This paper presents a new approach to the study of pragmatic functions in corpora. In contrast to forms, functions cannot be retrieved automatically, which makes function-to-form approaches notoriously difficult. Exploiting the fact that pragmatic functions are not evenly distributed across corpora, but instead tend to co-occur, this study shows that it is possible to retrieve text passages that are particularly suitable and relevant for a detailed qualitative analysis. This method is demonstrated with the example of stance expressions. A set of 20 lexical items that are used to express epistemic and evidential stance were tagged in four Early Modern English corpora. A sample of 300-word passages with a high density of the tagged lexical items was then analysed manually to illustrate the kinds of observations that can be made based on such data. The findings show that this method can lead to new insight into stance marking in Early Modern English. For instance, it can be used to identify previously unstudied stance markers, (con)textual factors that deserve further investigation, and problems for the interpretation of quantitative findings. As such, the method complements existing approaches to the study of stance in corpora.