The "file drawer problem" is common one. Suppose you have a null hypothesis. You are unlikely to get research published which merely confirms a null hypothesis. Suppose you want to prove that seeing television violence makes little kids more violent, or more insensitive to violence. A researcher might have done 8 experiments, but only one with a result debunking the null hypothesis. That study is published, and the results from the other seven go in the file drawer. So the desired result shows a significant correlation, but the other data is never even seen.
In the humanities this is not a problem, because the data is always cherry-picked. So suppose I want to show that Lorca had a significant impact on US drama. The null hypothesis is that he didn't. I can prove this false by accumulating enough evidence, enough isolated examples, to make an interesting 20 page article. I can show that at least five award-winning dramatists have acknowledged Lorca in very direct ways: Baraka, Kennedy, Albee, Shepard, T. Williams, N. Cruz. The fact that someone has won some awards is rather trivial, right? But that is a measure of recognition that is outside of my subjective judgment. It is someone else, not just me, saying that these are playwrights of note.
It is up to me not just to count examples, but to construct an argument, finding patterns. Once again, I can ignore all the Arthur Millers who have not been influenced by Lorca. Sometimes I feel I am doing some special pleading, but in the end I have to be honest with myself too, and figure out whether I believe my own argument, or how strongly I want to push it. When I began I wanted to show how Lorca was not influential in the American theater, and why. It would be an interesting article, and unusual in arguing for a null hypothesis. As I learned more, however, I discovered I was wrong. There was more to the story than I had suspected. This article is not for the file drawer.