This paper explores the semantic-pragmatic intricacies of two related types of interactional humor. The two phenomena under scrutiny, hyper-understanding (Veale et al. 2006) and misunderstanding, are categorized as responsive conversational turns as they connect to a previously made utterance. In the first part of the paper, an analytical model is developed that provides a unified account of both phenomena, using insights from Clark's notion of layering and Fauconnier's mental spaces theory. Hyper-understanding revolves around a speaker's ability to exploit potential weak spots in a previous speaker's utterance by playfully echoing that utterance while simultaneously reversing the initially intended interpretation. Misunderstanding, on the other hand, involves a genuine misinterpretation of a previous utterance by a character in the fictional world. Both cases, however, hinge on the differentiation of viewpoints, yielding a layered discourse representation. The second part of the papers presents a corpus-based study of hyper- and misunderstanding in the staged interactions of the British television series Blackadder. The corpus analysis reveals the spectrum of different pivot elements can serve as a trigger for hyper- and misunderstanding. Common to all instances, it is argued, is a mechanism of figure-ground reversal. (Geert Brône)