The Coding Flexibility of Radical Position in Chinese
Abstract
Although there are theories of word recognition/reading predicting a transposed letter (TL)-type effect in Chinese character recognition, specifically, a transposed radical (TR) effect, no empirical demonstrations of TR effects have been reported to this point. As a result, instead of adopting a position-general assumption of radical processing, a position-specific assumption of radical processing has been adopted in most Chinese character recognition models. In the present Experiment 1, computational models were created in order to determine whether models that do not make the position-general assumption can account for any TR effect if one were to be found. In Experiment 2, 3, and 4, the masked priming technique was used to investigate whether there are TR priming effects in Chinese character recognition, as well as whether such effects would interact with radical type (character radical vs non-character radical) or character structure (left-right vs top-bottom). Experiment 2 used the lexical decision task (LDT) and Experiment 3 and 4 used the presumably more orthographically oriented same-different matching task (SDMT). Event-related potentials (ERPs) were analysed in Experiment 2 and 3 in order to better examine the time course of any effects. The results of Experiment 1 showed that computational models can produce TL-type effects, but only if trained with orthographic inputs that follow the position-general view of radical processing. Experiments 2 and 3 (using left-right structure characters) showed that TR priming effects do emerge in LDT and SDMT tasks. In the SDMT task, the ERP data showed that the TR effect emerged in a slightly later time window than the repetition priming effect, implying that the TR effect was produced at the orthographic processing stage rather than at an earlier feature processing stage or a later semantic processing stage. In Experiment 4, the TR effect was found to have a different pattern for top-bottom structure characters than for the left-right structure characters of Experiments 2 and 3, indicating an impact of character structure. These results provide support for the position-general view of radical representation, the noisy position coding models of orthographic processing, and the idea that position flexibility of orthographic representations and their processing are shaped by the language environment of the reader.