There are many stories about Sherlock Holmes. Or are there? There never was such a person as Sherlock Holmes---that is part of what makes those stories fiction. On the other hand, of course the fictional character Holmes exists, because Arthur Conan Doyle created him. I explore three problems related to this.

First, how are fictional objects created? John Searle, Gregory Currie, and a few other theorists sketch accounts in terms of fiction-makers' intentional actions or general propositions that are true in the relevant works. These fail for many kinds of fiction, because fictional objects need not be created intentionally, and need not correspond uniquely to such general propositions. A general account of creation must appeal to the function of fiction, namely to guide our imaginative activity. Kendall Walton uses that to sketch a better answer: works generate fictional objects by guiding us to imaginatively simulate de re mental representation. I develop this in terms of the psychology of mental simulation and de re representation, and argue it is superior to other accounts.

Second, do such works that generate fictional objects also represent them? In general, do representations apparently about fictional objects really represent them? Surprisingly, some theorists accept fictional objects but deny they are represented in many cases. I reconstruct and criticize their arguments, concluding that they have not yet found an obstacle to such representation.

Third, Peter Geach raises a puzzle: analyze the sentence 'Hob thinks a witch has blighted Bob's mare, and Nob wonders whether she (the same witch) killed Cob's sow' as true even if there are no witches. Dozens of philosophers and linguists see this as a problem in the semantics of anaphora. But based on careful exegesis, I argue that Geach misrepresents the puzzle: the sentence cannot be true if there is no relevant witch, and so the puzzle is to account for its seeming truth without witches. I thus reduce it to the problem of apparent representation of fictional (or mythical) objects. This rules out nearly every published response, but leaves open a few possible paths, which I sketch.