語音轉碼對臺灣大學生閱讀歷程的影響：來自眼動實驗的證據

The Effects of Phonological Recoding on Taiwanese Students' Reading Processes: Evidence from Eye-Tracking Experiments

官英華(Ying-Hua Guan)

《教育心理學報》 52卷2期 (2020/12) Pp. 459-488

https://doi.org/10.6251/BEP.202012_52(2).0010

摘要

語音在中文閱讀中的角色長久以來是心理語言學界爭論的話題。雖然一些證據顯示中文母語者閱讀中文時，語音可以在字詞辨識早期被激活，語音在字詞辨識和閱讀理解中的角色仍有爭議。本研究進行了兩個眼動實驗，探討是否中文成人的讀者需要依賴語音轉碼來理解短篇論說文，以及是否當讀者沒有與特定領域文章內容相關的背景知識時，更需要依賴語音轉碼來理解文意。實驗一請30位大學生在5個閱讀條件（語音抑制、逐字朗讀、聽不相干語音、聽文章語音以及安靜閱讀）下閱讀20篇短篇論說文。受試者在每篇文章後的閱讀測驗分數以及他們的眼動行為都被記錄下來，之後針對每個依變項進行數個（廣義）線性混合模型分析，以找出最佳配適度的模型。結果顯示語音抑制明顯的降低閱讀理解且造成較大的回視率和目標詞的重讀率，表示語音轉碼在中文閱讀理解和字詞辨識的重要性。實驗二沿用同樣的實驗法另請30位大學生在上述5個閱讀條件下閱讀15篇物理學專業的短文，其中15位主修人文社會科學，15位主修自然科學。實驗結果顯示只有背景知識對閱讀理解有顯著效果，但閱讀條件以及它與背景知識的交互作用都不顯著。背景知識無法調節受試者閱讀理解時對語音轉碼的依賴程度。整體來說，語音轉碼對特定知識領域短文的理解並不那麼重要。語音抑制並未顯著降低物理學專業短文的閱讀理解表現。逐字朗讀會使閱讀變慢並降低回視率，但不能提升閱讀理解。聽文章語音會增加閱讀短篇論說文中目標詞的重讀率，也會增加閱讀物理學專業短文的閱讀時間和平均凝視時間。聽不相干語音不影響閱讀理解表現，只增加物理學專有名詞的重讀率以及延長沒有背景知識受試者的首次凝視時間。

關鍵字

中文閱讀 ； 眼動實驗 ； 語音轉碼

並列摘要

Phonological recoding is a psychological process in which readers convert printed words into sounds (i.e., phonological codes) during silent reading. The role of phonological recoding in reading has long been debated in the field of psycholinguistics. Most relevant studies have mainly focused on examining the role of phonological recoding in word recognition. Three major theoretical approaches have been proposed for lexical access. First, the phonological mediation model assumes that the meaning of a word is accessed through its phonology; that is, the sequence of the word recognition process is orthography phonology semantics. Phonological recoding of a word occurs prior to the access of its meaning (Lukatela & Turvey, 1994; Rayner, Pollatsek, & Binder, 1998; Van Orden, Johnston, & Hale, 1988). Second, the direct access route model assumes that the meaning of a word can be directly accessed through the transition from orthography to semantics without the mediation of phonology (M. Coltheart & Coltheart, 1997; Taft & Van Graan, 1998). The model, however, does not deny the possibility of phonological activation. Rather, it assumes that phonological recoding occurs concurrently to or later than the access of the word's meaning. Third, the dual route model suggests that two parallel running routes can achieve lexical access: one is the phonologically mediated route, and the other is the direct access route (M. Coltheart, 2000). The route that achieves lexical access depends on the route that runs the fastest (Barron, 1986; Paap, Noel, & Johansen, 1992). McCusker, Hillinger, and Bias (1981) indicated that the difficulty of material, subjects' reading fluency, the frequency of occurrence of items, and task demands can determine the use of phonological recoding. Chinese is a logographic language and is also regarded as a phonologically opaque one. Nevertheless, studies investigating the time course of phonological activation in Chinese word recognition have yielded inconsistent results. Although some evidence has suggested that phonology can be activated very early in the word recognition process when Chinese readers read Chinese words or sentences (Guo, Peng, & Liu, 2005; Tan, Hoosain, & Peng, 1995; Zhang & Perfetti, 1993), it remains controversial whether phonological recoding plays a crucial role in word recognition and text comprehension. Few studies have measured Chinese subjects' reading comprehension under the condition of articulatory suppression, which has been proven to interfere with phonological recoding (Lu & Zhang, 2007; Ding & Wang, 2006). However, the results of such studies have been inconsistent, and whether phonological recoding is important for Chinese text comprehension remains unclear. This study conducted two eye-tracking experiments by using the same experimental paradigm implemented by Guan (2015, 2020) to investigate whether adult Chinese readers need to rely on phonological recoding to understand short expository texts and also whether readers without the background knowledge relevant to domain-specific texts must rely more on phonological recoding to comprehend them. In Experiment 1, 30 university students in Taipei were recruited whose mother tongue was Mandarin Chinese. They were asked to read 20 short expository texts (each one was 350 characters in length) under five different reading conditions: articulatory suppression, reading aloud, concurrent reading and listening to an irrelevant text (IRS), reading while listening to the same text (RWL), and silent reading. The topics of the texts included, for example, ＂advice for new graduates when entering workforce,＂ ＂the impact of 3C products on human cognition,＂ ＂water pollution,＂ ＂the color of eggs,＂ ＂the crisis of water resources,＂ ＂the function of sun light,＂ ＂the philosopher and the frog,＂ ＂Mr. thought,＂ ＂experiment with monkeys,＂ ＂the stereotypes of teachers in Taiwan,＂ and ＂the influences of watching TV on small children.＂ Subjects' reading time and comprehension score for each text as well as their eye movements were recorded, which were then analyzed by several linear or generalized linear mixed models to find the best fitting model for each dependent variable. The results revealed that articulatory suppression significantly impaired reading comprehension and caused a larger regression ratio in text reading as well as a higher rereading rate on target words, whereas the effects of the other reading conditions on comprehension did not differ from each other. These results suggest that phonological recoding is essential for Chinese expository text comprehension and word recognition. Reading aloud, compared with silent reading, significantly increased the overall reading time, mean fixation duration, and number of fixations in a text but substantially reduced the mean saccade length and regression ratio. In word-level processing, reading aloud led to a much longer first fixation duration, gaze duration, rereading time, and total reading time on target words selected in the texts. Through word-by-word reading, reading aloud seemed to facilitate integrated sentence meaning, but it failed to significantly enhance overall reading comprehension. Moreover, RWL caused higher rereading rates on target words than did silent reading, and it did not improve reading comprehension. In line with results found under the reading aloud condition, the null effects of reading aloud and RWL on reading comprehension could indicate that both reading conditions make the covert phonological recoding during silent reading overt and therefore imposed similar effects on reading comprehension to those exerted by the phonological recoding in silent reading. Experiment 2 was conducted to further examine whether subjects' background knowledge regarding the text content would moderate their reliance on phonological recoding in understanding domain-specific texts, such that subjects without much background knowledge would rely more on phonological recoding in the processing of unfamiliar terms and text content. To this end, 15 excerpts from journal articles in Scientific American were used as experimental materials. The topics of these texts were all about physics, such as the Higgs boson, quantum computing, loop quantum gravity, antimatter, charge-coupled devices, and 3He superfluid. The same experimental paradigm as in Experiment 1 was used in Experiment 2, and a different sample of 30 university students in Taipei were recruited as participants. Half majored in humanities and social sciences and had no background knowledge in physics. The other half of the students majored in natural sciences, such as physics, atmospheric physics, and materials science; therefore, they had background knowledge in physics. Participants were asked to read the texts and answer four comprehension questions after reading each text. The results revealed that only background knowledge had a significant effect on reading comprehension, whereas reading conditions and their interaction with background knowledge did not exhibit any significant effects. Participants with relevant background knowledge performed significantly better than did their counterparts. However, background knowledge did not moderate participants' dependency on phonological recoding in reading comprehension. It seems that phonological recoding did not play a crucial role in understanding short domain-specific texts. Analysis of participants' eye movements indicated that reading aloud could slow their reading speed and reduce their regression ratio, but it did not enhance reading comprehension. RWL increased the rereading rate on target words in short expository texts, and it increased reading time and the mean fixation duration in domain specific texts. IRS did not affect comprehension performance; it only increased the rereading rate on domain-specific target words and prolonged the first fixation duration for participants without relevant background knowledge. In sum, the results of this study suggest that phonological recoding provides critical support for Chinese adult readers to comprehend short expository texts, but it only plays a minor role in understanding domain-specific texts. When participants read domain-specific texts, it was their background knowledge that dominated their reading comprehension performance.

並列關鍵字

Chinese reading ； eye-movement experiments ； phonological recoding

參考文獻

李俊仁、柯華葳（2009）：台灣學生聲韻覺識作業之聲韻表徵運作單位。教育心理學報，41（1），111–124。https://doi.org/10.6251/bep.20081117 [Lee, J.-R., & Ko, H.-W. (2009). Phonological representation unit in the phonological awareness task among Taiwanese students. Bulletin of Educational Psychology, 41(1), 111–124. https://doi.org/10.6251/bep.20081117]

官英華（2015）：檢視語音轉碼在英語為外語閱讀歷程中扮演的角色。英語教學期刊，39（3），107–140。https://doi.org/10.6330/etl.2015.39.3.05 [Guan, Y.-H. (2015). Investigating the role of phonological recoding in English-as-a-foreign-language reading processes. English Teaching & Learning, 39(5), 107–140. https://doi.org/10.6330/etl.2015.39.3.05]

陳惠茹、張鑑如（2011）：指讀及文字討論之共讀方式對幼兒認字的影響。教育心理學報，43（2），377–396。https://doi.org/10.6251/bep.20100611 [Chen, H.-J., & Chang, C.-J. (2011). Effects of print referencing in joint storybook reading on preschoolers’ word recognition. Bulletin of Educational Psychology, 43(2), 377–396. https://doi.org/10.6251/bep.20100611]

Elbro, C. (2018). Knowledge-based inference making for reading comprehension: what to teach and what not. Bulletin of Educational Psychology, 49(4), 701–713. https://doi.org/10.6251/bep.201806_49(4). 0009

Guan, Y.-H. (2020). Exploring the effects of phonological processing on foreign-language reading: An eye-tracking study for Chinese learners of Japanese-as-a-foreign-language. Bulletin of Educational Psychology, 51(3), 483–504. https://doi.org/10.6251/bep.202003_51(3).0007