To naturalize philosophy of science radically and thoroughly, we must re-examine the ontological character (not ontological commitment) of theories and ask how theories are realized in this physical world. Paul M. Churchland dare to answer to this question and claims that a theory is a partition across activation space which is realized by a specific pattern of synaptic weights in a brain. He also tries to justify some Feyerabendian strategies for doing science well in terms of neurocomputational functions of our brains. In this paper, Churchland's project to naturalize philosophy of science is defended against some criticisms. Then, a minor deficiency of his theory is pointed out and a way-out is suggested.
