Children solve explicit false belief (FB) tasks around age 4 but implicit tasks (e.g. helping) in infancy (e.g. Buttelmann et al., 2009). Nativist accounts claim early conceptual competence, masked by performance factors (e.g. Leslie, 2005); sceptical accounts deny competence before age 4 (Perner & Ruffmann, 2005). A recent two-system-theory (Apperly & Butterfill, 2009) provides an alternative explanation: an early mindreading-system (1) tracks simple forms of mental states and a later flexible capacity (2) allows cognitively demanding inferences based on a fully-developed concept of belief. Because system 1 operates on relational rather than propositional attitudes it has clear flexibility-limits: it can represent FB’s about object location but not FB’s about identity. We contrasted 2.5-year-olds’ helping behavior in a 2 (identity/location) X 2 (false/true belief) design. Results suggest limited performance in the identity compared to the location task. Implications of this finding are discussed and follow-up studies will be presented.