The Role of Word Relation in Chinese Character Priming Chih-Wei CHEN 1, 2 and Jei-Tun WU 1, *

1. The Role of Word Relation in Chinese Character Priming Chih-Wei CHEN1, 2 and Jei-Tun WU1, * 1National Taiwan University, Taiwan, 2University of Washington, U.S.A. e-mail: jtwu@ntu.edu.tw ICPEAL 2012 Session Number: 26P-05 • Introduction • Manipulating relations between prime and target characters in Chinese character priming tasks is an important research method to explore Chinese character recognition process. • In the Chinese writing system, three types of relations between prime and target characters should be carefully manipulated or controlled. • Semantic relation: Prime and target characters are semantic related. For example, 叔 (uncle) and 姨 (aunt). • Associative relation: According to association norms, prime and target characters are associated. For example, 禪 (Zen) and 佛 (Buddha). • Word relation: Prime and target characters can form a Chinese word. For example, 綁 (tie, bind) + 匪 (gangster) → 綁匪 (kidnapper). • In previous research, word relation was often confounded with semantic and associative relations. Therefore, the role of word relation in the Chinese character recognition process is not yet clear. • The purpose of the present research is to investigate whether the word relation is involved in automatic character recognition process or is only used strategically as a contextual clue to predict target characters. • Method • Experiments • Experiment 1: Single-character priming paradigm with lexical decision task (LDT). • Experiment 2: Single-character priming paradigm with naming task. • Sub-experiments • Each experiment was divided into three sub-experiments according to three stimulus onset asynchronies (SOAs) which were 50 milliseconds, 150 milliseconds, and 250 milliseconds. • Procedures • Controlled Variables • Semantic and associative relations (including forward and backward association) between prime and target characters were avoided as much as possible when the stimulus material was selected. • Character frequency of prime and target characters. • Neighborhood size of target characters. • Participants: In each experiment, 90 valid participants were assigned randomly to three sub-experiments. • Rationales • If word relation is indeed an important theoretical relation, priming effects should still be found in the word pairs when semantic and associative relations were eliminated. • If word relation is involved in the automatic recognition process, priming effects should be found at short SOAs. Word frequency should also make difference between high and low word frequency groups when word frequency represents natural reading experience. • If word relation is only involved in strategic processes, priming effects should only be found at long SOAs. Neighborhood size of prime characters will probably make difference between large and small neighborhood size groups because it may affect the accuracy of predicting target characters. • Results Experiment 1 Mean correct RTs (ms), error rates (in parenthesis), & priming effects Experiment 2 Mean correct RTs (ms), error rates (in parenthesis), & priming effects • Manipulated Variables • Word relation: A prime character can/cannot form a word with the target character (Word pair vs. non word pair). • Word frequency (WF): High-frequency word vs. low-frequency word. • Neighborhood size of prime characters (NSP): Large neighborhood size vs. small neighborhood size. For example, the prime character 棄 (abandon) in the large neighborhood size group can be the constituent character of 29 different words (e.g. 棄嬰(foundling), 放棄 (give up)). On the other hand, the prime character 綁 (tie, bind) in the small neighborhood size group can only be the constituent character of 13 different words (e.g. 綁匪 (kidnapper)). • Examples • Discussion • Automatic Recognition Process • In the Experiment 1 and 2, the results showed that the word relation yielded priming effects at an SOA of 50 milliseconds. It evidenced that word relation is involved in the automatic character recognition process and it is an important theoretical relation. • Surprisingly, larger priming effects were found in the low word frequency trials at the SOA of 50 milliseconds. It may imply that, in the high word frequency trials, the processing of target character was interfered by the automatic processing of high frequency word. • Strategic Processes • In the Experiment 1,the patterns of priming effects were affected by neighborhood size of prime characters at an SOA of 150 milliseconds. Strategic processes may involve at the circumstance. • In the high word frequency and small neighborhood size trials, participants would have higher accuracy of using word relation as a clue to predict target characters and gain more priming effects. • In the Experiment 2, the patterns of priming effects were quite the same at the SOA OF 50 and 150 milliseconds. It showed that participants did not take the strategy in the naming task. • Two counterbalanced sets of stimulus trials were used in both experiments to avoid repeated characters. Each counterbalanced set contained 40 trials, and another 40 character-non-character trials were added in the experiment 1.