The paper argues that corpus linguistics can make useful contributions to the descriptive inventory of literary stylistics. The concept of local textual functions is employed as a descriptive tool for the stylistic analysis of a corpus of texts by Charles Dickens. It is suggested that clusters, i.e. repeated sequences of words, can be interpreted as pointers to local textual functions. The focus is on five-word clusters and five functional groups are identified: Labels, Speech clusters, As If clusters, Body Part clusters and Time and Place clusters. The analysis draws on the identification of key clusters comparing the Dickens corpus with a corpus of nineteenth-century fiction, it identifies links to literary criticism and it gives specific attention to the group of Body Part clusters to illustrate the functional variation of clusters.