We focus in this paper on the potential uses of Ludics, a logical framework based on recent developments in proof theory, for modeling natural language dialogues. The aim is to be able to grasp and to account for some aspects of incoherence in discourses. Among the properties relevant for studying dialogues, two features of Ludics are of interest for our purpose: the duality of viewpoints of two interlocutors as well as the possibility of ruptures. We illustrate first the potential usefulness (...) of such a model for accounting and analyzing incoherence in dialogues due to cultural gap. We then focus on dialogues with schizophrenic individuals to try to answer the following questions: to what extent does incoherence play a role in schizophrenic discourse? And, would a more precise characterization of “non-standard” rationality help explain the range of potential complications of schizophrenia? (shrink)