A new approach to the measurement of mentalizing is introduced. Instead of measuring the presence or absence of mentalizing capacity, the current study aimed at developing a mentalizing task that focuses on investigating biases in mentalizing through the use of ambiguous peer-related social scenarios. The response consistency of 659 children was investigated in a community sample of children aged seven to 11. Confirmatory latent class analysis allocated children to three groups on the basis of their responses: an overly positive style (OP), a rational/neutral style (R) or an overly negative style (ON). Children classed as OP showed a greater likelihood of being above cut-off on a population screen for externalizing disorder. Over a two-year follow-up period, the children who were classified as R were most likely to remain so whereas the OP and ON children were found to change group. The results are discussed in the context of social-cognitive research in this age group. Further studies should examine the external validity of the mentalizing task because the results reported here concern only internal construct validity.