224 Commentary We Also Need Complete Behavioral Models

I am very pleased to see Donahoe, Palmer, and Burgos foster this important discussion with a very clear and cogent paper. I find their reasoning inescapable that environment–behavior relations are what is always conditioned. My own computer model of operant behavior is built on the same assumptions (Hutchison, 1984, 1985, 1995, 1997; Stephens & Hutchison, 1993), and the many simulations I have done are uniformly supportive of their claim that a single model can produce discriminated and undiscriminated conditioning in a wide variety of situations (e.g., Hutchison, 1997). I hope I can now come out of the closet without having to wear a scarlet S-R on my chest. I also strongly agree with Donahoe et al. regarding the value of computer modeling as a scientific tool. Donahoe and Palmer (1994) describe computer simulations as formal interpretation (p. 128), automatically ‘‘reasoning’’ or predicting from a set of assumptions to their implications. All scientists are constantly making predictions from theoretical formulations, both in designing experiments and in judging the congruence of data with predictions from various formulations. Just as we use a computer spreadsheet to ask ‘‘what if?’’ questions when data get too complicated to manage on paper, we can use a computer model to ask ‘‘what if?’’ questions of complicated behavioral relations. Our subject matter concerns environment–behavior systems that complicate enormously as the history of the organism develops. In my opinion, computer models are essential for further progress in our discipline, and they imply no necessary compromise of our radical behavioral philosophy of science. Quantitative behavior analysts are increasingly using these tools in valuable ways; for example, simulating how molecular conditioning processes that occur