The notion of content selection refers to a seemingly simple issue—which parts of a text are question-worthy. Many rule-based and neural question-generation methods presume that a target answer has been preselected in a text. But how do we select a target answer, the focus of the question? In educational settings, a typical reading passage for school reading assignments may consist of a hundred to couple of thousand words. Wikipedia and news articles also contain from tens to a few hundred sentences. Other textual materials, like book chapters or instruction manuals, can be even longer. How should an automated system be set up to select content for questions?