analogy,
we would not presume that a falling object calculates its terminal velocity
via a differential equation, although a physical scientist could model it by
using a differential equation.
Does any of this help?
Gavan
Gavan Lintern
Aviation Research Laboratory
University of Illinois at Urbana Champaign
# 1 Airport Rd., Savoy, IL 61874
TEL: (217) 244-8637/6905 FAX: 244-8647