Burgess, si; Bizzi, no!

[Avery Andrews 931007.1355]
  (Bill Powers (931006.1230 MDT))

>>A sensory template is a learned representation within the
>>central nervous system of the sensory input that will occur
>>in the course of a movement if the movement is carried out as
>>intended.
>
>Several problems here. First, the intention is connected to the
>output instead of the input, as if the output is planned, then
>the result is compared with the template to see if the plan was
>right. No concept that the system could automatically adjust the
>output on the basis of the error.

I think you're misreading Burgess. Look at the picture on p. 721
with about five loops, with labels such as `Signals adjusting selected
E(ffort)-D(esired)A(pproach)P(rofile) to match selected E-DAP. It
seems to me that kinesthetic percepts are being chosen on the basis that
they will produce the attainment of visual or otherwise devined ones,
and then themselves being attained by means of feedback control.

>Second, I don't buy the idea of a "template" which implies a
>reference pattern with which a perceptual pattern is directly
>Second, I don't buy the idea of a "template" which implies a

I'm not too happy about this bit either. What I'd say the templates
were is the output of nonlinear circuits - you activate the circuit
by sending in a signal, and a signal comes out that serves as a
reference contour. Whether this should count as `containing a learned
representation' I don't know, but it certainly isn't a picture.

So I'd still but Purgess on the good guy list.

Bizzi, on the other hand, is still at his devilry: on pg. 808, 2nd col,
mid, he talks about instability, and PD controllers, without bothering
to point out that integrators solve this problem.

Avery.Andrews@anu.edu.au