As Károly (2007) points out, the key to a successful study in discourse analysis lies in the choice of an appropriate number of texts to provide enough data for answering our research questions. The number of exemplars of both genres (RAs and theses), especially of research articles, is far too vast to hope for generalizable results. Instead, what this study can achieve is an exploration of typical cohesive features and typical errors in reference, and for these purposes, two corpora of 20 RAs and MA theses will suffice. The selection criteria will be described in more detail in Chapter 4 (Research design).