This paper describes disciplinary variation in university students’ writing, as it is reflected in the use of recurrent four-word sequences. In contrast to previous studies, disciplinary categories are not assumed at the outset of the analysis, but rather emerge from an initial analysis of variation across all writers in the corpus. Variation is presented in the form of a visual map representing degrees of similarity and difference between individual writers. Emergent disciplinary groupings are then used as the basis for a qualitative analysis of distinctive lexical bundles. Analysis reveals four main disciplinary groupings. A primary distinction appears between hard (science/technology) and soft (humanities/social sciences) subjects, with two further groupings (life sciences and commerce) being intermediate between these two. Evidence is also found of cross-group disciplines, which draw on a variety of influences, and of particular disciplines which are internally heterogeneous. A qualitative analysis of bundles which are distinctive of ‘hard’ and ‘soft’ disciplines is presented in order to characterize the discourse functions which mark these categories.