Most philosophical works on natural language semantics are done under the idealization that there exist shared stable linguistic conventions, and the mapping from sentences to propositions is common between all members of a community. While this idealization is fruitful for analyzing logical truth and logical consequence, it is not designed for the context of complex questions about natural language meanings. Whenever there is a debate about “What is X”, this idealization mandates that there is an absolutely correct answer, and at least one party must be wrong. But this idealization is, after all, an idealization. If we are willing to re-construe the question of “What is X” as “Has the meaning of ‘X’ changed if we apply it in such-and-such way in so-and-so situation?”, we will discover that speakers have different understandings of “X”. In the long run, it is plausible that the answer to “What is X” is a game-theoretical result of the interactions between different speaker’s Humpty Dumpty style understanding of the symbol “X”. Such an evolutionary understanding of semantics, if true, would have significant ontological implications. This big and exciting picture, however, requires many small and technical explications. Unfortunately, theoretical apparatuses for discussing semantic change are extremely impoverished and underdeveloped. In this dissertation I will strengthen the theoretical set-up for the type-token distinction, truth-condition theory of meaning, causal theory of reference, and 2-dimensional modal semantics, so that we have better tools to properly talk about semantic change. I apply these tools to argue for a more nuanced understanding of semantic change, verbal dispute, and analyticity (which are all fundamentally one and the same topic).