Abstract Heterodiegetic narrators are not present in the story they tell. That is how Gérard Genette has defined heterodiegesis. But this definition of heterodiegesis leaves open what ›absence‹ of the narrator really means: If a friend of the protagonist tells the story but does not appear in it, is he therefore heterodiegetic? Or if a narrator tells something that happened before his lifetime, is he therefore heterodiegetic? These open questions reveal the vagueness of Genette’s definition. However, Simone Elisabeth Lang has recently made a clearer proposal to define heterodiegesis. She argues that narrators should be called heterodiegetic only if they are fundamentally distinguished from the ontological status of the fictional characters: Heterodiegetic narrators are not part of the story for logical reasons, because they are presented as inventors of the story. This is, for example, the case in Johann Wolfgang Goethe’s novel Elective Affinities (1809): In the beginning of this novel the narrator presents himself as inventor of the character’s names (»Edward – so we shall call a wealthy nobleman in the prime of life – had been spending several hours of a fine April morning in his nursery-garden«). Based on that recent definition of heterodiegesis my article deals with the question whether such heterodiegetic narrators can be unreliable. My question is: How could you indicate that the inventor of a fictitious story tells something which is not correct or incomplete? In answering this question, I refer to some proposals of Janina Jacke’s article in this journal. Jacke shows that the distinction between homodiegetic and heterodiegetic narrators should not be confused with the distinction between personal and non-personal narrators or with the distinction between restricted and all-knowing narrators. If you make such differentiations, then of course heterodiegetic narrators can be unreliable: They can omit some essential information or interpret the story inappropriately. Heterodiegetic narrators of an invented story can even lie to the reader or deceive themselves about some elements of the invention. That means: A heterodiegetic narration cannot only be value-related unreliable (›discordant narration‹), but also fact-related unreliable. My article delves especially into this type of unreliability and shows that heterodiegetic narrators of a fictitious story can be fact-related unreliable, if they tell something which was not invented by themselves. In that case, the narrator himself sometimes does not really know whether he tells a true or a fictitious story. Such narrators are unreliable if they assert that the story is true, although they are suggesting at the same time that it is not. I call this type of unreliable narrator a ›fabulating chronicler‹ (›fabulierender Chronist‹): On the one hand, such narrators present themselves as chroniclers of historical facts but, on the other hand, they seem to be fabulists who tell a fairy tale. This type of unreliability occurs especially if a narrator tells a legend or a story from the Bible. My article demonstrates this case in detail with two examples, namely two novels by Thomas Mann: The Holy Sinner (1951) and Joseph and His Brothers (1933–1943). My article also discusses some cases where it is not appropriate or counter-intuitive to call a heterodiegetic narrator ›unreliable‹: i. e. the narrator of Thomas Mann’s novel The Magic Mountain (1924) and the narrator of Johann Wolfgang Goethe’s novel Wilhelm Meister’s Apprenticeship (1795/1796). On the one hand, these narrators show some characteristics of unreliability, because they omit essential pieces of information. On the other hand, these narrators are barely shaped as characters, they are nearly non-personal. However, in order to describe a narrator as unreliable, it is – in my opinion – indispensable to refer to some traces of a narrative personality: Figural traits of a narrator provoke the reader to identify all depicting, describing and commenting sentences of a narration as utterances of one and the same ›psychic system‹ (Niklas Luhmann). Only narrators who can be interpreted as such a ›psychic system‹, provoke the reader to assume the role of an analyst or ›detective‹, who perhaps identifies the narrator’s discordance or unreliability. In my article the unreliability of a narration is understood as part of the composition and meaning of a literary work. I argue that a narrator cannot be described as unreliable without designating a semantic motivation for this composition by an act of interpretation. Therefore, my suggestion is that a narration should be merely called unreliable if it encourages the reader not only to imagine the told story, but also to imagine a discordant or unreliable storyteller.