Brain Architecture for an Intelligent Stream of Consciousness - Brain Theory from a Circuits and Systems Perspective - page 22

Another form of logic, known as enabled logic, occurs for larger bodies of
capacitance, for instance, somas, as presented in a later chapter. As a result, a
complete set of Boolean logic (AND, OR, NOT) is available, a natural result of
synaptic activity in combination with active and passive regions in dendrites, soma,
and sometimes also axons. Dendritic logic is compared to enabled logic in a later
chapter.
Systems are typically expressed in terms of logical, not physical, structures, and
brain systems are no exception. Human LTM is associative, which hints that
location is not a prime consideration; things are recalled not according to where
they are, but according to what they are. In contrast, in a desktop computer, each
word in memory has a unique address, and a physical place. Addresses are used for
reading or writing to locations that may be chosen at random, so the choice is
random. Hence the term random access memory, or RAM. Human memory, in
contrast, does not have fixed addresses, so we know a lot less about where
memories reside.
Retrieval from subconscious LTM begins with cues. For example, seeing
someone's dog can bring forth a memory of your own faithful companion, and
might even bring forth details of face, color, size, and so on. Other such examples
are visiting your home neighborhood which brings forth memories of stores, houses,
rooms, passageways, and so on; seeing a topic that you used in college can bring forth
memories of what you learned in those days. Cues are essential to memory retrieval.
Cues need to be chosen judiciously from the current contents of conscious STM.
If the available cues do not result in a recall, editing is available to adjust the cues
for another try. Hopefully there will be a hit, which means an exact match to the
cues. Memories with attributes that exactly match the applied cues will be read for
possible entry into conscious STM. Cues applied to LTM have to be exact, since
LTM are called forth only for exact matches. Gratefully, old memories are dormant
until receiving an exact set of cues. Serious overload would result if we had to
experience all facts, even irrelevant ones, at all times.
Under the distributed model each attribute connects to specific locations in STM
and corresponding locations within words of LTM. These locations exist physically,
although current technology does not permit a knowledge of exact location. But
nevertheless each location in STM has corresponding locations in LTM that are
reserved for a particular feature. This is an efficient way to organize STM and LTM
because it avoids extra encoding and the passing of data to intermediary storage sites.
A given attribute might be said to have meaning not because it has any special
coding, but because of its physical location in a word. For example, the color red
might be in a certain location, say the 537th cell from the left in a delineation of
logical STM; it is also the 537th cell from the left in each of its replications in LTM.
Location in memory and the context in which the attribute is used contribute to
realizing the meaning of an attribute, as for example, when seeing a fire engine in
the context of a street, it is expected to be red.
The one-on-one distributed model implies that neurons are not multiplexed, that
is, different attributes are not communicated by the same neuron. Fortunately
neurons are quite plentiful and are conveniently dedicated to a given attribute as
described.