COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00006 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source
Commentary by John McCarthy, Artificial Intelligence Laboratory, Stanford, CA 94305.

	A comparison of animals and machines or computer programs may
suggests changing the emphasis from the command neuron to the behavior
commanded.
A %2unisource%1 behavior might be defined as one caused by an event
in a small part of a machine or animal whenever the rest of the
system satisfies certain general conditions that prevail an interesting
fraction of the time.  In designing a machine or computer program one
often faces decisions about whether a certain behavior should be produced
in a uni-source way, e.g. by a specific subroutine, or whether it must
arise from the interaction of many parts of the machine.  Thus
Fortran compilers give error messages on meeting ungrammatical
statements, e.g. a compiler may print "PARENTHESIS ERROR IN STATEMENT 45".
In some compilers, the printing of "PARENTHESIS ERROR IN STATEMENT" is
a %2unisource%1 event produced by a routine that detects parenthesis
errors.  In other compilers, it will be a multi-source event, namely
"ERROR IN STATEMENT" is %2unisource%1, being produced by a general part
of the error routine while the adjectival "PARENTHESIS" is generated
by the part that detected the parenthesis error.  In no reaonable compiler
will the whole event including printing "45" be %2unisource%1, because no-one
would make a subroutine specialized to detecting parenthesis errors in
statement 45.  If one found the word "PARENTHESIS" sometimes misspelled
in an error message, one would infer that there were at least two different
routines that handled parenthesis errors, although the inference would
not be completely conclusive.

	A second example comes from cryptography.  If one successfully
concludes that a certain feature of cryptograms in a given cipher is
the invariable consequence of the presence of a certain stock phrase
in the message, one is well on the way to solving the cryptogram.

	It would seem that the discovery that a given behavior in
a mollusc has a single source, e.g. is triggered by the firing
of a particular neuron, is important.  It raises the question of
whether %2unisource%1 behaviors in molluscs are usually triggered
by single neurons.