Reappraising Poverty of Stimulus Argument: A Corpus Analysis Approach

The debate between empiricism and nativism goes back to the very beginning of philosophy. More recently, the nature of linguistic structure has been the focus of discussion in the field of psycholinguistics. The poverty of stimulus argument for innateness of syntactic knowledge (Chomsky, 1980; Crain & Pietroski, 2001) is one of the most famous and controversial arguments in the study of language and mind. Although it has guided the vast majority of theorizing in linguistics for decades, claims about innate linguistic structure have provoked controversy and the argument is embroiled in dispute. The poverty of stimulus argument is based on the assumption that the information in the environment is not rich enough to allow a human learner to attain adult competence. This assumption has been previously questioned in that it is based on premature conclusions about the information present on primary linguistic input (e.g., Pullum and Scholz; 2002). Nativists have tended to dismiss a priori the idea that distributional information could play an important role in syntactic language acquisition. Nevertheless, recent studies show that distributional evidence is a potentially important source of information for word segmentation and syntactic bootstrapping (e.g., Christiansen, Allen and Seidenberg, 1998; Christiansen and Curtin, 1999; Lewis and Elman, 2001; Redington, Chater and Finch, 1999; Seidenberg and MacDonald, 2001). Moreover, infants are very sensitive to the statistical structure of the input (Saffran, Aslin and Newport, 1996), suggesting that they are able to pick up on distributional cues with a high level of accuracy. This growing body of work has provided support for the hypothesis that distributional properties of linguistic input could have a significant role on the acquisition of syntactic structure. The validity of the poverty of stimulus argument strongly relies on the premise of absence of sufficient information in the primary linguistic input for learning the grammar. A possible approach to address this assumption is to look for statistical evidence that allows a learner to decide between grammatical and ungrammatical hypotheses. In the present study we focus on one of the most used examples to support the argument, the claim concerning auxiliary fronting in polar interrogatives. Using a corpus analysis approach, we estimated the distributional information present in the primary linguistic input, and used it to decide between grammatical and ungrammatical examples of auxiliary fronting in polar interrogatives. We found that there is enough statistical information in the corpus to correctly decide between the grammatical and ungrammatical forms of aux-questions with a high level of accuracy.