This experiment was designed to investigate conceptual links directly through a word–picture matching task in children. Participants were asked to indicate between two pictures the one depicting the same concept as the newly learned L2 word (target). One of the two pictures was the target, while the other was either semantically related to it or was unrelated. To investigate whether learning methods modulate L2 word processing, two learning methods were compared: an L2-photograph method and an L1-L2 method. Results showed a semantic interference effect only with the L2-photograph method: Children responded more slowly in the related condition than in the unrelated one. The learning method modulated L2 word processing and L2 word meaning was not necessarily accessed through L1 mediation in the first steps of learning. The method used paves the way for the direct investigation of conceptual links.