Interesting point! Given the tools we have now to analyze corpus, I bet you could track the birth of memes - especially with an eye towards the question you posed.