Publication 2756

Excerpt: It is clearly a bit naive to consider trying to write the input-output transfer function for any middling-to¬-complex behavior. In order to make the problem simple enough to handle one must focus on very low- level definitions of input and output, with no guar-antee that the observed function is not contaminated. One can only hope that while the subject is tracking the target he does not change his mind about the importance of speed and accuracy, and does not start thinking about lunch, and is not reminded of some similar and unpleasant situation, and does not want to ruin the experiment. What we need is some way to find the right defini¬tions of input and output, definitions we can be rea¬sonably sure are appropriate to the behaving system and not gratuitous embellishments, scientistic irrele¬vancies or inappropriate anthropomorphisms. I think that we have such a way in a control-system model of behavior.