ABSTRACT
Extending an implicit assumption by Patrick Suppes, the chapter argues that automata may provide an adequate basis for a theory of complex learning. It is more natural to reduce a theory of this type to S-R associations in a way which is more consistent with Michael Arbib's axiom than with those assumed by Suppes as a basis for stimulus-sampling theory. At first glance, the S-R approach might appear to have certain advantages in that it pinpoints precisely which S-R instances need to be learned. Suppes' proof will prove useful only to the extent that: automata provide a useful concept for theorizing about complex human behavior and stimulus-sampling theory, or some other "conditioning" theory, provides the most natural reduction to S-R associations. In automata theory, there has been a general tendency to ignore the distinction which may be made between programs and machines which use them.
