Imagining being someone else from the inside is something relatively easy to do. In Williams (Imagination and the self, problems of the self: philosophical papers, p 26–45, 1973), for instance, one finds Williams’s famous imaginative scenario consisting in imagining being Napoleon from the inside at the battle of Austerlitz. However, providing an adequate analysis for imagination reports like “(1) Williams imagines being Napoleon (from the inside)” is no easy task, because the logical form of such imagination report is controversial. Following (...) Vendler (Revue de Métaphysique et de Morale 84(2):161–173, 1979), the logical form of statements “X imagines F-ing” typically involve a _PRO_ construction. Furthermore, it is generally acknowledged following Chierchia (Semant Contextual Exp 11:1–31, 1989) that _PRO_ constructions require a _de se_ reading. Consequently, (1) is argued to be an instance of _de se_ imagination (this is the “genuine _de se_” analysis of (1)). Yet, (1) is also crucially about Napoleon and, as forcefully argued for in Williams (Imagination and the self, problems of the self: philosophical papers, p 26–45, 1973), it is not even clear that it is about Williams. So (1) cannot be an instance of _de se_ imagination in the standard sense, because Williams does not self-ascribe the semantic content of the imagining episode (this is the “quasi-_de se_” analysis of (1)). In this paper, I vindicate the genuine _de se_ analysis, based on some new data involving nested imaginings. I then investigate some consequences of the view, which, I argue, are not available to the quasi-_de se_ theorists, including what the view says about failed imaginings. (shrink)