It is well known that modern philosophical controversies on analyticity originate in Kant's argumentation. For Kant, "analytic" was meant "true by virtue of vacuous predication." However, many philosophers since Kant have misleadingly identifyed this term with "(formal) logical truth," "a priori," "necessary truth," "linguistic truth," or "truth by virtue of meaning alone" and so forth. I have a feeling that such an identification has blurred the real issue on analyticity. Quite recently, J. J. Katz, J. A. Fodor and P. M. Postal have developed a semantic theory of natural language within the framework of N. Chomsky's generative grammar. Especially, Katz defined the term "analytic sentence" as one of the semantic properties and and relations within the framework of that semantic theory. In this paper, I discuss, first, in what respects Katz's definition is significant in comparison with such techniques as R. Carnap and other logical positivists have proposed. Second, I examine whether or not Katz's definition can be expected to surmount Quine's skepticism that the analytic-synthetic dichotomy is no more than dogma. Third, I discuss to what extent Katz's definition can refute the arguments of the so-called "gradualism" which has been supported by N. Goodman, M. White and many other analytic philosophers of today. I make it clear in this paper that Katz is successful in his definition of analyticity to a considerable extent, though he has some difficulties in it.