The aim of this paper is to show some of the limitations of the Theory of Mind approach to interaction compared to a conversation analytic alternative. In the former, mental state terms are examined as words that signify internal referents. This study examines children’s uses of ‘I want’ in situ. The data are taken from a corpus of family mealtimes. ‘I want’ constructions are shown to be interactionally occasioned. The analysis suggests that (a) a referential view of language does not (...) adequately account for how mental state terms are used in talk, (b) the dominant methodology for examining children’s understanding of ‘desires’ is based on several problematic assumptions. It is concluded that participation in interaction is a social matter, a consideration that is obscured by Theory of Mind and its favoured methods. (shrink)