Is I guaranteed to refer?

One claim about I, regularly made and almost universally endorsed, is that uses of the term are logically guaranteed to refer successfully (if they refer at all). The claim is only rarely formulated perspicuously or argued for. Such obscurity helps disguise the fact that those who profess to advance the claim actually turn out to support not a logical guarantee at all but merely high security through fortunate coincidence. This is not surprising. For we have no good reason to accept the claim – granted, any use of I is apt to refer successfully; but that can be explained by pragmatic features of its use. And we have some reason to reject the claim – it is notoriously difficult to see how genuine reference and guaranteed success do not exclude each other when considered as properties of the logic of any term