Abstract: When bilinguals switch from using one language to another (in laboratory tasks), there is generally a transient cost to performance, namely, the switching cost. Switching cost is the main aspect which is always mentioned in exploring the essence of the language switching mechanism. Such cost might be caused by different mechanisms. Therefore, the researchers focused on where this cost arises from. Two perspectives are involved in this issue: one argues that language switching cost emerges from the word-recognition system itself whereas the other holds the view that it is externally caused, i.e. according to the task control processes. If it is consistent to the externally-caused view, it might reflect that language switching cost is no different from task switching cost in general. In other words, the language switching process is the same as task switching process.
Up to now, most research focused on alphabetic characters and their orthographic specificity, i.e. letters unique to one language. This study focused on ideogram characters such as Mandarin characters and language specificity. Letters with language specificity contain a certain degree of semantic information(analogous to orthographic specificity, such as YiFu in Mandarin characters, i.e. “口” \ “扌”.). The study aimed to explore the mechanism of language switching at the sub-lexical level while Chinese-English bilinguals were performing semantic categorization tasks.
56 undergraduates participated in each experiment. The participants in experiment 1 were asked to judge if an action which was described by a verb presented on a screen was performed with mouth or with hand. In experiment 2, the participants were asked to classify the sex of the person described by a noun word presented as “male” or “female”. There were two kinds of words, either with-marker words (with a semantic radical)or without-marker words (without a semantic radical). All the participants were tested individually on computers, and each session lasted for 15~20 minutes. Reaction time for correct responses and error rates were analyzed by subject and item variance.
The results showed shorter reaction time and lower error rates with the with-marker Mandarin characters, no matter whether the word was a verb or a noun. It indicates that the semantic radical had an influence on the process of stimulus with a marker. However, the switching cost of with-marker words was not significantly different from that of without-marker words. This result indicates that the semantic radical did not affect the process of language switching. For alphabetic characters, the semantic radical also differentially affected the process of with-marker words and without-marker words; however, there is no influence on the switching cost of these two kinds of stimuli.
In short, the study showed that switching cost is not affected by semantic specificity. On a sub-lexical level, the switching cost does not come from the word-recognition system but is caused by external sources. Because language information did not affect the language switching cost, the mechanism of a language switch is the same as that of a task switch in general, both for Mandarin and alphabetic characters.

Key words: sub-lexical level, semantic radical, language switching, switching cost