Observations on a test of transitive inference in children