Brian Garrett has criticized my diagnosis of the paradox of self-consciousness. In reply, I focus on the classification of 'I'-thoughts, and show how the notion of immunity to error through misidentification can be used to characterize 'I'-thoughts, even though an important class of 'I'-thoughts (those whose expression involves what Wittgenstein called the use of 'I' as object) are not themselves immune to error through misidentification. 'I'-thoughts which are susceptible to error through misidentification are dependent upon those which are not. The (...) dependence here has to do with how a thinker understands what would defeat such thoughts. (shrink)