In this paper I provide an outline of a new kind of constitutive account of self-knowledge. It is argued that in order for the model properly to explain transparency, a further category of propositional attitudes-called "commitments"-has to be countenanced. It is also maintained that constitutive theories can't remain neutral on the issue of the possession of psychological concepts, and a proposal about the possession of the concept of belief is sketched. Finally, it is claimed that in order for a constitutive account properly to explain authority, it has to take a rather dramatic constructivist turn, which makes it suitable as an explanation of self-knowledge only for a limited class of mental states.Crispin Wright 1 has been one of the first theorists to propose a so-called "constitutive" account of self-knowledge. 2 Constitutive accounts are particularly apt to explain knowledge of our own propositional attitudes. For it is a tenet of this kind of models that having knowledge of one's first-order mental states is (at least) a necessary condition for having those mental states. Arguably, however, this would make little sense for sensations and perceptions, as well as for some kind of emotions, which we are generally happy to grant to creatures such as infants and higher-order mammals whom, however, we think lack knowledge of their own mental states.