類別化色彩知覺研究---色彩類別邊界、代表色相、命名及語意性歸類

行政院國家科學委員會

獎勵人文與社會科學領域博士候選人撰寫博士論文

成果報告

類別化色彩知覺研究：色彩類別邊界、代表色相、命名及

語意性歸類

核 定 編 號 ： NSC 98-2420-H-009-001-DR 獎 勵 期 間 ： 98 年 08 月 01 日至 99 年 07 月 31 日 執 行 單 位 ： 國立交通大學應用藝術研究所 指 導 教 授 ： 陳一平 博 士 生 ： 謝翠如 公 開 資 訊 ： 本計畫可公開查詢

中 華 民 國 99 年 08 月 05 日

⦚䵚ℳ抩⮶⸇

㑘䞷塬嫢䪣䴅㓏

◩



⭺



嵥



㠖



խ֮ۥ൑ဲნ֗፿ߢۥ൑ᣊ़ܑၴ

ColorTermsandLexicalColorCategorySpaceinMandarin

     

ࣴ!ز!ғǺᖴᆧӵ!

ࡰᏤ௲௤Ǻഋ΋ѳ!റγ!

!

!



₼ 噾 㺠 ⦚ ⃬ ◐ ⃬ ㄃ ⏼ 㦗

խ֮ۥ൑ဲნ֗፿ߢۥ൑ᣊ़ܑၴ

Color

TermsandLexicalColorCategorySpace

 inMandarin



ࣴ ز ғǺᖴᆧӵ                        StudentǺTuseiͲJuHsieh

ࡰᏤ௲௤Ǻഋ΋ѳ                        AdvisorǺDr.IͲPingChen



୯ ҥ Ҭ ೯ ε Ꮲ

ᔈҔ᛬ೌࣴز܌

റ γ ፕ Ў



A Dissertation

Submitted to Institute of Applied Art

College of Humanity and Social Science

National Chiao Tung University

in partial Fulfillment of the Requirements

for the Degree of Doctor of Philosophy

in

Art

Hsinchu, Taiwan, Republic of China





ύ๮҇୯ΐΜΐԃϤД

i

ύЎՅறຒ༼ϷᇟقՅறᜪձޜ໔

ࣴزғǺᖴᆧӵ ࡰᏤ௲௤Ǻഋ΋ѳ റγ ୯ҥҬ೯εᏢ ᔈҔ᛬ೌࣴز܌റγ੤ ᄔा Յறࢂ࿤ނ܌և౜рٰޑᗲܴຎ᝺੝ቻǴԶᇟق߾ࢂଯ฻ᇡޕфૈޑ߄౜Ǵ௖૸ፄ ᚇޑՅற࿶ᡍӵՖҗᇟقޑБԄ໺ၲǴၱԋࣁёᕕှӃϺЈඵᆶЎϯ࣬٩܄ޑख़ाᏢೌ ᝼ᚒϐ΋Ƕ߈ъঁШइаٰǴҗΓᜪᏢǵᇟقᏢǵՅறࣽᏢϷЈ౛Ꮲ฻ၠሦୱᏢޣӅӕ ໒௴ΑՅறޕ᝺کᇟقϕ୏ޑࣴزБӛǴԜ᝼ᚒჹܭ೭٤ᏢߐԶقԖ๱όӕ܄፦ޑख़ा ཀကǺჹЈ౛ᏢڗӛԶقǴՅறᇟقޑࣴزԋ݀ගٮᜢܭεတфૈᆶЎϯჹܭޕ᝺ቹៜ ޑှញǹԶჹܭΓᜪᏢϷᇟقᏢޣԶقǴՅறҔᇟޑፄᚇࡋ܈ёຎࣁޗ཮ԋዕࡋޑࡰ኱Ǵ ၠЎϯ໔ՅறҔຒว৖ޑӅ೯܄ΨඟҢΑΓᜪᇟقᄽϯޑᗦᙒೕࡓǹԶჹՅறࣽᏢԶقǴ аᇟقᜪձբࣁЈ౛ЁࡋޑՅறࡋໆБԄڀԖၨଯቫԛޑᇡޕཀကǴёᆶ໺಍ෳՅᏢ྽ ύаՅற୔ᒣ܈ຎ᝺ৡ౦⸡฻ၨࣁ୷ᘵޑՅறག᝺ൂՏϕࣁୖྣǴаံى௦ڗόӕᇡޕ ቫԛޑՅறག᝺Ёࡋܭ౛ፕ୷ᘵϷჴ୍ᔈҔ΢ޑज़ڋǶ ஒՅறག᝺аᇟقޑБԄጓዸޑՉࣁȐջՅறڮӜȑྍܭΓတჹܭᕉნڈᐟаᜪձ ϯޑБԄٰೀ౛ޑ੝܄ǶՅӀࢂፄᚇԶೱុޑڈᐟǴՅறޕ᝺ޑᐒڋҗவ୷ᘵޑғ౛ቫ ԛ໒ۈ൩аᜪձϯޑБԄஒၗૻᙁϯǵϩᜪǵ຾ԶֹԋِೲԖਏޑᒣ᛽ǶᇟقՅறᜪձ ࢂ᏾ঁՅறޕ᝺྽ύჹՅற࿶ᡍനཷۺϯޑ߄౜Ǵҭջךॺ཮ஒࢌᓎࢤޑՅӀǵ΋ဂё ᒣ᛽рፓηόӕޑ࣬߈Յறᘜࣁӕ΋ᜪձǴ٠а΋ঁՅறຒ༼఼ᇂඔॊǴٯӵआՅǶӧ ࣬ᜢሦୱ྽ύޑ΋ঁਡЈݾ᝼ࢂǺᇟقՅறᜪձޑ౜ຝࢂӃϺЪදၹޑǵᗋࢂ཮ڙډЎ ϯ܌ೕጄǻύЎ҆ᇟޣЈ္ޑȨआՅȩ΋ຒ܌఼ᇂޑՅறጄൎၟमᇟ҆ᇟޣЈ္ޑ“red” ࢂ΋ኬޑ༏ǻԜୢᚒЇଆ࣬྽ӭፕ៏Ϸჴቻၗ਑՘᛾Ǵٯӵଆۈܭ࢙լ๲εᏢᇟقᏢس Ꮲޣޑ୷ҁՅӜࣴزǵШࣚՅறᇟقፓࢗ(WCS)฻ǶฅԶǴၸѐ࣬ᜢЎ᝘аύЎࣁࣴز

ii ჹຝޑၗ਑ࠅᡉ๱όىǴԖ᠘ܭԜǴҁፕЎЬԑࣁаύЎ҆ᇟޣբࣁࣴزჹຝǴ௖૸ύ ЎᇟნΠޑՅறຒ༼аϷᇟقՅறᜪձޑ౜ຝǶ ҁፕЎख़ᗺԖΟǴϩձаჴቻፓࢗၲԋࣴزҞޑǴख़ᗺ΋ǺᕕှύЎՅறຒ༼ޑ٬ Ҕ౜ݩǺࣁၲԋԜࣴزҞޑǴ୺Չ΋ঁ 189 ՏύЎ҆ᇟޣୖᆶޑՅறຒ༼Ծҗӣགྷբ཰Ǵ ڗளऊϖίঁՅӜኬҁǴ٠а௶ॊ಍ीඔॊځख़ፄ܄ᆶӭኬ܄Ƕख़ᗺΒǺ୺Չ΋ঁҗ 36 ӜύЎ҆ᇟޣୖᆶޑՅறԾҗڮӜჴᡍǴа 121 ঁӧᑻჿև౜ǵೕࡓڗኬޑՅறࣁڈᐟǴ аՅӜǵϸᔈਔ໔ϷߞЈϩኧΟ໨ࣁ٩ᡂ໨ǴԜჴᡍҞޑӧܭᕕှӧύЎՅறڮӜՉࣁ ྽ύޑϸᔈ੝ቻǴаϷ࿶த೏٬ҔޑՅӜǶख़ᗺΟǺࡌҥаύЎ҆ᇟޣޑᇟقՅறᜪձ ӧ኱ྗՅறޜ໔(CIE1931 x-y diagram)΢ޑჹᔈጄൎǴ೭೽ϩޑჴᡍࣁа 12 ঁύЎ୷ҁ ՅӜբࣁᒧ໨ǵ461 ঁԵቾՅ࣬ǵறࡋᆶߝࡋ٠ЪೕࡓڗኬޑᑻჿՅறբࣁڈᐟǴᡣ 44 Տڙ၂ޣ຾ՉᒧՅჴᡍǴࣴز่݀ёբࣁԵໆᇟཀޑՅறس಍ϐࡌҥୖԵǴќѦКၨύ ЎᆶځдᇟقޑᇟقՅறᜪձጄൎӧϩթ΢ޑ౦ӕǴёբࣁՅறҔᇟࣴزޑٿঁ໺಍౛ ፕਣࢎǴȨදၹЬကȩ(Universalism)܈Ȩ࣬ჹЬကȩ(Relativism)ޑჴቻ᛾ᏵǶ ਥᏵҁፕЎޑΟঁسӈࣴز่݀ёᘜયраΠว౜Ǻ1.ӧԾҗӣགྷᆶԾҗڮӜჴᡍ ྽ύǴԖ 12 ঁύЎՅறຒ༼ڀԖଯࡋޑӣྉ౗Ϸख़ፄ܄Ǵхࡴआǵᐊǵ໳ǵᆘǵᙔǵ ๋ǵ໵ǵԪǵқǵڜଢ଼ǵણआǵਲआ฻Ǵ೭٤ຒ༼಄ӝ࣬ᜢЎ᝘྽ύޑ 11 ঁ୷ҁՅӜ ޑᜪձǴځύણआᆶਲआٿঁՅӜᗨឦӕ΋ᜪձǵЪΨό಄୷ҁՅӜޑȨൂຒȩ (monolexemic)ϐۓࡓǴՠٿޣӧᇟقՅறᜪձޜ໔ޑȨขᗺȩ(foci)Տ࿼ܴᡉԖ୔႖Ǵ ӧ౛ፕ΢܈ёຎࣁύЎϐ୷ҁՅӜǶ2.ԾҗڮӜ่݀ᡉҢதҔޑύЎՅፓঅႬຒ(tone modifier)хࡴߝǵསǵ఩ǵણǵభǵుǵୃаϷ҅Ƕ3.ᆶमᇟՅறԾҗڮӜࣴزޑ่݀ ࣬ၨϐΠǴύЎՅறڮӜޑՉࣁᖿӛܭ٬ҔၨӭޑፄӝՅӜǵၨϿޑ୷ҁՅӜϷൂຒՅ ӜǴԶЪύЎڮӜբ཰܌ሡޑϸᔈਔ໔ၨߏǴڙ၂ޣჹܭՅறڮӜޑѳ֡ߞЈࡋΨၨեǶ 4.ᆶӕኬᢀෳచҹޑВᇟՅறᜪձጄൎࣴز่݀࣬ၨǴҁࣴزޑᙔՅᆶᆘՅϐۓကጄൎ ᡉ๱౦ܭВᇟࣴز่݀ǴќѦԪՅᆶڜଢ଼ՅޑۓကጄൎӧύЎᆶВЎΨౣԖόӕǶ5.җ КၨҗԾҗڮӜаϷ 12 ՅӜᒧՅٿঁჴᡍޑ่݀ᡉҢǺڙ၂ޣଞჹϟܭᙔǵᆘϷ๋Յ ᜪձጄൎޑڈᐟ຾ՉڮӜբ཰ਔǴКځдڈᐟሡा׳ߏޑϸᔈਔ໔Ǵՠࢂӕኬጄൎޑڈ

iii ᐟӧᒧՅբ཰ޑϸᔈਔ໔ࠅࢂ࣬ჹอޑǶ೭߄ҢԜΟޣޑՅறᜪձޑղᘐᜤࡋ཮Ӣբ཰ ा؃ԶԖ܌ৡ౦Ǵӧமॐᒧ᏷(force-choice)ਔޑϸᔈਔ໔อǵբ཰ᜤࡋեǴԶЪჴᡍ่ ݀ޑ໣ύᖿ༈ၨଯǹԶӧԾҗڮӜբ཰ϐΠޑϸᔈਔ໔ߏǵԶЪڮӜ่݀໣ύᖿ༈࣬ჹ ၨեǴҗԜ౜ຝё௢ፕ׎৒ᙔǵᆘԿ๋ՅፓޑύЎՅӜӧ٬Ҕ΢ၨࣁፄᚇǴځᒧ᏷܈ӣ ྉਔޑᜤࡋၨଯǶ ᏾ᡏԶقǴҁፕЎаჴቻፓࢗϷჴᡍޑБԄගٮύЎՅறᇟقࣴزሦୱޑख़ाୖԵ ၗ਑Ǵхࡴ᏾ӝٿᅿՅӜፓࢗБݤ܌ள่݀ᘜય౜ՉޑύЎ୷ҁՅӜǴ೭٤ՅӜόӕܭ а۳Ў᝘ύ܌٬Ҕޑӷᜏڂύमᙌ᝿ǴᆶჴሞදၹޑҔᇟԖ܌рΕǶќѦҁࣴزϐ่݀ ΨёբࣁၠᇟقՅறᜪձࣴزޑύЎୖԵၗ਑Ǵ࣬ჹܭځдШࣚᇟقǵၸѐύЎޑჴቻ ࣴزኧໆࢂ࣬ჹีϿޑǶӆޣǴҁࣴزޑว౜ΨёբࣁύЎᆶᚑՅறᜪձࣴزޑ୷ᘵਣ ࢎǴ࣬ჹܭमЎޑ࣬ᜢࣴزޑس಍ᆶೕኳԶقǴύЎӧ୷ҁՅӜǵԛाՅӜ(secondary color term)ǵаϷՅӜᄽϯ่ᄬ฻Бय़೿ۘԖ຾΋؁௖૸ޑޜ໔Ƕ

iv

Color Terms and Lexical Color Category Space in Mandarin

Student: Tsuei-Ju Hsieh Advisor: Dr. I-Ping Chen Institute of Applied Art

National Chiao Tung University

Abstract

The lexical color categorization is a critical function of color perception which involves sorting visual responses to lights into certain color categories and coding them with language. The issue of verbalizing color experience, or color naming, had drawn many attentions from visual psychologists, linguistic anthropologists, and color scientists. Some anthropologists suspected that the amount of color vocabulary circulated within a language could be taken as an idex to the technological and cultural complexity held by the speakers. Although some data were reported in the pioneering work of Berlin and Kay (1969), the developmental status of Mandarin, i.e., the sophistication and the differentiation of its color vocabularies, remains unclear. Besides the theoretical impact on linguistic anthropology, the behavior of naming color experiences is also considered a mirror reflecting the cognitive structure of inner structure of color space. English color naming is a well-discussed topic, and there were over a hundred of different languages in previous extensive color naming survey (WCS). However, there is still a considerable vacancy of empirical color naming data in the relevant domain. The current study aims at establishing the groundwork of lexical color terms and categories in Mandarin by collecting empirical data from native speakers.The objectives to be achieved in this research include: 1. to investigate synchronic Mandarin color lexicon and the popularity of frequent color vocabularies. 2. to acquire behavioral data of color naming. 3. to determine Mandarin basic color terms by analyzing results of the empirical survey. 4. to locate Mandarin

v

speakers’ foci and boundaries of known lexical color categories in a standardized chromaticity diagram. The empirical works in the study includes: 1. a free-recall survey of prevalent color terms involving 189 informants who are native Mandarin speakers. The collected data would help establishing the color lexicon in the current cultural context. 2. a free color naming experiment with written color terms and response times as dependent variables. It is supposed that these variables provide not only the simple popularity counts of color terms, but also an index to the psychological link between color categorization and naming. 3. a 12-terms color sorting experiment. There are 461 color stimuli varying in hue, saturation and brightness in this experiment and participants were asked to sort them into twelve color terms.The results of the three empirical works showed that 1. there are twelve Mandarin color terms that are consistently recalled and named, आ(Hung), ᐊ(Ju), ໳(Huang), ᆘ(Lu), ᙔ(Lan), ๋(Zi), ໵(Hei), Ԫ(Hui), қ(Bai), ڜଢ଼(Ka-fei), ણआ(Feng-Hung) and ਲआ(Tao-hung). These terms correspond well to the eleven color categories found by linguistic anthropologists Berlin and Kay, and can be regarded as basic Mandarin color terms. 2. There are eight tone modifiers found to be frequently used in the free naming experiment,ߝ(bright),ས(dark), ఩ (pale), ણ(powder), భ(light), ు(deep), ୃ(-ish), and ҅(central, correct). 3. Compared to the results of English color naming study, Mandarin speakers use more compound color terms and fewer basic or monolexemic color terms. The response times of Mandarin color naming are longer, and participants’ confidence scores are lower. 4. Comparing the current results with Japanese color sorting experiment in a similar viewing condition, one finds that the location of blue-green boundary is quite different in the two studies. To sum up, this study conducted an exploratory survey on modern Mandarin color terms and color naming, and obtained the experimental data for constructing the space of Mandarin lexical color categories. These results form a good complement to the empirical vacancy in the related fields in world language community, and also serve as the backbone for further studies.

vi

ᇞᖴ

೭ҁᏢՏፕЎளаֹԋǴ२Ӄाགᖴৱৣഋ΋ѳറγǶӧᅺറ੤ΐԃය໔ഋԴৣޑ ࡰᏤϐΠǴךᕕှډ΋ঁᏢΓޑඵ᛽ڰฅёคज़ໆӦᘉкځቶࡋǴՠ׳ࣁਥҁޑࢂߥԖ ుࡋࡘԵᆶӳڻЈޑಞ܄Ǵ೭཮ࢂஒٰЍ࡭ךᝩុவ٣Ꮲೌࣴزޑᝊ຦ᄊࡋǶନΑࣴز ϐѦǴഋԴৣჹޕ᛽ϩ٦ޑ዗௃ǵჹᏢғ੿၈ޑᜢЈᆶᔅշǵჹӚᅿሦୱޑቶݱੋᘪǴ ჹךԶق೿ࢂനԖቹៜΚޑي௲ҢጄǶགᖴҁ܌ςࡺ஭ࡣ։Դৣӭԃ߻ޑ၉ȨϙሶӼ௨ ೿ࢂӳӼ௨ȩǴӧڗளᏢՏޑᅐߏၡ΢ǴόޕၰӭϿԛޑขቾᆶԾךᚶᅪ೿ӧӣᏫ஭Դ ৣޑܯВ၉ᇟ྽ύᕇள๤጗ǴᡣךԖߞЈᝩុ኷ᢀӛ߻ǶךΨགᖴᔈ᛬܌ځдٌ༇ठΚ ܭ௲Ꮲࣴزޑ௲௤ॺޑࡰᏤᆶЍ࡭ǴᆶԴৣॺޑҬࢬ૸ፕᗺрךࣴز΢ޑޓᗺǴᡣፕЎ රӛ׳ڬӄԋዕޑБӛว৖ǶགᖴറγፕЎޑα၂ہ঩Ǻ݅ࠔക௲௤ǵ׵ϺҺ௲௤ǵ৊ ቼЎ௲௤аϷҗम୯ӣѠα၂ޑᛥܴ௲௤ǴፏՏԭԆϐύऐЈᕕှᏢғޑፕЎϣ৒ǵ٠ ኘϧᆿᖏα၂ගٮ஑཰ޑཀـǴගଯԜፕЎֹԋࡋ೚ӭǴ׵௲௤ᆶᛥ௲௤ٿՏ஑཰Յற Ꮲ߻፸ჹፕЎޑЍ࡭ႴᓰǴᡣךుགᄪ۩ǶԜѦǴགᖴӕืұՔॺܭᔌǵඵ౺ǵৎჱǵ ϐᆢǵ᰾ةǵ၃൛аϷځдᏢ׌ۂॺǴεৎόӕޑ஑ߏӧഋԴৣჴᡍ࠻྽ύᐟᕏрόϿ Ԗ፪ޑԋ݀Ǵ۶ԜӧᏢಞᆶғࢲ΢ޑϕ࣬ᜢЈ೿ࢂзΓᜤבޑӣᏫǶ གᖴךޑՔߧഋࡏדۈಖӵ΋ޑངᆶбрǴךॺӧӚԾڗளറγᏢՏޑၡ΢ค࡜ค ৷Ӧ࣬ϕן࡭Ǵ܌Ԗޑ഻৹ᆶᑃግ೿ࢂᆶգӅ٦Ǵךགډߚத۩ၮǶགᖴךޑР҆ᒃӭ ԃ߻ӧЈ౛ᆶ࿶ᔮ΢Ѝ࡭ךܫకচԖ௲ᙍԶ፯຾ҬεԋࣁӄᙍࣴزғǴΨགᖴךޑϦϦ இஇаϷ܌ԖৎΓჹךคచҹޑߞҺᆶх৒ǴٰԾৎ৥ޑᜢངᆶྕཪഉՔ๱ךӼЈوၸ ೭ࢤਓำǶ ᖴᆧӵ ᙣᇞ ୯ҥҬ೯εᏢᔈҔ᛬ೌࣴز܌ 2010/6/20

vii

Contents

Chinese Abstract i English Abstract iv Acknowledgment vi Contents vii List of Figures ix List of Tables xi Chapter 1 Introduction 1 1.1 Backgrounds ……… 1 1.2 Scope ……… 4 1.3 Structure ……….. 6

Chapter 2 Literature Review 8 2.1 Color discrimination versus color categorization ……… 8

2.2 Cognitive linguistic debates: Universalist and relativist views ………. 11

2.3 Color naming methods ………. 13

2.4 Basic and Secondary color terms ………. 15

2.5 World Language Survey (WCS) ………. 16

2.6 Color naming in Mandarin ……….. 17

Chapter 3 Lexicon Survey in Mandarin 20 3.1 Purpose ………. 20

3.2 Method ………. 20

3.3 Results ………. 21

viii

3.3.2 Classification of the collected color terms ……… 23

3.3.3 Popularity of the monolexemic color terms ………. 26

Chapter 4 Mandarin Color Naming Space 31 4.1 Purpose ……… 31

4.2 Method ……… 32

4.3 Results ………. 35

4.3.1 Classification methods ………. 35

4.3.2. The characteristics of different classes ……… 36

4.3.3. The structure of color naming map ………. 43

Chapter 5 Structural Formation of Lexical Color Categories 48 5.1 Purpose ………. 48

5.2 Method ………. 48

5.3 Results ………. 53

5.3.1 Zone map of color categories ……… 53

5.3.2. Response times and the boundary definition ……….. 65

5.3.3. Summary ………. 68

Chapter 6 General Discussion 72 6.1 Comparing popularity ranking of color terms between studies ……… 72

6.2 Comparing with other studies’ free-naming results ………. 74

6.3 Comparing free naming and sorting maps within this study ………. 75

6.4 Future works ………. 78

ix

List of Figures

2.1 Color encoding levels from stimuli and discrimination to lexical encoding…….. 10

2.2 Stimulus array used in WCS………... 17

2.3 Foci color data gained form Mandarin informants (Berlin & Kay, 1969)………. 18

3.1 Histograms of frequency count of total amount (a), frequent use (b) and the extends (c)………... 23

3.2 A hierarchical classification of Mandarin color terms based on the collected data in the recalling study………... 24

3.3 The frequency count histogram of monolexemic color terms……….... 29

4.1 Stimuli plotted on CIE x-y diagram at 50 cd/m2 luminance level………. 33

4.2 The structure of classified color terms………... 36

4.3 From histogram of top to bottom: occurrence (%), mean response times, mean confidence rating, and numbers of Chinese characters of collected data. More details are given in the text………. 37

4.4 The color naming map derived from the results of color naming experiment…... 44

4.5 The naming of the typical stimulus of blue, brown, red and orange……….….... 45

4.6 The contour map of mean confidence rating……….. 47

4.7 The contour map of mean response time (ms)………... 47

5.1 Stimuli plotted on the CIE1931 L-x-y color space………... 51

5.2 The stack histogram for presenting normalized frequency distribution of each condition………. 54

5.3 Upper six x-y diagrams of different L levels using circle color and size to present color category and mode size, respectively. The lower diagram combines all results and differentiates the modes of six conditions with open circles that decrease in size………. 57

5.4 Contour line map showing the formation of red, orange and yellow...………….. 61

5.5 Contour line map showing the formation of green, blue and purple……….. 62

x

5.7 Contour line map showing the formation of gray, white and black………... 64 5.8 Zones of color categories in six luminance conditions……… 67 5.9 Contour maps presenting RTs in six luminance conditions………... 68 5.10 Line plot of the mean RTs of 11 color categories (white is excluded) in their

frequently identified luminance conditions……… 68 6.1 The lexical color category map overlaying the free-naming and sorting results

(I)……… 77 6.2 The lexical color category map overlaying the free-naming and sorting results

xi

List of Tables

1.1 Structure of the thesis………. 7 3.1 Descriptive statistics of the color term recall………. 22 3.2 Rank list derived from monolexemic Chinese color terms……… 27 4.1 Statistic summary of collected terms (each occurrence percentage over 2%)…... 39 5.1 The descriptive overview of categorical sorting results in different L level

conditions……… 53 6.1 A comparison of the popularity of color terms elicited with free recall task

between current study and two previous studies(Lu, 1997; Stanlaw, 1997)…….. 73 6.2 A comparison of free naming results between the current Mandarin study and

1

CHAPTER 1. INTRODUCTION

This chapter gives an overview of the general background and scope of this dissertation. Session 1.1 addresses the brief history, orientation and impact of cross-language color naming studies. Session 1.2 focuses on the scope, purpose and expected contribution of this dissertation. Session 1.3 gives the structure and completed works in the dissertation.

1.1 Backgrounds

Color is the most salient feature in the visible world. The ability of making good use of color information greatly enhance the chance of survival and the quality of life of human and animals alike. Color also carries the potential of arousing aesthetical emotions, and therefore plays a key role in visual art and modern display industry. Due to its significance in human life, color study has a long history and involves diverse research disciplines. Interests in color can be found in academic areas ranging from philosophy, visual arts, culture studies, to physics, psychology, and a good variety of industrial applications. The remarkable variety of color research was discussed in certain review works such as Color For Science, Art and

Technology (Kurt, 1998). Among various perspectives of color related studies, an

extraordinary research line regarding the verbalization of color perception is relevant to the present dissertation.

Some anthropologists suspected that the amount of color vocabulary circulated within a language could be positively related to the technological and cultural complexity held by the speakers (Berlin & Kay, 1969). Besides serving a probable index of cultural maturity, the behavior of naming the color experience is also considered a mirror reflecting the cognitive structure of inner color space, or even the fundamental mechanism of color vision.

There is a critical mechanism involving sorting visual responses of lights into certain color vocabulary, which is the lexical color categorization. This mechanism reveals the nature of information processing of human cognition. Fundamentally, the mind processes vast

2

amount of continuously changing sensory data into more manageable discrete pieces of information, that is, to sort many similar or related events into a finite number of accessible categories. The categorization phenomenon is an information reduction procedure found in the whole cognitive system, and particularly underlies the process of color perception. We literally “see” colors in a categorical fashion. For instance, people lump various discriminable or indiscriminable bluish shades together into a single ‘blue’ category, and we also use linguistic labels, such as “blue” in English and “៴” in Chinese, to convey the visual quality of that whole color category. Common color terms like red, yellow, and green are also color categories in that they cover a set of color stimuli instead of referring to a specific monochromatic stimulus.

The study on color verbalization soon picked on a heated debate over whether color category is governed by color vision, which is supposed to be universal, or by language, which is relative and culture dependent. Will the concept of any color category be similar between different populations using different color terms, provided they share the same visual physiology?

Around half century ago, two linguistic anthropologists Brent Berlin and Paul Kay (hereafter B&K) were intrigued by the ease with which common color terms could be translated between languages form locales as diverse as Tahiti and Mesoamerica (Hardin & Matffi, 1997). This seems to be inconsistent with the culture relativity theory in which languages are thought to divide color space arbitrarily (e.g. Whorf), and shape the way their speakers perceive colors (Whorf, 1956). B&K afterwards conducted a cross-languages color naming survey on U.C. Berkley campus and published their empirical observation in Basic

Color Term (Berlin & Kay, 1969). B&K proposed two general hypotheses about the naming

of perceptual color categories: (1) there is a restricted universal inventory of naming of these categories; (2) a language adds basic color terms in a constrained order, interpreted as an evolutionary sequence. The universal and evolution features within their earlier findings were

3

improved and confirmed by the subsequent World Color Survey (hereafter WCS) with a more comprehensive scope and systematic method.

In the rich history of color-vision research, color scientists dedicated themselves to psychophysical aspects of color, such as color matching and discriminating, adaption, and the measurement of thresholds, but seldom addressed the categorical structure or linguistic expression of color perception. However, since B&K and the followers initiated the investigation on the color categories and their naming in various human languages, the linguistic expressions of perceptual color categories are considered to interact with the introspective side of color vision. In the traditional domain of color-linguistic studies, the interaction between color perception and languages was framed into two fundamentally separate questions (Regier, Kay, Gilbert, & Ivry, in press): (1) Are semantic distinctions in languages determined by largely arbitrary linguistic convention? (2) Do semantic differences cause corresponding cognitive perceptual differences in speakers of different languages? In the past decades, many linguistic anthropologists and visual psychologists produced a considerable amount of evidence and arguments to answer the questions. Chapter 2 would track the research on these questions in detail.

There were over a hundred of different languages in previous extensive color naming survey. However, the status of Mandarin regarding color categories and the naming could not be fully clarified based on the existing data. Specifically, the latest WCS report started from 1970s and surveyed 110 different languages do not cover Mandarin. Dating back to 1960s, in the survey of 20-unrelated languages carried out by B&K (1969), the data that contributed by a restricted number of Mandarin informants was insufficient to affirmatively determine the evolutionary color naming stage to which Mandarin belongs. B&K then roughly treated Mandarin as an example of evolutionary Stage V, in that contains only six color terms, black, white, red, yellow, green, and blue, while Japanese, Korean and Cantonese were assigned into Stage VII which fully contains eleven basic color terms. Practically most native Mandarin

4

speakers would find the terms in use are surely much more than the six terms that were concluded. A later duplicate survey in Mandarin color naming conducted by Taiwanese color researcher Lu Ching-Fu was motivated by this debatable discrepancy (Lu, 1997). In general, although Mandarin is a complex language used by more native speakers than any other language, there is a considerable vacancy of empirical color naming data in the relevant domain.

In addition to the theoretical impact of color categories and naming issue, knowledge about this issue could lay the groundwork in applied researches of color. It is known that one’s color experience in color-science lab and in real life is qualitatively different. We process, memorize and communicate colors in categorical form with linguistic labels, while color vision research reports psychophysics data based on very basic discrimination response. Most color researches focused on the psychophysics of color, instead of the cognition of color. However, in the application of visual communication and visual design involving colors, the color knowledge at cognitive level would be necessary. A landmark color order system Natural Color System (NCS) was developed on the cognitive level of color perception (Sivik, 1997) .

1.2 Scope

The issue of language-dependent color naming had drawn many attentions from visual psychologists, linguistic anthropologists, and color scientists. However, previous studies did not collect enough observations from local native Mandarin speakers. Seeing that, this study aims at establishing the groundwork of Mandarin color categories and terms by collecting color naming data from native speakers who resident in one of the areas currently using Mandarin and traditional Chinese characters, Taiwan. This fundamental-orientated study holds specific objectives listed as below:

5

vocabularies.

(2) Acquiring behavioral data of color naming and establishing the cognitive model of color naming space.

(3) Locating Mandarin speakers’ foci and boundaries of known major color categories in a standardized chromaticity diagram.

(4) Sieving out frequently-used secondary color terms and their chromaticity structure relative to basic color terms.

(5) Discussing the impact of color naming research on applied domain such as visual communication and color design.

The listed objectives in this dissertation were approached through empirical works involving methods of color lexicon survey, color naming and sorting experiments. Comparing with typical color naming works following B&K’s paradigm, this study holds some methodological distinctions:

(1) Most of previous cross-language survey, including WCS, adopted reflective color stimuli such as color chips sampled form Munsell system, and the viewing conditions were not strictly controlled. Actually, the appearance of reflective color could drastically affected by multiple factors, e.g. lighting, or even more subtle factors like stimulus size or viewing distance. The color stimuli in the study are LCD-display and under well-controlled condition.

(2) In the previous anthropological survey, the foci color (the best example within a color category) was determined by simultaneously presenting many color samples to the viewers. This method is direct and efficient, but the juxtaposition would lead to color contextual effects like color assimilation or color contrast, which might alter the appearance of color samples. Instead, this study locates foci color through behavioral statistics, e.g. frequency count of sorting and naming, and response times (hereafter RTs).

6

(3) Color samples used in B&K and WCS were in the highest saturation level and varied along Munsell Hue and Value axis, i.e. perceptual dimension of hue and brightness. Colors of middle and low saturation were excluded in those works, while most colors in the real world should be in these shades. The stimuli in this study are evenly-sampled from the CIE1931 chromaticity diagram and cover several distinct luminance levels. The stimulus sampling was sweeping the available gamut of a typical LCD display. Thus the stimuli are a set of finely sampled colors varying along hue, saturation and brightness. Moreover, the results can be transformed to other color spaces or color appearance models.

1.3 Structure

The structure of this thesis is given in Table 1-1, which presents the major works surrounding each theme and how they are arranged into separate chapters. Chapter 2 gives an overall review of literature related to the current study. The first empirical work in Chapter 3 is a free-recall survey of prevalent color terms involving 189 informants who are native Mandarin speakers. The gathered data would help establishing color lexicon of current cultural context. Chapter 4 presents a free color naming experiment with written color terms and response times as dependent variables. It is supposed that these variables provide not only the simple popularity counts of color terms, but also an index to the psychological links between color categorization and naming. Chapter 5 is a 12-terms color sorting experiment. There are 461 color stimuli varying in hue, saturation and brightness in this experiment and participants were asked to sort them into twelve color terms corresponding to B&K’s eleven basic color categories. The cross comparison with the current results and the results in previous studies would be given in Chapter 6. The last chapter also organizes the findings in the study. The further extension works and the links with application domain would be discussed too.

7

Table 1.1. Structure of the thesis

Structure of the thesis

Theoretical grounds Methods and experiments Results and discussion

Chapter 2. Literature review 2.1. Color discrimination

versus color categorization 2.2. Cognitive linguistic

debates: Universalist and relativist views

2.3. Color naming methods 2.4. Basic and Secondary color

terms

2.5. World Language Survey (WCS)

2.6.Color naming in Mandarin

Chapter 3.

Color Lexicon Survey in Mandarin A Mandarin color terms survey

with method of free-recall Chapter 4.

Mandarin Color Naming Space

A free color naming experiment

Chapter 5. Structural Formation of

Main Color Categories

A 12-terms color sorting experiment

Chapter 6. General Discussion 6.1.Comparing popularity

ranking of color terms between studies 6.2. Comparing with other

studies’ free-naming results

6.3 Comparing free naming and sorting maps within this study

6.4. Future works

8

CHAPTER 2. LTERATURE REVIEW

This chapter introduces the theoretical background of the thesis. It briefly reviews various aspects of lexical color category discussed in the following disciplines: psychophysics, perceptual cognitive, linguistic psychology and anthropological linguistics. Also, the study of color naming in Mandarin will be reviewed.

2.1. Color discrimination versus color categorization

One of the core issues in the long history of color science is the inquiry of the functional unit of color perception. Human eyes perceive distinct hues in the visible radiant spectrum ranged from 400 to 700 nanometers, as Isaac Newton observed three centuries ago. Phenomenally, the apparent chromaticity quality derived from visible lights varies region by region instead of wavelength by wavelength. The attempt of segregating perceptual regions of color had been made intensively by classic color-scaling experiments ( Boynton, 1975). Color coding involves multiple levels starting from the wavelength continuum to color discrimination, and to color categorization and finally the act of naming. Within this complex process of internal percept transformation, the discrimination and categorization are two different stages that serve different purposes and adopt different information processing strategies.

The discrimination capability is associated with how many different wavelengths observers can tell apart. The paradigm of color discrimination experiment is to juxtapose two spectral fields and alter one relative to another systematically across the spectrum to derive a JND ((just noticeable difference, or īŊ) function, which means the degree of wavelength change required to elicit a just noticeable differences in color as a function of the reference wavelength ( Bornstein, 1990). With brightness and saturation controlled, there are approximately 120-150 JNDs among color-normal observers ( Bornstein, 1990), or 200 distinguishable steps (Gouras, 1991a), across the visible spectrum. If considering other

9

perceptual dimensions of brightness and saturation variation in the real world, human can theoretically discriminate millions of colors. Ecologically, the benefit of high resolution in color discrimination affects mainly the object recognition from chaotic and complicated background.

Another stream of processing chromatic light is categorization. This is an information reduction strategy that entails rapid coding of colors. There is evidence supporting the model of parallel processing of continuous and categorical (discrete) information in color perception ( Bornstein & Korda, 1984). Studies on color categorizing questions concern whether observers perceive qualitative similarities of hues among spectral wavelengths (Bornstein, 1990). The illustration in Figure 2-1 gives three psychological levels of color encoding. The ’grain size’ of color codes in these levels, the physical level of visible wavelength, the psychophysical level of color discriminations, and the linguistic-psychological level of color naming, are getting coarser and coarser in that order. Physically, the possible values of wavelengths combination across the visible spectrum are infinite. Psychophysically, the number of discriminable steps of chromatic stimuli are constrained by the visual system, especially for hue discrimination (Gouras, 1991a). The observer’s color sensations only change when the stimulus light cuts across the division between two different wavelength regions; therefore the perceptual color space is virtually partitioned into distinct JND bands instead of a sensory continuum.

10

Figure 2-1. Color encoding levels from stimuli and discrimination to lexical encoding.

The categorical color perception at higher order cognitive level always links with the behavior of lexical color encoding, i.e., using verbal description to represent certain color shades. Language is a highly developed cognitive function affecting many aspects of human behaviors. In fact, the formation and the structure of color categories are tightly bounded with language. The color category is conceptually more abstract on the linguistic level than that on the discrimination level. Consequently, the quantity of distinguishable categories on the linguistic level is drastically reduced. The number of frequently used color terms depends on the linguistic evolution stage. In the well-developed stage, a language usually contains no less than 11 basic color terms (Berlin & Kay, 1969). The endeavor of mapping color terms across languages on the color space leads to the finding that the corresponding chromaticity range of the equivalent color terms may vary across different cultures.

Discriminable wavelengths are categorized into a group due to the fact that they appear perceptually similar and share a dominant hue quality. The categorization effect affects human performance in various tasks ranging from discrimination, recognition to memorization.

11

Between-category color discriminations are more accurate and efficient than equivalent within-category discriminations ( Bornstein & Korda, 1984; Boynton, Fargo, Olson, & Smallman, 1989; Goldstein, Davidoff, & Roberson, 2009). Color categorization is also found to enhance the performance in recognition tasks (Dale, 1969; Goldstein, et al., 2009; Kimball & Dale, 1972; Ostergaard & Davidoff, 1985). This categorical effect is also apparent in color memorization ( Boynton, et al., 1989; Heider, 1972; Seliger, 2002; Uchikawa & Shinoda, 1996). The fundamentality of categorical color perception is supported by empirical evidence of early infant vision studies ( Bornstein, 1985; Bornstein & Kessen, 1976) and recent neurophysiological studies (Franklin, Drivonikou, Bevis, et al., 2008; Holmes, Franklin, Clifford, & Davies, 2009; Roux, Lubrano, Lauwers-Cances, Mascott, & Demonet, 2006). Categorization of color seems to be an innate function, that is, a universal cognitive phenomenon shared by all human beings. However, these findings apparently oppose a language dominancy viewpoint proposed by anthropological linguist Benjamin Lee Whorf (Whorf, 1956). The following session discusses the major debate over color terminology.

2.2. Cognitive linguistic debates: Universalist and relativist views

The regular pattern of sorting continuous lights into discrete categories has provokes researchers’ interest on pondering the relations between world and brain, language and perception. Lexical color category and its implication became an important issue in anthropological linguistics and cognitive science, and induced a controversy lasting for half of a century (Jameson & D'Andrade, 1997; Regier & Kay, 2009; Regier, et al., in press). This is a classic debate on the relation between perception and language. Two opposing stances in linguistic anthropology— universalist and relativist— engaged in this intense debate regarding the dominant hierarchy of thought and language (Dedrick, 1998; Kay & Regier, 2006; Regier, et al., in press).

12

repertoire of thought of the world. The universalist view holds that language is a limited semantic palette shaped and restricted by human cognition. The range of color categories are projected from the universal color foci and therefore located in similar positions in color space across world languages. The first systematic cross-culture color naming survey provides consistent evidence for this view. Berlin and Kay established a pioneering theory of basic color terms by conducting this anthropological survey; they proposed 11 common color terms that are widely used across cultures (Berlin & Kay, 1969). The universal color terms in English are black, white, red, green, yellow, blue, orange, purple, pink, brown and grey. The sequence is based on the developmental order of the terms. This universalist view of the usage of color terms, involving a belief in profound common ground that connects human cultures and minds, has been observed in various types of studies, including cross-culture surveys (Goldstein, et al., 2009; Lin, Luo, MacDonald, & Tarrant, 2001a; Kay & Regier, 2007; Lindsey & Brown, 2006; Lu, 1997; Regier, Kay, & Cook, 2005) , free color-naming tasks (Guest & Van Laar, 2000; Lin, et al., 2001a; Sturges & Whitfield, 1997), developmental studies (Bernasek & Haude, 1993; Bornstein, 1985; Karpf, Gross, & Small, 1974; Pitchford & Mullen, 2002), and psychophysics and physiological experiments ( Boynton & Gordon, 1965; Boynton, Maclaury, & Uchikawa, 1989; Boynton & Olson, 1990; Boynton, Schafer, & Neun, 1964; Holmes, et al., 2009; Ingling, Scheibner, & Boynton, 1970; Roux, et al., 2006; Sakurai, Ayama, & Kumagai, 2003; Sturges & Whitfield, 1997), despite the continued existence of opposing, relativist arguments (Dedrick, 1998; Ozgen, 2002; Roberson, Davies, & Davidoff, 2000) .

In contrast, the relativist view denies the universal foci theory and argues that language shapes thought. The relativist view holds that the human perception of the world is shaped by the semantic categories of the native language, which means the content of mental categories is defined by cultural conventions and would vary across languages. This view is often associated with anthropologist Whorf who hypothesized that language organizes attributes of

13

the world and that linguistic organization in turn influences perception (Whorf, 1956). According to this view, color categories are defined at their boundaries by local language conventions and the corresponding range of a given color category may vary widely across cultures (Roberson, et al., 2000) .

Over the years, the dominant view has swung back and forth between these two poles. The latest authoritative review addressing this issue gives a compromised solution (Regier & Kay, 2009). It concludes that Whorf, the representation of the relativist view, was half right, in two different ways: (1) Language influences color perception partially in the right half of the visual field due to the left hemisphere’s dominance for language (Franklin, Drivonikou, Bevis, et al., 2008; Franklin, Drivonikou, Clifford, et al., 2008). (2) Color naming across languages is shaped by both universal and language-specific forces. Generally, the landmark basic terms, red, yellow, green and blue, behave more universal than other basic color terms.

Besides the traditional frames of universalist versus relativist, there is another theory intends to explain the phenomenal pattern of cross-languages color naming. It proposed that color naming reflects optimal or near-optimal divisions of the irregularly shaped perceptual color space (Jameson & D'Andrade, 1997; Regier, Kay, & Khetarpal, 2007). This hypothesis seems to be supported by tests of the hidden consensus of WCS (world color survey) (Regier, et al., 2007) .

2.3. Color naming methods

Color naming is a frequently used paradigm in related studies. The observers participating color naming experiment may be exposed to chromatic stimuli and were asked to name (or identify to group) the spectral light to derive color-naming functions, that is the percentage of times basic color names are applied to different wavelength. Boynton and Gordan found only four color terms – red, yellow, green and blue – and their combinations are sufficient to describe the perceptual color space exhaustively. These color terms thus were considered as the psychological color elements: unique red, unique yellow, unique green and

14

unique blue ( Boynton & Gordon, 1965). The usage of other color terms such as purple or orange was inconsistent among participants and was considered to be less reliable. It suggests that there are only four basic color labels were regularly and satisfactorily needed when observers were to partition color spectrum.

As reviewed previously, the thought and language of colors has been under extensive investigations. Color terminology and its primary mechanism of categorical color perception are heated topics in part because color is a salient visual feature in most human cultures. Additionally, the ‘thought’ of color can be scientifically defined in the chromaticity space through standard measurement techniques. In other words, the appropriate experimental survey can convert the color semantics from the linguistic domain to the physical domain. The color naming method is widely applied to make attempts to gauge the chromaticity range that corresponds to color term. Color naming method was adopted from the early applied investigation of signal lights ( Halsey, 1959a; Halsey, 1959b), to theoretical issues of visual psychophysics, such as precisely verifying the psychologically primary hues (Sternheim & Boynton, 1969) and characterizing the Bezold Brücke hue shift ( Boynton & Gordon, 1965).

In addition to color naming method, free naming and color sorting task are often used in lexical color studies. The free naming method is good for collecting a large, diverse amount of color name data (e.g. Grant, 1980; Guest & Van Laar, 2000; H., et al., 2001a; Lu, 1997; Sturges & Whitfield, 1995, 1997) , while the sorting method focuses on the corresponding chromaticity range of the color terms in question ( e.g. Lin, Luo, MacDonald, & Tarrant, 2001c; Lu Ching-Fu, 1997; Shinoda, Uchikawa, & Ikeda, 1993). In a typical sorting task, also called the constrained method in some studies,(e.g. Lin, et al., 2001c) observers are given a set of color terms (traditionally, Berlin and Kay’s 11 color terms) as options for sorting the presented color stimuli. This method employs forced-choice tasks and systematic stimulus sampling, which can efficiently bridge each color term and its corresponding area in the chromaticity coordinate. Both free color naming and color sorting tasks would be adopted in

15

the experiments of this thesis.

Besides those task-orientated naming methods, in psycholinguistic researches, the free-recall method without involving any stimuli and naming or matching task is considered to be the most faithful representation of the reference relationships that are peculiar to a given language (Lenneberg, 1967).

2.4. Basic and Secondary color terms

Basic color terms

Every culture or language has indefinitely diverse expressions that denote the perceptual experience of colors. A patch of color is capable of arousing various associations and descriptions. For instance, a simple blue stimulus could elicit the expressions such as watercolor blue, bright blue, turquoise, light sapphire, the color of clear sunny sky…. and so on. However, psychologists and anthropological linguists have long operated with a concept of basic color terms which include simple forms like black, white, red, and green. A basic color term is thought to exhibit the following four features.

(1) It is monolexemic: that is, basic color terms are simplex lexemes; they are lexemes whose meanings are not determinable from the meanings of internal components. ( Casson, 1997; Conklin, 1962).

(2) Its signification is not included in that of any other color term. For example the term scarlet in English or ߧ(chi) in Chinese can be replaced by red or દ(Hong) for most people.

(3) It is general. Its application must not be restricted to a narrow class of object. It should be general if it applied to diverse classes of objects and its meaning is not subsumable under the meaning of another term.

(4) It is psychologically salient. It should be readily elicitable, occurs in the idiolects of most speakers, and is used consistently by individuals and with a high degree of consensus

16

among individuals.

Within the eleven well-tested basic color terms, the uniqueness and affirmed psychophysical foundation of four terms: red, yellow, green and blue are known as the

Landmark basic color terms in related studies. The rest orange, purple, pink, brown, grey are

other basic color terms. Secondary color terms

Secondary color terms, or non-basic color terms, are defined as all color expressions excluding basic color terms. It by definition includes simplex and complex lexemes ( Casson, 1997; Casson, 1994). English terms such scarlet, indigo, rose, emerald and turquoise are simplex lexemes, yellowish orange, light green, orange-red, wine-red and bright pinkish violet are complex lexemes; the above color terms are secondary color terms.

However, in the empirical survey of current study, the huge amount of secondary color terms are further classified into sub-classes such as simplex (monolexemic secondary color terms), complex (polylexemes) and other forms with specific combination pattern, in order to finely organize and analyze the obtained data.

2.5. World Language Survey (WCS)

The WCS is a massive study started in 1976 and carried out by University of California at Berkeley, International Computer Science Institute, Berkeley and University of Chicago. It was designed for two major purposes: to assess the general hypotheses advanced by B&K against a broader empirical basis, and to deepen the knowledge regarding universals, variation, and historical development in basic color-term systems (Hardin & Matffi, 1997). The methods and some initial results of the WCS are reported in Kay, Berlin, and Merrifield (1991). A large number of comparable data on naming ranges and focal choices for basic color terms was collected on 110 languages, but the survey did not cover Mandarin. A methodological departure of the WCS from the method used by B&K was that chip-naming judgments were

17

obtained on individual chip presentations, rather than the full array of stimuli. Judgments of best example (foci color) were obtained in the same way as in the original study, by requesting selection of the chip or chips that best represent each basic color word of the native language from an array of 330 color patches, representing 40 equally spaced Munsell hues at 8 levels of lightness (at maximum saturation plus 10 levels of lightness of neutral colors black, grey, white). The WCS stimuli and the denotation of stimuli are shown in Figure 2-2.

All of the WCS data and detailed method is open to public for comparative studies (available in http://www.icsi.berkeley.edu/wcs/data.html). The valuable information would be a footstone of comprehending color terminology in all languages.

Figure 2.2. Stimulus array used in WCS

2.6. Color naming in Mandarin

The linguistic class that Mandarin belongs to is Sino-Tibetan, Han Chinese. It was originated from northern China and spoken widely by Modern Chinese around the world. Mandarin is the official language in China, Taiwan, Hong Kong and Macaw. The early color term study conducted by B&K classified Mandarin as a language in the developmental Stage V, which holds six color terms, white, black, red, yellow, green and blue. The foci color of these terms can be found in the origin data plot shown in Figure 2-3 (Berlin & Kay, 1969) .

cond varie deve terms cases a ter non-n plann this l the v the c Chap appro thorn regio Figu However, ducted in 19 ety of color lopment of s, their uni s” (in the se rtiary term, native spea ned to obtai language, e variety and t current stud pter 6. In addition opriate tran ny problem ons, times a ure 2.3 Foci co this conclu 990s refuted r terms in u f color terms iversality an ection 2.5, p , due to th akers. Altho in more dat either. Fortu the range of dy. The com

n to the em nslation bet

in the prese and speakers

olor data gain

usion remai d this findin use (Lu, 19 s in Mandar nd evolutio p41-42). On he inconsis ugh B&K t a for this la unately a fe f Mandarin mparison of mpirical an tween Engl ent study, d s. There is c 18 ned form Mand

ins controv ng and made 997). Actua rin. In their on, Mandari ne of the un stency of th tentatively t anguage in f ew recent st color terms f current re nd theoretic lish color t due to the w currently no darin informan versial for l e an attemp ally, B&K l report on c in was trea nsolved issu he data co treated Man further rese tudies did r s and, thus, esults with cal vacancy terms and M wide varianc o consensus nts (Berlin & local color pt to provide left many o cross-langua ated as one ues was whe ollected from

ndarin as St arch, the lat recruit nativ provided so these studi y of Mand Mandarin c ce in Manda s on the con Kay, 1969). researcher e evidence o open questi age survey B of the “pro ether grey i m limited tage V at th ter WCS di ve speakers ome referen ies would b darin color color terms arin color te nvention of rs. A study of a greater ons for the

Basic color

oblematical s a basic or number of hat time and d not cover s to address nce base for be given in terms, the is another erms across color terms y r e r l r f d r s r n e r s s

19

in contemporary Mandarin. Therefore, it is inappropriate to assign basic color names simply by following an existing dictionary. Additionally, the definitions of color terms in Chinese are more ambiguous than in English. There are diverse wording choices to describe the same color category. For example, the brown category can be conveyed by distinct terms like࠼೽

Ka-fei (the phonic translation of coffee), ཝ Tsong (palm fiber, coir) or ᓣ He (tan). Similarly, multiple color categories can be expressed with identical color terms. The ancient polysemous term ॹ Ching can refer to blue, green and sometimes purple and black. Though there are some studies concerning the usage of basic Mandarin color terms ( Lin, et al., 2001a, 2001c; Lu, 1997), the translations were unfortunately not consistent, particularly for non-landmark basic terms. The term pink can be translated in two different ways: ృદ Fenhong ( Lin, et al., 2001a) and௒ Tao (Lu, 1997). Brown can be both ཝ Tsong ( Lin, et al., 2001a) and ᓣ He (Lu, 1997), and orange can be ᖪ Ju ( Lin, et al., 2001a) and ᖨ Cheng (Lu, 1997).

20

CHAPTER 3. COLOR LEXICON SURVEY IN MANDARIN

This chapter presents the local survey of color lexicon with the method of free-recall task. The whole survey was designed to compare with the similar survey conducted in 1990s, but with modified procedure. The purpose and methodological details is given in Section 3.1 and 3.2. The statistical results are summarized in Section 3.3.

3.1. Purpose

The free-recall survey was conducted for the purpose of obtaining the data of color vocabulary which is synchronic (present linguistic phenomenon without concerning the factor of time, contrast of diachronic). Specifically, the obtained data would be the representation of these aspects: (1) popular color terms currently in use by native Mandarin speakers resident in Taiwan, (2) The popularity rank of those color terms, and (3) the diversity of those color terms. A similar survey employed systematic anthropological paradigm was conducted and published in the 1990s (Lu, 1997). The current results would be compared with that in the work and another Japanese study in Chapter 6. A historical change and cultural difference of the popularity of color terms can be expected in that comparison.

In addition to the purpose of surveying color term lexicon, another purpose of this experiment is to the collect conventional Mandarin color terms that would serve as options in the later color sorting experiments in Chapter 5 i.e., 12-terms sorting experiment. There is no consistent translation of Mandarin color terms in previous studies. To solve this problem, the color terms in Chinese Mandarin written form that were to be used in the sorting experiment and future works involving Mandarin color terms would be filtered by this survey. Only the terms that emerged most frequently from the free-recall task were employed.

21

This survey was executed without any reference resources in order to elicit the most intuitive and tangible color terms currently in use. There were totally 189 informants performed this free-recall task. The voluntary participants are native Mandarin speakers aged 16-45, education level spanned from high school to PhD, opportunity sampled from undergraduates, postgraduates, engineers, businesses, labors, designers, public officials, home makers, academic researchers, school staffs and teachers. The percentage of visual arts or design professionals was under 10%. The informants were provided with a blank sheet, and a pencil or pen. The task instruction was to “write down color terms/ vocabularies you frequently use, hear and read.” After the frequent terms recalling was done, the author encouraged the informant to freely recall color terms as many as he/she could and also write them down on the sheet. The informants were allowed to use any written forms in Chinese as long as the written color terms are recognizable to the researcher. Most informants used traditional Chinese characters, and simplified characters and notional phonetic alphabet were occasionally used. There was no time limit for informants, but they were invited to recall color terms as hard as possible. The recalling task typically took 5-20 minutes. The whole data collection works started from May 2009 to Feb. 2010, and each informant performed the task individually with the instruction of the researcher.

3.3. Results

There were 5102 color terms produced by 189 informants. Several types of color terms were considered to be insufficient for representing both the frequency and diversity of color terms, thus were removing form the raw data. One of the discarded type is those that uses common tone-modified adjectives, e.g. light red, dark red, bright red, vivid red and so on. Another discarded type is a compound of any two or more common color terms, e.g. red-yellow, blue-purple and so on. Although those color terms are valid for describing real color experience, all of them can be classified or decomposed into known color categories.

22

Moreover, some other terms were also excluded if they were judged to be inadequate in describing the perceptual quality of a simple color, e.g. transparency, fluorescence, neon and so on. Some terms referring to a certain series of colors, e.g. rainbow, candies, sunset, autumn leaves and so on, were eliminated, too. A few unrecognizable color terms were screened out as well. After filtering out those color terms, there are 4961 color terms left, containing 2493 items that were of frequent use and 2468 items that were recalled less consistently.

3.3.1 Descriptive overview

The descriptive statistics of the results are given in Table 3-1. The informants were able to produce 26.25 valid color terms on average, but the individual difference is noticeable in the performance of the later task that solicits more color terms. For the task of recalling terms in frequent use, the central tendency indexes and the STD is relatively small. For visualizing the distribution of the counts of collected terms, Figure 3-1 shows the histograms of the frequency counts of total amount (a), frequent use (b) and extended color terms (c). The results of Gaussian curve fitting capturing these three distributions are presented as well. In the case of frequent use (c) there is a peak at the value of 12, while in the case of extends (b), the curve appears to be leaning leftwards. The distortion of total frequency count (a) roughly fits a normal distribution model but leaves a long right-side tail. In general, these descriptive statistics indicate the capacity of color lexicon produced from the sampled population. There is a significant individual difference in the amount of less frequent terms.

Table 3.1 Descriptive statistics of the color term recall

Count Mean Mode Median STD Max Min Total 4961 26.25 20 25 10.7 58 8

Frequent use 2493 13.19 12 12 3.69 26 6 Extends 2468 13.05 8 10 9.12 42 2

23

3.3.2 Classification of the collected color terms

The free recall survey gathered a great quantity of color terms with an extraordinary diversity. These samples include the basic color terms representing common color categories mentioned by previous lexical color category related studies. Also, equipped with the flexibility and complexity that Chinese language holds, the informants produced a lot of ornate color expressions. These various color terms were classified into different types in a hierarchical manner as Figure 3-2 shows. The chart presents the organization of Mandarin color terms classification. The numbers are the proportion of each class relative to the total amount (4961) of collected terms, and some block stands for subclasses. The classifying method was extended from some of the previous color naming studies with different tasks, e.g. Guest’s free-naming experiment (Guest & Van Laar, 2000).

Figure 3.1. Histograms of frequency count of total amount (a), frequent use (b) and the extends (c).

(a) (b)

24

Figure 3.2. A hierarchical classification of Mandarin color terms based on the collected data in the recalling study.

The collected terms were first sorted into either monolexemic or compound classes. The criterion of the monolexemic color term is that it uses only single color vocabulary or noun, or lexeme in the linguistic term. Actually, the participants in many previous color naming studies were constrained with using only monolexemic color terms to generate response (Sturges & Whitfield, 1995, 1997). Some terms contain two or more Chinese characters but indicate a single color category are considered belonging to this class, e.g. ࠼೽ۥ(brown) or ࿫ᢅᥞۥ (violet). The compound color terms are defined as those consisting of two or more monolexemic color terms, or terms with the adjective. However, if the modifier of the compound is used to describe different shades of same color category, examples in Mandarin

Mandarin colorterms monolexemic 71.38% basic 38.8% landmarkbasic 15.86% red,green, yellow,blue otherbasic 22.94% e.g.orange, purple,brown antique 9.1% e.g.ߧ,Chi secondary 41.66% naturalobjects 33.43% e.g.rice (ۏۥ, MiͲse) artifitialobjects 5.11% e.g.brick (ᗤۥ,ZuanͲse) loan 3.12% e.g.Champagne (ଉឳۥ,XiangͲbin) compound 28.62% nounͲbasic 18.14% e.g.skyͲblue pigmentͲbasic 8.43% e.g.Cobaltblue commercialͲ basic2.05% e.g.TiffanyͲblue

25

are ෡, ෍, ᖺ, ෉, ॽ, ᄆ, (deep, light, dense, pole, bright, dark), that compound is discarded. The subclasses under the monolexemic and compound are defined by the criteria of basic-or-not and origin. There are four subclasses under the monolexemic class:

(1) The basic color term (38.8%): includes four landmark basic terms red, green, yellow and blue, and 7 other basic terms, black, white, grey, purple, pink, orange and brown. (2) The antique color term (9.1%): includes color terms that are found in classic Chinese

literature or used by ancient Chinese. Some of them are still in use in more formal occasions or in idiomatic phrases. For instance ߧ Chi and ڹ Zhu represent red, and ॹ Ching represents blue, green or black.

(3) The secondary color term (58.4%): by definition this class should contains all the color terms except for the basic color terms. But in current context the class is defined as monolexemic color terms except for basic and antique color terms. These terms can be further distinguished as deriving from either natural or artificial materials. The color terms in this subclass demonstrates many culture signatures compared to the previous reports of secondary color terms in English. For instance, the term ಁ Cha(tea) represents a reddish dark brown and could be similar to walnut in English, andᢎ Ou (lotus root) represents a unique shade of pale, pinkish grey. Some man-made materials, e.g. ᗤ Zhan (brick), represent brownish red or orange. Also there are many secondary color terms that were transliterated from foreign languages and do not have local origins, such as ࠼೽(café) and ଉឳ(champagne). They represent the general brown category and a glimmering light yellow respectively.

The sub-classification of compound color terms is rather challenging since their variability is larger than that in monolexemic class. But they are all comprised of a modifier of various origins and one common basic color terms. There are four distinguishable subclasses:

(1) Noun-basic color terms (18.14%): includes compound terms consisting of a noun referring to a specific object and a basic color terms, e.g. sky-blue, tomato-red,

26

lake-green, grass-green, ivory-white, wine-red and so on. Many of these terms are modified with common objects thus can be directly translated into English without much distortion. Such terms can convey more specific color quality than a single basic, or a single secondary color term.

(2) Pigment-basic color terms (8.43%): includes compound terms used in the context of painting, glazing or dyeing industry, and the specialized modifier in front of the basic term is related to the chemical or cultural origin, e.g. ሣ៴ (cobalt-blue), ᢏ႓ (gamboges yellow), Ւ ۘ ࠡ ៴ , (Turnkey-blue), ੉ દ (Western-red), ཏ ᕙ Փ ៴ (Prussian-blue). Some terms originated from the pigments of Chinese painting but many others are loanwords or translated ones.

(3) Commercial-basic color terms (2.05%) : These are compound terms comprised of the name of a celebrated brand or well-known commercial image, including four consensus terms Tiffany blue, Ferrari red, Kitty pink, and Hermes orange. These terms were rarely recalled, but to whom exposed to these commercial images, they are precisely linked to that very unique color shade representing the whole image of the brand or the product series.

The above classification principles are set for clarifying the common composition of Mandarin color terms. Two terms fits the characteristics of compound but stands for the color category. They are pink, i.e., Fen-Hong (ృદ, directly translated as powder-red) and Tao-Hong (௒ દ , directly translated as peach-blossom-red), which are classified into monolexemic class due to the shortage of a simplex lexeme representing pink in Mandarin. Also, the sub-divided types under the monolexemic class are not mutually exclusive. Specifically the antique are overlap with the secondary type.

3.3.3 Popularity of the monolexemic color terms

27

monolexemic terms.

Table 3.2. Rank list derived from monolexemic Chinese color terms. R

k.

Ct Ch. Ph.Trans .

English Freq.use Extends Subclass ct % ct %

1  18 દ Hong red 182 97.3 5 2.67 Landmarkbasic 2 18 ៴ Lan Blue 184 99.4 1 .54 Landmarkbasic 3 18 ࿫ Zi Purple 176 95.6 8 4.35 Otherbasic 4 18 ጸ Lu Green 178 97.8 4 2.20 Landmarkbasic 5 17 ᖪ Ju Orange/tangerin 174 97.2 5 2.79 Otherbasic 6 17 ႓ Huang Yellow 169 94.9 9 5.06 Landmarkbasic 7 17 ృદ Fenhong Lightpink 145 82.8 30 17.1 Compound/Otherbasic 8 16 ࠼೽ KaͲFei Brown/Coffee 123 75.9 39 24.0 Otherbasic

9 15 ᖨ Cheng  Orange 105 66.0 54 33.9 Otherbasic

1 15 ᓣ He Brown/tan 94 59.8 63 40.1 SecondaryͲnatural/Otherbasic 1 14 ۊ hui Gray/ash 126 85.1 22 14.8 Otherbasic

1 13 ௒દ Taohong Darkpink 107 78.1 30 21.9 Compound/Otherbasic 1 13 ཝ tsong Brown/palm 88 65.6 46 34.3 SecondaryͲnatural/Otherbasic 1 13 ᙨ Dian deepblue/indigo 62 47.3 69 52.6 SecondaryͲartificial

1 11 ॹ Ching Green/Blue/black 55 47.4 61 52.5 Antique

1 10 ᓅ Fu Skin 89 84.7 16 15.2 SecondaryͲnatural 1 93 ८ Jin Gold 51 54.8 42 45.1 SecondaryͲnatural 1 89 ۏ Mi Rice/Beige 64 71.9 25 28.0 SecondaryͲnatural 1 88 ಁ Cha Tea 52 59.0 36 40.9 SecondaryͲnatural 2 85 ߧ Chi Red 7 8.24 78 91.7 Antique

2 84 Ꭼ Yin Silver 56 66.6 28 33.3 SecondaryͲnatural 2 51 ࿫ᢅᥞ Zilolan Violet 2 3.92 49 96.0 SecondaryͲloan  2 39 ႕ Hei Black 39 100. 0 .00 Landmarkbasic 2 38 ׬ࠡ KaͲqi Khaki 32 84.2 6 15.7 SecondaryͲartificial 2 36 ᕊ Tuo Camel 17 47.2 19 52.7 SecondaryͲnatural 2 35 ػ Bai White 35 100. 0 .00 Landmarkbasic 2 34 ޖ Xing Apricot 15 44.1 19 55.8 SecondaryͲnatural 2 32 ᖬᨛ Ganlan Olive 12 37.5 20 62.5 SecondaryͲnatural 2 32 Ւ Tu Soil/Earth 14 43.7 18 56.2 SecondaryͲnatural 3 30 फጇ Meigui Rose 11 36.6 19 63.3 SecondaryͲnatural 3 29 ᗤ Zhuan Brick 20 68.9 9 31.0 SecondaryͲartificial 3 28 ᢎ Ou Lotusroot 18 64.2 10 35.7 SecondaryͲnatural 3 26 ଉឳ Xiangbin Champagne 12 46.1 14 53.8 SecondaryͲartificial/loan 3 25 ڹ Zhu Red/Cinnabar 4 16.0 21 84.0 Antique/SecondaryͲ artificial 3 24 ՛ຽ Xiaomai wheat 9 37.5 15 62.5 SecondaryͲnatural

28

3 23 ߞ Yu Taro 10 43.4 13 56.5 SecondaryͲnatural

3 20 ᕠ Mo Ink/black 8 40.0 12 60.0 Antique/SecondaryͲ artificial 3 18 ؊ई NaiͲyou Cream 6 33.3 12 66.6 SecondaryͲnatural

3 13 ፈፇ FeiͲcui Emerald 1 7.69 12 92.3 SecondaryͲnatural 4 10 ײᎭ Gutong Bronze 3 30.0 7 70.0 SecondaryͲnatural 4 9 خ Xuan Dark/Black 0 .00 9 100. Antique

4 8 ੱᅔ Shanhu Coral 2 25.0 6 75.0 SecondaryͲnatural

4 7 ጘ Bi Jasper 1 14.2 6 85.7 Antique/SecondaryͲ natural 4 7 ᔈ Zhe Sienna/ocher 0 .00 7 100. Antique/SecondaryͲ natural 4 6 ੴఇ Zhenzhu Pearl 1 16.6 5 83.3 SecondaryͲnatural

4 2 ⋁ Gan Darkpurple 0 .00 2 100. Antique/SecondaryͲ natural

The circulation and distribution of monolexemic color terms are the most representative information for probing the color lexicon within a language. Table II provides the statistical results of collected monolexemic terms.The columns from top left to right are the ranking order, denoted as Rk., the frequency count (Ct.) of each, the color terms in written Chinese (Ch.), in phonic transliteration (Ph.Trans.), in English, the frequency counts (ct) and percentage (%) of the term recalled in the prior task regarding color terms in frequent use (Freq. use), the same data in later task of recalling more extends, and the last column denotes the subclass each term belongs to, some terms are qualified for more than one subclass. For visualizing the quantity and proportion of the monolexemic terms, the histogram shown in Figure 3-3 represents the frequency count of each recalled monolexemic color term, with black and white fill to distinguish either the term was mentioned in the frequent use task or the extends task. The total count of each term are positively correlated to that count in the frequent use task (Pearson correlation coefficient=0.95). The level of 100%, 75%, 50% and 25% threshold are marked with horizontal dash-lines.

29

Figure 3.3. The frequency count histogram of monolexemic color terms.

In general, the results shown in above table and histogram reveal the popularity and the degree of consensus of the collected monolexemic color terms. The top 15 terms encompass B&K’s universal color categories, though some overlapping terms in the same categories are evident. For instance, the term 8th, 10th and 13th all refer to brown category, and 7th and 12th can stand for pink category, and 5th and 9th for orange category. These rank order data provide the base for selecting proper Chinese translations in subsequent color-term-sorting experiments. Black and white were not frequently recalled as they ranked in the 23th and the 26th respectively, perhaps due to many informants account these two achromatic terms.

The more popular terms are distinct color concepts featuring in hue identity, e.g., red, blue, purple, green, orange and so on, and all these are frequently used in modern society and are legible in both colloquial and writing. The count of terms decreases with its

30

distinctiveness in color appearance, but those less frequently recalled terms can represent a delicate tone by using a specific object as a metaphor. There is a plunge between the 22th and the 23th, all terms from there on were counted fewer than 25%, and half of the collected monolexemic terms are below this level.

It is worth noting that many different terms refer to a same brown category. The informants produced multiple secondary terms, e.g. ࠼೽(coffee), ᓣ(Tan), ཝ(Palm), ಁ (Tea), ᕊ(Camel), Ւ(Soil/Earth), that are linked to colors of brown category. However, from the current recalling data, it is difficult to tell whether these could indicate different shades within brown category, or they are just different individual idioms for describing a similar color range. This question would be clarified in the free-naming experiment presented in Chapter 4.