In this paper, I combine pragmatic theorizing with empirical interaction analysis to analyze conversational humor. By taking dialogic structures and the level of performance (utterance formulation, conversational sequencing and emergent genre construction) into consideration, I also hope to offer insights for cognitive linguistics. One key to performance analysis is that utterances always provide information, often called contextualization or framing procedures, on how they should be understood. Pragmatic modeling should take framing procedures and the reconstruction of various forms of knowledge invoked in normal, everyday communication into account to explain inferencing in humorous interaction. In most humorous activities, non-standardized inferencing forms the core of the humorous potential. It arises either from a sort of script opposition (Attardo 1994) or from playing with formulation standards and expected ways of speaking within conversational sequencing. To frustrate expectations is to invite unusual associations. I distinguish between comical interaction modality (keying and framing) and punch-line humor. Both invite listeners to make complex inferences. Besides this, they work with allusions. The listener must draw on life-world knowledge in order to understand the allusions that are central to both sorts of humor. I favor retaining the Gricean cooperation principle and its maxims for the analysis of humor and try to link the Gricean pragmatics of cooperation to performance analysis (framing, keying, layering, contextualization). Clark's (2004) plea to take into account the distinction between primary and collateral signals is highly suggestive of possible ways to grasp the functioning of emergent humorous interaction. In this paper, I apply pragmatic theories to two examples of emergent conversational humor not belonging to well-known conventional humor genres and consequently hard to classify using conventional folk taxonomies. Using these examples, I show what a pragmatics of performance can and should contribute to explaining how interlocutors co-construct humor and process various layers of meaning. Only when combined with a performance analysis that takes into account the layering of meaning and social and cultural knowledge can pragmatics explain the joint production of humor.