Here is my second answer to a question posed to me by a reader of this blog about how to begin a dissertation.
You need to know disciplinary requirements and expectations. For example, a lot of us 20 years ago had dissertations on a single author, where I notice now that that's become more and more rare. It's a dumb prejudice, but a lot of people now think you are too narrow if you work on a single author, even an important one. We have a lot of dissertations here with one author from Brazil and the others from Spanish speaking countries, or three chapters on novels and one on film.
So several authors, from the same region or overlapping areas. There should be both temporal and geographic parameters, but it shouldn't be a topic that strikes the observer as esoterically narrow. Think of having to explain the topic briefly in an interview; if it isn't a topic that you can foresee explaining orally it is going to be harder to conceptualize.
The topic has to name a critical problem, with a how and a why, not just a what or who.
People often don't think to look at existing dissertations in the library of their institution. Usually, all the previous dissertations in your university are going to be housed in the library. Look at disses in the last 5 years in your own department. Who has directed them? Which ones look most interesting? Which of those students is now having a successful career? Which dissertations became books quickly?