Studies with infants show divergence between performance on theory of mind tasks depending on whether "direct" or "indirect" measures are used. It has been suggested that direct measures assess a flexible but cognitively demanding ability to reason about the minds of others, whereas indirect measures assess distinct processes which afford more efficient but less flexible theory of mind abilities (Apperly & Butterfill, 2009). This leads to the prediction that performance on indirect measures should be subject to "signature limits." The current study tested whether the Level-1/Level-2 distinction might constitute one such limit. The study adapted a task that has shown evidence of Level-1 perspective-taking on both direct and indirect measures (Samson, Apperly, Braithwaite, Andrews, & Bodley-Scott, 2010). The aim was to test Level-2 perspective-taking in a sample of 6- to 11-year-olds (N = 80) and adults (N = 20). Participants were able to make Level-2 judgements on the direct measure. In contrast with the findings from Level-1 perspective-taking, there was no evidence of automatic processing of Level-2 perspectives on the indirect measure. This finding is consistent with the view that theory of mind abilities assessed by indirect measures are subject to signature limits. The Level-1/Level-2 distinction, suitably refined, marks one way in which efficient but inflexible theory of mind abilities are limited. (Contains 2 figures.)