再探語意預視效應：中文雙字詞處理 - 政大學術集成

全文

(1)國立政治大學語言學研究所碩士論文 National Chengchi University Graduate Institute of Linguistic Master Thesis. 指導教授：蔡介立博士 Advisor: Dr. Jie-Li Tsai. 學. ‧ 國. 立. 政治大. 再探語意預視效應：中文雙字詞處理. ‧. REVISITING SEMANTIC PREVIEW BENEFIT: EVIDENCE FROM PROCESSING OF CHINESE TWO-CHARACTER WORDS. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. 研究生：李孟璋撰 Student: Bing-Tsiong Li 中華民國一百零四年八月 August, 2015. v.

(2) REVISITING SEMANTIC PREVIEW BENEFIT: EVIDENCE FROM PROCESSING OF CHINESE TWO-CHARACTER WORDS. 學. ‧ 國. 立. 政 BY 治大 Bing-Tsiong Li. ‧. n. er. io. sit. y. Nat. al. Ch. i n U. A Thesis Submitted to the Graduate Institute of Linguistics In Partial Fulfillment of the Requirements for the Degree of Master of Arts. engchi. August, 2015. v.

(3) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. Copyright © 2015 Bing-Tsiong Li All Rights Reserved. i n U. v.

(4) Acknowledgement. 終於到了論文完成、開始著手寫謝辭的這一天了。回想這一路上的顛顛簸簸，能走到今天，我首要要感謝指導我完成這本論文的蔡介立老師。從 2012 年伊始，承蒙老師厚愛得以進入蔡老師的實驗室，跟著老師寫論文。身為一個充滿雄心壯志卻不切實際的初生之犢，我要感謝老師三年來不厭其煩的訓練、引導我的研究，讓原本的一直無法底定的研究議題得以定下根基、發展、終於成為現在這本論文。這一切的一切，都要感謝蔡老師這些年的提攜和砥礪。其次要感謝在百忙之中，犧牲暑假時間撥冗替我口試的徐嘉慧老師、李佳穎老師及李佳霖老師。老師們從各個不同的研究領域出發所給的意見，對最後論文的完整性來說，都是不可或缺且彌足珍貴的。除了專業意見之外，也謝謝老師們給我論文的鼓勵，它們對我來說一樣珍貴。另外也要感謝兩位老師—劉昭麟老師及 Reinhold Kliegl 老師。謝謝劉老師在論文提案口試給我的建議，以及後來不斷的鼓勵。謝謝 Kliegl 老師的工作坊，以及對我論文統計上提供的意見；如果沒有它們，我沒有辦法自己完成這份研究的所有統計。除此之外，我也要感謝所上的栽培，感謝每一位為我上課的老師：黃瓊之老師、詹惠珍老師、蕭宇超老師、徐嘉慧老師、何萬順老師、戴志偉老師、萬依萍老師、莫建清老師，謝謝老師們提供的專業知識和討論。我也要感謝實驗室的先進雅嵐學姐、婉柔學姐、和婉雲學姐，是你們帶我熟悉陌生的實驗室。不論是在研究上還是實驗室的各種疑難雜症，與你們的一席話總是一掃我的彷徨。也要謝謝家興學長，在程式跟統計上都提供了很多見解。謝謝前後兩位實驗室助理易男和冠慧，關於實驗室的大大小小事，都謝謝有你們罩。謝謝翊倫，有幸跟你一起準備口試、一起抱怨各種麻煩。最最要感謝的當然是柏亨，從實驗程式開始就一直受惠於你留下來的細心的筆記。關於所上的事情、我們同學的事情、各種八卦，總是只有你最清楚。你一直是最好的聆聽者，也總是能給我最好的鼓勵。一路走到最後的口試這一關，沒有哪一步可以沒有你。最後，我很感謝實驗室這個大家庭的每一個人，在苦悶、踟躕或是進度報告的各種崩潰的時候，不管是關於理論或統計的討論、或是跟大家一起八卦閒扯、一起買蝦醬炒飯，每一聲哀嘆或歡笑都能振奮我繼續跟論文奮鬥下去。另外我要感謝所上的同學：賈彬、柏亨、宇彤、高珍、怡婷、聖瑋、孝晨、郁萱、家昱、心怡。100 級的大家都是別具特色，還有那些一起修課、討論、乃至發瘋大笑的時光，謝謝你們帶給我不一樣的研究生活。我要特別謝謝賈彬，不論是實務上的各種道具，或是各種加油打氣，或是我最後崩潰的時候的一通電話，. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i. i n U. v.

(5) 你一直是我最依賴的支持。我要再次感謝何老師，除了在課堂和工作上的指導外，與老師的談話總是能給我許多正面能量，不論是在談學術、或是只是閒談。特別感謝助教學姐在這一段時間的精神支持，謝謝學姐願意傾聽我們的煩惱。在許多茫茫然不知所措的時刻，都要謝謝學姐的指引和鼓勵。我要感謝台大電機系，不論是在系裡的同儕，或是系上的訓練，對現在的我或現在的研究來說，是一樣重要的。其中特別要感謝我的導師陳中平教授，如果沒有老師的幫忙輔導，也許我現在還在讀學士班。另外也謝謝台大外文系的胥嘉陵老師，在我灰濛的轉跑道的路上鼓勵我踏出第一步，走自己規劃的方向。謝謝陳品蓁、黃志儒、羅盛華。謝謝你們從小看著我長大，到後來轉換跑道，到今天論文終告完竣，都很謝謝有你們的陪伴和支持。我要特別感謝兩位十年以上的好友－子培和雅淳。在這條未知卻又特別的路上，除了論文的壓力，還要忍受周遭親友的眼光。在我退縮猶疑、否定自己的時刻，你們的建議總是那麼中肯，你們的陪伴和鼓勵總是那麼溫暖。在攀上低峰、小有成就的時候，你們也不吝於讓我跟你們分享喜悅。謝謝你們陪我走過這一切的起起伏伏，沒有更多的言語可以表達我對你們的謝意。最後，我要謝謝我的家人－李碧宗先生、陳素琴女士，還有親愛的弟弟祐棠，你們不只是我最重要的家人，也都是我的榜樣。. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ii. i n U. v. 2015.8.13.

(6) TABLE OF CONTENT. Acknowledgements……………………………………………………………………i List of Tables…………………………………………………………………………vi List of Figures………………………………………………………………………vii Chinese Abstract……………………………………………………………………ix English Abstract……………………………………………………………………xi Chapter 1 Introduction…………………………………………………………1 1.1 General background………………………………………………………1 1.2 Semantic preview benefit in Chinese………………………………………3 1.3 Research questions…………………………………………………………5. 立. 政治大. ‧ 國. 學. ‧. Chapter 2 Literature Review……………………………………………………6 2.1 Chinese compound words…………………………………………………6 2.1.1 Characteristics of written Chinese……………………………………6 2.1.2 Chinese two-character compounds……………………………………7 2.2 Parafoveal processing………………………………………………………9 2.2.1 Parafoveal vision and perceptual span………………………………9 2.2.2 Preview benefit………………………………………………………10 2.2.3 Modulation of preview space and preview time……………………13 2.2.4 Reading models and semantic preview benefit………………………15 2.2.5 Semantic preview benefit in English and other languages…………20 2.3 Parafoveal processing in Chinese…………………………………………24 2.3.1 Scope of effects………………………………………………………24 2.3.2 Preview benefit in reading Chinese…………………………………25 2.4 Processing of Chinese two-character compounds during reading…………28 2.4.1 Prominence of compounds…………………………………………28 2.4.2 Parafoveal processing of two-character words………………………30 2.5 Transposition………………………………………………………………31. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. Chapter 3 Experiment One: Parafoveal semantic preprocessing of Chinese two-character compounds……………………………………………………35 3.1 Method……………………………………………………………………36 3.1.1 Participants…………………………………………………………36 3.1.2 Materials……………………………………………………………36 iii.

(7) 3.1.2.1 Norming study: Meaning relatedness, sentence predictability, and sentence plausibility………………………………………………………39 3.1.3 Apparatus……………………………………………………………42 3.1.4 Procedure……………………………………………………………43 3.2 Data analysis………………………………………………………………45 3.3 Results……………………………………………………………………49 3.3.1 Pretarget region………………………………………………………49 3.3.2 Target region…………………………………………………………51 3.3.2.1 Modulation of preview time and preview space………………53 3.3.3 Posttarget region……………………………………………………57 3.4 Discussion…………………………………………………………………57 3.4.1 Parafoveal-on-foveal effect…………………………………………57 3.4.2 SR preview cost………………………………………………………58 3.4.3 Semantic preview benefit……………………………………………59 3.4.3.1 Plausibility effect on semantic preview benefit………………62. 立. 政治大. ‧. ‧ 國. 學. Chapter 4 Experiment Two: Semantic preview benefit with plausible pre-context………………………………………………………………………65 4.1 Method……………………………………………………………………66 4.1.1 Participants…………………………………………………………66 4.1.2 Materials……………………………………………………………66 4.1.2.1 Norming study: Meaning relatedness, sentence predictability, and sentence plausibility………………………………………………………69 4.1.3 Apparatus……………………………………………………………70 4.1.4 Procedure……………………………………………………………70 4.2 Data analysis………………………………………………………………71 4.3 Results……………………………………………………………………72 4.3.1 Pretarget region………………………………………………………72 4.3.2 Target region…………………………………………………………73 4.3.2.1 Modulation of preview space and preview time………………74 4.3.3 Posttarget region……………………………………………………78 4.4 Discussion…………………………………………………………………79 4.4.1 Parafoveal-on-foveal effect…………………………………………80 4.4.2 Semantic preview benefit……………………………………………80 4.4.3 SR preview cost………………………………………………………82. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. Chapter 5 General Discussion…………………………………………………83 5.1 Parafoveal-on-foveal effect………………………………………………83 5.2 Semantic preview benefit…………………………………………………85 5.2.1 Modulation of preview time and preview space……………………87 iv.

(8) 5.2.2 Implication to reading models and contextual processing in reading…………………………………………………………………………89 5.2.3 Contextual constraint as an alternative to plausibility effect………92 5.3 SR preview cost……………………………………………………………94 5.4 Limitations and suggestion for future studies……………………………96 Reference…………………………………………………………………………99 Appendices……………………………………………………………………105 A. Materials of Experiment 1………………………………………………105 B. Materials of Experiment 2………………………………………………107. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. v. i n U. v.

(9) List of Tables. Table 1. Means and standard errors of word frequency per million, C1, C2, and total strokes of the two constituents of each preview condition in Experiment 1……38 Table 2. Means and standard errors of rated results of the norming study for Experiment 1…………………………………………………………………………………39 Table 3. The fixation patterns and corresponding calculation of each eye movement measures used in the present study……………………………………………46 Table 4. Results of analyses of pretarget fixation duration in Experiment 1 and the interactions with preview space (p.s.) pretarget log word frequency (w.f.) ……50 Table 5. Results of analyses on the target region in Experiment 1, with preview space (p.s.), preview time (p.t.) and the interaction (both) of the both as covariates…52 Table 6. Results of analyses on the posttarget region in Experiment 1, with posttarget log word frequency (w.f.) as covariate……………………………………………57 Table 7. Means and standard errors of word frequency per million, C1, C2, and total strokes of the two constituents of each preview condition in Experiment 2……67 Table 8. Means and standard errors of rated results of the norming studies for Experiment 2…………………………………………………………………………………68 Table 9. Results of analyses on the pretarget region in Experiment 2, with pretarget log word frequency (w.f.) and preview site (p.s.) as covariates……………………72 Table 10. Results of analyses on the target region in Experiment 2, with preview time (p.t.), preview space (p.s.), and the interaction (both) of the both as covariates………74 Table 11. Results of analyses on the posttarget region in Experiment 2, with posttarget log word frequency (w.f.) as covariate……………………………………………79 Table 12. Main effects of non-identical preview costs from both experiments…………86 Table 13. Trends of interaction between semantic preview effects and preview time under different preview space conditions in the two experiments……………………88 Table 14. Selected results of non-identical-to-identical effect from various studies that match in their designs as close to those in current study………………………95. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. vi. i n U. v.

(10) List of Figures. Figure 1. Display change in boundary paradigm with an invisible boundary between express and connects…………………………………………………………11 Figure 2. Semantic preprocessing predicted by E-Z Reader…………………………19 Figure 3. Eye-tracking experiment procedure in this study……………………………44 Figure 4. Estimated FFD (left panel), SFD (right panel), and GD (middle panel) on pretarget words of Experiment 1 to preview space, with pretarget wordfrequency-related effects and random effects of participant and item removed. Errorbands showed 95% of confidence intervals……………………………51 Figure 5. Estimated FFD (left panel), SFD (right panel), and GD (middle panel) on target words of Experiment 1 to preview space, with only random effects of participant and item removed. Errorbands show 95% of confidence intervals……………………………………………………………………54 Figure 6. Estimated FFD (left panel), SFD (right panel), and GD (middle panel) on target words of Experiment 1 to preview time, with random effects of participant and item removed. Errorbands show 95% of confidence intervals……………………………………………………………………55 Figure 7. Estimated GD on target words of Experiment 1 to preview time under large preview space condition (within 1.05 characters, left panel) and under small preview space condition (beyond 1.05 characters, right panel) with random effects of participant and item removed. Errorbands show 95% of confidence intervals……………………………………………………………………56 Figure 8. Estimated FFD (left panel), SFD (right panel), and GD (middle panel) on target words of Experiment 2 to preview space, with random effects of participant and item removed. Errorbands show 95% of confidence intervals……………………………………………………………………76 Figure 9. Estimated FFD on target words of Experiment 2 to preview time under large preview space condition (within 1.21 characters, left panel) and under small preview space condition (beyond 1.21 characters, right panel), with random effects of participant and item removed. Errorbands show 95% of confidence intervals……………………………………………………………………77. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. vii. i n U. v.

(11) Figure 10. Estimated SFD on target words of Experiment 2 to preview time under large preview space condition (within 1.2 characters, left panel) and under small preview space condition (beyond 1.2 characters, right panel), with random effects of participant and item removed. Errorbands show 95% of confidence intervals……………………………………………………………………78. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. viii. i n U. v.

(12) 國. 立. 政. 治. 大. 學. 研. 究. 所. 碩. 士. 論. 文. 提. 要. 研究所別：語言學研究所論文名稱：再探語意預視效應：中文雙字詞處理指導教授：蔡介立研究生：李孟璋. 政治大. 論文提要內容：(共一冊，20370 字，分 5 章 20 節). 立. 本研究探討中文讀者對於中文雙字詞的早期語意處理，特別要探討的是中文. ‧ 國. 學. 讀者是否能在凝視一個中文雙字複合詞之前，即可提取該詞的語意資訊。在中文. ‧. 的閱讀研究中曾經發現單字的語意資訊可以在該單字被凝視之前提取，不論是成. Nat. io. sit. y. 詞的單字或是多字詞其中的組成單字。也有證據顯示中文雙字詞或雙字複合詞呈. er. 現在中央視野時的處理方式是整詞處理。由於事實上就詞長來看，雙字詞的詞類. al. n. v i n Ch 數量為中文詞類的最大宗，也是最常被使用的詞類，因此中文雙字詞語意處理的 engchi U. 時間歷程便是本研究的研究目標。實驗一旨在檢視雙字詞的語意資訊是否能和單字一樣，在被凝視之前即被提取。本實驗採用邊界典範（Boundary paradigm, Rayner, 1975），除了目標詞預視（identical preview）之外，本實驗亦包含語意相關預視（semantic-related preview）、語意無關預視（semantic-unrelated preview）及非詞預視（nonword preview）。實驗發現語意相關預視能促進目標詞的處理。然而和預視空間（preview space）及 ix.

(13) 預視時間（preview time）的交互作用則顯示語意預視效益（benefit）在預視空間較大的時候會隨著預視時間拉長而增加，在預視空間較小的時候，則會隨著預視時間漸減。在實驗一中，語意關聯性的高低和該詞合理性（plausibility）的高低是共變的，因此這兩個要素構成一個混淆的因子。語意相關預視和語意無關預視之間的效果有可能來自語意關聯性的差異，或是來自在句中合理性的差異。合理性同時. 政治大. 也能解釋在為何在實驗一中，不合理的語意無關預視在目標前詞（pretarget）上. 立. 造成較短的閱讀時間。為了解決這項混淆的因子，實驗二因此採用對目標詞預視、. ‧ 國. 學. 語意相關預視、及語意無關預視都合理的句子。結果發現，語意預視的主要效果. ‧. 消失。然而交互作用的模式則顯示出，語意預視效益在較長的預視時間、較大的. Nat. io. sit. y. 預視空間下仍會存在。但在較小的預視空間下，會隨著較長的預視時間而轉為耗. er. 損（cost）。實驗二的結果因此提供證據支持在沒有合理性的交互作用下，中文. al. n. v i n C h ，兩個實驗的差異顯示在中文閱讀中存在合雙字詞的語意預視效應仍會發生。最後 engchi U. 理性的預視效益，此結論和之前中文閱讀的研究結果一致。. x.

(14) Abstract. The present study investigates the early semantic processing of Chinese two-character words by Chinese readers. Specifically, whether Chinese readers are able to extract semantic information of an up-coming two-character compound as a. 治政 whole when the word is yet being fixated. In Chinese,大 it has been demonstrated that 立 ‧ 國. 學. semantic information can be extracted from a single character, whether it is a word or part of a word, before the character is being fixated. There is also evidence for whole. ‧. word processing of foveally presented two-character compounds/words. Since. sit. y. Nat. io. n. al. er. two-character words actually constitute the majority of word type and are used most. i n U. v. frequently in total, the time course of processing the meaning of such combination of. Ch. engchi. characters during reading is then the goal of this study. The first experiment aimed to examine whether semantic information of a two-character word can be extracted before it is fixated, as what have been found for single characters. Boundary paradigm (Rayner, 1975) was used, with identical, semantic-related, and semantic-unrelated words, as well as nonwords as preview. Semantic-related preview did facilitate target word processing. The interaction pattern of the effects with preview space and preview time, however, showed that semantic xi.

(15) preview benefit could increase with preview time with small preview space, but decrease with preview time under large preview space. A possible confounding factor in the first experiment was the overlap between semantic relatedness and plausibility. The effect between semantic-related and semantic-unrelated previews could be of semantic or plausibility nature. Plausibility may also explain the shortened fixation duration found in Experiment 1 when. 政治大. implausible semantic-unrelated preview was presented parafoveally. Experiment 2. 立. then solved this confounding by using sentence frames which are plausible for. ‧ 國. 學. identical, semantic-related, and semantic-unrelated previews. In Experiment 2, main. ‧. effect of semantic preview benefit disappeared, while the interaction patterns showed. Nat. io. sit. y. that such benefit existed for large preview space with long preview time, but became. er. cost for small preview space with long preview time. The results of Experiment 2 thus. al. n. v i n C h effect of Chinese provide evidence for semantic preview e n g c h i U two-character words without the interaction with plausibility. Finally, the discrepancies between the two experiments indicate the existence of plausibility preview benefit, which previous studies have suggested to exist in Chinese.. xii.

(16) Chapter 1. Introduction 1.1 General background During reading, the reader’s final goal is to retrieve the meaning from each word in the text and eventually integrate all the information into the meaning of the sentence, then of the text. While most of the information was retrieved during relative steady fixation from the fixation point, it has been demonstrated that lexical. 政治大. information such as phonology or orthography can be extracted from the more. 立. peripheral vision called parafovea and benefit the ensuing processing. However, the. ‧ 國. 學. extraction of semantic information parafoveally has been an issue in reading studies,. ‧. not only because of its relative elusiveness to other preview benefit, but also because. Nat. io. sit. y. of its language-dependency. While alphabetic English and German, for examples,. er. exhibit different behaviors and different degrees of effect size regarding semantic. al. n. v i n logographic C Chinese exhibit different h e nalso gchi U. preview benefit,. reading patterns in. semantic preview benefit when plausibility comes into play or when different units, such as semantic radicals, characters in words, single-character words, or two-character words are of interest. Previous studies have attributed the diverse findings to different processing load required by the different designs of the writing systems. The rigid spelling system in German requires fewer resources to resolve the texts to sounds, while English, as a writing system without rigorous correspondence 1.

(17) between spelling and sounds, retain fewer resources to recode the phonological information into meanings. The proposed explanation following this rationale for positive evidence of semantic preview benefit in Chinese then usually states that in Chinese, the orthography-phonology-semantic route is not followed. Rather, Chinese readers apply the strategy of orthography-semantic route, which requires fewer resources.. 政治大. Aside from the diverging findings from different writing systems, there are also. 立. theoretical implications regarding the existence of semantic preview benefit. There. ‧ 國. 學. have been models proposed to describe reading behaviors, with mechanisms. ‧. determining reading times or fixation locations or such. Two most prevailing groups. Nat. io. sit. y. of models are serial attention shift (SAS) models and attentional gradient (GAG). er. models. The key difference related to this issue is the different saccade target selection. al. n. v i n between the two models. WhileCSAS h emodels h i Udeterministic saccade targeting, n g cassume which may exclude words that have been semantically processed not to be the target, target selection of GAG models is not deterministic and allows semantic preview benefit to occur. The issue of reading behavior brings up another line of investigation related to preview benefit. It has been proposed that preview benefit observed in previous studies may have resulted from preview cost brought by previewing an unrelated 2.

(18) preview, in addition to preview benefit, which is brought by previewing a related preview. This is supported by recent findings that the size and direction of preview effect may depend on how long the preview is parafoveally presented, how close the previewing site is to the preview, or how much foveal load is there during preview. The elusiveness of main effect of semantic preview benefit may be attributed to the interactions with these variables. While it may be necessary to investigate semantic. 政治大. preview effect with these variables, one may also gain insight into how these variables. 立. come into play during reading with such investigation.. ‧. ‧ 國. 學. 1.2 Semantic preview benefit in Chinese. Nat. io. sit. y. Previous studies of Chinese semantic preview effects have been focusing on the. er. processing of single character. It has been found that semantics-related information. al. n. v i n C hwriting unit can beUextracted parafoveally. These from different levels of Chinese engchi include sub-character radicals, non-word single characters, single-character words, and homographic characters that bear different morphemic meanings when embedded in different two-character words. In language use of Chinese, however, although characters are the basic units of writing, words can consist of only one character, or be composed of more than one characters. In most alphabetic writing systems, words are separated by spaces in between, but in Chinese, such multi-character words are not 3.

(19) visually separated. Among these words of different lengths, two-character words are the majority of word type and word token. The time course of resolution of meaning of such a word is thus of high importance, not only because these words are the majority of actual Chinese use, but because they lack explicit visual boundary when embedded in the sentences. Yang (2013) addressed the issues with transposable two-character words, which,. 政治大. once transposed, are still words and may bear similar or different meaning to its. 立. original form. The reason for her choice of materials was to minimize orthographic. ‧ 國. 學. difference between the targets and the transposed previews. The findings, however,. ‧. suggested that Chinese two-character words exhibit plausibility effect rather than. Nat. io. sit. y. semantic preview effect. On the one hand, plausibility preview effect, which has also. er. been demonstrated for single-character words (Yang, Wang, Tong, & Rayner, 2012),. al. n. v i n C hprocess as early asUduring parafoveal preview. On may indicate an ongoing integration engchi the other hand, the effect observed in her study may suffer from other interference.. First of all, the use of transposable words risks blurring the effect when the lexical representation of the transpose is activated. Second, her conclusion was drawn from comparing the effects between identical and transpose preview across different groups of materials, rather than directly from the benefit brought by previewing a related item compared to the unrelated one. Lastly, transposable two-character words only 4.

(20) represent a small part of Chinese lexicon or of Chinese two-character words. It remains unknown whether the effect found in the study is applicable to general two-character words. (Schotter, Lee, Reiderman, & Rayner, 2015; Yang, 2013; Yang et al., 2012).. 1.3 Research questions. 政治大. The present study therefore presents two experiments in investigating the time. 立. course of semantic access of Chinese two-character words and its interaction with the. ‧ 國. 學. context. The study addresses the specific research questions as follows:. ‧. (1) Can semantic information be extracted from a general Chinese two-character word,. Nat. sit. n. al. er. io. the word?. y. rather than the specific type of transposable words, prior to the fixation of eyes on. Ch. engchi. (2) Does semantic preview benefit stand when. iv n the pre-context U. supports both. semantically related and unrelated previews? The inference from the answer of the questions above will help us answer a third more general question of: (3) How do Chinese readers process parafoveally the meaning of a two-character word during reading? That is, what is the time course of two-character word processing during parafoveal preview? 5.

(21) Chapter 2. Literature Review 2.1 Chinese compound words 2.1.1. Characteristics of written Chinese Chinese is commonly written in square-like units called characters, mostly from. left to right, with equal-size small spaces between each character. Characters can be inseparable combination of strokes by itself, or they can be composed of more than. 政治大. one radical, which either conveys part of the character’s meaning or gives a clue to its. 立. pronunciation. Except for some stylistically manipulated characters, almost all. ‧ 國. 學. Chinese characters correspond to one syllable in the oral language. Characters are. ‧. therefore said to be the smallest writing unit of Chinese. The smallest unit to. Nat. io. sit. y. constitute a meaningful sentence, however, is word, like the other languages of the. er. world. Chinese words can consist of one or more than one characters/syllables. While. al. n. v i n C hare among the most some of the single-character words e n g c h i U frequently used words, such as 我 “I”, 的, associative marker, and 是, copula, taking about 45.1% of occurrence in total (1922255 in 4264322), the majority of different word types go to. two-character words, which account for 51.2% of word types (65228 in 127524) in Chinese, and 48.1% of word occurrence in use (2049601 in 4264322), as estimated from Academia Sinica Balanced Corpus of Modern Chinese (Academia Sinica Balanced Corpus, 2004). 6.

(22) Another characteristic relevant to this study is the disconnection between orthographic form and pronunciation of Chinese character, and thus of multi-character words as well. Beside the correspondence of one character to one syllable, characters have a less correspondence from their form to their pronunciation, unlike alphabetic writing systems. This characteristic has been mostly argued in previous studies to be the reason why semantic information in Chinese is easier to access. A more direct link. 政治大. has been proposed from orthography to meaning, with phonological representation as. 立. by-product during lexical access (Hoosain, 1991; Zhou, Marslen-Wilson, Taft, & Shu,. ‧ 國. 學. 1999). However, the fact is that most of the characters in Chinese are phonograms (形. ‧. 聲), which encode their pronunciation loosely with a phonological radical. This. Nat. io. sit. y. resembles the situation in English, which also has a loose orthography-phonology. er. correspondence. As will be argued the sections below, phonological information, like. al. n. v i n C h a role during Chinese what is found in English, also played e n g c h i U reading. 2.1.2. Chinese two-character compounds Most of Chinese words are composed of more than one characters. Their. meanings can either be compositional from the constituent characters, or be independent of its components. For example, a two-character word in Chinese can be monomorphemic word, whose constituent characters have no independent meanings, 7.

(23) like 蟋蟀 “cricket” or 佝僂 “stooped”. They can as well be composed of two characters, each with its own meaning. For example 字典 “dictionary” is composed of 字 “character” and 典 “canon”. In the case of two-character compounds, although both characters are termed morphemes, the connection between the combination of morphemes and the meaning of the word can be transparent, like the example above, or opaque, like 津貼 “subsidy” from 津 “ferry” and 貼 “to stick”.. 政治大. For most of the time however, they are somewhere in between the two extremes.. 立. Since the only spacing in written Chinese is between characters rather than words,. ‧ 國. 學. these meaningful component character in compounds can as well be demarcated as. ‧. single-character words, though it may not lead to plausible parsing of the sentences.. Nat. io. sit. y. This arouses the question of how such compounds are processed and when the. er. meaning of the entire compounds is accessed during reading. In research of such. al. n. v i n C h parafoveal processing question as time course of processing, has been used to probe engchi U the time course and the qualitative nature of the processing. Specifically, the process occurs when the visual presentation of the word falling in parafoveal vision during reading. The following section provides an introduction to this approach and its link to Chinese two-character compounds in this study.. 8.

(24) 2.2 Parafoveal processing 2.2.1. Parafoveal vision and perceptual span During reading, as well as other visual task, most of the textual information is. extracted from the foveal vision, which is a narrow area around that fixation point within a visual angle of about only 2° degree (Levi, Klein, & Aitsebaomo, 1985). Our eyes fixate at the fixation point relatively steadily, until a rapid saccade, during which. 政治大. our vision is mostly suppressed, that brings the eyes to another fixation point. When. 立. the eyes fixated at a point, visual information with less acuity is also extracted from. ‧ 國. 學. an area of 2° to 5° degree from the fixation point. Such parafoveal vision results from. ‧. physical conditions of our eyes and provides visual cues, for example the position of. Nat. io. sit. y. the next word, to guide eye movement. Meanwhile, this parafoveal vision is. er. symmetrical around the fovea and is universal to all human. The range of peripheral. al. n. v i n C h is further modulated textual information utilized in reading e n g c h i U by allocation of attention, which is adapted to fit different writing systems. Such range is termed perceptual span, which, unlike parafoveal vision, is asymmetrical due to the direction of reading and differs according to different writing systems. By limiting the text available from a window around fixation point (moving-window paradigm, McConkie & Rayner, 1975) and measuring whether reading is impeded, one can determine the perceptual span in a certain writing system. For example, it has been found that in English, perceptual 9.

(25) span extends 14 to 15 letters to the right and 3 to 4 letters to the left of the fixation, in accord with the fact the English reads from left to right (McConkie & Rayner, 1975, 1976; Rayner, Well, & Pollatsek, 1980). Since Hebrew reads from right to left, the pattern of perceptual span reverses, extending further to the left and shorter to the right (Pollatsek, Bolozky, Well, & Rayner, 1981). In Chinese, and in Japanese as well, with the use of characters, which are more condensed spatially, especially horizontally. 政治大. in comparison to alphabetic writing systems, the perceptual span is found to be 2 to 3. 立. characters to the right and 1 character to the left of the fixation in Chinese (for. ‧ 國. 學. Japanese, see: Ikeda & Saida, 1978; Inhoff & Liu, 1998). The presence of such a span. ‧. implies that at least part of the information in this area is crucial to fluent,. Nat. io. sit. y. non-interrupted reading. The ensuing questions regarding parafoveal previews would. n. al. er. then be the depth of processing of such extracted information and its role in fluent reading.. 2.2.2. Ch. engchi. i n U. v. Preview benefit In order to investigate such questions, boundary paradigm (Rayner, 1975) has. often been implemented. In this paradigm, an invisible boundary is set at certain position in the sentence. The target at the right side (when the writing system reads from left to right) of the boundary is replaced by a preview until the eyes saccade 10.

(26) across the boundary (see Figure 1). When vision is suppressed during saccade, the preview switches rapidly back to the target word, and so the reader would not detect any changes. By doing so, although the reader reads the target correctly, s/he has received lexical information of another word from this region when fixating text prior to the boundary. If processing of a word starts as early when the word falls in parafoveal vision, using a non-identical preview would interrupt the preprocessing. 政治大. and resulted in longer fixation duration when the target is fixated. Such identical. 立. preview benefit is indeed widely reported in many studies and is regarded as an. ‧ 國. 學. indication of existence of parafoveal preprocessing.. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. Figure 1. Display change in boundary paradigm with an invisible boundary between express and connects. Further, lexical processing can be decomposed into stages of processing or activation of different lexical properties, such as phonology, orthography, or meaning. As the terminal goal of reading is to access the meaning of each word and to integrate 11.

(27) them from merely visual input, orthographic information, for example, must first be processed in order to activate other representation, such as phonology, and to eventually access the meaning. By manipulating the overlap of lexical property between the preview and the target, one can infer from any ease of processing, as compared to an unrelated preview, about the depth of parafoveal processing. For example, a homophonous preview may yield a shorter fixation time on the target in. 政治大. comparison to a non-homonymous preview. This means that phonological. 立. representation of a word can be accessed when the word falls in parafoveal vision and. ‧ 國. 學. thus shortens the ensuing foveal process. Such phonological preview benefit is well. ‧. documented for example in English (Ashby, Treiman, Kessler, & Rayner, 2006),. Nat. io. sit. y. French (Miellet & Sparrow, 2004), and even Chinese (Liu, Inhoff, Ye, & Wu, 2002;. n. al. er. Tsai, Lee, Tzeng, Hung, & Yen, 2004), which does not seem to encode phonological. Ch. engchi. representation as systematically as alphabetic. iv n writing U. systems. Aside from. homophonous preview benefit, preview benefit for onset syllable structure (Ashby & Rayner, 2004), orthography (McConkie & Zola, 1979; Rayner, McConkie, & Zola, 1980 among others), initial letter string (Inhoff, 1989), and morphology (Deutsch, Frost, Pollatsek, & Rayner, 2005). Just like perceptual span, however, the presence of certain types of preview benefit varies across different writing systems. Since different writing systems encode 12.

(28) language information in different ways, reading strategy and depth of parafoveal preprocessing differs as well. The above-mentioned morphological preview benefit, for instance, is only found for Hebrew but not for English and Finnish, since constituent letters of morphemes in Hebrew can be interwoven across a words (Schotter, Angele, & Rayner, 2012). While those in English and Finnish are basically linearly aligned, such morphological process is inevitably overlapped with. 政治大. phonological or orthographic process.. Modulation of preview space and preview time. 學. 2.2.3. ‧ 國. 立. ‧. However, some may question whether the effects observed in the experiments. Nat. io. sit. y. were real preview benefit or preview cost, or they were the results of both effects at. er. work. The notion of preview cost is that when a reader is previewing an unrelated. al. n. v i n C hinterfere with the reading word or random symbols, it may e n g c h i U since this is not a usual. reading situation. However, simply measuring the difference in reading times for different previews does not distinguish between the two effects. Kliegl, Hohenstein, Yan, and McDonald (2013) then discussed the methodological possibilities to investigate into this issue with preview space and preview time. When the eyes fixate closer to the target before they foveally fixate it, larger portion of the target would fall into parafoveal vision, and the target would be closer to the foveal vision. The reader 13.

(29) can thus theoretically receive a better preview of the target. Preview time, on the other hand, is the time the reader spends during previewing. Operationally it can be defined as single fixation duration (SFD, the only first-pass fixation duration on a certain region) or gaze duration (GD, sum of all first-pass fixation durations on a certain region) of the pretarget under the situation where first-pass fixation on the target follows immediately the first-pass fixation on the pretarget. When better preview is. 政治大. provided, by means of closer preview space or longer preview time, increase in. 立. reading times for unrelated preview or random control indicates preview cost, while. ‧ 國. 學. decrease for identical or related preview would be preview benefit.. ‧. There had been studies investigating the influences of these factors on preview. Nat. io. sit. y. effects. McDonald (2006) compared the preview benefit between identical previews. er. and random letter strings in two conditions, in one of which the boundary was set at. al. n. v i n Cinhusual boundary paradigm the end of the pretarget word, as e n g c h i U experiments. In another condition the boundary was set between the 4th and 5th letters of pretarget word, which were all 9- or 10-letter long. He found that preview type was only a predictor for target fixation durations when the boundary was at the end of the pretarget. As for preview time, Yan, Risse, Zhou, and Kliegl (2012) investigated preview benefit in Chinese, following the design of Yan, Richter, Shu, and Kliegl (2009), which will be discussed later. They found that under longer pretarget SFD, as their operational 14.

(30) definition of preview time, increased the preview benefits. In another study by Tsai, Kliegl, and Yan (2012), where the operational definition of preview time was pretarget GD, preview benefits also increased with longer preview time. Details of some of the studies mentioned above will be discussed later. For now, it seems clear that preview effects can be modulated by the factors of preview time and preview space. Further, according to how these factors influence the preview. 政治大. effects – by increasing the reading times with unrelated preview or by decreasing the. 立. reading times with related or identical preview, one could conclude whether such. ‧ 國. 學. preview effects are due to preview benefit from related features or preview cost from. ‧. unrelated features. Before further examining the studies in semantic preview benefit,. Nat. io. sit. y. the theoretical importance of such effect regarding the reading models will be. n. al. er. discussed in the next section.. 2.2.4. Ch. engchi. i n U. v. Reading models and semantic preview benefit With accumulation of eye movement data, models have been proposed to fit. and explain the data from our reading behavior. The two most successful and popular models are E-Z Reader (Rayner, Li, & Pollatsek, 2007; Reichle, Liversedge, Pollatsek, & Rayner, 2009; Reichle, Pollatsek, Fisher, & Rayner, 1998; Reichle, Pollatsek, & Rayner, 2006, 2007; Reichle, Warren, & McConnell, 2009) and SWIFT (Engbert, 15.

(31) Longtin, & Kliegl, 2002; Engbert, Nuthmann, Richter, & Kliegl, 2005), which belong respectively to serial attention shift (SAS) models and attentional gradient (GAG) models. According to E-Z Reader, attention is allocated at each word during reading, while in SWIFT, attention is distributed around the fixation point, where the level of activation of each word being attended rises simultaneously, since they are processed in parallel.. 政治大. Although detailed specifications of each type of models are not the issue here,. 立. descriptions of their architectures are discussed for further discussion in semantic. ‧ 國. 學. preview benefit. In E-Z Reader, detailed mechanism has been proposed about the time. ‧. course and sequence of stages of visual processing, lexical processing, of attention. Nat. io. sit. y. shift, and of saccade planning. In E-Z Reader, lexical processing is assumed to be. er. word-wise serial. A word is being processed when attention, not fixation point, is. al. n. v i n C two allocated to that word. There are h estages i U processing, L1 and L2. L1 is n g cofhlexical the early stage of lexical processing and is generally associated with processing of phonological and orthographic information. The completion of L1 will initiate L2, where deeper processing such as semantic processing takes place. The speed of each. stage is influenced by factors such as word frequency, contextual information, and deviation of the word from the fixation point. While lexical processing is strictly serial, eye movement control in E-Z Reader is modulated by the state of lexical 16.

(32) processing. E-Z Reader assumes two cascading stages of saccade planning as well. The labile stage M1 can be cancelled and replaced by new saccade planning, but when this saccade planning enters the non-labile stage M2, the planned saccade will have to be executed at the moment when M2 is done, before any new saccade planning can take place. A new saccade planning is triggered by completion of L1 of a certain word n, termed L1n. This new saccade planning aims at the next word n+1,. 政治大. termed M1n+1. Following the architecture of E-Z Reader above, when the eyes fixate. 立. on a certain word n but the first stage of lexical processing on the next word n+1. ‧ 國. 學. (L1n+1), for example, completes before first stage of saccade planning to the next. ‧. word (M1n+1) is done, M1n+1 will then be canceled and replaced by M1n+2, to the even. Nat. io. sit. y. next word. However, when L1n+1 completes at the time when the saccade planning to. er. the next word has entered the second stage (M2n+1), it cannot canceled this saccade. al. n. v i n C his carried out. According is done and the saccade to E-Z Reader, the first engchi U. until M2n+1. situation is the case of skipping (of word n+1), and in the second situation, preview benefit occurs. In the second situation, lexical processing of word n+1 completed before the eyes fixate, but not skip, the word, and that shall reduce the time spent on word n+1. SWIFT model, on the other hand, assumes distributed attention, and the words within the attention window are being processed at different rate. The speed of word 17.

(33) processing, which results in the rises of what SWIFT terms activation levels, is decided by various factors, including the level of attention distributed to it. What differs SWIFT most from E-Z Reader is the temporal and spatial decision regarding saccade planning. In E-Z Reader, saccade planning has a specific goal at the beginning of M1, which is triggered by level of lexical processing (completion of L1). In SWIFT, however, saccades are autonomous generated without specific targets. Two. 政治大. stages of saccade programming are also suggested by SWIFT, first one labile. 立. followed by the non-labile stage. If there is no new saccade programming that. ‧ 國. 學. intervenes and cancels the current labile saccade programming, at the end of the labile. ‧. stage, saccade target will then be decided based on the activation level of each word.. Nat. io. sit. y. As for temporal variation in SWIFT, the base for saccade latency is the stochastic. n. al. er. process in saccade generation, modulated by the intended saccade amplitude. Since. Ch. engchi. saccade target is decided with the completion. iv n ofUlabile stage,. this modulation. influences only the length of non-labile stage. Furthermore, saccade latency is modulated by foveal inhibition, which stems from difficult foveal words and aims to lengthened current fixation for further processing. An important issue here is the difference in their predictions to certain phenomena, one of which would be semantic preview benefit. In E-Z Reader model, L1 only accounts for a low-level process of the word. In order to access semantic 18.

(34) representation from word n+1, L1n+1 would have been done for L2n+1 to reach a certain level of completion. Under this circumstance, M1n+1 would have been canceled and a saccade to skip word n+2 is being planned, resulting in skipping of word n+1 without any detectable benefit on that word. Aside from mislocation of saccade, the only occasion for such preview benefit to appear is that by the time L1n+1 is done, saccade planning has reached M2n+1 (see Figure 2). The time for L2n+1. 政治大. processing before the eyes saccade to word n+1 is then shorter than M2n+1, which is. 立. regarded as “formidable” by (Hohenstein & Kliegl, 2014). SWIFT model, on the. ‧ 國. 學. other hand, does not exclude the possibility that a word whose semantic information. ‧. has been activated becomes a potential target of saccade, since the selection is. Nat. io. sit. y. probability-based and is not only a function of activation level but also of other. n. al. er. physical and lexical properties.. Ch. engchi. i n U. Figure 2. Semantic preprocessing predicted by E-Z Reader 19. v.

(35) Another reason for the hot debate for semantic preview benefit is its elusiveness in English. As stated above, a certain type of preview benefit can be a writing-system-dependent phenomenon, such as morphological preview benefit for Hebrew. Semantic preview benefit, however, has been disputed in English, the most-studied writing system. The following section will review the studies that provide evidence, null or positive, for such effect in English and in other languages as well.. 立. ‧ 國. 學. 2.2.5. 政治大. Semantic preview benefit in English and other languages. ‧. Rayner, McConkie, et al. (1980) conducted an experiment investigating. Nat. io. sit. y. parafoveal semantic processing when there was no foveal load. Subjects were asked. er. to name a word (e.g. table) initially presented in parafovea while semantic (e.g. chair). al. n. v i n C hinitially took theUplace of the target word. The or unrelated (e.g. chore) preview engchi reaction times were the same for both types of previews, providing no evidence for semantic preview benefit. Even when the common boundary paradigm was used, studies in English (Rayner, Balota, & Pollatsek, 1986) and Finnish (Hyona & Haikio, 2005; White, Bertram, & Hyönä, 2008) provided no positive evidence for such preview benefit. The experiment manipulation in White et al. (2008) did yield some semantically related preview effects. In the study, the target was the second 20.

(36) constituent of long, two-constituent compounds, for example vaniljakastike “vanilla-sauce”.. Semantic-related. “vanilla-mustard”,. and. preview,. semantic-unrelated. for preview,. example. vaniljasinappi. such. vaniljarovasti. as. “vanilla-priest”, along with identical and nonword preview (vaniljaseoklii “vanilla-nonword”) were included in their experiment manipulation. Semantic related preview did not yield shorter fixation durations on the second constituent but resulted. 政治大. in longer go-past time (i.e. time from the first fixation on the target in first pass. 立. reading until the eyes fixated at the region to the right of the target) when the whole. ‧ 國. 學. compound region was analyzed. There is no knowing what kind of process this rather. ‧. late measure on the whole word rather than on the target area reflected. In another. Nat. io. sit. y. study conducted by Altarriba, Kambe, Pollatsek, and Rayner (2001), native Spanish. er. speakers with fluent English ability were recruited in a boundary experiment, where. al. n. v i n previews,Ccognates from terminology h e n g(different chi U. there were identical. in historical. linguistics, cognate here is defined as orthographically and semantically similar word pair in different languages, e.g. crema – cream), pseudo-cognates, (i.e. orthographically similar word pairs with unrelated meaning, e.g. grasa “fat” – grass), non-cognate translations (i.e. semantically but not orthographically related word pairs, e.g. dulce – sweet), and unrelated previews. Preview benefits were found for identical, cognate, and pseudo-cognate preview but not for non-cognate translation preview. In 21.

(37) sum, preview benefit here in this study did not go beyond the effect of orthographic preprocessing. Recently, though, Schotter (2013) has demonstrated the presence of preview benefit in synonymous preview condition with neutral pre-context, while Schotter et al. (2015) found preview benefit for general semantic-related preview under special manipulation of the pre-context. The reason why only synonyms exhibit preview benefit under general neutral context was that semantic relatedness can be. 政治大. categorized into different sort of relatedness. For example antonyms like happy and. 立. sad may not seem related but they belonged to the same category of emotion. Train. ‧ 國. 學. and track are related due to their high co-occurrence in the same schema of train. ‧. transportation. Synonyms, although some argue that there are no real synonyms,. Nat. io. sit. y. formally refer to the same concept and therefore can be regarded as the most. er. semantically related word pair/group. In the two studies (Schotter, 2013; Schotter et. al. n. v i n C hsemantically related, al., 2015), identical, synonymous, e n g c h i U and unrelated previews (e.g.. begin – start – ready – check) were used, with different degrees of constraint in pre-context. In Schotter (2013), she not only found the synonym preview benefit, but also replicated previous results that non-synonymous semantic-related previews yield no preview benefit. In Schotter et al. (2015), however, they found that merely semantic-related previews provide preview benefit when the target was contextually predictable. 22.

(38) However, not all the alphabetic languages lack or have only elusive semantic preview benefit. Hohenstein, Laubrock, and Kliegl (2010) found semantic preview benefit in German with fast-priming paradigm, where semantic un-/related preview at target word n was presented only during a short time window since word n-1 was fixated. After the limited preview time window, the parafoveal region was once again replaced by the target word. This then was the first study that demonstrated. 政治大. facilitation by previewing a semantic related word in alphabetic writing system. Later. 立. on, Hohenstein and Kliegl (2014) confirmed this results by using normal boundary. ‧ 國. 學. paradigm without resorting to synonyms (e.g. Riese “giant” with semantic related. ‧. preview Zwerg “dwarf”). Furthermore, Korean, an alphabetic writing system with one. Nat. io. sit. y. syllable written in one square-like box, also exhibits an effect similar to semantic. er. preview benefit. Kim, Radach, and Vorstius (2012) manipulated the consistency of. al. n. v i n C h and semantic relations case marker, which indicate syntactic in Korean. They found engchi U. an effect of correct and incorrect case marker in Korean. Although it was not an ordinary boundary paradigm experiment, where content words are used, it demonstrated that the syntactic function of the preview, which is of neither orthographic nor phonological nature, can be extracted from the preview. To sum up, for alphabetic languages, English and Finnish have no semantic preview benefit. This may also include Spanish, as inducted from the lack of semantic 23.

(39) preview benefit in the bilingual study by Altarriba et al. (2001). On the other hand in German and Korean, various kinds of semantic preview benefit have been demonstrated. An explanation has been proposed (see Schotter, 2013 among others) to reconcile with the discrepancies. It has been argued that since English has a more irregular spelling, which adopts a loose correspondence between the orthography and phonology of the language, more resources are dedicated to process orthographic. 政治大. information to obtain phonological representation during lexical processing, including. 立. previewing. German, on the other hand, has a more rigid spelling system and thus. ‧ 國. 學. more resources are available for semantic preprocessing. Such an explanation cannot. ‧. explain the findings in Finnish and Spanish, since they also adopt rigid spelling. Nat. io. sit. y. system. However, it is tempting for such studies in Chinese, since semantic access in. er. Chinese, as argued above, does not require the phonological stage, which spares more. al. n. v i n C h processing. The resources for the follow-up semantic e n g c h i U following section will then review the findings for preview processing in Chinese.. 2.3 2.3.1. Parafoveal processing in Chinese Scope of effects As described above, Chinese scripts can be decomposed into different levels of. unit. Sentences are composed of words, words of characters, characters possibly of 24.

(40) radicals, and radicals of strokes. Except for the characters in monomorphemic words, which do not have their own character meaning, most of the characters have its own meaning and can stand alone as single-character word or serves as morpheme in multi-character words. Take a two-character word 可口 /ké kŏu/ “tasty” for example, each character 可 /kĕ/ “worthy” and 口 /kŏu/ “mouth” has its own meaning related to but distinct from the compound. And although orthographically a compound is. 政治大. identical to the combination of its constituents, there were experiments suggesting a. 立. distinct representation of compound word other than those of its constituents (e.g. 丁. ‧ 國. 學. 国盛 & 彭聃龄, 2006). When considering single-character word, these properties of. ‧. character level and of word level are then overlapped. Therefore, when considering. Nat. io. sit. y. which property of the preview is extracted, the scope of the property has to be. er. well-defined. While there may be other studies that defined “lexical” property at. al. n. v i n C hat word level in this character level, the term is defined e n g c h i U study and is thus applied to the property of two-character or single-character word as a whole but not constituent character of the two-character compounds. With clear definition of level of effects, the following section will introduce the preview benefits found in reading Chinese.. 2.3.2. Preview benefit in reading Chinese What seems to be universal to all the writing systems of the world are 25.

(41) phonological and orthographic preview benefit, as noted above. These effects have been found in English and other languages. Chinese is no exception. Phonological and orthographic preview benefit has been found (Liu et al., 2002; Tsai et al., 2004). As for the controversial semantic preview benefit, since Chinese is believed to have a more direct link from orthography to semantic. It is predicted that semantic preview benefit should be easier to observe in Chinese. The situation, however, becomes. 政治大. complicated when Chinese is being discussed, since both character and word can have. 立. its own meaning, plus the morphemic status of constituent character in multi-character. ‧. ‧ 國. 學. words.. Most of the studies so far have focused on preview of single character. Yan et al.. Nat. io. sit. y. (2009) and Tsai et al. (2012) selected integrated characters (i.e. characters that cannot. n. al. previews respectively in. er. be further decomposed into radicals, for example 羊 “sheep”) as their targets and. v i n C h and traditional simplified e n g c h i U Chinese.. Their studies. demonstrated semantic preview benefit along with orthography and phonology preview benefit in Chinese. Their findings, however, were character-based since their target characters were embedded as the first character of two-character compounds. Such a morpheme preview benefit is further confirmed in the second experiment of Yen, Tsai, Tzeng, and Hung (2008). The experiment used homographic single-character morphemes that can only be disambiguated in the context of matrix 26.

(42) compound. For an example in their design, the target word 戒菸 (“quit” and “smoking”) “to quit smoking” would have preview with the same morpheme such as 戒除 (“quit” and “get rid of”) “to give up a hobby”, or the one with a different morpheme 戒備 (“guard against” and “prepare”) “to guard against”. Their results showed that readers could disambiguate the morphemes when there was enough preview time. As for word-level semantic preview benefit, Yang et al. (2012) used. 政治大. single characters as their targets, but they were single-character words when. 立. embedded in the sentence frame, such as the experiment sentence in their study. (1) 陳健拎著一箱鞋來到我經營的小店裡。 “Chen carried a box of shoes to the store I’m running.”. ‧. ‧ 國. 學. below:. Nat. io. sit. y. Furthermore, their materials included both integrated and compound characters (i.e.. er. characters that can be further decomposed into sound- or meaning- bearing radicals,. al. n. v i n C h of semantic radicals for example 悲 /bei/ “sad”, composed e n g c h i U 心 “heart” and phonetic radicals 非 /fei/). By doing so, they found semantic preview benefit, but only when the preview fit plausibly into the context. They also found plausibility effect, which was the contrast of semantic-unrelated-implausible preview to the average of semantic-related-plausible and semantic-unrelated-plausible previews. Another study by Yan, Zhou, Shu, and Kliegl (2012) investigated the influence of semantic radical’s influence on semantic preview benefit. They found that transparency of mapping 27.

(43) between the meaning of semantic radical and the meaning of the character can influence the size of semantic preview benefit. Their results, in addition to Tsai et al. (2004), demonstrated that sub-character information, either phonological or semantic, influences preview as well. However, the targets of the study were also characters embedded as the first character of two-character words. On a strict standard, except for Yang et al. (2012), the aforementioned studies were all sub-lexical preview benefit. 政治大. since the benefit came from components of words. The following section will then. 立. introduce the studies investigating into processing of two-character compounds per se.. ‧. ‧ 國. 學. 2.4 Processing of Chinese two-character compounds during reading. Nat. y. Prominence of compounds. io. sit. 2.4.1. er. The reason why studies of normal reading have been focused on character-level. al. n. v i n Cinhwritten Chinese, Uthe spaces are used to separate processing is conceivable, since engchi. characters but not to demarcate words. However, there were studies that showed the special status of two-character words during reading. Tsai, Lee, Lin, Tzeng, and Hung (2006) found effects of word frequency and neighborhood size of Chinese two-character compounds. Words of higher frequency or with more neighbors, which are defined in the study as words that share the first character, were read faster. Furthermore, skipping rate was also higher with words with large neighborhood size. 28.