The paper considers cohesion of texts and discourse in the perspective of corpus linguistics. Being one of the main features of texts, it has a sophisticated character and is a challenge for corpus annotating, both automatic and manual. The research deals with cohesion of interrogatory discourse and is aimed at setting up problems for its special tagging. The authors of the paper have designed a probable variant of an algorithm to analyze cohesion in interrogatory contexts, specified three main tasks correlating with key stages and including subordinate tasks. A set of indices to be applied in tagging has been worked out as well. The authors illustrate the algorithm with examples of interrogatory contexts from the novel by Graham Swift “Waterland” which is characterized by high frequency of interrogatory utterances, on the one hand, and a highly dense cohesion web, on the other. It is explained how to identify limits of interrogatory contexts in order to formalize their analysis. The classification of indices for tagging lingual means of cohesion and its kinds is provided. The interrogatory utterance is analyzed as a means of cohesion itself as well. Parallel, consequential and adjoining kinds of cohesion are illustrated, accompanied with a system of indices for their tagging.