Taking Mandarin Possessive Construction (MPC) as an example, the present study investigates the relation between lexicon and constructional schemas in a quantitative corpus linguistic approach. We argue that the wide use of raw frequency distribution in traditional corpus linguistic studies may undermine the validity of the results and reduce the possibility for interdisciplinary communication. Furthermore, several methodological issues in traditional corpus linguistics are discussed. To mitigate the impact of these issues, we utilize phylogenic hierarchical clustering to identify semantic classes of the possessor NPs, thereby reducing the subjectivity in categorization that most traditional corpus
linguistic studies suffer from. It is hoped that our rigorous endeavor in
methodology may have far-reaching implications for theory in usage-based
approaches to language and cognition.