Returning for a moment to Stage 3 of this research, it was mentioned in the discussion of the results that in the process of the analysis we found it difficult to deal with errors in a consistent way. When one deals with student writing, the analysis of reference is encumbered with all sorts of problems: referring items without presupposed items or more referents than necessary; items in intended ties that are supposed to match according to the context but do not, and so on. Ting (2003), who used Halliday and Hasan’s taxonomy, warns that the topic of cohesive errors has not received enough attention in the study of cohesion – as inappropriately used cohesive ties are counter-productive in making a piece of writing cohesive, a qualitative analysis is necessary for any study using this taxonomy to be valid. To achieve this purpose, this stage of our study intends to systematize the analysis of student errors based on corpus data.