In this paper, I present a conceptual connection between fictional disagreements and thought experiments. Fictional disagreements happen when two readers disagree about a fictional detail. The “great beetle debate” is a paradigmatic case. Nabokov once argued that Gregor Samsa, in The Metamorphosis, metamorphosed into a beetle. Yet many critics and readers imagine Gregor to be a big cockroach. Analysing a fictional disagreement is interesting because it exhibits the informational structure which is common to all fictions. First, it shows the distinction between the fictional foreground (what is expressed by the narrator) and background (what the reader automatically infers from the narration). Second, it shows how the fictional background is filled with the reader’s representations of reality and other shared conventional representations. The fictional background is a sophisticated mixture of traceable fictional and non-fictional bits of information. I argue that one can use this complex informational structure to explain how it is possible to extract new information originating in fiction for non-fictional purposes. The possibility of “learning from fiction” has led to a long-standing philosophical debate. However, everyone agrees on the possibility of extracting fictional information: this corresponds to drawing a moral from a given fiction. This possibility is, I argue, analogous to performing a thought experiment. I show that thought experiments and fictional disagreements exploit the same informational structure. Instead of filling the fictional background, one informs one’s non-fictional representations using the same informational channels in reverse direction.