In article <ERIC.93Oct26174737 at thing4.nntp-read.bu.edu>,
Eric Schwartz <eric at thing4.nntp-read.bu.edu> wrote:
>>One (hopefully) helpful hint: please don't confuse (as is increasingly
>common) the field of computational neuroscience with " modeling
>neurons", as opposed to computational modeling at all levels of the
This is closely related to another distinction, that of
"connectionism" vs "computational neuroscience". Making the
definitions at the lowest level of abstraction (i.e., comparing
"realistic" single neuron models with "connectionist" single "neuron"
models) helps to disambiguate between the two, in part because its the
level of abstraction where the fields are most different. 
But kick up a few levels of abstraction and its less clear where the
boundaries are -- models of biological function at a level too
abstract to allow "realistic" single-neuron modeling have a lot of
mathematical machinery in common with "connectionist" models for
Vision modeling and motor-control modeling are two good examples ...
all that's left to differentiate between engineer and scientist is the
intention, and when those intentions are best summarized as, say
"understanding how the brain computes robust visual motion" and
"making an engineering system that computes robust visual motion, by
taking inspiration from neurophysiology and psychophysics", we're
splitting hairs ... take, for example, Nowlan and Sejnowski's paper at
NIPS last year, "Filter Selection Model for Generating Visual Motion
Signals", the paper would be equally at home in an engineering or a
computational neuroscience journal.
 Although there are fringes of connectionism that are quite interested
in single-neuron models that share some of the dynamics of real neurons.