One of the goals of cognitive science is to develop theories that are strongly equivalent with respect to to-be-explained systems. This requires that evidence be collected to defend the claim that the model and the to-be-explained system are carrying out the same procedures to compute a function.
One type of evidence that can be used to support this claim is intermediate state evidence (Pylyshyn, 1984).. This involves observations of the intermediate steps, and/or the intermediate states of knowledge, that the two systems pass through as they move from being given a problem to providing an answer.
For example, if one was using a Turing machine as a model, then an immediate source of intermediate state evidence would be what the machine does to its tape with each processing step.
In studying human subjects, intermediate state evidence is not directly available. However, one method that might provide some evidence about these intermediate states is protocol analysis. For instance, when Newell and Simon (1972) compare a production system model to a human problem solver by comparing the problem behavior graphs generated by the two agents, they are relying upon intermediate state evidence.
- Newell, A., & Simon, H. A. (1972). Human Problem Solving. Englewood Cliffs, NJ: Prentice-Hall.
- Pylyshyn, Z.W. (1984). Computation and cognition. Cambridge, MA: MIT Press.
(Revised February 2010)