On a Technical Problem for Behaviorism

Today's post is a little wonkish, I'll admit, so if behaviorism in general doesn't interest you you may want to scroll down and see some of the other posts. Supposedly a problem for behaviorism is that mental terms infest behavioral analyses of psychological descriptions, for example "He likes chocolate" means, among other things, that he'll move towards chocolate providing he doesn't want to fool us, and providing he doesn't hope to lose weight, etc. The mental terms don't wash out. I don't think that this is necessarily a problem. If the behaviorist was right to say that the psychological descriptions were properly analyzed as observable dispositions to behave, then all of the mental terms that keep cropping up in the analyses could also be so analyzed. We could do something along the lines of a Ramsey sentence, using DBx, for "Disposition to Behave x," instead of Fx. That doesn't mean that there is no problem. Carnap, as I recall, diagnosed the real problem here: the analyses into behavioral dispositions would result in indefinitely long, perhaps even infinitely long, descriptions, and that is a problem, one that defeats behaviorism, in fact. But it looks like functionalism is going to have the same problem. The function of any given state/process is going to be defined by that state's role in the larger functional economy, but that economy is open-ended. This is similar to Davidson's argument for the impossibility of psychophysical laws: the functional property, like the intentional property (same thing under a different description) is defined in reference to the larger functional/intentional "web," but the physical state is not.