Computational models trained on realistic child-directed corpora successfully acquire structure-dependent rules without innate grammatical constraints, as Pullum and Scholz demonstrate.
?Rate how convincing each reason is below to see the overall strength.