ABSTRACT
We understand how nature works only through models. The models may be very simple, like our model of the visual world or Copernicus’s model of the solar system. Or they may be very complex, like the model of DNA. The issues for behaviorists are whether they need models at all (perhaps some kind of unidimensional “strength” notion is sufficient?), and if so, what kind should they be. Using habituation in nematodes as an example of the simplest kind of learning, I have shown that response “strength” by itself is not sufficient: a model with at least two dimensions is needed to explain experimental results. Obviously, the much more complex kinds of learning shown by higher animals will require models at least as complex as this. The necessity of modeling means that sticking to observables, to behavior, is not an option. The concept of internal state cannot be avoided.
What kinds of models should we look for? I argue here that ruthless parsimony is our best guide. Neither subjective experience nor biological plausibility nor a kind of Erector-set approach using neurobiological ingredients is a reliable guide. The unique contribution of the theoretical behaviorist is to find the simplest possible process to explain behavioral data.
