AbstractTo enable a clear and detailed description of narrative features, narratological concepts should be well-defined and unambiguous. My paper aims to contribute to this by analyzing, reworking, and relabeling two models of ›unreliable interpreting‹. I show how these two models developed by Phelan/Martin and Margolin are based on relevant intuitions yet have problematic aspects. Building upon these approaches, it is reasonable to integrate two concepts into our theory of unreliable narration: that of ›unreliable belief‹, which, like unreliable reporting, concerns the realm of facts but affects the narrator’s cognition instead of his utterances, and that of ›inference-related unreliability‹, which is concerned with the validity of the narrator’s reasoning on both the utterance and cognitive level. By integrating these two concepts, this approach makes the theory of unreliability more robust and helps provide better answers to common questions in the field of unreliable narration.