The purpose of this paper will be to show that the 'merge' model of mental files is to be preferred to the 'link' model of mental files because some of the problems that arise for the 'merge' model also arise for the 'link' model, and because the problems for the 'merge' model can be argued against via counterexamples or explained away. This will be shown by giving a detailed description of what both the link model and the merge model of mental files entail. The flaws of the mental file merge model will be presented, followed by the advantages of the link model, ending with several arguments as to why the merge model is better suited to explain our intuitions about identity than the link model.