[Comp-neuro] Discussion - Kuhn - and brief comments

[James Schwaber]

This post provides a view of the disagreement about whether there is a 
comp-neuro paradigm.

It argues that there is, but that in light of new data we can no longer 
believe the paradigm's core assertion, around which the field is 
organized, that a structural foundation underlying input-output 
electrophysiology is the fundamental unit for brain analysis.

The comp-neuro paradigm is well-entrenched in this forum and in the 
homogeneous culture of communities that gather for the CNS meeting, 
CRCNS reviews, as NEURON-GENESIS users, etc. The paradigm formalizes the 
data and ideas developed in the classical era of neuroscience, extending 
into the early 80's. This era was dominated by experiments on 
tract-tracing/connectionism, neurophysiology and channels, and these 
were motivated by the ideas of that time as to 'how the brain works', as 
in the following 3 premises:

(1) The neuron is the fundamental unit for analysis, a neuron's 
input-output membrane electrophysiology defines its function, and this 
arises from the detailed, precise structure of the neuron's somatic 
morphology, dendrite morphology and synapse-channel locations on same.

(2) These specific neuronal structures are essentially structurally 
fixed, and are located in functionally specific connectional circuit 
architectures.

(3) Neuronal structure and circuit architecture are the fingerprint of 
natural selection, the result of evolution's genetic program and thus it 
all "matters" - the brain is "Kolmogorov Complexity Complete" (KCC) - at 
a first approximation brain function (traits, behaviors, minds) arises 
from the naturally selected, structurally fixed, functionally specific 
structures as instinctive-ethological type outputs, approaching fixed 
action patterns.

The comp-neuro paradigm has been immensely productive, arguably the most 
successful use of modeling within biology. However, equally arguably, 
progress has greatly slowed. Worse, its agenda now, at the limit, calls 
for 'KCC complete' type goals that are heroic if not completely out of 
bounds. Worst of all, the value in completing these goals appears 
questionable in light of more contemporary approaches such as systems 
biology, which for example shows that great 'biological variability 
(noise)' is expected, generates robust systems, and is normal among and 
within neurons.

Even more important than the above difficulties within the comp-neuro 
paradigmatic structure itself is the fact that, from the early 80s on, 
neuroscience has undergone a technical revolution that has produced a 
wealth of new data and approaches, and these new data and ideas are not 
accounted for in the comp-neuro paradigm. For example:

(1)  Work in contemporary neuroscience emphasizes receptor complexity 
leading to combinatorially derived phenotype or function, including 
non-synaptic inputs and dominated by inputs not coupled to ion channels, 
without direct electrophysiological effects. By these mechanisms 
environmental experience is read and encoded. A neuron's receptor 
profile has downstream effects on signaling networks, gene networks and 
function. All these layers are interconnected by feedback, and each 
layer (signaling, genes, receptors, neurons etc.) is organized in 
networks that have nonlinear dynamics. The functional impact includes 
channel state, type, density, location but also has many dimensions 
beyond electrophysiology. If all this is true, then the first premise 
above is not.

(2) Directly related to this is the known plasticity in modulators, 
networks and cellular physiology such that neurons are constantly 
changing, not in a structural steady state. This produces flexibility 
and variability in structural organization and in function. Similarly, 
at a higher level, the work of Mezernich and of Greenough and others on 
brain plasticity-remodeling comes to mind here, as well as the emerging 
literature on adult stem cells. If all this is true, then the second 
premise above is not.

(3) The evidence from functional genomics and signaling systems likewise 
is not consistent with a structurally committed brain, but rather one 
using diverse strategies and ongoing remodeling, nor are recent 
epigenomic developments, suggesting rapid, current evolution of traits. 
The kind of genetic determinism inherent in the comp-neuro paradigm is 
naïve. If all this is true, then the third premise is not.

A new computational paradigm for neuroscience arguably not only has to 
account for these and other new data but also needs to develop a guiding 
concept of 'how brains work' that organizes the data, putting adaptive 
remodeling and fungible interactions of nonlinear dynamical networks in 
functional process in a central role.

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.neuroinf.org/pipermail/comp-neuro/attachments/20080905/16e55863/attachment-0001.html