The current study used a forced choice pointing paradigm to examine whether English children aged 2 ; 1 can use abstract knowledge of the relationship between word order position and semantic roles to make an active behavioural decision when interpreting active transitive sentences with novel verbs, when the actions are identical in the target and foil video clips. The children pointed significantly above chance with novel verbs but only if the final trial was excluded. With familiar verbs the children pointed consistently above chance. Children aged 2 ; 7 did not show these tiring effects and their performance in the familiar and novel verb conditions was always equivalent.