Mass Action in the Nervous System

Harry Erwin erwin at trwacs.fp.trw.com
Sat Dec 4 14:53:35 EST 1993


I've been working my way through Walt Freeman's 1975 book, Mass Action in
the Nervous System (Academic Press) and am currently reading section 4.5,
Divergence and Convergence in Neural Masses, where he discusses the way
activity diverges into larger and larger masses as it progresses away
from the olfactory bulb. Why am I reminded forcefully of passive sonar
arrays (including the systolic dynamics)? Perhaps Freeman's models would
be better expressed in terms of time evolution of signal power, rather
than in time evolution of signal amplitude. That allows the carrier
waveform to be ignored temporarily. If I do that, I discover that the
granule cells, with their dendro-dentritic synapses, act to sharpen the
signal that the periglomular cells amplify and the tufted/mitral cells
distribute over the OB. In addition, if I stimulate the granule cells
externally, I shut down the corresponding regions of the OB. How do I use
this in conjunction with the AON and PPC to condition the OB? Now it
starts to become clear. The signal power I will be using to condition the
OB starts as a high power unfocused signal generated by the PPC. This I
run through the AON, where I also apply a control signal through the
granule-cell equivalents to sharpen the focus. This sharpened signal then
emerges from the AON and passes over to the 'wet end' of the OB, where it
serves to act as a conditioning signal for specific patterns of sensory
inputs.

This would seem to suggest that the olfactory system is dual-mode. In mode
1, associated with a strong signal generated in the PPC and focussing in
the AON, the 'wet end' of the OB can be painted with 'semantically
significant' patterns to be learned for matching to the afferent sensory
signals. In mode 2, the OB/PPC/AON system then does processing against
afferent sensory data, with 'chaotic resonance' used to amplify the
typically 'ratty' sensory data by proposing alternative pattern completions.
These pattern completions are then matched against the stored patterns of
interest.

The AON may be dual function even during normal operation. This is because
it appears that there are independent paths required for the two modes. This
may be the reason for the separate L1 and L2 paths, one synapsing on the
periglomular neurons (synthetic sensory data) and the other on the OB
granule cells (focussing instructions). Since the granule cells synapse
onto the tufted/mitral cells dendrodendritically, they may be able to go
beyond inhibition and actually prevent conditioning by intervening between
the conditioned stimulus and unconditioned stimulus. Since synapses
apparently habituate automatically if conditioning does not occur
(whether or not the neuron actually fires, due to the failure of the NO(?)
feedback mechanism), this might then facilitate habituation that would
otherwise not occur.

(Please note that this is 'speculation in progress.' Comments happily
accepted.)

Cheers,
-- 
Harry Erwin
Internet: herwin at cs.gmu.edu or erwin at trwacs.fp.trw.com
Working on Freeman nets....



More information about the Neur-sci mailing list