Peter Carruther's Interpretive Sensory-Access (ISA) theory of self-knowledge is an interesting account of the opaqueness of our own minds that draws upon a wide range of theories from cognitive science and philosophy. In the present paper, I argue that the theory's assumptions support the conclusion that the available perceptual evidence massively underdetermines all of an agent's second-order beliefs about her own beliefs. Such a result is far more negative than the ISA's well-known pessimism regarding self-knowledge. Furthermore, I also argue that, from the same assumptions, it is possible to build an argument to the effect that cognitive scientists trying to determine an agents' true behavior-causing attitude face similar underdetermination problems. Toward the end of the paper, I suggest that the theory's problems arise from a conflation of two different ways in which terms denoting propositional attitudes, such as 'belief', are used in its formulation. Distinguishing between the two usages of these terms, in turn, leads to a further distinction between two different senses in which we can talk about the 'opaqueness' of our own minds.