🚀 go-pugleaf

Speculative!

The sensory minicolumns in the mammalian olfactory bulb (OB) run perhaps
3000 neurons. These are labelled 'glomeruli'. The number of glomeruli range
from 300 in mouse to 3000 in man to 30000 in pig. In man, the lateral
olfactory tract (LOT) has about 500000 axons, and there are about
3000000 excitatory tufted/mitral cells in the OB, so one t/m cell in 6
communicates from the OB to the Prepyriform Cortex (PC).

* This works out to 170 axons per glomerulus transiting the LOT.

* Each axon synapses about six times with a given superficial pyramidal
  cell in the PC. This results in perhaps 500 sp cells hearing from
  each axon in the LOT and each sp cell hearing from about 500 axons.

* Hence each sp cell hears from about 500 glomeruli of 3000 in man and
  from about 500 of 30000 in pig.

* Assume the OB is organized into activation networks by smell and that
  the OB keeps about 100 smells in memory at one time. These activation
  networks are made up of t/m cells interconnected by potentiated
  axo-dendritic synapses. If activation networks compete to respond to
  a given pattern of sensory data, then the average sp cell would hear from
  5 cells representing the winning activation network. There is reason
  to believe this is too few. For smell-oriented animals, the average
  sp cell could be hearing from less than one cell in the OB.

* Note that there is evidence for a lack of invariance in the efferent
  signal of the OB. In other words, there are no 'grandmother' cells
  for smells in the PC.

* The implication is that the signal of the winning activation network
  in the OB has to be amplified (in the OB) before it can be detected
  in the PC. You get the following:

1. Glomeruli are activated by the sensory data.
2. Activation networks are organized via potentiated long-distance
   connections between specific tufted/mitral cells.
3. The activation networks compete and one wins.
4. The winning activation network entrains the remaining cells in
   the associated glomeruli. This amplifies their synchronized signal to
   the point that the PC can detect it. However, the losing activation
   networks remain silent during this final stage. (How is hard to
   imagine unless the large majority of the tufted/mitral cells in each
   glomerulus are independent of the activation networks.)

* Note that a similar argument might be applied to the association 
  cortex. Activation networks need to compete to respond to an
  afferent pattern, after which the winning network has its synchronized
  signal amplified by the great mass of excitatory cells in the system.

Comments?

--
Harry Erwin
Internet: herwin@gmu.edu 
Working on Katchalsky nets....