A Trap in Corpus Linguistics: The Gap between Corpus-based Analysis and Intuition-based Analysis

Abstract

The main concern of this paper will be to demonstrate corpus-based analysis and informants’ intuition as native speakers being mutually essential for a thorough understanding of authentic language use. Corpus-based analysis has been utilised by many linguists to determine appropriate word or phrase usage. However, in such analyses, the statistically imperfect properties of the corpora in question were not considered. Thus, the possibility that the corpora indeed may not appropriately represent language use was ignored. On the other hand, other linguists advocate the validity of informants’ intuition as native speakers, not paying sufficient heed to the deviation of linguistic intuition from actual language use. To clarify the significance of the collaboration of corpus-derived information and informants’ intuition, it will be shown by thorough investigation of ‘NP1 + promise + NP2 + to-clause,’ ‘ill + personal noun,’ and ‘just now’ with the present perfect aspect that the results of corpus-based quantitative analysis and intuition-based qualitative analysis can differ significantly. The matter of corpus representativeness will then be discussed, dealing with the frequency distributions of ‘subject-to-subject raising and extraposed that-clauses controlled by likely,’ ‘major modal auxiliaries’ (will, would, can, could, may, might, should and must) and three synonymous adjectives (big, large and great). Finally an investigation of the use of ‘lest …(should)… ’and ‘introductory which’ will bring to light the deviation of linguistic intuition from actual language use.