Spatial specification in Chinese character recognition: The role of structural configuration and radical similarity

Abstract

Introduction: About 34% of 4-letter words in English and French can form other words by rearranging their letters. Therefore, spatial specification of letter position is vital to correct word recognition and production (see Grainger & Van Heuven, 2003). The orthographic form of Chinese characters differs dramatically from alphabetic scripts as character components (radicals) are arranged in a square shape of constant size. Whether spatial information of these components is necessary in the Chinese lexicon can also be asked given the existence of transposable characters, e.g. 呆 and 杏 (Taft, Zhu, & Peng, 1999). Hence, how characters are structurally represented has led to at least two proposals. The Multilevel Interactive- Activation model (Taft & Zhu, 1997) postulates that radicals are specified for position of occurrence within a character (e.g. 嫁 constitutes of a left radical 女 and right radical 家). In contrast, the Lexical Constituency model (Perfetti, Liu, & Tan, 2005) assumes a role of structural configuration in characters. This refers to the relative spatial relationship among radicals in a character and is conceptualized as ‘spatial slots’ that radicals occupy (e.g. left-right: 嫁; top-bottom: 爸; enclosing: 鬧or困). While the former finds evidence from facilitative priming effects on lexical decision latency when prime and target share the same radical in the same position, unambiguous support for the latter is lacking. Methods. Thirty-three right-handed Cantonese speakers performed a lexical decision task where each trial began with a prime (48ms) sandwiched between a forward (100ms) and backward (16ms) mask, followed by the target character. Targets were 30 low frequency (<80 per million, Leung & Lau, 2011) characters with left-right configuration and 30 filler pseudo-characters. Primes were pseudo-characters related to corresponding targets by Radical Similarity (same vs. different radical) and Configuration (same vs. different), giving four prime types: (1) same configuration-same radical, (2) different configuration same radical, (3) same configuration-different radical, and (4) different configuration-different radical. Same radical primes never shared the same position to their targets. All pseudocharacters were stroke matched to their corresponding real word targets. Electrophysiological data was collected simultaneously with the behavioral reaction time and accuracy data. Results and Discussion. Behavioral results showed facilitative effects for shared radicals whereby participants were faster to identify targets that shared the same radical as the prime. A radical-byconfiguration interaction was also found where radical similarity facilitation was observed when the prime and target shared the same configuration. ERP results reflected a greater contribution of configuration as significant effects of configuration were found at the N1 component (80-140ms) for all electrodes in the occipital region. Targets sharing the same configuration as their primes elicited a more negative N1 than ones with different configuration. In contrast, the radical similarity effect was only observed in the left occipital electrode, where greater negativity was found for targets that shared the same radical as their primes (Figure 1). The findings suggest that configuration plays an early role in Chinese character recognition and challenges models that do not assume such features in the orthographic lexicon.