語言相對論：中文分類詞之事件相關腦電位研究

全文

(1)國立臺灣師範大學英語學系碩士論文 Master’s Thesis Department of English National Taiwan Normal University. 語言相對論：中文分類詞之事件相關腦電位研究 Linguistic Relativity: An ERP Study on Mandarin Classifiers. 指導教授：詹曉蕙博士 Advisor: Dr. Shiao-Hui Chan 研究生：胡世強 Student: Shih-Chiang Hu. 中華民國一百零八年六月 June 2019.

(2) 摘. 要. 語言相對論（linguistic relativity，又稱薩丕爾-沃夫假說，Sapir-Whorf hypothesis）主張語言結構會影響語言使用者的世界觀，不同語言的使用者也因此擁有不同的思考模式。語言相對論的相關文獻並未呈現一致的研究結果，而這樣的不一致可歸因於行為實驗研究方法的侷限性。即便是支持語言相對論的實證研究也常因使用行為實驗而備受詬病，因為那些“無趣老套”的實驗發現通常都是在有語言干擾─可能是語言的實驗刺激或是受試者的內在獨白（internal monologue）─的情況下產生（如 Pinker 2007 [1994]）。 Thierry (2016) 則建議，若要檢驗此假說，研究語言相對論的學者應該測量構基於無意識感知處理的自發大腦活動。本研究旨在利用事件相關腦電位（event-related potentials，ERP）探討在不使用語言的情形下，中文母語者的視覺感知是否仍會受中文分類詞影響。受試者為中文母語者與英文母語者各十四名。實驗採用特異刺激典範（oddball paradigm），而受試者的作業則是在察覺目標刺激時按鍵。實驗刺激分成標準刺激（四種使用相同分類詞的物體照片，出現概率 80%）、特異刺激（一種使用不同於標準刺激分類詞的物品照片，出現概率 15%）與目標刺激（一張貓的照片，出現概率 5%）。根據分類詞的分類，特異刺激與標準刺激的差異可能造成維度內違反（within-dimension deviant），或維度間違反（betweendimension deviant）。實驗結果發現，與標準刺激相比，只有中文母語者的維度內違反特異刺激誘發了顯著的 P2 效應（150-250 ms），顯示出分類詞影響了中文母語者的物體概念分類與感知。本研究結果異於前人的相關研究，因前人多以語言作為實驗刺激，且發現分類詞的影響只存在於非快速的（unspeeded）實驗作業中（Huang & Chen 2014; Imai & Saalbach 2010; Saalbach & Imai 2007）。雖然分類詞的效果非常精細，但這樣的效果揭示了在沒有語言信息與語言處理的情況下，語言仍影響著其他非語言的認知功能。. 關鍵詞：語言相對論，中文，分類詞，視覺感知，特異刺激作業，事件相關腦電位，P2. i.

(3) ABSTRACT Linguistic relativity, also known as Sapir-Whorf hypothesis, upholds that the structure of a language influences its speakers’ worldview and speakers of different languages might therefore think differently. Research on linguistic relativity has obtained inconsistent results, which might be due to the methodological limitation of behavioral studies. Moreover, if adopting behavioral methods, empirical studies yielding positive results were usually castigated for showing ‘banal’ findings resulting from the contamination of explicit verbal cues or inevitable internal monologue during the experimental tasks (e.g., Pinker 2007 [1994]). As suggested by Thierry (2016), relativists are encouraged to test the hypothesis by measuring spontaneous brain activities underlying the unconscious perceptual processes. Using the event-related potentials (ERP) technique, this study aimed to explore whether Mandarin Chinese speakers’ visual perception would be modulated by the Mandarin classifier system. Fourteen native speakers of Chinese (CS) and fourteen native speakers of English (ES) participated in the experiment. Participants were instructed to look attentively at the stimuli and to perform a target detection task in an oddball paradigm. The standard stimuli (probability 80%) were photos of four (repeated) objects sharing the same classifier, and the deviant stimuli (probability 15%) were photos of one object not categorized by the standards’ classifier, which were either dimensionally congruent (within-dimension deviant) or dimensionally incongruent (between-dimension deviant) with the standards in terms of classifier categories. The targets were photos of a cat (probability 5%). The results revealed that when compared with standards, within-dimension deviants induced a significant P2 effect (150-250 ms) only in the Chinese group, suggesting that classifiers affected Chinese speakers’ object conceptual categorization and perception. Our findings were qualitatively different from previous research on classifier relativity where classifier effects only appeared in unspeeded tasks with verbal cues (Huang & Chen 2014; Imai & Saalbach 2010; Saalbach & Imai 2007). Although the classifier effect was intricate, it demonstrated that language affects non-verbal cognitive systems, even when linguistic processing is not needed.. Keywords: linguistic relativity, Mandarin Chinese, classifier, visual perception, oddball paradigm, ERP, P2. ii.

(4) ACKNOWLEDGMENT 我覺得完成一篇碩論絕對是要付出很多心力，但沒有這麼偉大啦。寫論文與做實驗的甘苦談細節是不會看到了，但該有的 credit 還是要說一下，所以在這主要感謝在論文研究與整個研究所求學階段幫助我的老師與朋友。感謝曉蕙老師擔任我的指導教授。曉蕙老師給學生最大的空間探索自己的研究興趣。能在學術研究上做自己喜歡的主題，而且老師還願意指導，似乎是得來不易的幸福（？）老師總是用盡全力幫助自己的學生。很多時候研究碰到了死結，老師不厭其煩地與我一再討論，找到解決問題的方法後甚至比我還興奮（Yeah!!!）。除了老師真正地熱愛研究之外，我知道老師是想振奮死氣沉沉的我，讓我對自己的研究有信心。老師的專業與人格特質是天生的學者與師者，是我學習的模範！感謝葉素玲老師與陳振宇老師擔任我的論文口試委員。雖然是口試委員，但兩位老師對於我的論文有如指導教授般地投入。從提論文大綱到最後論文口試，兩位老師不吝提出許多專業的建議，讓我的研究細節更有條理，也讓我用更宏觀的視野去看自己的研究。論文口試還歷時兩個多小時，宛如一場完整精實的 seminar。如此近距離感受老師們對研究的嚴謹、熱情與想像，這種經驗彌足珍貴。感謝碩班教導過我的老師們，Lindsey 老師、丁仁老師、妙玲老師、妙霞老師、俐馨老師、席瑤老師、純音老師、甄儀老師、颯楊老師、曉虹老師、曉蕙老師、蕙珊老師、靜蘭老師。我有現在一點點的知識儲備與能力，都是因為有老師們課上與課後的教導。還要特別感謝以下幾位老師的幫助。感謝丁仁老師在我休學服替代役的期間特別寫信給我，給我繼續完成碩班的信心。感謝妙霞老師從我大學時期就一直特別照顧我，儘管碩一表現讓老師失望，老師非常理解我，還好幾次打電話鼓勵我。感謝席瑤老師，和老師非常有緣份，除了修過老師四門課，大學也當老師助理，跑遍台北各大旅宿業與觀光景點，學到很多社語田調訪問的技巧，2017 年還奇妙地和老師同台演出。很可惜遇上老師出國一年，否則第一個（應該）合唱社會語言學的研究可能就會誕生了。論文發表若能搭配師生合唱應該會很酷。感謝純音老師，要不是老師簽了休學單又撕了休學單、抓著我在丹堤把修課行程搞定、督促我一學期狂修五門課、鞭策我 “just let go”、安排許多樁腳在我旁邊撐我，我也不會現在就能畢業。感謝甄儀老師一直很關心我的求學狀況，也給我許多建議。大學語概、碩班語通都是上老師的課，很多語言學的基本概念都是老師教給我的。感謝蕙珊老師，因為老師的推薦與指導，我才能順利得到科技部補助，前往香港研討會發表音韻學論文。感謝靜蘭老師在我對研究產出迷惘的時候，以亦師亦友的態度分享了很多自身的經驗，教會我放下心裡的不安，也包容我的拖延與舉棋不定。感謝師大華語文教學系的孫懿芬老師。大學修華語文教學學程而認識老師，老師 iii.

(5) 是我重要的語言教學與人生導師之一。我的研究因參與實驗的條件嚴格，英文母語受試者格外難尋，超過一半的受試者是幸得老師大力相助，動用各種人脈關係才收到，最後終於擠出足夠可分析的人數。讓老師費心了。感謝師大神經語言學實驗室的夥伴：理克 Aymeric、定武 Terry、宜琪 Vicky、張哲 Ronald，還有實驗室的前輩：家萱 Gracie、耿育 Ken、李全 Matt（現在全部在美國讀博士，望塵莫及）。有你們的協助、意見還有陪伴，我才能順利地完成論文。特別感謝 lab manager 理克帶我投營曉蕙老師的團隊，在招募受試者、實驗進行、資料分析處理、統計，還有各～～方各～～面的大～～力相助，沒有你我早就退學八百次了。特別感謝 Ken 從無到有教我設定電極帽、收集腦波資料，你是我 EEG 實戰的啟蒙老師。特別感謝家萱學姐傳授我搞定腦波的無上功力，任何受試者的腦波都會聽你的話，乖乖的。感謝我從大學到碩班的同學―羽辰 Eliza，我們都是在實習後才復學，可以一起修課、討論、抱怨真的很棒。在我每次要放棄的時候都有你為我煞車。我在泰國服役時，你也叫我早早準備研討會論文發表，我才能在剛復學的那個學期就解決一大障礙。感謝一樣是大學到碩班的同學―淞筌 Nick，雖然不常一起，但我們都彼此支持。感謝我從大學到碩班的超強學弟妹―惠真 Stephanie、堡升 Mark、洪寬 Howard。因為我先去服役的關係，跟你們還比較像同學，一起修了好多課、做了好多報告。身為學長，我沒有拉著你們，反而是你們拉著我，非常感謝你們！還要感謝韋伶 Eileen 和前面三位一起陪伴我，每次看到你都笑嘻嘻。感謝健民 Kelvinn 總不忘記我，通知我系上的大小事，超級暖心就是你。以上的人都比我早畢業不知道多久，在各領域發光發熱了，優秀！感謝江妍 Anita，從我還是小大一你是大三學姐擔任話劇比賽主持人的時候就對你印象深刻，沒想到因為你回師大讀博班而建立起關係，而且這麼合（應該），還分享了那麼多研究與非研究的事情。你也是我的標竿！感謝幾個大學同學。感謝孟憲 Neal，無酬幫我校對論文看到眼睛瞎，然後一起說貶低自己、奉承對方的屁話（但我是實話）。感謝小群組于芳 Ashley、亭寧 Iris、涓如 Isabell、趙瑩 Lily，除了討論教學、研究之外，還要一起砲轟教育界、砲轟學術界，砲轟這個世界，轟轟轟！感謝慕涵助教耐心仔細地處理研究生的各種教務工作，每次都被我煩還是很溫柔地回答我、幫我解決，讓我不用擔心繁瑣的行政問題。還要感謝豪谷助教和羽立助教，碰到好的助教不容易，好的助教帶你上天堂。最後要感謝幫助我做前測、找受試者的朋友與同學，也感謝我所有的受試者。實證研究靠資料說話，是你們讓我有話說。話說完，好畢業。. iv.

(6) TABLE OF CONTENTS. 摘. 要......................................................................................................................................... i . ABSTRACT...............................................................................................................................ii ACKNOWLEDGMENT.......................................................................................................... iii TABLE OF CONTENTS ........................................................................................................... v LIST OF TABLES .................................................................................................................... vi LIST OF FIGURES .................................................................................................................vii 1 Introduction ............................................................................................................................. 1 1.1 Linguistic relativity ...................................................................................................... 1 1.1.1 Neurolinguistic relativity .................................................................................. 7 1.2 Mandarin Chinese as a classifier language ................................................................ 12 1.3 Research question ...................................................................................................... 18 2 Methodology ......................................................................................................................... 19 2.1 Participants................................................................................................................. 19 2.2 Materials .................................................................................................................... 19 2.3 Procedure ................................................................................................................... 22 2.4 Behavioral and EEG recordings................................................................................. 24 2.5 Data Analysis ............................................................................................................. 25 3 Results ................................................................................................................................... 28 3.1 Behavioral data .......................................................................................................... 28 3.2 ERP data ..................................................................................................................... 28 4 Discussion ............................................................................................................................. 46 5 Conclusion ............................................................................................................................ 53 References ................................................................................................................................ 54 Appendix A. Experimental stimuli .......................................................................................... 70 Appendix B. Results of the pretest and posttest on experimental stimuli ............................... 71 Appendix C. Instructions for the ERP experiment .................................................................. 72 Appendix D. The questionnaire for the pretest on experimental stimuli ................................. 74 . v.

(7) LIST OF TABLES. Table 1. Tests for C/M distinction (adapted from Her & Lin 2015:59-60) ............................. 17 Table 2. Experimental conditions in a sample list ................................................................... 21 Table 3. Numerical facts of stimuli ......................................................................................... 22 Table 4. Summary of the degrees of freedom, F values, and p values of repeated measures ANOVAs for the midline and laterality analyses of the P2 effect (150-250 ms) ...... 40 Table 5. Summary of the degrees of freedom, F values, and p values of repeated measures ANOVAs for the left and right analyses of the P2 effect (150-250 ms) .................... 42 Table 6. Summary of the degrees of freedom, F values, and p values of repeated measures ANOVAs for the right CS and right ES analyses of the P2 effect (150-250 ms) ...... 44 . vi.

(8) LIST OF FIGURES. Figure 1. Classes and subclasses of hypotheses on how language might affect thought (Wolff & Holmes 2011:254). ................................................................................................ 2 Figure 2. Sample of Thierry et al.’s (2009) experimental stimuli ............................................. 8 Figure 3. Sample of one block ................................................................................................ 24 Figure 4. ERP responses elicited by targets, standards, and deviants across deviant types, dimensions, and language groups. ........................................................................... 29 Figure 5. ERP responses to standards, within- and between-dimension deviants in the Chinese group. ......................................................................................................... 30 Figure 6. ERP responses to standards, within- and between-dimension deviants in the English group. ....................................................................................................................... 31 Figure 7. Grand averaged ERP responses to standards and within-dimension deviants......... 33 Figure 8. Grand averaged ERP responses to standards and between-dimension deviants. .... 34 Figure 9. Difference waves of within-/between-dimension deviants minus standards. .......... 36 Figure 10. Topographic maps of the deviancy effects (deviant minus standard) (P2 effect: 150-250 ms). ............................................................................................................ 37 . vii.

(9) 1 Introduction 1.1 Linguistic relativity From the 1950s onward, research on the language-thought interface has been thriving. Whorf’s (1956) discussion on language and thought, albeit not exactly the first one, serves as a seminal work in the field of linguistic relativity. In crude terms, linguistic relativity, also known as Whorfian hypothesis or Sapir-Whorf hypothesis, concerns whether speakers of different languages think differently. To be more specific, the immense diversity in linguistic expression across languages can result in alternative ways of perceiving the world as a consequence of the constant and consistent exposure to and repetitive use of a particular language, which direct its speakers’ attention to certain aspects of experience and therefore reinforce their sensitivity to certain dimensions. After decades of exploration and disputation, issues pertaining to linguistic relativity have gone beyond just seeking yes-or-no answers; instead, this contentious and unsettled hypothesis sustains researchers’ interest in the degree of the linguistic effect on thought and in the way language and thought interact. Although the strongest version of linguistic relativity—namely, linguistic determinism— is unequivocally rejected on empirical grounds (Ameel et al. 2005; Gennari et al. 2002; Malt et al. 1999; Munnich, Landau & Dosher 2001; Papafragou, Massey & Gleitman 2006), the moderate presumption that language influences one’s worldview has never been abandoned. A vast body of empirical research demonstrated that language has a profound effect on thought. 1.

(10) in various ways indeed; nonetheless, the strong-weak continuum is no longer sufficient to illustrate linguistic relativity in different contexts. As Wolff and Holmes (2011:253) stated in their review article, “Linguistic relativity can now be said to comprise a ‘family’ of related proposals that do not necessarily fall along a single strong-to-weak continuum.” The authors, in turn, provided a family tree, which well summarizes all the related proposals supported by some previous studies (see Figure 1).. Language affects thought. Thought is language. (Language as language-of-thought). Thought is separate from language.. Thought and language are structurally parallel. (Linguistic determinism). Thinking before language (Thinking for speaking). Linguistic representations compete with nonlinguistic representations. (Language as meddler). Thought and language differ structurally.. Thinking with language. Linguistic representations extend/enable nonlinguistic representations. (Language as augmenter). Thinking after language. Language makes certain properties highly salient in nonlinguistic thinking. (Language as spotlight). Language primes certain types of processing in nonlinguistic thinking. (Language as inducer). Figure 1. Classes and subclasses of hypotheses on how language might affect thought (Wolff & Holmes 2011:254).. With the agreement that language affects thought, there are two major lines of reasoning: one vs. two; that is, language and thought can be considered either one single entity or two. 2.

(11) distinct systems. If thought is equated with language, it is immediately self-evident that thought varies across languages (language as language of thought). This line is generally refuted, since it is not always effortless to express oneself precisely. In other words, thoughts cannot be entirely represented in natural language; otherwise, troubles like “I don’t know how to put it” would never come. Another argument against absolute identity between language and thought is that people understand linguistically ambiguous expressions, such as Time flies like an arrow; fruit flies like a banana. The ability to recognize and understand linguistic ambiguity suggests a level of representation separate from the linguistically encoded one. If thought is language, words expressing new concepts are not possible either because people can only think in words originally existing in their languages (Pinker 2007 [1994]). Moreover, many a study showed that sophisticated forms of cognitive processing could be observed in infants and nonhuman primates lacking the resource of language (Bloom & Keil 2001; Cheney & Seyfarth 1990; Hare, Call & Tomasello 2001; Hespos & Spelke 2004; Needham & Baillargeon 1997; Penn, Holyoak & Povinelli 2008; Phillips & Santos 2007; Spelke 2003). All the aforementioned arguments and findings lead to the conclusion that thought is predominantly autonomous, not language. If language is detached from the conceptual system, the shaping roles of language could be manifold. Different, but not necessarily contradictory, versions of linguistic relativity have been suggested in many empirical studies: (1) linguistic determinism: language might determine the basic categories of thought and blind its speakers to everything except the. 3.

(12) concepts that are encoded in language; (2) language for speaking: the type of thinking immediately prior to language production might be influenced by language, since speakers must attend to certain language-specific aspects before they speak (e.g., Gennari et al. 2002; Papafragou, Hulbert & Trueswell 2008; Slobin 1996); (3) language as meddler: language might facilitate or impede thinking, depending on whether linguistic and nonlinguistic codes are consistent or in conflict with each other during nonlinguistic processes where linguistic processes are activated along (e.g., Davidoff, Davies & Roberson 1999; Gilbert et al. 2006; Papafragou, Hulbert & Trueswell 2008; Roberson et al. 2005; Roberson, Davies & Davidoff 2000; Roberson & Hanley 2010; Winawer et al. 2007); (4) language as augmenter: linguistic representations serve as an additional conceptual kit, without which people would have grave difficulty in accomplishing certain tasks of reasoning, analogy, and categorization (e.g., Christie & Gentner 2012; de Villiers & de Villiers 2003; Dehaene et al. 1999; Frank et al. 2008; Gentner 2003; Gordon 2010; Loewenstein & Gentner 2005; Milligan, Astington & Dack 2007; Pyers & Senghas 2009; Waxman & Markow 1995); (5) language as spotlight: the long-term use of and exposure to a language, where certain aspects of the world are accentuated in the words and constructions, makes its speakers’ perception more sensitive to those highlighted properties (e.g., Boroditsky, Schmidt & Phillips 2003; Gallistel 2002; Haun et al. 2006; Imai & Gentner 1997; Imai & Mazuka 2003; Levinson et al. 2002; Levinson 1999; Newcombe 2005; Vigliocco et al. 2005); and (6) language as inducer: the long-term use of and exposure to a. 4.

(13) language primes a particular mode of conceptual processing, which might persist, even without linguistic codes being recruited (e.g., Holmes & Wolff 2010). However, many skeptics deprecate the attainment of research on linguistic relativity mentioned above. For instance, Pinker (2007 [1994]) declared that it is hardly surprising that some of the experiments which tested “weak” versions of the Whorfian hypothesis actually worked, and considered their findings banal. What is the bone of contention? Most studies were based on behavioral measures, and thus were suspected of drawing false conclusions from verbally contaminated responses resulting from silently verbalizing all along the supposedly nonverbal tasks. If diverse worldviews or concepts are not fundamentally due to differently structured minds of speakers of different languages, but rather are just somehow displayed in speakers’ behavior nudged by temporary online processing of grammar and terminology in their languages, then that is nothing but a predictable plain story (at least to the skeptics). Up to this point, it seems that linguistic determinism, on the one hand, is too strong a claim to be true, and that the weaker claims, on the other hand, are too obvious to be intellectually fascinating and inspiring enough. More importantly, results of behavioral studies, though suggesting the versions of linguistic relativity in nonlinguistic context (i.e., language as spotlight and language as inducer), fail to convince the opposing side because of the inherent and fatal defect in their experimental design—incapacity to avoid verbal reference entirely. Research on linguistic relativity must get out of this predicament, or just stops here.. 5.

(14) Encouraging news is that neuroscientific studies in the last decade have broken the deadlock. The game is not over yet. Get back a bit to the language-thought relation. Although it might not be true that language is thought, language and thought are intrinsically bound together from the point of view of neurophysiology and cognitive neuroscience. Making a distinction between language and other cognitive faculties implies a modular human brain where independent brain areas subserve domain-specific cognitive functions (Thierry 2016). This implication, however, is essentially contrary to connectionism, which believes that the human brain actually works as a network consisting of interlinked functional subnetworks (Barsalou 2008; Pulvermüller 1999), and the evidence of language-specific regions in the human brain has little been corroborated by anatomical and neurophysiological studies (e.g., Démonet, Thierry & Cardebat 2005; Price, Thierry & Griffiths 2005). Besides, ample empirical evidence did point to interaction between language and other perceptual-cognitive processes, such as decision-making, emotion, or inhibitory control (Costa et al. 2014; Wu & Thierry 2012; Wu & Thierry 2013; among others). With the convergent findings from research of neurophysiology and cognitive neuroscience demonstrating a strong language-thought binding contingency, relativists can be fairly positive that language inevitably influences thought. But how? As urged by Imai and Saalbach (2010), Thierry (2016), and many other recent relativist researchers, research on linguistic relativity has to depart from proving the existence and establish the nature as well as the conditions of. 6.

(15) language effects on thought. In order to test potential linguistic effects on cognition (again, not the ones accused of being verbally contaminated in behavioral studies), researchers should aim to investigate unconscious, automatic, and sufficiently early phases of mental processing, the phases that are not subject to the direct influence of online language processing. “There is thus no alternative but to test the hypothesis using physiological correlates of perception that derive from brain activity” (Thierry 2016:693). In the following section, several neurolinguistic studies are presented to show how neuroscientific methods effectively tackle the Whorfian question.. 1.1.1 Neurolinguistic relativity To eschew possible recruitment of linguistic codes during nonverbal tasks, one should turn to information of neural activity reflecting unconscious mental processing of nonverbal stimuli. Thierry and colleagues (2009) recorded event-related brain potentials in English and Greek native speakers when the participants were presented with a stream of shapes one by one and were asked to discriminate, by button pressing, infrequent squares (target, probability 20%) from regular circles (probability 80%), in which most of the circles are light blue (standard, probability 70%) and the rest are dark blue (deviant, probability 10%), as illustrated in Figure 2 (or the other way around; i.e., more dark blue circles and fewer light blue ones). This design was to divert participants’ attention from color deviancy. Greek differentiates light and dark. 7.

(16) shades of blue in its basic color terms (ghalazio and ble), but English does not. This terminological distinction led to a visual mismatch negativity (vMMN) effect greater in the Greek participants than in the English participants, but the greater deviancy effect was not observed for green (the control condition) because both Greek and English have only one basic color term for green.. standard. target. deviant. time. Figure 2. Sample of Thierry et al.’s (2009) experimental stimuli. Usually expected in an oddball paradigm, the MMN is an automatic and unconscious brain response elicited by perceptual deviant stimuli, even the stimuli outside of the focus of attention (Czigler 2014; Näätänen, Kujala & Winkler 2011). In Thierry et al. (2009), this preattentive processing started before 200 milliseconds, earlier than the online access to specific lexical information reported in the literature (Costa et al. 2009; Strijkers, Costa & Thierry 2010; Strijkers, Holcomb & Costa 2011). In other words, it could be forcefully argued that there was minimal chance of verbal contamination of the participants’ responses to the critical stimuli. The findings suggested that repeated exposure to language-specific categorization (i.e., lexical distinction in color) affects perceptual distinction (i.e., color hue discrimination). Put simply, speakers of different languages perceive the world differently. 8.

(17) In a follow-up study (Athanasopoulos et al. 2010), the Greek participants were regrouped according to different lengths of stay in the UK. A markedly reduced vMMN effect was observed in the long-stay Greek participants, indicating that the long-stay Greek participants perceived light and dark blues as more similar than the short-stay Greek participants. Different lengths of exposure to English, where there is no blue contrast, shaped the Greek speakers’ perception to different extents. The results of the follow-up study not only consistently supported linguistic relativity but also suggested that the shaping effect of language on thought is eminently plastic in nature. Following the investigations in the domain of color, Boutonnet and colleagues (2013) tested Whorfian hypothesis in the domain of object categorization. In English, cups and mugs are two objects terminologically differentiated while they are both called taza in Spanish. With an experimental design similar to that in Thierry et al.’s (2009) color study, the participants were asked to detect target pictures of bowls in a sequence of standard and infrequent deviant pictures of cups and mugs. As expected, the deviants elicited a deviant-related negativity (DRN) of greater magnitude in the English speakers than in the Spanish speakers. The DRN is the earliest modulation of the negative component closely comparable to vMMN (Csibra, Czigler & Ambrò 1994; Czigler, Balázs & Pató 2004; Czigler, Balazs & Winkler 2002; István Winkler et al. 2005; Turatto et al. 2002). Peaking around 160 milliseconds and substantially earlier than N2 components elicited by overt cognitive control (Folstein & Van Petten 2007), the identified. 9.

(18) DRN effect served as an index of automatic, pre-attentive, and pre-lexical cognitive mechanism (Costa et al. 2009; Strijkers, Costa & Thierry 2010; Strijkers, Holcomb & Costa 2011). The findings established an effect of language on high-level perceptual processing, the effect that should not be nullified by potential verbal interference or top-down strategies. Interestingly, the P1 peak of ERPs around 100-150 milliseconds, elicited by the contrast of cups and mugs, was of similar amplitude in two language groups. The P1 component was associated with early discrimination of object categories (Dering et al. 2011; Thierry et al. 2007). The insignificant P1 difference, coupled with the DRN results presented above, revealed that the participants of both language groups did perceive a cup and a mug as different objects (similar P1 amplitude), but those who use two terms to specify the two objects seemed to enjoy the privilege of distinguishing the two objects in a more spontaneous and effortless fashion than those who use only one term (greater DRN in the English speakers). The obtained ERP results not only captured the fact that speakers of different languages would not be blinded by language-specific terminologies, but meanwhile demonstrated linguistic relativity effects. Like terminology, grammar can modulate one’s unconscious cognition as well. In the study by Boutonnet, Athanasopoulos, and Thierry (2012), the participants (native English speakers and Spanish-English bilinguals) were presented with triplets of pictures, and asked to judge whether the semantic category of the third picture of a triplet is the same as that of the first two pictures by pressing buttons. What the participants were not informed of was that the. 10.

(19) grammatical gender of the third picture name in Spanish was incongruent with that of the first two in half of the trials. Through the recorded ERPs along the tasks, the authors found a semantic priming effect (N400) in both groups and a negative modulation (left anterior negativity; LAN) by gender inconsistency exclusively in the Spanish–English bilinguals. The LAN, an index of morphosyntactic processing (i.e., grammatical gender in question) (Friederici & Jacobsen 1999; Friederici, Pfeifer & Hahne 1993; Hahne & Friederici 1999; Thierry, Cardebat & Démonet 2003), was elicited in an all-in-English semantic categorization task where gender information was not required and signaled; moreover, awareness of gender manipulation was never reported by the participants. Hence, the retrieval of task-irrelevant gender information was argued to be unconscious and implicit (Thierry & Wu 2007; Strijkers, Holcomb & Costa 2011; Wu & Thierry 2010). The results suggested that conceptual object categorization could be affected by grammatically encoded information on an unconscious level, which corroborated linguistic relativity. To sum up, the effective and convincing neurolinguistic results restore confidence in the research on linguistic relativity. The ERP components (e.g., vMMN, DRN, and LAN), unlikely to be contaminated by online language processing as in behavioral experiments, are solid neurophysiological evidence to show that language influences thought. That being the case, speakers of different languages might have differently structured minds. Language does not determine principal categories of thought, but the universality of human conceptual system. 11.

(20) does not preclude the possibility of fundamental language effects on cognition in general, either.. 1.2 Mandarin Chinese as a classifier language A classifier language like Mandarin Chinese requires classifiers for almost all nouns when quantifying them, whereas a non-classifier language like English requires a unit of quantification only for mass nouns (M. Hsieh 2008). Consider the following examples in (1):. (1) a.. 一. *(顆). yi *(ke) one CL ‘one apple’. b. 兩. 蘋果 pingguo apple. *(本). liang *(ben) two CL ‘two books’. 書 shu book. The examples above demonstrate the mandatory status of classifiers in a numeral phrase in a classifier language. The numeral phrases in the absence of classifiers are ungrammatical in Mandarin Chinese while the mere numeral-noun combination works perfectly in English as shown in the corresponding English translations. It is of interest that one classifier can categorize a class of nouns according to some shared salient perceptual properties, which implies a one-to-many correspondence between a classifier and nouns. For instance, the classifier zhang (張) can co-occur with objects like zhi (紙) ‘paper’, chuang (床) ‘bed’, and zui (嘴) ‘mouth’. The objects, by intuition, are not semantically related to each other in many aspects. Even so, since the objects categorized by the same classifier. 12.

(21) zhang (張) denoting a flat surface form a single class, they might be stored ‘closer’ to each other than to the other nouns in Chinese speakers’ mental lexicon. Indeed, some empirical studies based on behavioral methods have revealed that classifier categories affect perception of objects. Schmitt and Zhang (1998), Saalbach and Imai (2007), Imai and Saalbach (2010), and Huang and Chen (2014) among others showed that native Chinese speakers in some tasks, compared to native speakers of non-classifier languages (English and German), had a magnified, though minuscule, sensitivity to the properties constituting classifier classes in Chinese and therefore perceived the objects sharing the same classifier as more similar. Examined closely, classifiers can be put into categories, such as shape and (in)animacy (Gao & Malt 2009; Tai 1994). Each category is further divided into subcategories. For example, the subgroups of ‘shape’ concerns the number of dimensions of an object (e.g., zhi (支) and gen (根) are one-dimensional, zhang (張) and mian (面) are two-dimensional, and ke (顆) and li (粒) are three-dimensional). Hence, not only the objects sharing the same classifier but also the objects whose classifiers belong to the same classifier (sub)category might be perceived as more semantically associated owing to the cognitive-based linguistic categorization of objects (i.e., classifier system). Note that the indispensable role of classifiers cannot fully define Mandarin Chinese as a classifier language. In fact, classifiers (hereafter C) should be distinguished from measure words (hereafter M). M, also massifiers and mensural classifiers, to name a few, is universal 13.

(22) across all natural languages. On the contrary, C, also individual classifiers and sortal classifiers, is a unique part of classifier languages. It is C that sets classifier languages apart (Her & Hsieh 2010). The frequently quoted passage below from Tai and Wang (1990:38) generally characterizes the distinction:. “A classifier categorizes a class of nouns by picking out some salient perceptual properties, either physically or functionally based, which are permanently associated with entities named by the class of nouns; a measure word does not categorize but denotes the quantity of the entity named by noun.”. Making a precise distinction between C and M is especially important in the current study for several reasons. First, the use of C is a language-specific phenomenon in a small number of classifier languages, and this idiosyncrasy, which might flex human perception, is of particular interest to research on linguistic relativity. Second, C, rather than M, is related to language-specific categorization of objects, which might lead to distinct conceptual representations. If M is mistaken for C when experimental materials are created, the results will be distorted and thus invalid. The distinction from Tai and Wang (1990) is more descriptive without explicit tests being offered to help distinguish C and M accurately. In order to devise appropriate experimental stimuli, the current study adopts the tests for C/M distinction from 14.

(23) Her and his colleagues. Her and Hsieh (2010) delineated semantic characterization for C and M, and refined previously proposed tests for C/M distinction. They concluded that C is semantically null while M is semantically substantive. C just highlights some inherent property of the noun without contributing additional meaning, whereas M specifies the quantity of the noun and blocks quantification and modification to the noun (e.g., one box(M) of apples ≠ one apple; big apples ≠ big boxes(M) of apples). The semantic characterization and tests for C/M distinction were subsequently endorsed by Her’s (2012) other study which examined the C/M distinction from a mathematical perspective. He argued that both C and M function as a multiplicand in a mathematical sense. For example, the equation for three dozens of apples is 3 ൈ 12 = 36, and 12 is the multiplicand. The difference between C and M lies in the value they have. C’s value is necessarily 1 while M’s is not (i.e., C = 1, M ≠ 1) because M is semantically substantive. If the idea of C/M is “ൈ x”, the C/M distinction can be represented in the formula below (2).. (2) [Num K N] = [Num ൈ x N], where K = C iff x =1, otherwise K = M (Her 2012:1679).. Her and Lin (2015) summarized the ideas and the tests from Her and Hsieh (2010) and Her (2012) (see Table 1), and then investigated and revised the inventory of C in Taiwan Mandarin previously listed by Her and Lai (2012) on the basis of Mandarin Daily Dictionary of Chinese 15.

(24) Classifiers (Huang, Chen & Lai 1997). In sum, Mandarin Chinese is a classifier language where the classifier system organizes objects in a language-specific way. The C/M distinction is meaningful for the current study to investigate whether speakers of a classifier language (Mandarin Chinese in question) have a different perception of the world because of the linguistic categorization. This study relies on the tests, developed on the basis of semantic and mathematical characteristics of C and M, to precisely make the distinction.. 16.

(25) Table 1. Tests for C/M distinction (adapted from Her & Lin 2015:59-60) Test A: [Num C N] = [Num N] ≠ [Num M N] (A1) 一顆蘋果 = (A2) 一蘋果1 yi ke pingguo one C apple ‘one apple’. ≠ (A3)一箱. yi pingguo one apple ‘one apple’. yi xiang pingguo one M-box apple ‘one box of apples’. Test B: [Num C CN/*UN] vs. [Num M CN/UN] (B1) 三顆蘋果/*果汁 (B2) 三 san ke pingguo/*guozhi three C apple/juice ‘three apples/*juice’. 蘋果. 斤. 蘋果/果汁. san jin pingguo/guozhi three M-kilo apple/juice ‘three kilos of apples/juice’. Test C: [Num A-C N] = [Num C A-N] vs. [Num A-M N] ≠ [Num M A-N] (C1) 三大顆蘋果 = (C2) 三顆大蘋果 san da ke pingguo three big C apple ‘three big apples’ (C3) 三大箱蘋果 san three. da xiang big M-box. ≠ (C4). pingguo apple. san ke da pingguo three C big apple ‘three big apples’ 三箱大蘋果 san xiang da pingguo three M-box big apple ‘three boxes of big apples’. ‘three big boxes of apples’. Test D: #[Num A1-C A2-N] vs. [Num A1-M A2-N] (A1 and A2 are antonyms) (D1) #三大顆小蘋果 (D2) 三大箱小 #san three. da ke big C. xiao small. pingguo apple. san da xiang xiao pingguo three big M-box small apple ‘three big boxes of small apples’. Test E: [Num C N] = [Num ge N] ≠ [Num M N] (E1) 一顆蘋果 = (E2)一個蘋果 yi ke pingguo one C apple ‘one apple’ Test F: [yi M/*C de N] (F1) 一箱的. yi xiang de pingguo one M-box DE apple ‘one box of apples’. 1. ≠ (E3)一箱. yi ge pingguo one C apple ‘one apple’ 蘋果. 蘋果. (F2) *一. 蘋果. yi xiang pingguo one M-box apple “one box of apples” 顆. 的. 蘋果. yi ke de pingguo one C DE apple ‘one box of apples’. A numeral phrase without the occurrence of a classifier is marginal usage in modern Mandarin Chinese. 17.

(26) 1.3 Research question This study aimed to explore whether Mandarin classifier system influences Chinese speakers’ perception, and, if yes, how is the effect reflected by neuronal activities? We tackled this question by using the event-related potentials (ERP) technique, hoping to find out if such an influence could be detected at a pre-lexical stage. Our hypothesis was straightforward: if the classifier system does influence how Chinese speakers perceive the world, a language effect should appear in early time windows; otherwise, similar ERP responses should be observed in both Chinese and English speakers.. 18.

(27) 2 Methodology 2.1 Participants Fourteen Mandarin Chinese speakers (CS; 8 females, 20-26 years old, mean age = 22.86, SD = 2.07) and fourteen native English speakers (ES; 10 females, 20-32 years old, mean age = 24.64, SD = 4.38) in Taiwan with normal or corrected-to-normal vision were recruited to participate in the ERP experiment. The participants from both groups were right-handed and had no history of neurological surgery or disorder. They were neither simultaneous bilinguals nor speakers with a high level of proficiency in a second language, since having a high level of competence in another language is likely to affect one’s perception as suggested in the literature previously discussed. More specifically, the participants had never passed the intermediate level of any standardized foreign language proficiency test (CEFR B1/GEPT intermediate level/TOEFL 42/TOEIC 550/TOCFL level 2/HSK level 4, etc.) and they were not fluent in a second language according to their self-report. For the CS group, the participants were born and bred in Taiwan and had not spent more than one year in a foreign country. The experimental protocol was approved by the Research Ethics Office of National Taiwan Normal University. Participants were paid NT$150 per hour as compensation for their time.. 2.2 Materials This study explored linguistic relativity with pictorial stimuli using the oddball paradigm. 19.

(28) We manipulated two factors pertaining to classifier categories: dimension (one-dimensional and two-dimensional) and deviancy type (within- and between-dimension violations) to create four conditions: (1) 1W: One-dimensional standards and Within-dimension deviant, (2) 1B: One-dimensional standards and Between-dimension deviant, (3) 2W: Two-dimensional standards and Within-dimension deviant, and (4) 2B: Two-dimensional standards and Betweendimension deviant (see Table 2). All the standards within the same dimension used the same classifier while the deviant used a different one, but the deviants can be either dimensionally congruent or dimensionally incongruent with the standards. For example, in Condition 2W (Two-dimensional standards and Within-dimension deviant), the standards and the deviant cooccur with two different classifiers zhang (張) and pian (片) respectively, but the deviant is a within-dimension deviant, for both zhang (張) and pian (片) are two-dimensional. In contrast, the deviant in Condition 2B (Two-dimensional standards and Between-dimension deviant) is categorized by the use of the one-dimensional classifier tiao (條), which is dimensionally inconsistent with the classifier zhang (張) and thus is a between-dimension deviant. Importantly, the between-dimension deviant was selected from one of the standards in the other category (e.g., the deviant co-occurring with zhang (張) in 1B was selected from the two-dimensional conditions), with the photo being conceptually similar but visually different (e.g., two kinds of SOFA). This was to examine if the perception of an object would be influenced if its associated classifier was previously used with standards. 20.

(29) Four blocks of 360 stimuli were created (1440 stimuli in total). The standard stimuli were photos of four (repeated) objects sharing the same classifier (probability 80%), and the deviant stimuli were photos of one object not categorized by the standards’ classifier (probability 15%). Targets were photos of a cat (probability 5%) and were included to distract participants’ attention from the standards and deviants and to make sure that they stayed alert during the experiment. Two sets of standards belonging to the same dimension were created so that the standards in one condition would not be repeated in the other. The block where the betweendimension deviants occurred was preceded by another one or two blocks where the deviants served as standards. There were a total of 4 lists and the combination of the standard set and the deviant type were counterbalanced across lists (see Table 2 for a sample list and Appendix A for the complete stimuli; see Table 3 for the numerical facts of stimuli).. Table 2. Experimental conditions in a sample list Condition/ Block 1W 1B 2W 2B. Dimension one-dimensional two-dimensional. Deviant type within-dimension between-dimension within-dimension between-dimension. standard (80%) tiao (條) tiao (條) zhang (張) zhang (張). deviant (15%) zhi (枝) zhang (張) pian (片) tiao (條). target (5%) cat cat cat cat. 1W: One-dimensional standards and Within-dimension deviant; 1B: One-dimensional standards and Between-dimension deviant; 2W: Two-dimensional standards and Withindimension deviant; 2B: Two-dimensional standards and Between-dimension deviant. 21.

(30) Table 3. Numerical facts of stimuli. Stimuli per block Stimuli per dimension Total stimuli Types per block Types per dimension Frequency of one type per block. Standard (80%). Within-/Betweendimension deviant (15%). Target (5%). Sum (100%). 288 576 1152 4 8 72. 54 108 216 1 2 54. 18 36 72 1 1 18. 360 720 1440. Grayscale photos of real-world objects were used as materials. The photos were mainly derived from the Bank of Standardized Stimuli (BOSS) by Brodeur et al. (2010) and Brodeur, Guérard and Bouras (2014). When a proper photo for a concept was not available, photos from other resources were adopted and edited to supplement the materials (see Appendix A: DB, SC1 from Konkle et al. 2010, SD3 from the CardFactory.co.uk). To establish whether the objects were named as intended and frequently collocated with the classifiers of interest, the selected photos of stimuli were pretested by asking twenty-five native Chinese speakers, none of whom participated in the ERP experiment, to come up with a numeral phrase ‘one-classifiernoun’ for a total of 40 photos of objects (21 formal stimuli and 19 fillers). The results showed that all the objects were named consistently and collocated with the classifiers of interest (criterion: above 80%; see Appendix B).. 2.3 Procedure The experiment was conducted in the Neurolinguistics Lab at National Taiwan Normal 22.

(31) University in Taiwan. Participants had to sign a consent form first and filled in a demographic questionnaire before the experiment. An eletro-cap to measure brainwaves was put on the participants and they were seated in front of a computer monitor at a distance of 90 to 100 cm. With the distance, stimuli were presented at the center of the screen, subtending a visual angle of about 5° (object size ≈ 10 cm2). A one-minute practice session prior to the formal experiment was given to familiarize the participants with the presentation and tasks. The stimuli used in the practice session would not appear in the formal experiment, and they were not categorized by any of the classifiers for the experimental manipulation. Block order was randomized. Presentation order within each block was pseudorandomized so that deviant stimuli or target stimuli would appear at least after eight standard stimuli at the beginning of a block, and two deviant stimuli would never appear in immediate succession. Each block started with a fixation cross at the center of the screen for 1000 ms followed by a series of photo stimuli. Each photo appeared for 350 ms with a jittered interstimulus interval (ISI) randomly selected from 400, 450, 500, 550, and 600 ms (mean = 500 ms). The participants were instructed to press a button on a joystick (Logitech F310 Gamepad) with their right or left index finger (which was counterbalanced across participants) as soon as possible when detecting a target (a cat). The participants also had to say dadadada… silently all along the stimulus presentation in an attempt to minimize the potential influence of conscious naming of objects in their mind. Participants were allowed to take a break between blocks. After the oddball experiment, the 23.

(32) participants were asked to name each object like the pretest on the materials (A few objects were slightly below the criterion 80%: 3 Chinese participants did not use the intended classifiers for SA1, SB1, DB, and SC1, and 3 English participants did not name SC2 as intended; see Appendix B). Figure 3 illustrates the stimulus presentation within a block.. Figure 3. Sample of one block. 2.4 Behavioral and EEG recordings The E-prime 2.0 software (Psychology Software Tools Incorporated) was used to present the experimental materials, to record participants’ behavioral response, including reaction times and accuracy rates, and to send event codes to the electroencephalogram digitization computer. Electroencephalogram (EEG) was recorded from 32 electrodes according to the 10/20 system. The average of the left and right mastoids (A1 and A2) was used as the reference for each channel in both online and off-line analyses. Four additional electrodes (HEOL and HEOR on the outer canthus of each eye and VEOU and VEOL on the upper and lower ridge of the left eye) were placed to monitor eye movements (blinks and saccades). The impedance of 24.

(33) all the electrodes was kept below 5kΩ. The sampling rate was 1000 Hz and the amplifier rate (Gain) was 19, corresponding to an input range of +/−131.5 mV. The EEG signals were filtered between DC to 100 Hz (NuAmps, NeuroScan Incorporated).. 2.5 Data Analysis The behavioral data of the target (a cat) were analyzed by submitting reaction times and accuracy rates to Welch’s independent samples t tests between (language) groups. Since few participants were able to respond to the target within the short duration of stimulus presentation (350 ms), the blank right after the target was also included in the analysis. The raw EEG data were processed with EEGLAB (Delorme & Makeig 2004) and ERPLAB (Lopez-Calderon & Luck 2014) in MATLAB (Math-Works 2005). First, continuous EEG data were loaded into MATLAB’s workspace using the EEGLAB toolbox, and the 4 monopolar eye-movement channels (VEOU, VEOL, HEOL, HEOR) were converted into 2 bipolar ones (VEOG for vertical eye movement, HEOG for horizontal movement) using EEG Channel operations (ERPLAB toolbox). The EEG data were filtered by an Infinite Impulse Response (IIR) filter, with the high-pass value set at 0.1 Hz, 12 dB/oct, and excessive artifacts were manually removed before the performance of an independent component analysis (ICA) (see Luck 2018, a suggestion posted on the blog of ERP INFO). After the ICA was computed, components associated with either eye- or body-movement were identified and rejected. 25.

(34) Channel interpolation was applied after the component rejection if a channel was severely contaminated by artifacts or noise. An event list was then created to sort the EEG data, which were epoched from 100 ms before and 600 ms after the stimulus onset. Baseline correction was applied with the pre-stimulus −100−0 ms interval. Note that ERPs elicited by the standard stimuli of the same dimension were averaged, whereas ERPs elicited by the deviants of different conditions were averaged respectively. Contaminated data were manually rejected by visual inspection of each epoch by reference to the automatic artifact detection, moving window peak-to-peak amplitude and step-like artifacts (see Luck 2014:196-198). The overall rejection rate of the data was 8.45%. The averaged ERPs were computed and then filtered with the lowpass value set at 30 Hz, 12 dB/oct. Finally, the grand average ERP brainwave for each experimental condition was obtained by averaging all the participants’ ERP data. For the statistical analysis of the ERP data, 9 representative channels (F3, C3, P3, FZ, CZ, PZ, F4, C4, P4) were selected for the N1 (100-150 ms) effect and the P2 effect analyses (150-250 ms) (e.g., Gratton, Evans & Federmeier 2009; Luck & Hillyard 1994). To better capture the scalp distribution of the brain activity, the mean amplitudes of N1 and P2 were submitted to the following two repeated measures ANOVAs: (1) a four-way midline analysis (FZ, CZ, PZ), with Stimulus Type (standard, within-dimension deviant, between-dimension deviant), Dimension (one-dimensional, two-dimensional), and Anteriority (front, middle, back) as within-subject factors and Language (Chinese, English) as a between-subject factor; (2) a 26.

(35) five-way laterality analysis (F3, C3, P3, F4, C4, P4), with Laterality (left, right) as an additional within-subject factor. Also, to examine whether the ERP responses to targets were different between the language groups, the mean amplitudes of P2 were submitted to two repeated measures ANOVAs: (1) a three-way midline analysis (FZ, CZ, PZ), with Stimulus Type (standard, target) and Anteriority (front, middle, back) as within-subject factors, and Language (Chinese, English) as a between-subject factor; (2) a four-way laterality analysis (F3, C3, P3, F4, C4, P4), with Laterality (left, right) as an additional within-subject factor . When the Mauchly’s test of Sphericity was violated, Greenhouse-Geisser correction was applied to adjust the p-values. Follow-up paired t-tests were performed and Bonferroni corrected when interactions were observed.. 27.

(36) 3 Results 3.1 Behavioral data One English participant was excluded from the analysis of behavioral data because of the unexpected program error (Chinese: N = 14; English: N = 13). The accuracy rates of the target (a cat) detection task were above 94% in all participants, and the averaged accuracy rate was 99% (Chinese: Mean = 0.99, SD = 0.01; English: Mean = 0.99, SD = 0.02), showing that the participants were paying attention to the stimuli presented on the screen during the experiment. There were no significant differences between groups on target detection accuracy rates nor reaction times (accuracy: t(27) = .57, p = .58; reaction time: t(27) = −1.3, p = .2).. 3.2 ERP data General description of the brainwaves. Figure 4 depicts the brain response to standards, deviants and the target, which shows that the ERP responses elicited by targets are much stronger than those by standards and deviants.. 28.

(37) Figure 4. ERP responses elicited by targets, standards, and deviants across deviant types, dimensions, and language groups.. 29.

(38) Grand averaged ERPs to standards, within- and between-dimension deviants are plotted in Figures 5 and 6 for Chinese and English speakers, respectively.. Figure 5. ERP responses to standards, within- and between-dimension deviants in the Chinese group. 30.

(39) Figure 6. ERP responses to standards, within- and between-dimension deviants in the English group.. 31.

(40) As Figure 5 shows, in the Chinese group, the ERP response to standards is similar to that to between-dimension deviants in early latency, while the response to within-dimension deviants diverges from the other two stimulus types, especially between 100-250 ms. In contrast, Figure 6 shows that the ERP responses to all stimulus types in the English group generally merge together between 100-200 ms. It is not until 200 ms that the response to between-dimension deviants starts to diverge from the other two stimulus types. A quick glance at Figures 5 and 6 reveals that ERP responses collected from the occipital sites (O1, OZ, O2) are different in polarity between 100-200 ms: the brainwaves generally went in the opposite direction with negative-going brainwaves at the anterior sites but positive-going one at the occipital sites. Occipital electrodes were thus excluded from further analysis to resolve the statistical problem that the amplitudes would otherwise cancel each other out because of the inverse relation between ERP components within the same time window. Figures 7 and 8 show the grand averaged ERPs to standards and deviants in both language groups, with Figures 7 and 8 centering on within-dimension and between-dimension deviants respectively.. 32.

(41) P2 effect. Figure 7. Grand averaged ERP responses to standards and within-dimension deviants. CS: Chinese Speakers; ES: English Speakers.. 33.

(42) Figure 8. Grand averaged ERP responses to standards and between-dimension deviants. CS: Chinese Speakers; ES: English Speakers.. 34.

(43) Figures 7 and 8 above show that the ERP components are more fronto-centrally distributed with relatively small amplitudes observed over the posterior sites. An early component N1 (peaking around 150 ms) can be detected in all conditions, but it seems much smaller for the within-dimension deviants in the Chinese group. The N1 then shifts to a positive-going wave, a P2 which peaks between 200 and 250 ms. This P2 component in the Chinese group is notably greater in amplitude for within-dimension deviants compared with the corresponding standards. Following the P2, a negativity is found with an amplitude generally larger in the English group than in the Chinese group, for both standards and deviants. All the brainwaves then start to converge between 300 and 400 ms except for the betweendimension deviant in the Chinese group, which slightly diverges to another positive-going wave until 400 ms. Figures 9 and 10 illustrate the difference waves and the topographic maps of the P2 deviancy effects (deviants – standards) between 150-250 ms, where a sharp distinction between language groups and a more frontally oriented distribution can be visually identified.. 35.

(44) P2 effect. Figure 9. Difference waves of within-/between-dimension deviants minus standards. CS: Chinese Speakers; ES: English Speakers.. 36.

(45) μV. CS. ES. Figure 10. Topographic maps of the deviancy effects (deviant minus standard) (P2 effect: 150250 ms). Top: Chinese Speakers (CS). Bottom: English Speakers (ES). Rightmost: channel locations of 32 electrodes. Positivity is painted in red and negativity in blue.. As shown in Figure 9, the within-dimension deviancy effect in the Chinese group (black line) is conspicuous in the P2 time window while the between-dimension deviancy brainwaves of Chinese and English groups largely merge together (red line vs. green line), except for the more positive-going brainwave in the Chinese group from 300 to 400 ms. Figure 10 presents the topographic maps, showing that the P2 effect is frontally distributed, and that it is salient in the within-dimension condition in the Chinese group, but not in the between-dimension condition or in the English group.. 37.

(46) The target effect. Mean amplitudes between 150-250 ms of the targets were measured and submitted to two repeated measures ANOVAs for the midline and the laterality analyses, respectively. The midline analysis was carried out with two within-subject factors of Stimulus Type (standard, target) and Anteriority (front, middle, back) and one between-subject factor of Language (Chinese, English); the laterality analysis was carried out with an additional withinsubject factor of Laterality (left, right). Both analyses revealed a main effect of Stimulus Type (F(1,27) = 50.41, p < .0001, ηp2 = .66), showing that the P2 effect was stronger for targets than for standards (D(target-standard) = 2.86, p < .0001)2. However, Stimulus Type did not interact with Language (F(1,27) = .05, p = .82, ηp2 = .002), suggesting that there was no difference between language groups with respect to processing targets (cat) among standards. The N1 effect. Mean amplitudes between 100-150 ms of the non-target stimuli were measured and submitted to two repeated measures ANOVAs for the midline and the laterality analysis, respectively. The midline analysis was carried out with three within-subject factors of Stimulus Type (standard, within-dimension deviant, between-dimension deviant) and Anteriority (front, middle, back) and one between-subject factor of Language (Chinese, English); the laterality analysis was carried out with an additional within-subject factor of Laterality (left, right). Neither of the analyses revealed a main effect of Language or an. 2. D stands for the microvolt mean differences. 38.

(47) interaction with Language, showing that the classifier effect in the Chinese group was not yet significant during the selected N1 window. The P2 effect. Mean amplitudes between 150-250 ms of the non-target stimuli were measured and submitted to two repeated measures ANOVAs for the midline analysis and the laterality analysis, respectively. The midline analysis was carried out with three within-subject factors of Stimulus Type (standard, within-dimension deviant, between-dimension deviant), Dimension (one-dimensional, two-dimensional), and Anteriority (front, middle, back) and one between-subject factor of Language (Chinese, English). The results revealed a significant main effect of Stimulus Type (F(2,27) = 23.81, p < .0001, ηp2 = .478), and such effect further interacted with Language (F(2,27) = 5.86, p < .01, ηp2 = .184), showing that the P2 effect was stronger for the within-dimension deviancy, but not for the between-dimension one, in the Chinese than in the English participants (CS: D(within-standard) = 1.57, p < .0001; D(between-standard) = -.27, p = .72; ES: D(within-standard) = .37, p = .48; D(between-standard) = -.37, p = .35). A significant Anteriority effect was also found (F(2,27) = 28.3, p < .0001, ηp2 = .521). Its interaction with Stimulus Type indicated that the within-dimension deviancy effect was more widespread (but maximal in the fronto-central region, reflected by the mean differences) (F(4,27) = 7.35, p < .005, ηp2 = .22; front: D(within-standard) = 1.12, p < .0001; middle: D(within-standard) = 1.18, p < .0001; back: D(within-standard) = .62, p < .01). In addition, its interaction with Dimension showed that the difference between 1D and 2D stimuli was significant in the parietal region (F(2,27) = 7.85, p. 39.

(48) < .01, ηp2 = .232; front: D(1D-2D) = .1, p = .64; middle: D(1D-2D) = -.08, p < .67; back: D(1D-2D) = -.44, p < .05). No other main effects and interactions were found (Language: F(1,27) = .56, p = .46, ηp2 = .021; Dimension: F(1,27) = .58, p = .45, ηp2 = .022). See Table 4 for a summary of the ANOVA for the midline analysis.. Table 4. Summary of the degrees of freedom, F values, and p values of repeated measures ANOVAs for the midline and laterality analyses of the P2 effect (150-250 ms) MIDLINE ANALYSIS Main effect Language Stimulus Type Dimension Anteriority 2-way interaction Language × Stimulus Type Language × Dimension Language × Anteriority Stimulus Type × Dimension Stimulus Type × Anteriority Dimension × Anteriority 3-way interaction Language × Stimulus Type × Dimension Language × Stimulus Type × Anteriority Language × Dimension × Anteriority 4-way interaction Language × Stimulus Type × Dimension × Anteriority LATERALITY ANALYSIS Main effect Language Stimulus Type Dimension Anteriority Laterality 2-way interaction Language × Stimulus Type 40. dfs. F. p. 1 2 1 2. 0.56 23.81 0.58 28.30. 0.46 <.0001 0.45 <.0001. 2 1 2 2 4 2. 5.86 2.28 1.24 1.05 7.35 7.85. <.01 0.14 0.28 0.35 <.005 <.01. 2 4 2. 0.75 0.49 0.08. 0.47 0.59 0.82. 4. 2.01. 0.14. 1 2 1 2 1. 0.18 11.08 3.35 45.02 0.34. 0.68 <.0005 0.08 <.0001 0.56. 2. 4.88. <.05.

(49) Language × Dimension Language × Anteriority Language × Laterality Stimulus Type × Dimension Stimulus Type × Anteriority Stimulus Type × Laterality Dimension × Anteriority Dimension × Laterality 3-way interaction Language × Stimulus Type × Dimension Language × Stimulus Type × Anteriority Language × Stimulus Type × Laterality Language × Dimension × Anteriority Language × Dimension × Laterality 4-way interaction Language × Stimulus Type × Dimension × Anteriority Language × Stimulus Type × Dimension × Laterality 5-way interaction Language × Stimulus Type × Dimension × Anteriority × Laterality. dfs 1 2 1 2 4 2 2 1. F 1.55 1.33 1.38 3.32 27.09 0.09 17.68 2.27. p 0.23 0.26 0.25 0.05 <.0001 0.80 <.0001 0.14. 2 4 2 2 1. 0.90 0.75 0.44 0.06 0.04. 0.40 0.47 0.54 0.83 0.85. 4 4. 0.78 0.12. 0.47 0.82. 4. 5.60. <.005. Following the midline analysis, the laterality analysis was carried out with an additional within-subject factor of Laterality (left, right). As also shown in Table 3, the analysis was beset with a five-way interaction of Language, Stimulus Type, Dimension, Anteriority, and Laterality (F(4,27) = 5.6, p < .005, ηp2 = .18). Due to this significant five-way interaction, the data were then split by Laterality (right, left) to conduct two four-way repeated measures ANOVAs with three within-subjects factors of Stimulus Type, Dimension, Anteriority, and one betweensubject factor of Language. Similar to the midline analysis, the left analysis showed a significant Stimulus Type effect (F(2,27) = 12.45, p < .0001, ηp2 = .324) and its interaction with Language (F(2,27) = 4.31, p 41.

(50) < .05, ηp2 = .142), with the P2 effect stronger for the within-dimension deviancy in the Chinese group (CS: D(within-standard) = 1, p < .0001; D(between-standard) = -.12, p = 1; ES: D(within-standard) = .18, p = 1; D(between-standard) = -.18, p = 1). The main effect of Anteriority was found (F(2,27) = 56.03, p < .0001, ηp2 = .683), and such effect also interacted with Stimulus Type (F(4,27) = 28.11, p < .0001, ηp2 = .519) and Dimension (F(2,27) = 15.57, p < .0005, ηp2 = .375), respectively. For the interaction of Anteriority and Stimulus Type, follow-up t-tests suggested a fronto-centrally distributed within-dimension deviancy effect (front: D(within-standard) = .98, p < .0001; middle: D(within-standard) = .78, p < .0001; back: D(within-standard) = .02, p = 1). For the interaction of Anteriority and Dimension, follow-up t-tests suggested a significant difference between 1D and 2D stimuli at the posterior site (front: D(1D-2D) = .09, p = .67; middle: D(1D-2D) = -.16, p = .37; back: D(1D-2D) = -.67, p < .005). No other main effects and interactions were found (Language: F(1,27) = .48, p = .5, ηp2 = .018; Dimension: F(1,27) = 2.03, p = .17, ηp2 = .072). See Table 5 for a summary of the ANOVA for the left analysis.. Table 5. Summary of the degrees of freedom, F values, and p values of repeated measures ANOVAs for the left and right analyses of the P2 effect (150-250 ms) left dfs Main effect Language Stimulus Type Dimension Anteriority. 1 2 1 2. 42. F. right p. F. p. 0.48 0.50 0.01 0.91 12.45 <.0001 7.52 <.005 2.03 0.17 4.68 <.05 56.03 <.0001 31.04 <.0001.

(51) left dfs. F. right p. F. p. 2-way interaction Language × Stimulus Type Language × Dimension Language × Anteriority Stimulus Type × Dimension Stimulus Type × Anteriority Dimension × Anteriority. 2 1 2 2 4 2. 4.31 <.05 4.10 <.05 1.51 0.23 1.43 0.24 0.87 0.37 1.56 0.22 2.50 0.10 3.60 <.05 28.11 <.0001 13.98 <.0001 15.57 <.0005 14.56 <.0005. 3-way interaction Language × Stimulus Type × Dimension Language × Stimulus Type × Anteriority Language × Dimension × Anteriority. 2 4 2. 1.01 0.29 0.24. 0.36 0.76 0.68. 0.64 1.14 0.03. 0.52 0.32 0.90. 4-way interaction Language × Stimulus Type × Dimension × Anteriority. 4. 0.09. 0.91. 3.29. <.05. Unlike the left analysis, the right analysis revealed a four-way interaction of Language, Stimulus Type, Dimension, Anteriority (F(4,27) = 3.29, p < .05, ηp2 = .112). The data were then split by Language (right CS, right ES) to conduct two three-way repeated measures ANOVAs with only within-subject factors of Stimulus Type, Dimension, and Anteriority. In the right CS analysis, but not in the right ES analysis, two main effects were found: Stimulus Type (CS: F(2,27) = 10.23, p < .005, ηp2 = .44; ES: F(2,27) = .34, p = .57, ηp2 = .026) and Dimension (CS: F(1,27) = 8.76, p < .01, ηp2 = .403; ES: F(1,27) = .52, p = .53, ηp2 = .038). The Stimulus Type effect in the Chinese group further interacted with Anteriority (F(4,27) = 12.05, p < .0005, ηp2 = .481), suggesting an anterior P2 effect for the within-dimension deviancy (front: D(within-standard) = 1.51, p < .0001, D(between-standard) = -.3, p = .91; middle: D(withinstandard). = 1.42, p < .0005; D(between-standard) = -.004, p = 1; back: D(within-standard) = .62, p = .22; 43.

(52) D(between-standard) = .33, p = .7). It is noted that the Chinese within-dimension deviancy effect also marginally interacted with Dimension (F(2,27) = 3.11, p = .07, ηp2 = .193), which showed that within-dimension deviancy effect was larger in 2D and which might indicate the sensitivity to dimension in the right hemisphere (1D: D(within-standard) = .60, p = .26; 2D: D(within-standard) = 1.76, p < .0005). Both CS and ES analyses revealed an Anteriority main effect and the interaction of Anteriority and Dimension; however, follow-up t-tests showed that significant 1D vs. 2D difference in the centro-parietal region was only found in the Chinese group (front: D(1D-2D) = -.23, p = .26; middle: D(1D-2D) = -.47, p < .05; back: D(1D-2D) = -.98, p < .005), not in the English group (front: D(1D-2D) = .14, p = .67; middle: D(1D-2D) = -.05, p = .85; back: D(1D-2D) = -.56, p = .08). No three-way interaction was found in both language groups. See Table 6 for a summary of the ANOVAs for the right CS and right ES analyses.. Table 6. Summary of the degrees of freedom, F values, and p values of repeated measures ANOVAs for the right CS and right ES analyses of the P2 effect (150-250 ms) right CS. right ES. dfs. F. p. F. p. Main effect Stimulus Type Dimension Anteriority. 2 1 2. 10.23 8.76 31.98. <.005 <.01 <.0001. 0.34 0.52 13.61. 0.57 0.53 <.001. 2-way interaction Stimulus Type × Dimension Stimulus Type × Anteriority Dimension × Anteriority. 2 4 2. 3.11 12.05 7.79. 0.07 <.0005 <.01. 1.11 3.66 6.80. 0.34 0.05 <.05. 3-way interaction Stimulus Type × Dimension × Anteriority. 4. 0.77. 0.46. 2.89. 0.06. 44.