As digital methods allow more and more quantitative approaches even with respect to historical research, corpus studies are gaining immensely in significance. But how to create corpora that allow robust hypotheses and eventually reliable conclusions? How many sources do we need? Ho must they be selected? What mistakes need to be avoided? The selection of sources has a b influence on the research results and needs, therefore, to be reflected. From a historical point of view, the construction of a corpus is not just a starting point but an essential part of the research itself. My talk will address these issues from an historiographical perspective.