The Acquisition of Chinese Characters: Corpus Analyses and Connectionist Simulations

Abstract

Empirical evidence has accumulated that regularity and consistency in orthography-to-phonology correspondence both affect the processing and acquisition of Chinese characters, and that they interact with character frequency in complex ways. However, despite growing interests in the Chinese orthography, few previous studies have systematically analyzed these effects on a large scale, or have modeled the acquisition process using connectionist networks. In this study, we set out to analyze a realistic character corpus as learned by children in Chinese elementary school, in order to evaluate the degree to which corpus analysis can inform us of the roles that regularity, consistency, frequency, and their interactions play in childrens acquisition of Chinese characters. We further modeled character acquisition in a selforganizing connectionist network, developing orthographic representations on the basis of our analysis of character properties in a large-scale character corpus. Our model is able to faithfully capture the orthographic similarities of Chinese characters, and moreover, display effects of regularity, consistency, character frequency, and the complex interactions among them, matching up well with available empirical evidence on childrens acquisition of characters during the elementary school years.