This paper is a new defense of the old orthodox view that self-consciousness requires self-concepts. My defense relies on two crucial constraints. The first is what I call Bermúdez’s Constraint, that is, the view that any attribution of content must account for the intentional behavior of the subject that reflects her own way of understanding the world. The second is the well-known Generality Constraint of Evans, which is also termed the recombinability constraint. The claim I want to support in this (...) paper is the following: Since whether and to what extent we can attribute to non-linguistic creatures and prelinguistic infants genuine knowing self-reference or de se contents is an open empirical question, the proponents of the nonconceptual self-consciousness face a dilemma. If we are convinced that the available empirical evidence is overwhelming, I argue--based on Evans’s Generality Constraint--that these self-representations are nothing but primitive prelinguistic self-concepts. However, if we are convinced that the available empirical evidence is not persuasive, I maintain--relying on Bermudez’s Constraint--that we do better by assuming that the subject is not self-represented. The content of her experiences and thoughts are best modeled as simple selfless propositional functions that are true or false relative to the subject of these experiences and thoughts. I refer to this as self-concernment without self-reference. Thus, against the recent ingenious work of Peacocke, I claim that there is no compelling reason for postulating nonconceptual middle level self-representation, between self-concernment and conceptual self-reference. However, as I hope to make clear, my claim is quite different from those of other recent oppositions to the idea of nonconceptual self-consciousness. According to the thesis of self-concernment without self-reference, the contents of experiences and thoughts are selfless propositional functions, true or false relative to the bearer of the respective mental states. (shrink)