Abstract : Choice reflects the interplay among the facts of human cognition and especially human limitations, out intuitions concerning the principles to be found in simple cases, and pragmatic considerations of the sort last mentioned. Probability gives us a handle on these things, but of course probability is one of them itself. Is it more fundamental? Sure. Does that cut much ice? Not much. To the extent that we can limit our considerations to what is common to a number of views of probability, we may find that we have a basis for choice. What we need is an account of defeasible reasoning -- that is, of a kind of reasoning that will give us practical certainties, but that is not monotonic. It will be a kind of reasoning that allows us to change our minds in response to new evidence. This principle of specificity seems to work well in a limited domain of cases, namely the cases in which we should be looking at subsets. Construed broadly, as logical strength, the principle seems incontrovertible, but it also seems devoid of itself. Construed as something in between the recommendation to choose a superset as a reference class, and the general heuristic recommendation to use all the evidence you have, it is not clear how to apply the principle.