Long-Term Memory Neural Circuits, Fast and Precise - Brain Theory from a Circuits and Systems Perspective - page 128

Chapter 8
Long-Term Memory Neural Circuits,
Fast and Precise
Introduction
Long-term memory has a great many logical properties, so it follows that its
neurons most likely possess a logical structure. Perhaps the most obvious logical
property is associability, since fragments in consciousness result in associative
recalls from long-term memory. Fragments in consciousness refer to collections
of attributes, that is, basic colors, shades, shapes, edges, and other components of a
mental image.
Cues are selected from conscious short-term memory (STM) and sent “down”
into long-term subconscious associative memory; images with attributes that
exactly match all cues are returned for possible updating of conscious STM. It is
emphasized that images are called forth only for exact matches. Gratefully, old
memories are dormant until receiving an exact set of cues since serious overload
would result if we had to see all returns, even those totally irrelevant.
A neuron never has to multiplex two or more attributes, and so will not have to
endure incessant jumping between unrelated signals in the same axon. Multiplexing
is excluded in the model of this topic partly because neurons are plentiful and can be
dedicated to given attributes, and partly because dedicated neurons are necessary in
order to define a plausible neural system.
It follows that attributes that have their own dedicated column of neurons and are
distributed. In the model below, neural clusters are dedicated to a given attribute;
each attribute connects to specific locations in a word of STM and to corresponding
locations in words of long-term memory. Location is thought to define the meanings
of attributes. Locations obviously exist physically, although we do not yet know
exactly where.
For example, the color fire engine red might be the 37th neuron from the left in
the word of STM; it is also the 37th neuron from the left in each word of long-term
memory. Of course, this 37th neuron will be active only when red is needed, such as
when recalling a red fire engine.