ABSTRACT
Michael Arbib's comments on the stimulus-response theory of finite automata raise in a somewhat different form issues that continue to divide cognitive and stimulus-response psychologists. This chapter shows how the theory of finite automata and the theory of test-operate-test-exit (TOTE) hierarchies in the sense of Miller and Chomsky could be subsumed in a rigorous way within an explicitly and precisely stated stimulus-response theory. Arbib alleges that he has given a much simpler derivation of my main theorem: given any connected finite automaton, there is a stimulus-response model asymptotically isomorphic to it. Arbib argues in several places that learning simply could not take place according to stimulus-response conceptions because the number of states required is too large and the amount of time needed for conditioning is too restricted. He expresses clear preference for TOTE hierarchies without considering these hierarchies as special cases of stimulus-response models, according to the first corollary of main theorem.
