The paper shows that certain well established arguments widely used in the philosophy of language to discredit some semantic theories are, in fact, defective (from the methodological viewpoint). In particular, the direct reference theory is usually rejected, because (i) it is impossible to substitute co-referential proper names for each other in epistemic contexts and (ii) true identity sentences of the form “a = b”, where a and b are proper names, have to be both necessary and a posteriori. Both arguments are based on what a competent user of a given language is willing to admit or refuse to admit. It is shown that the language user invoked in these arguments has to be perfect in a certain sense. Otherwise, the arguments are inconclusive. However, if perfect speakers enter the picture, these arguments become theory laden in a certain way and thus fail to be neutral means which can be used to distinguish correct theories from incorrect ones. In this sense, they cannot be used to discredit neither the direct reference theory, nor any other rival theory.