84One of the most obvious blind spots in corpus approaches to discourse studies is the fact that communication is multimodal. Language-only corpora will tell us some of the story, but they will not tell us the whole story, as confirmed by van Leeuwen (2011: 668), who states that ‘many forms of contemporary written language cannot be adequately understood unless we look, not just at language, but also at images, layout, typography and colour’. Here van Leeuwen is talking about examining the distinct potential for meaning of each semiotic mode (image, gesture, speech, architecture) involved in a communicative artefact or event, or, as Kress (2010: 28) calls it, a modal ensemble. This is a complex task, and one that involves multiple methods (e.g. corpus linguistics plus multimodal discourse analysis) and careful sampling techniques. It also requires agility: in finding researchers with the necessary expertise for collaboration; compatibility and communication across various software for analysis; sufficient time to conduct analyses; and space (in publications) to be able to explicate the complexity of the findings. These are points already noted by other scholars, for example Egbert and Baker (2016: 204) in relation to triangulation, and Adolphs and Carter (2013: 178), who note that size remains a particular limitation of multimodal corpora given the time and effort involved in aligning different streams of data.