Assuming the existence of encoding synapses that record presynaptic axonal
'on-off' pattern as the content of memory, the author has presented a brain
model. In this brain model, the synapses of a neuron work like a static ra
ndom access memory (RAM) that may encode 2 the power of 10 000 'on-off' pat
terns, the cell body like a central processing unit (CPU) that produces a s
ignal of 1 or 0 in response to different presynaptic axonal 'on-off' patter
ns, and the axon like a data bus to form synapse with another neuron. Accor
dingly, the brain is analogous with a computer made of serial static RAMs a
mid 14 billions of parallel processing CPUs. Such a brain model converges d
ata with each tier of computation, because there are always more input pres
ynaptic 'on-off' patterns than output axonal 'on-off' patterns in a cortica
l area. The more computations of the data from the primary perceptive corti
ces, the more likely the data involving the synapses of central cortices, a
nd the more abstract the content of the memory-hence, in the retrieval of m
emory, parts of the 'on-off' patterns of the original stimulus may lead to
the converged, abstract memory. This can be the mechanism of pattern recogn
ition in the brain. (C) 1999 Harcourt Publishers Ltd.