Bigram-based learning and the richness of the stimulus for language acquisition

Recent challenges to Chomsky’s poverty of the stimulus thesis for language acquisition suggest that children’s primary data may carry ‘indirect evidence’ about linguistic constructions despite containing no instances of them. Indirect evidence is claimed to suffice for grammar acquisition, without need for innate knowledge or specialized learning mechanisms. We report experiments based on those of Reali & Christiansen (2005), who demonstrated that a simple bigram language model can induce the correct form of auxiliary-inversion in certain complex English questions. We investigate the nature of the indirect evidence that supports this learning, to assess how generally it might be available. Results confirm the original finding but show that the model’s success is highly circumscribed, resting on one particular word pair in the grammatical test sentences. The model performs poorly on inversion in related constructions in English and Dutch, which evidently do not afford effective cues accessible to the learning model. Other more powerful statistical models have so far been shown to succeed only on the same very limited subset of cases as the bigram model, which are clearly not indicative of broader success. The ‘richness of the stimulus’ for auxiliary-inversion thus remains unsubstantiated by this line of research to date.