The first analysis of 10 RAs (RA1–10) revealed that to have a more valid tool for the analysis of cohesive reference, it is necessary to account for the workings of all those lexical items in the text that are easily mistaken for referential items unless otherwise specified. It is not enough if the researcher tries to identify cohesive relationships, because without seeing the full picture, several items may be overlooked, or ambiguity may lead to different analyses. Therefore, after a brief explanation of the method, this chapter will first describe the problem areas encountered in the cohesion analysis of the first 10 RAs, and then it will propose a taxonomy to account for non-cohesive uses of reference items.