Analyzing a thesis for mistakes or errors is clearly not straightforward. Of course, it is partly a simple accuracy or well-formedness check on the level of sentences. However, there is much more to it: analyzing referential ties and chains involves building up a mental image of the whole text, which means keeping track of the referential ties and chains that are present in the text. Sometimes there is an obvious break in the chain, which is very easy to notice. At other times, there might be several interpretations of a particular item. Sometimes errors go unnoticed until the analysis of reference is carried out, because readers tend to fill in missing information with their own ideas – often unconsciously and not always correctly – and do not notice where exactly something went wrong in the text. An analysis of reference cannot be carried out by a computer because it needs text interpretation using information involving background knowledge and experience. One must also notice relationships between sets of items or the likelihood of co-occurrence patterns. This also means that the reliability analysis of references largely depends on the reader, as does the interpretation of some of the mistakes. Besides clear cases, there are ambiguities that need to be resolved one way or another, and this resolution may differ person-by-person, because all readers bring different experience and background knowledge to the understanding of a text – as readers, we construe cohesion and coherence differently.

 

The categories of learner errors that emerged from the sentence-by-sentence cohesive and non-cohesive reference analysis of the MA thesis corpus are summarized in Figure 17 below. The labels represent types of errors for which at least 20 instances were found in the corpus of 20 MA theses (frequencies are shown in Figure 18).

 

 

The following four subsections explain the concepts covered by the labels in Figure 17 and illustrate the types of errors with examples from the MA thesis corpus.