Syntactic category of constituent components in the processing of compounds: evidence from noun compounds in Mandarin
Quansheng XIAa, Wenxiao GONG b, Yong LY b, a Nankai Univeristy; b Tianjin Normal University
Introduction: The question of whether nouns and verbs are dissociable in the brain has been heavily investigated. The studies on Mandarin mainly focus on the comparison of disyllabic nouns and verbs. However, the effect of the syntactic category of the constituents on the compound processing has not received much attention. In Mandarin, the syntactic category of whole words is not always the same as that of their components. For example, noun compounds might be composed of noun and noun elements (e.g. __ ji1 chang3, airport), verb and noun elements (e.g. __ fei1 yu2, flying fish), or noun and verb elements (e.g. __ fang2 zu1, house rent). Studies on aphasic patients (Bates et al., 1991; Chen & Bates 1998) and normal people (Hsu et al., 1998) revealed noun-verb dissociation at the sub-lexical level, in which the syntactic information of the constituents had an effect on the processing of the whole word.
However, in the previous studies, the factors of semantic transparency and internal structure information might have not been taken into consideration. In this study, we controlled these two variables. There were three types of noun compounds in the study: words with noun and noun constituents (NN), words with verb and noun constituents (VN), and words with noun and verb constituents (NV). These three types of words were semantically transparent words and were all in the attribute-head structure. We compared the processing of these three types of words to investigate whether the syntactic information at the sub-lexical level influences the lexical processing.
Furthermore, we also explored the roles of attribute and head in word processing. Bloomfield (1931) argued that the word class of compounds depends upon that of the head, which is known as centering theory. In contrast, the information-stress principle proposed by Duanmu (1999, 2007) claimed that attribute is more important than head in the word because attribute carries more information and thus should be stressed. In this study, if NN and VN are processed faster than NV, the centering theory will be supported; if NN and NV are processed faster than VN, the information-stress principle will be supported.
Subjects: Twenty-four native speakers of Mandarin were paid to participate in the experiment (12 male, 14 female; mean age=22). All the participants were right-handed, with normal or corrected-to normal vision and no reported history of neurological illness. None was majoring in linguistics, psychology or any other related discipline.
Materials: The stimuli consisted of three types of nouns. There were 26 NN, 26 VN, and 26 NV. The whole words and their N, V constituents were mostly unambiguous, and were chosen with the criterion that frequency of occurrence in the target word is at least ten times greater than its second most frequent usage (word frequency from Modern Chinese Balanced Corpus, Language Committee of China, 2012). Several variables, such as word frequency, number of strokes, semantic transparency, AoA, familiarity, neighborhood size, and imageability were matched among the three types of compounds (ps > 0.05). The internal structure of all the word types was attribute structure, in which the head was always the second component of the words. Seventy-eight disyllabic words for both the adjective and adverb categories were included as fillers.
Design: Each trial started with a fixation presented in the screen center for 500ms. After the offset of the fixation, the target word was shown on the screen for 1500ms or less if a response was given sooner. The interval between trials was 1500ms and a five-minute break was given between blocks. There were two blocks in the experiment. The participants were instructed to judge whether or not the words were nouns by pressing different keys. The whole experiment lasted 20 mins.
Results and Discussion: The results are shown in Table 1. One way repeated-measures ANOVA of reaction time revealed a significant main effect of syntactic category of constituent (F(2, 46)=27.725, p<0.001). Post-hoc Bonferroni tests showed that NN is accessed faster than VN and NV (ps<0.001), and NV has a shorter response latency than VN (p=0.05). The same analysis on accuracy also indicated a significant main effect of syntactic category of constituent (F(2, 46)=25.421, p<0.001). Post-hoc Bonferroni tests showed that the accuracy of NN is higher than that of the other two compounds (ps<0.001), and the accuracy of NV is slightly higher than that of VN, but such difference did not reach significance (p>0.05).
Table 1 Reaction time (RT) and Accuracy (ACC.)
|
Condition |
RT (ms) |
ACC. (%) |
|
NN |
665±45 |
95.15±8.26 |
|
NV |
729±68 |
84.68±16.95 |
|
VN |
756±91 |
83.8±15.29 |
Note: The results were shown in the form of mean (stand deviation).
In this study, the reaction time suggested a largest facilitation effect for NN among all types of compounds, and a larger facilitation for VN than for NV. These findings, along with the previous studies, indicated that the syntactic category of constituents could influence the processing of compounds in Mandarin. Furthermore, neither the centering theory nor the information-stress principle is fully supported in this study. NN had the fastest reaction time and was erred on least, indicating that the syntactic information of both head and attribute is used in whole word processing. However, the effects of head and attribute on the word class processing of compounds might not be equal. The reaction time indicated a better performance of NV than VN. This suggests that attribute plays a more important role than head in whole word processing, at least in Mandarin. This study calls attention to the importance of the syntactic category of constituents and internal structure information in word processing in Mandarin.