What is right with the miracle argument: Establishing a taxonomy of natural kinds

IT CERTAINLY strikes us as one of the most remarkable types of scientific achievement when apparently disparate phenomena are unified theoretically. What appeared to be disparate to the untutored eye turns out to arise from the same underlying mechanism and thus to be identical in kind. When a door slams because the windows are open and it is windy outside, this happens due to the same cause and according to the same mechanism that makes a plane lift off the ground. The prima facie conclusion is that science succeeds in going beyond the specious distinctions of the senses. It teaches us what things are truly alike. My aim in this paper is to examine the viability of this popular view. And the result will be that the view is basically correct. More precisely, I will try to show, first, that in some distinguished cases science arguably manages to induce the right classification or taxonomy among the phenomena, and that, second, this is the only access to reality that science is justifiably able to gain. Accordingly, what I am aiming to do is to support a particular and comparatively weak form of scientific realism. Scientific realism contends that claims about certain unobservable ‘items’ which emerge from the theoretical or experimental activity of scientists are literally true; these claims faithfully refer to what is, as it were, going on behind the scenes. There is some quarrel, however, about what these ‘items’ are legitimately supposed to be. The leading brand of this doctrine is theoryrealism. According to this position, the successful theories of mature science are approximately true. That is, these theories correctly portray the notdirectly-observable processes and mechanisms that make the phenomena occur the way they do. A more attenuated version of scientific realism is entity-realism. On the one hand, entity-realism is an immediate consequence of theory-realism. The truth of a theory implies the existence of its theoretical entities. On the other hand, there is also a more autonomous type of entity-realism which is advanced on the basis of experiment-centered arguments. In this version, entity-realism says that the capacity to manipulate certain unobservable entities, and, in particular, to manipulate them in order to experiment on something else, gives