Debate on the locus of dysfunction in psychotic speech or thought disorder continues. Research in the area tends to reflect researchers' assumptions about the relationship between language and thought, and these determine the nature of the methodologies used. The circularity of argument produced in this way is inevitable, while fully contextualized comprehensive descriptions of interactions with psychotic patients are lacking. This article outlines a method of analysis which will yield basic descriptions, and produce data readily comparable to naturally occurring conversation. It includes consideration of context, conversation analysis, cohesion analysis, and tone-unit analysis. A crucial feature of the method is serious attention paid to the role of the listener in the creation of misunderstandings in conversations with psychotic speakers.