華語母語者對粵語聲調之範疇類比研究

全文

(1)國立臺灣師範大學英語學系碩士論文 Master Thesis Graduate Institute of English National Taiwan Normal University. 華語母語者對粵語聲調之範疇類比研究 Categorical assimilation of Cantonese tones by native speakers of Mandarin. 指導教授：曾. 金. 金. 博士. Advisor: Dr. Chin-Chin Tseng 研究生：劉. 又. 慎. Student: Yu-Shen Liu. 中華民國一○二年八月 August, 2013.

(2) 中文摘要本研究針對華語母語者接觸非母語聲調的初期情形進行跨語言調查，招募來自中國大陸北方的普通話母語者與台灣的國語母語者，了解受試者對非母語之粵語聲調的感知與產出情形，並透過受試者對同一非母語聲調的不同範疇類比調查普通話與國語受試者的母語聲調範疇差異。本研究包含標記、辨認與發聲三項實驗；在標記實驗中，調查零起點華語母語者對粵語聲調的印象，了解受試者將粵語聲調類比至母語聲調的情形，之後透過辨認與發聲實驗，研究受試者區辨非母語聲調的難易程度和混淆情形，並於實驗結束後對受試者進行問卷調查，了解其語言背景與對實驗和粵語聲調的感想。除了報告華語母語者在學習粵語過程中可能出現的難點與混淆之外，本研究也試圖以 Best 的感知範疇模型來為結果提出解釋（Perceptual Assimilation Model，PAM），顯示受試者語言變體與方言背景的差異可以對非母語聲調的感知與產出造成影響，應在進行相關研究時納入考量。. 關鍵字：跨語言聲調感知、範疇類比、粵語、普通話、國語 . II.

(3) ABSTRACT The present study is a cross-language study on the non-native tonal experience of native speakers of Mandarin in the initial stage. The study takes possible dialectal influences into account and recruits speakers of two varieties of Mandarin, Putonghua and Taiwan Mandarin, to conduct a two-way investigation into how subjects of both groups perceive and produce the non-native Cantonese tones, and how they interpret their own native tone categories during the categorical assimilation of non-native tones. A series of three tasks, Labeling, Identification, and Vocalization, are conducted in the study: the Labeling task is an impressionistic investigation on naïve native speakers of Mandarin to see how they assimilate non-native Cantonese tones to native Mandarin ones. Subjects are then primed with a brief introduction to the Cantonese tonal system before performing Identification and Vocalization tasks, where they discern and vocalize the newly-learnt Cantonese tones. Stimuli of these tasks comprise sets of tonal minimal pairs that differ in vowels and syllable structures, and are embedded in a carrier sentence. Subjects fill out a questionnaire in the end with regard to their language backgrounds, impression of Cantonese tones, and selfestimation of their performance. Results of the study show that variance in subjects' varieties and dialectal backgrounds, even within the same language, may have its influence on the perception and production of non-native tones and should be taken into account when undergoing similar investigations. Apart from reporting possible difficulties and common confusions that native speakers of Mandarin may encounter when learning Cantonese tones, the study also attempts to account for the results on the basis of Best’s Perceptual Assimilation Model (PAM). Keywords: cross-language tonal perception, categorical assimilation, Cantonese, Putonghua, Taiwan Mandarin III.

(4) ACKNOWLEDGEMENTS I would like to express my deepest gratitude to my thesis adviser, Dr. Tseng ChinChin, and the commission members of my oral defense, Prof. Zhu Chuan and Dr. Kwock-Ping Tse. Without your criticism and support, this thesis would never come to exist. I would also like to thank my family—in particular my incredibly talkative younger brother—for your forbearance, as well as my classmates and teachers in the English Department, the good people in our Department Library, my Putonghua and Taiwan Mandarin subjects and all the kind people who are willing to participate in the research for your assistance and encouragement. Many thanks go to Dr. An-Guor Wang, Dr. Chia-Hua Yeh, Dr. En-Mei Wang, Yan-Zhi Chen, and Hong Kong moviemakers. Finally, I am truly indebt to my soulmates and roommates for the solace and mirth that you bring with your very existence. Thank you, all my sassy companions on this rough journey. I treasure every moment we spent together.. . IV.

(5) 1. TABLE OF CONTENTS. CHINESE ABSTRACT .............................................................................................II ABSTRACT............................................................................................................... III ACKNOWLEDGEMENTS .................................................................................... IV TABLE OF CONTENTS ..........................................................................................V LIST OF TABLES ..................................................................................................VII LIST OF FIGURES .............................................................................................. VIII CHAPTER ONE: INTRODUCTION ........................................................................1 1.1 Learning Cantonese tones..........................................................................................1 1.2 Research questions .....................................................................................................2 1.3 Thesis organization.....................................................................................................2 . CHAPTER TWO: LITERATURE REVIEW...........................................................4 2.1 Cantonese, Mandarin, and Southern Min tones ......................................................4 2.1.1 Cantonese tones.....................................................................................................4 2.1.2 Mandarin tones ......................................................................................................5 2.1.3 Taiwanese tones ....................................................................................................7 2.1.4 Entering tones in Mandarin, Cantonese, and Taiwanese ......................................7 2.2 Models of Cross-language perception.......................................................................8 2.3 Studies on cross-language tone perception.............................................................10 . CHAPTER THREE: METHODOLOGY ...............................................................13 3.1 Overview ....................................................................................................................13 3.2 Subjects......................................................................................................................13 3.3 Materials....................................................................................................................15 3.4 Procedures .................................................................................................................18 3.4.1 Labeling...............................................................................................................18 3.4.2 Priming ................................................................................................................18 3.4.3 Identification .......................................................................................................19 3.4.4 Vocalization ........................................................................................................20 3.4.5 Questionnaire ......................................................................................................22 3.4.6 Administration.....................................................................................................22 3.5 Instrumentation ........................................................................................................23 3.6 Data analysis .............................................................................................................23 . CHAPTER FOUR: RESULTS .................................................................................25 4.1 Labeling .....................................................................................................................25 4.1.1 Overview .............................................................................................................25 4.1.2 Subject groups .....................................................................................................28 V.

(6) 4.1.2.1 Categorization of Cantonese tones ............................................................................28 4.1.2.2 Tone labels between subject groups..........................................................................31 . 4.2 Identification .............................................................................................................37 4.2.1 Overview .............................................................................................................37 4.2.2 Subject groups .....................................................................................................39 4.2.2.1 Accuracy....................................................................................................................39 4.2.2.2 Confusions.................................................................................................................40 . 4.3 Vocalization ...............................................................................................................46 4.3.1 Overview .............................................................................................................46 4.3.2 Subject groups .....................................................................................................48 4.3.2.1 Accuracy....................................................................................................................48 4.3.2.2 Confusions.................................................................................................................51 . 4.4 Questionnaire ............................................................................................................57 4.4.1 Tasks....................................................................................................................57 4.4.2 Tones ...................................................................................................................58 . CHAPTER FIVE: DISCUSSION.............................................................................61 5.1 Subjects of Putonghua versus subjects of Taiwan Mandarin...............................61 5.1.1 Difference between varieties: Mandarin Tone 3 in Putonghua and Taiwan Mandarin..........................................................................................................................61 5.1.2 Differences between dialectal backgrounds: Taiwanese and northern vernaculars 62 5.1.2.1 The presence and absence of entering tones .............................................................62 5.1.2.2 F0 sensitivity: Contour tones versus level tones .......................................................63 . 5.2 Assimilation of tonal categories...............................................................................64 5.2.1 Categorical assimilation of Cantonese tones.......................................................64 5.2.2 Non-native tonal contrasts on the basis of tonal assimilation .............................66 5.2.3 Tonal categories in production ............................................................................68 . CHAPER SIX: CONCLUSIONS .............................................................................70 6.1 Summary ...................................................................................................................70 6.2 Contribution of the study.........................................................................................71 6.3 Limitations and future research..............................................................................71 . REFERENCES...........................................................................................................73 APPENDICES ............................................................................................................77 . VI.

(7) LIST OF TABLES. Table 1.. Perceptual assimilation patterns of non-native contrasts in PAM..... 10. Table 2.. Cantonese tonal minimal pairs.......................................................... 26. VII.

(8) LIST OF FIGURES. Figure 1.. Tone contours of six Cantonese tones .............................................. 4. Figure 2.. Tone contours of four Mandarin tones ..............................................6. Figure 3.. Labeling of Cantonese tones by Mandarin-speaking subjects .......... 26. Figure 4.. Labeling of Cantonese tones by Mandarin-speaking subjects (nonentering tones and entering tones) .................................................... 27. Figure 5.. Categorization of Cantonese tones by Putonghua subjects............... 28. Figure 6.. Categorization of Cantonese tones by T Mandarin subjects.............. 29. Figure 7.. Labeling of Cantonese tones by Putonghua subjects......................... 30. Figure 8.. Labeling of Cantonese tones by T Mandarin subjects....................... 30. Figure 9.. Labeling of Cantonese non-E Tone 1 by both subject groups........... 32. Figure 10.. Labeling of Cantonese E Tone 1 by both subject groups................... 32. Figure 11.. Labeling of Cantonese Tone 2 by both subject groups...................... 33. Figure 12.. Labeling of Cantonese non-E Tone 3 by both subject groups........... 33. Figure 13.. Labeling of Cantonese E Tone 3 by both subject groups................... 34. Figure 14.. Labeling of Cantonese Tone 4 by both subject groups...................... 34. Figure 15.. Labeling of Cantonese Tone 5 by both subject groups...................... 35. Figure 16.. Labeling of Cantonese non-E Tone 6 by both subject groups........... 36. Figure 17.. Labeling of Cantonese E Tone 6 by both subject groups................... 36. Figure 18.. Overall accuracy in Identification task...............................................38. Figure 19.. Overall accuracy in Identification task (non-E tones and E tones) ... 38. Figure 20.. ID accuracy between subject groups.................................................. 39. Figure 21.. ID accuracy of non-entering tones between subject groups............... 40. Figure 22.. ID accuracy of entering tones between subject groups...................... 40. Figure 23.. Answers for Tone 1 stimuli from Putonghua subjects....................... 41. Figure 24.. Answers for Tone 1 stimuli from T Mandarin subjects..................... 41. Figure 25.. Answers for Tone 2 stimuli from Putonghua subjects....................... 42. Figure 26.. Answers for Tone 2 stimuli from T Mandarin subjects..................... 42. Figure 27.. Answers for Tone 3 stimuli from Putonghua subjects....................... 43. Figure 28.. Answers for Tone 3 stimuli from T Mandarin subjects..................... 43. Figure 29.. Answers for Tone 4 stimuli from Putonghua subjects....................... 44. Figure 30.. Answers for Tone 4 stimuli from T Mandarin subjects..................... 44. Figure 31.. Answers for Tone 5 stimuli from Putonghua subjects....................... 45 VIII.

(9) Figure 32.. Answers for Tone 5 stimuli from T Mandarin subjects..................... 45. Figure 33.. Answers for Tone 6 stimuli from Putonghua subjects....................... 46. Figure 34.. Answers for Tone 6 stimuli from T Mandarin subjects..................... 46. Figure 35.. Overall performance as task proceeds................................................ 47. Figure 36.. Overall performance in Vocalization task.......................................... 47. Figure 37.. Overall performance in Vocalization task (non-E tones and E tones) ............................................................................................................ 48. Figure 38.. Vocalization performance by Putonghua subjects as task proceeds.. 49. Figure 39.. Vocalization performance by T Mandarin subjects as task proceeds 49. Figure 40.. Vocalization performance between subject groups............................ 50. Figure 41.. Vocalization performance of non-E tones between subject groups... 50. Figure 42.. Vocalization performance of E tones between subject groups.......... 51. Figure 43.. Production performance for Tone 1 stimuli of Putonghua subjects... 52. Figure 44.. Production performance for Tone 1 stimuli of T Mandarin subjects. 52. Figure 45.. Production performance for Tone 2 stimuli of Putonghua subjects... 53. Figure 46.. Production performance for Tone 2 stimuli of T Mandarin subjects. 53. Figure 47.. Production performance for Tone 3 stimuli of Putonghua subjects... 54. Figure 48.. Production performance for Tone 3 stimuli of T Mandarin subjects. 54. Figure 49.. Production performance for Tone 4 stimuli of Putonghua subjects... 55. Figure 50.. Production performance for Tone 4 stimuli of T Mandarin subjects. 55. Figure 51.. Production performance for Tone 5 stimuli of Putonghua subjects... 56. Figure 52.. Production performance for Tone 5 stimuli of T Mandarin subjects. 56. Figure 53.. Production performance for Tone 6 stimuli of Putonghua subjects... 57. Figure 54.. Production performance for Tone 6 stimuli of T Mandarin subjects. 57. Figure 55.. Impression and accuracy among tasks............................................... 58. Figure 56.. Impressions on perception and production performance of Cantonese tones................................................................................................... 59. Figure 57.. Perception and production performance of Cantonese tones............. 59. IX.

(10) CHAPTER ONE: INTRODUCTION. 1.1. Learning Cantonese tones. Cantonese, a language originated in the southern China, is widely spoken in Guangdong Province, Hong Kong and Macau. Owing to the intensive interaction of the prosperous Guangdong area with the world, Cantonese is one of the dominant languages in overseas Chinese communities in Southeast Asia, the United States, Canada, and Australia. Apart from immigrant heritage, the economic and media influence of Hong Kong and Macau area has also given the Cantonese language considerable significance around the world. Since Cantonese is widely used in Hong Kong, where people from across the Taiwan Strait constantly interacts, Mandarin speakers have a great chance encountering and learning Cantonese, giving studies on Cantonese acquisition its necessity and practical value. However, as compared to Mandarin, the Cantonese language is less well studies in cross-language studies, which gives rise to the present study the motivation to investigate the auditory perception and production of Cantonese tones by native speakers of two different varieties of Mandarin, namely Putonghua (Mainland China) and Taiwan Mandarin (Taiwan). It is hoped that the present study may provide empirical evidences for future writing of teaching materials, serve pedagogical needs as well as enhance the mutual understanding of languages and peoples across the strait. Teaching materials of the Cantonese language is relatively common among Chinese languages, including those written for foreigners and those written for native speakers of Mandarin. However, in most of these materials, the tone system is either briefly touched or totally ignored. For a language with a relatively more complex tonal system (at least in the number of tones) than Mandarin, a slight deviation in tonal heights may result in drastic change in word meaning and implication. Given the 1.

(11) considerable chances for an innocent utterance to go astray and become profanity (Pang 2007), it is proposed that the teaching of Cantonese tones should be treated with greater emphasis.. 1.2. Research questions. The present study intends to investigate the following aspects: First, on the basis of Best’s Perceptual Assimilation Model (PAM), the study intends to investigate how native speakers of Mandarin, a tone language, perceive and produce Cantonese tones, a foreign tone system different from their native one, in the initial stage. Second, given the differences in varieties of Mandarin, the study intends to look into whether speakers of Putonghua and Taiwan Mandarin perceive and produce Cantonese tones differently? To what extent does varying Mandarin varieties influence their perception and production of Cantonese tones? Finally, will the bilingualism of Mandarin and Southern Min, another southern language that is prevalent in Taiwan and contains seven tones, has any impact on subjects' perception and production of Cantonese tones?. 1.3. Thesis organization. The thesis contains six chapters. Chapter One introduces the motivation and purpose of the present study. Chapter Two reviews previous studies in cross-language perception as well as tonal systems of the languages in questions. Details of investigations conducted in the present study are provided in Chapter Three and the 2.

(12) results are reported in Chapter Four. Chapter Five discusses findings of the experiments and their implications with regard to the research questions. Chapter Six summarizes the study and further discusses its implications and limitations.. 3.

(13) 2. 2.1. CHAPTER TWO: LITERATURE REVIEW. Cantonese, Mandarin, and Southern Min tones. 2.1.1 Cantonese tones Similar to Mandarin, Cantonese is a lexical tone language, where differences in fundamental frequencies are applied to contrast lexical meanings (Yip 2002). There are six phonemic tones in Cantonese1, including three level tones High Level [55] (Tone 1), Mid Level [33] (Tone 3), and Low Level [22] (Tone 6), two rising tones High Rising [35/25] (Tone 2) and Low Rising [23/13] (Tone 5), and one falling tone Low Falling [21/11] (Tone 4)2 (Fok Chan 1974, Matthews & Yip 1994, Bauer & Benedict. 1997,. Figure 1.. 1. Zeng. 2002,. Rao. et. al.. 2009).. Tone contours of six Cantonese tones. Here refers to Hong Kong Cantonese, the variety of Yue dialect that is widely used in the area.. 2. Some descriptions of the language portray the tone as [11]. In this case, Cantonese would be a language with four level tones. 4.

(14) It has been argued whether High Level (Tone 1) and High Falling tone are allotones or the latter being a distinctive tone itself. In the case of Hong Kong Cantonese, the two versions are interchangeable, with a tendency to use a high level tone most of the time (Matthews & Yip 1994), whereas in Guangzhou area, speakers of Cantonese tend to use the falling tone (Zeng 2002). In Matthews & Yip (1994), High Rising and Low Rising are reported as having similar onset of tonal heights, with former having a steeper change in contour, which results in the occasional confusion of the Low Rising tone with the Mid-Level tone. The two tones are even merged in Malayan Cantonese (Killingley 1985). Matthews & Yip (1994) also reported the difficulties of foreign learners in distinguishing the Low Falling tone (Tone 4) from the Low Level tone (Tone 6) on account of the relatively minor scale of the falling contour.. 2.1.2 Mandarin tones There are four tones in Mandarin, traditionally described as High Level [55] (Tone 1), High Rising [35] (Tone 2), Dipping [214] (Tone 3), and High Falling [51] (Tone 4).. 5.

(15) Figure 2.. Tone contours of four Mandarin tones. The Dipping tone (Tone 3) has an allotone [21] that is used when the tone is followed by a different tone, that is, Tone 1, Tone 2, or Tone 4, in a phrasal word (Li & Thompson 1989). In Tseng (2008: 27), speeches of news reporters in Mainland China and Taiwan are compared and analyzed, revealing a series of contrasts in standard speech of the two varieties. The study shows a different preference between Putonghua and Taiwan Mandarin speakers with regard to the pronunciation of the Dipping tone (Tone 3). While Taiwan Mandarin speakers tend to pronounce the tone as [21], Putonghua speakers apply the full length of the tone as [214], resulting in an averaged time span more than twice in length than those of the Taiwan Mandarin speakers.. 6.

(16) 2.1.3 Taiwanese tones Taiwanese, or the Taiwan Southern Min, is the variety of Southern Min that is commonly used in Taiwan. Taiwanese in general contains seven tones: High Level [44] (Tone 1), High Falling [53] (Tone 2), Low Falling [31] (Tone 3), Low [22/32] (Tone 4), Rising [13] (Tone 5), Mid-Level [22] (Tone 7), High [33] (Tone 8) (Yang 1991). Tone 6 are commonly merged with either Tone 2 or Tone 7 in Taiwan (Ang 1985), hence the absence of Tone 6 in the system.. 2.1.4 Entering tones in Mandarin, Cantonese, and Taiwanese Entering tones, or checked syllables, is one of the traditional tone classes in Chinese, characterized by having unreleased plosives in coda position and being relatively short in time span. After a series of splits and mergers, the eight tonal categories in Middle Chinese have evolved into the four tones in modern Mandarin, where entering tones, along with the syllabic structures, have melted into other tones and disappeared in the language (Hu 1962). Nevertheless, entering tones can still be found in southern Chinese dialects, such as Cantonese, Hakka, and Hokkien, with varied number and tone values. Taiwanese, for instance, contains two entering tones: Low [22/32] (Tone 4) and High [33] (Tone 8). There are three entering tones in Cantonese, which renders the language as having nine tones in its traditional description. Though traditionally categorized as individual tones, the three entering tones, High [5] (Tone 7), Mid [3] (Tone 8), and Low [2] (Tone 9), are considered as abbreviated versions with similar f0 heights of the three level tones High Level [55] (Tone 1), Mid Level [33] (Tone 3), and Low Level [22] (Tone 6), by modern scholars (Chao 1947, Matthews & Yip 1994, Bauer & Benedict 1997).. 7.

(17) 2.2. Models of Cross-language perception. Studies on cross-language speech perception have mainly focused on segments and proposed various models to account and investigate how people perceive of and eventually acquire non-native phones and how their language experiences may have interacted with their speech. Among them are Flege’s Speech Learning Model (SLM) and Best’s Perceptual Assimilation Model (PAM). Since SLM aims at the end product of L2 speech acquisition, that is, the pronunciation of bilinguals who have spoken their L2 for an extended period of time, the following paragraph will focus on PAM, a model that looks into how adult listeners perceive foreign non-native phones and contrasts. In PAM, it is assumed that people tend to assimilate non-native segments to allophones of the most similar native segment. The perceptual assimilation of nonnative segments can be classified as one of the following cases: 1) The non-native phone is assimilated to a specific native segmental category, where the non-native phone may be perceived either as an ideal, acceptable, or deviant exemplar of the category; 2) The non-native speech phone is assimilated as a speech sound that fall outside any particular category as uncategorized speech; 3) The non-native phone is not assimilated as a speech and thus perceived as a non-speech sound. Following the classification, assimilation of non-native contrasts (between non-native phones) and their discriminability to L2 perceivers are predictable, as illustrated in Table 1. The assimilation patterns include:. 8.

(18) Two-category assimilation (TC): The two non-native segments are assimilated to different native categories. (Discrimination: excellent) Category-goodness difference (CG): The two non-native segments are assimilated to the same native category and only differ in their category goodness. (Discrimination: depending on how different the two segments are from each other in their category goodness.) Single-category assimilation (SC): the two non-native segments are assimilated to the same native category and are equally acceptable or deviant from the ideal exemplar. (Discrimination: poor) Both uncategorizable (UU): both non-native segments are recognized as speech but cannot be assimilated to any native category. (Discrimination: poor to very good) Uncategorized versus categorized (UC): one non-native segment is assimilated to a native category while the other is not. (Discrimination: very good) Nonassimilable (NA): both non-native segments are perceived as non-speech sounds. (Discrimination: good to very good). 9.

(19) Table 1.. Perceptual assimilation patterns of non-native contrasts in PAM Categorized. Uncategorized. good. acceptable. deviant. speech. Good. SC. CG. CG. UC. Acceptable. CG. SC. CG. UC. Deviant. CG. CG. SC. UC. Uncategorized Speech. UC. UC. UC. UU. Categorized. Non-. Nonspeech. NA. speech. Since non-native contrast assimilation is predictable from non-native phone assimilation, and it is possible to discover the assimilation pattern of a non-native phone to native categories with identification and categorization tasks, PAM seems to be a desirable basis for the investigation and prediction of the present study.. 2.3. Studies on cross-language tone perception. Cross-language studies on tone perceptions mostly focus on comparing how native speakers of tonal and non-tonal languages discriminate and identify non-native tones, for instance Thai, Mandarin, Cantonese, and synthesized tones. Previous studies have pointed out the influence of language background on speech perception, indicating that native speakers of tonal languages have the advantage of using changes in fundamental frequency as cues of lexical contrasts (Gandour & Harshman 1978). It 10.

(20) has been pointed out, however, that native knowledge of a tonal language does not necessarily facilitate acquisition of foreign tones (So & Best 2010, Chiao et al. 2011). Studies on non-native tone acquisition or perception by native speakers of tonal languages have been relatively rare, with most of them focusing on error analysis (Chen 2010) or comparison between standard and accented varieties of Mandarin (Jin 2008) with the aim to teach Mandarin to native speakers of various dialects. From time to time, Cantonese tones are involved in tone perception studies. One example is Qin & Mok (2011), where the perception of Cantonese tones by native speakers of Mandarin, English, and French is investigated. Also in Chung (2008), the researcher looks into the perception and production of Cantonese tones by Thai and Philippine students in Hong Kong secondary schools. Nevertheless, there have been few studies onhow Mandarin speakers perceive and produce Cantonese tones (Ku 2010). In those studies that involve a comparative analysis of Cantonese and Mandarin, researchers have reached differing analysis conclusions and results. One example is the analysis of the low falling tone (LF) in Cantonese. In Chen (2010), LF is assimilated to the falling tone in Mandarin, with reduced falling scope. The assimilation is also found in Leung (2008), while it is refuted by the results. In Francis et al. (2008), the Mandarin subjects often identified LF as low rising or high rising, while in Ku (2010), LF can be identified as easily as high level and high falling tones. Since f0 height and direction are found to be the primary distinctive features of tone categories and the weights put to each of the dimension are different among languages (Gandour 1983), prediction on how listeners assimilate non-native tones to native tones should be made on account of the preferred distinctive features of the listener. Hence, the low-falling tone in Cantonese may be assimilated to the dipping tone in Mandarin, given the existence of the low-falling allotone for the dipping tone, while 11.

(21) LF may also be confounded with mid-level or low-level tones, since Mandarin speakers are relatively insensitive to f0 heights. As a result, it could be beneficial to the cross-language study of tone perception if empirical evidence on how native speakers of Mandarin (a contour-based tonal language) perceive Cantonese tones (a tonal language that focus equally on f0 height and contours) could be provided. Also, from the pedagogical perspective, a survey of Mandarin speaker’s perception and production of Cantonese tones not only helps Mandarin speakers learn Cantonese, a language with a more complex tonal system, but also contribute to the Mandarin teaching of Cantonese speakers through advancing studies on tone perception. Furthermore, since the influence of L1 transfer on production by native speakers of tonal languages is rarely studied, the present study intends to collect production data as well and look into the correlation between cross-language perception and production, and between L1 transfer and pronunciation.. 12.

(22) 3. 3.1. CHAPTER THREE: METHODOLOGY. Overview. A series of three tasks (Labeling, Identification, and Vocalization) are conducted in the present study. The first two tasks, Labeling and Identification, investigate the perception of Mandarin speakers of Cantonese tones. In the Labeling task, subjects listen to a series of Cantonese tones and assign Mandarin tonal values to the audio stimuli. The answers of the subjects are collected with the hope to understand how they assimilate these non-native tones to their native tones and how easy or hard each decision is. The other perceptual task is Identification, where subjects listen and identify what Cantonese tones the stimuli are. In the Vocalization task, subjects read out romanized Cantonese syllables on screen. A priming procedure is provided after the Labeling task to familiarize subjects with Cantonese tones, along with a questionnaire at the end of all tasks on their linguistic backgrounds and their impression of the tasks in general. Recordings from the Vocalization task are then rated and identified by one native speaker of Cantonese.. 3.2. Subjects. A total of twelve native speakers of Mandarin are recruited in the present study, among them six subjects from Maintain China (the Putonghua group) and six subjects from Taiwan (the Taiwan Mandarin group). The subjects, including 4 males and 8 females, aged between 28-35 except for one 19-years-old subject, are all devoid of profound hearing and visual impairments. Since prior musical trainings, apart from L1 prosodic backgrounds, has been proposed as one of the possible confounding factors in tone perception (Gottfried & Riester 2000), candidates who have received formal 13.

(23) musical trainings are avoided. In addition, those enrolled are surveyed with regard to their prior musical experience. All Taiwan Mandarin subjects are college students who also speak Taiwanese to a certain degree. Subjects of the Putonghua group are all Mainland China students in Taiwan, who came from northern regions of China, mostly Hebei and Shaanxi, and do not speak Taiwanese. None of the subjects have any prior knowledge of the Cantonese language. Most of them can speak English. Some have learnt other foreign languages, such as French, German, Korean, Japanese, and Spanish, while none of them reported themselves as fluent speakers of these languages. To maintain homogeneity of dialectal backgrounds, all subjects in the Taiwan Mandarin group are self-reported as bilingual speakers of Taiwanese and Taiwan Mandarin. Owing to the obligatory Mandarin education that has come to existence for decades in Taiwan, the younger generations of Taiwanese-speaking people in Taiwan are mostly also Mandarin speakers, and their self-rated fluency of the Taiwanese language are either similar or (mostly) inferior to that of their Taiwan Mandarin, and sometimes even worse than English. Instead of monolinguals of Taiwan Mandarin, the present study recruited speakers of Taiwanese and Taiwan Mandarin with varied levels of bilingualism because even monolinguals of Taiwan Mandarin may be under the dialectal influence of Taiwanese on account of the fact that the variety Taiwan Mandarin is heavily influenced by Taiwanese and these monolinguals, even though they do not speak Taiwanese themselves, are surrounded by the language environment of Taiwan Mandarin. Similarly, a number of subjects in the Putonghua group reported themselves as speakers of northern vernaculars. Given the northern origin of modern Mandarin, the line between language and accent is further blurred and it can be troublesome to 14.

(24) determine whether the subjects speak dialects of northern Chinese or accented Mandarin in their hometown. Since all subjects reported Mandarin as their main language and the present study does not intend to look into the performance of monolingual versus bilingual subjects but that of Putonghua (in northern regions of Mainland China) versus Taiwan Mandarin subjects, the distinction of monolingualism and bilingualism is not pursued in the study.. 3.3. Materials. To investigate the topic to a fuller extent, the present study employed vowels and syllable structures as two factors that may burden subject performance, and diversifies stimuli with respect to the two features. However, given the limited size of data and contradicting effects, these factors are only introduced here as an account for the choice of materials. Results with regard to vowels and syllable structures will not be presented and discussed in Chapter Four3. Two Cantonese vowels, /a/ and /œ/4, are used throughout the three tasks. The vowel /a/ is widely found in most languages in the world and is considered to pose least burden to subjects from both groups, where /œ/ is a vowel foreign to speakers of Mandarin and absent in many dialects of Chinese languages. It is assumed that subjects will have greater difficulty with regard to /œ/ than /a/, and the difference may be reflected in their performance.. 3. Results from the Labeling task show that the vowel /œ/ has rising effect for non-entering tones, where subjects tend to perceive the tone with a higher f0. Accuracy for syllables with /œ/ is lower than those with the vowel /a/ when these syllables are combined with Tone 1, Tone 3, Tone 4, and Tone 6, all of which are level or nearly level tones.. 4. Including [œ] and its allophone [ɵ]. 15.

(25) Three syllable structures, syllables with no coda (open syllables), nasal coda /m/, and unreleased plosive codas (entering tones) are used in the study. Since open syllables are commonly found in Mandarin, this type of structure should not pose any difficulty to speakers of Mandarin. On the other hand, the phonotatctics of Mandarin only allow two nasal codas, /n/ and /ŋ/; syllables ended with /m/ are illegitimate in the language. Even though the nasal /m/ is present in Mandarin phonology, its placement in the coda position may still sound foreign to native speakers of Mandarin. Finally, unreleased plosive codas (entering tones) are a type of structure totally absent in modern Mandarin, and subjects may encounter great difficulty when dealing with these syllables. Since nasal coda /m/ and entering tones are still common in some dialects of Chinese languages, it is worthwhile to investigate the influence of subjects’ language backgrounds, which are to be surveyed in the questionnaire. Hence, for the vowel /a/, syllables /ma/, /ham/, and /pak̚/ are used with their tonal minimal pairs. Similarly, for the vowel /œ/, syllables /sœŋ/, /sœk̚/, /tʃœk̚/, /sɵt̚/, /tʃɵt̚/ are used. Since some combinations of syllables and tones are limited in number or absent in the language itself, the four syllables /sœk̚/, /tʃœk̚/, /sɵt̚/, /tʃɵt̚/ are chosen to ensure there are equal numbers of stimuli for each entering tone. Still, the choice of stimuli is asymmetrical for non-entering tones, since an open syllable with the /œ/ nucleus is relatively rare and does not have a set of minimal pairs with all six tones. Also, for vowel /a/, it is hard to find tonal minimal pairs that share the same onset and differ only in codas. It should be emphasized that the study does not intend to include all combinations of syllables. Diphthongs, for example, are not included in the study, along with many other monothongs. The purpose of stimuli selection, as discussed in this section by now, is to pose levels of challenges to speakers of Mandarin and. 16.

(26) investigate possible influences of language backgrounds. Variations in vowels and syllable structures are means to serve the purpose, not the ends here. Table 2. Nucleus Tone. Coda. High Level. Cantonese tonal minimal pairs. /a/. /œ/. Ø. Nasal. Stop. Nasal coda. /ma/. /ham/ 嵌. /pak̚/迫. /sœŋ/. 孖 “twins”. ”embed”. ”pressing”. 雙 “pairs”. /-œk̚/. /-ɵt̚/ /sɵt̚/ 恤 “shirt” /tʃɵt̚/ 卒 "soldier". High Rising. /ma/. /ham/ 豃. /sœŋ/. 媽“mother”. "steep". 想 ”want”. /ma/. /ham/. /pak̚/百. /sœŋ/. /sœk̚/. 嘛 particle. 喊 “cry”. ”hundred”. 相”look”. 削 ”slim”. Mid-Level. /tʃœk̚/ 雀 "sparrow". Low Falling Low Rising. Low Level. /ma/. /ham/. /sœŋ/常. 麻“numb”. 鹹“salty”. “common”. /ma/. /ham/ 濫. /sœŋ/上. 馬“horse”. "flood". ”go for…”. /ma/. /ham/餡. /pak̚/白. /sœŋ/上. /tʃœk̚/ 着. 罵“scold”. “stuffing”. ”white”. ”up”. "manifest". /sɵt̚/術 “technique”. Materials used in this study are from Chinese Character Database with wordformations phonologically disambiguated according to the Cantonese dialect, Research Centre for Humanities Computing, Chinese University of Hong Kong, within which audio recordings are pronounced by a native speaker of Cantonese and are provided with reference to their syllables and tones. The word lists are for reference only.. 17.

(27) 3.4. Procedures. 3.4.1 Labeling In the Labeling task, subjects listen to one single word at a time and label it with either one of four Mandarin tones, which they think the speaker has just said. Two Mandarin syllables, /ma/ and /sɔŋ/, that are combined with four Mandarin tones and spoken in two variations of Mandarin, Putonghua and Taiwan Mandarin, are added as fillers, resulting in 48 tokens (24 Cantonese tokens and 24 Mandarin fillers). This impressionistic survey is conducted for both theoretical and pedagogical reasons. First, it is hoped to assess how naïve Mandarin listeners assimilate the non-native tones to their native tones, so as to predict the assimilation patterns of non-native tonal contrasts. Second, the task aims to provide an empirical basis for an analogical approach commonly found in Cantonese teaching materials and comparative analyses that mostly based on the writers’ arbitrary association. The assigned labels to each tone are obtained from this task, which demonstrate how native speakers of Mandarin relate Cantonese tones to their native tones and their patterns of perceptual assimilation for these non-native tones, which in turn may account for the difficulties and confusions people have in non-native tone acquisition.. 3.4.2 Priming After the Labeling task, subjects are introduced to a brief material on Cantonese tones. The materials are excerpted from the self-learning materials on a Cantoneseteaching website, held by the Independent Learning Centre of the Chinese University of Hong Kong. The materials include a brief introduction of the Cantonese tone system, diagram of Cantonese tones, and audio files of examples words. Subjects are 18.

(28) encouraged to read through the material, listen to all the audio examples, and repeat after them, if desired. The purpose of the procedure is to prepare subjects with some brief knowledge of Cantonese tones, thus they can be ready to perform subsequent tasks. The procedure also aims to endow linguistic implication to f0 variations and direct subjects to treat stimuli as segments of speech instead of musical sounds.. 3.4.3 Identification In the Identification task, subjects hear one single word at a time and judge which tone the word bears. The word is presented in combination with a carrier sentence in Cantonese, meaning “(Someone) knows the word __”. The sentence is given in IPA on screen, along with a diagram of all six Cantonese tones. Chinese characters are avoided lest Mandarin pronunciation of these Chinese characters should confound the perception of non-native tones. The diagram is provided with the hope to ease the memory load that the subjects may have when encountering a foreign and relatively complex tone system. One concern that is distinctive in tone studies is the relative nature of tones. That is, tones need other tones to facilitate its recognition of relative f0 height. It is especially true in Cantonese, where there are multiple level tones that could bear the same f0 height in a sentence, depending on its relative contexts (So 2001). Therefore, a carrier sentence is provided here as a context for relative f0 heights. It is necessary to state that the context provided in tokens served only as reference points for one another, and multiple-word phrases or sentences are not intended here, so as to avoid the influence of co-articulation and focus on the perception of tones. Also, since order of presentation is reported to have language-specific effects in perception of lexical 19.

(29) tones (Francis & Ciocca 2003), all tokens contain the same carrier sentence, providing an identical tonal environment for all stimuli. The Identification task is conducted to investigate how subjects perceive non-native tones after their brief exposure to these tones. On account of the Priming procedure provided beforehand and the tone diagram and carrier sentence presented during the task, subjects are expected to focus on the perception of audio stimuli as non-native tones and be exempt from memorizing at this stage a tonal system that they are unfamiliar with. Based on the assimilation patterns of non-native tones to native ones that are obtained from the Labeling task, patterns of assimilation could be assigned to each of non-native tonal contrast in Cantonese. As a result, the data of accuracy, reaction time, and confusions that are collected in the Identification should indicate subjects’ discriminability of these contrasts, and in turn verify whether it is indeed predictable from the patterns of perceptual assimilation, as stated by Best.. 3.4.4 Vocalization In the Vocalization task, subjects are instructed to pronounce stimuli, one word at a time, when they see the IPA transcription of the word on screen. For each set of words, there are two phases of pronunciation. First half of the task is imitation, where phonetic transcriptions, tone diagram, and audio demonstration of the stimuli are presented to subjects and subjects are instructed to repeat after the audio example, six tones in succession for three rounds. In the second half of the task, only IPAs are provided and subjects are instructed to read out the transcriptions on screen without audio examples. The stimuli are presented in due order of tones in the first round, then they are given in random order in the second round. The process is repeated in all syllables in the order of /ma/, /ham/, /pak̚/, /sœŋ/, /sœk̚/, and /sɵt̚/. When they think 20.

(30) they have finished vocalizing the word, subjects can press the button and proceed to the next token. If the subject has made multiple tries, the final attempt is taken as valid. The production data, including imitation and read-out sections, are recorded and the results are listened and rated by a native speaker of Hong Kong Cantonese, who is instructed to judge whether the imitation is accurate and identify which of the six tones the subjects say in the production phase. If the native speaker finds the answer indistinguishable, question marks are used. Some concerns on possible confounding factors have led to the design of the task. First, the partially shared orthographical systems of Cantonese and Mandarin have made the use of Chinese characters in elicitation under the risk of confusing Mandarin-speaking subjects, and therefore avoided. Second, on account of the fact that they are likely to possess little knowledge of IPA, subjects are provided with audio demonstration of the tokens for imitation to inform them of the correct pronunciation. Also, given the relatively complex tonal inventory of Cantonese, it is almost impossible to train naïve subjects without increasing their memory loads. Nevertheless, by using the same sets of stimuli throughout tasks, subjects are expected to have some degree of familiarity with these Cantonese tones after the Priming procedure and the Identification task. The preceding practice with a nonnative tonal system, along with repeated imitation of the production stimuli, is utilized to reduce cognitive loading of the Vocalization task, with the hope to increase their performance and chances in concentrating on production itself. While the acquisition of an L2 speech often starts from imitation, as noted in Rochet (1995), and it will certainly be less perplexing to have subjects repeat after audio stimuli, the study attempts to obtain autonomous production data from near-naïve subjects with its 21.

(31) design. On account of the criticism of imitation as doubtful representation of actual speech production (Beddor and Gottfried 1995), a task design that involves both imitation and elicited production may shed more light on the early stage of L2 tonal acquisition.. 3.4.5 Questionnaire Subjects are surveyed with regard to their language backgrounds and impressions of the tasks and their own performance. A short discussion follows to learn their opinions about the investigation, which part of the Cantonese tones do they find the most difficult, and the strategies they apply to distinguish these tones. A copy of the questionnaire is attached in Appendix I.. 3.4.6 Administration Subjects all put on headphones during the experiment. Volumes are adjusted before the tasks begin to ensure they can all properly hear the audio files in the study. Subjects are informed that the experiment consists of two parts. In the first part, they are instructed to perform the Labeling task. The priming material is presented in PowerPoint format at the beginning of the second part. The Priming procedure is selfpaced and subjects are free to flip forward and backward through the slides. The Priming procedure is mostly completed within ten minutes, followed by the Identification task and the Vocalization task. Afterwards, the subjects complete a questionnaire on their basic information, linguistic backgrounds, and language attitudes and have a short discussion on their opinions over the experiment and the Cantonese language.. 22.

(32) 3.5. Instrumentation. The tasks are conducted in a quiet room on university campuses with a Macintosh notebook computer running OS X 10.5.8 and a pair of in-ear headphones. The headphones (frequency response 20 Hz to 20,000 Hz) contain an integrated microphone (frequency responses 100 Hz - 10,000 Hz) and audio recordings are made during the Vocalization task at a sampling rate of 44100Hz, 32-bit float. The Vocalization task is also recorded with an audio recording app on an Android smartphone for redundancy. All recordings on the computer side and audio editing are made with Audacity, a software for recording and editing sounds. To present stimuli and collect data, tasks are programmed and generated using Psychopy, an open-source application written in Python for the design and conduction of neuroscience and experimental psychology (Peirce 2007). Materials in the Priming procedure are presented with Microsoft PowerPoint for Mac. Tone graphs are analyzed with Praat, and statistical analyses are conducted with SPSS.. 3.6. Data analysis. With the aim to investigate how naïve Mandarin subjects perceive of Cantonese tones, the tasks are designed on the basis of Best’s PAM, with the Labeling task assessing the assimilation patterns of each non-native tone to native tones, and the Identification and Vocalization tasks assessing the ease or difficulty of contrasting non-native tones in perception and production. Therefore, results of the Labeling task are expected to provide clues to account for perception and production performances in the initial stage of L2 tone acquisition. It is assumed that the category goodness of a non-native tone can be indicated by the 23.

(33) label distribution of each tone. That is, if the non-native phone is highly similar to a native phone, the subject will tend to label the phone with the native tone. Similarly, if the subject finds the decision very hard to make, the answers should diverse greatly. On the other hand, the discriminability of two non-native contrasts can be illustrated by analyzing the accuracy and confusions of results in Identification and Vocalization tasks. If a certain tone is constantly confused with some other tone in the Identification task despite of the priming procedure and the help of a tone diagram, subjects are considered as having difficulty contrasting non-native tones.. 24.

(34) 4. 4.1. CHAPTER FOUR: RESULTS. Labeling. 4.1.1 Overview In the Labeling task, subjects are presented with single Cantonese words that are composed of tones and syllables that may be foreign to them. They are yet to learn to identify or produce the said phones, but simply label these unfamiliar sounds using the familiar Mandarin tones to categorize and describe how they perceive these phones and what these tones sound like in their own terms. It should be noted that even though some tones or sounds are categorized in the same groups, it does not necessarily mean that the subjects cannot distinguish one tone from another within the category. What one can infer from these results is that tones categorized within one category may share some similarities, at least to the listener, and they are more likely to be considered as distinctive or different from tones within another category. In general, Mandarin-speaking subjects in both groups tend use Mandarin Tone 1 to label level tones. They also label Mandarin Tone 2 for the two rising tones Tone 2 and Tone 5, and Mandarin Tone 3 for the low-falling Tone 4. Mandarin Tone 4, the falling tone, is used to label Tone 1, Tone 3, Tone 4, and Tone 6. Distribution of these labels indicates three contour-based categories: level tones, rising tones, and falling tones. Within the level tone category, Tone 1 and Tone 3 are much more similar to each other than to Tone 6.. 25.

(35) 100% 90% 80% 70% . 0 11 . 2 2 9 . 3 9 . 0 12 11 . 15 . 15 . 60% . 24 . 50% 59 . 40% 30% . 4 . 49 . 12 . Mandarin Tone 1 . 20% . 33 . 3 . 10% . 5 0 Tone 1 . Figure 3.. Tone 2 . Mandarin Tone 3 Mandarin Tone 2 . 23 . 21 . 0% . Mandarin Tone 4 . Tone 3 . Tone 4 . 2 Tone 5 . Tone 6 . Labeling of Cantonese tones by Mandarin-speaking subjects. Since Cantonese do not contain high falling tones, except for allotones of Tone 1, which is not included in the stimuli, it is bewildering to see subjects choose Mandarin Tone 4, a high falling tone, to represent what they hear. While one can propose that Mandarin-speaking subjects are sensitive to the slight falling contour of Cantonese Tone 4, thus the choice of Mandarin Tone 4 for Cantonese Tone 4, the choice of Mandarin Tone 4 for Cantonese level tones remain unaccounted. After all, it seems unlikely that these subjects are wired with the phonological knowledge of a language they are foreign to and consider high level tones and high falling tones, which are contrasted phonemically in their native language, belong to the same tonal category. More is revealed when entering tones are separated from their counterparts. As shown in the figure below, except for Tone 4, the labeling of Mandarin Tone 4 on Cantonese level tones are exclusively for entering tones, especially Entering Tone 1 and Entering Tone 6. The results show that while entering tones are considered as abbreviated versions of their level tone counterparts, they do not necessarily leave similar 26.

(36) impression on their listeners and it is worthwhile to look into them separately in the following discussion.. 100% 90% 80% . 0 2 7 . 70% . 0 11 . 0 2 5 . 2 0 4 . 15 . 11 . 9 . 24 29 . 27 22 . 30 . 6 . 5 0 . Mandarin Tone 4 Mandarin Tone 3 Mandarin Tone 2 . 21 3 . 10% . 11 . 7 . 23 . 21 . 20% 0% . 1 . 5 . 50% 30% . 0 . 1 2 . 60% 40% . 4 . Mandarin Tone 1 12 . 2 . non‐E E Tone Tone 2 non‐E E Tone Tone 4 Tone 5 non‐E E Tone Tone 1 1 Tone 3 3 Tone 6 6 . Figure 4.. Labeling of Cantonese tones by Mandarin-speaking subjects (non-entering tones and entering tones). Hence, some sketches can be obtained as to how native speakers of Mandarin may perceive of the set of non-native tones: they may find that Tones 1, 3, and 6 sound fairly similar and are akin to Mandarin Tone 1; Tones 2 and 5 sound like Mandarin Tone 2; and Tone 4 sounds like Mandarin Tone 3. Within the level tone category, the lowest one, Tone 6, is slightly different from the other two level tones. Furthermore, entering tones are perceived differently from level tones and have a quality reminiscent of Mandarin Tone 4.. 27.

(37) 4.1.2 Subject groups 4.1.2.1 Categorization of Cantonese tones A closer look reveals that there are some diverse in tendencies with regard to Putonghua and Taiwan Mandarin speakers' labeling criteria for the same set of sounds.. M Tone 4 . 2 4 1 6 . 8 . M Tone 3 2 1 . M Tone 2 . 11 . 6 . M Tone 1 . 2 . 10 . 0 . Figure 5.. 1 . 7 . 7 . 4 1 2 . 7 . 10 20 non‐E Tone 1 non‐E Tone 3 Tone 5 . 10 . 7 . 1 1 . 3 4 . 13 . 15 . 30 40 E Tone 1 E Tone 3 non‐E Tone 6 . 5 1 . 50 . 13 . 60 Tone 2 Tone 4 E Tone 6 . 7 . 70 . Categorization of Cantonese tones by Putonghua subjects. First of all, Putonghua and Taiwan Mandarin speakers seem to have a varied preference in the use of Mandarin Tone 2 and Mandarin Tone 3 labels. As shown in the figure below, Taiwan Mandarin speakers tend to label the two rising tones, Tone 2 and Tone 5, with Mandarin Tone 2, and they almost exclusively label the low falling Tone 4 with Mandarin Tone 3.Contrarily, Putonghua subjects label Tone 2 and Tone 5 with Mandarin Tone 2 and Mandarin Tone 3 alike, and they do not have a particular preference to the labeling for Tone 4. One may infer, from these results, that Taiwan Mandarin subjects have a different interpretation concerning what Mandarin Tone 3 should be like from their Putonghua counterparts. It seems that, to Taiwan Mandarin. 28.

(38) subjects, Mandarin Tone 3 is defined to a certain extent by the low falling contour in the dipping process and the rising that follows is less significant, whereas the falling and the rising are of equal salience to Putonghua subjects.. M Tone 4 . 3 5 . M 0 4 1 Tone 3 . 17 . M 1 Tone 2 . 14 . M Tone 1 . 17 0 . 10 . 1 . 1 3 1 . 5 . 16 . 2 3 . 15 20 . non‐E Tone 1 non‐E Tone 3 Tone 5 . Figure 6.. 8 . 16 30 . 40 E Tone 1 E Tone 3 non‐E Tone 6 . 15 50 . 1 . 8 . 60 . 70 . 5 80 . Tone 2 Tone 4 E Tone 6 . Categorization of Cantonese tones by T Mandarin subjects. Also, Putonghua subjects apply Mandarin Tone 4 label nearly three times more than their T Mandarin counterparts. Tones that are labeled with Mandarin Tone 4 are mostly Entering Tones 1 and 6, and sometimes Tone 4, while Taiwan Mandarin subjects label Entering Tone 1 and 6 as Mandarin Tone 4 only occasionally. It seems that Mandarin Tone 4 is prominent not only for its falling contour but also for a short time span or a sudden burst of force. This tendency is not commonly found in Taiwan Mandarin subjects, who tend to label entering tones as other level tones and prefer Mandarin Tone 1. On account of these differences in labeling criteria, the labeling of Cantonese tones by the two groups of subjects is rearranged as followed for further comparison.. 29.

(39) 100% 90% . 0 2 . 0 . 80% 70% . 4 . 8 . 6 . 0 1 . 2 0 1 . 4 . 4 13 . 15 . 7 . 0% . 7 . 5 . 10% . Putonghua Tone 3 Putonghua Tone 1 . 7 . 7 . Putonghua Tone 4 Putonghua Tone 2 . 13 . 2 . 10 . 20% . 1 . 7 . 2 . 40% . 6 . 10 . 1 . 50% . 1 1 3 . 11 . 60% . 30% . 0 . 1 . 0 . non‐E E Tone Tone 2 non‐E E Tone Tone 4 Tone 5 non‐E E Tone Tone 1 1 Tone 3 3 Tone 6 6 . Figure 7.. 100% . 0 1 . Labeling of Cantonese tones by Putonghua subjects. 0 3 0 . 90% 80% . 4 . 0 1 1 . 0 3 . 0 . 0 1 . 0 5 8 . 70% 60% 50% 40% . 17 . 17 15 . 14 . 16 . 5 16 . 2 . 15 . 3 . 30% 8 . 20% . 5 . 10% 0% . 0 . 1 0 . T Mandarin Tone 4 T Mandarin Tone 3 T Mandarin Tone 2 T Mandarin Tone 1 . 1 . non‐E E Tone non‐E E Tone Tone non‐E E Tone Tone 2 Tone Tone 4 5 Tone Tone 1 1 3 3 6 6 . Figure 8.. Labeling of Cantonese tones by T Mandarin subjects. By considering Mandarin Tone 3 as a low falling tone, Taiwan Mandarin subjects use the label mostly for tones with lower f0, while Putonghua subjects use it evenly for the falling Tone 4 and the rising Tone 2 and Tone 5.This extra tonal attribute of low 30.

(40) falling has a few manifestations. First of all, the rising feature is now reassigned to Mandarin Tone 2, which may explain why Cantonese Tone 2 and Tone 5, both rising tones, are largely labeled with Mandarin Tone 2 by Taiwan Mandarin subjects. The low falling feature assigned to Mandarin Tone 3 also results in the exclusive labeling of Tone 4 with Mandarin Tone 3. The feature is prominent as well on the labeling of Tone 6 and Entering Tone 6. When encountered with low level tones, Taiwan Mandarin subjects use the Mandarin Tone 3 label almost as equally as the high level Mandarin Tone 1, whereas Putonghua subjects prefer Mandarin Tone 1. Entering Tone 6 is rather obscure for both groups of subjects; still, there is a higher proportion of Mandarin Tone 3 labeling for Taiwan Mandarin subjects.. 4.1.2.2 Tone labels between subject groups The following section reviews results in the Labeling task, one Cantonese tone at a time, for a more detailed compare the two groups of subjects. Tone 1 (HL). As shown in the figure below, subjects of both groups tend to label Tone 1 with Mandarin Tone 1. While Taiwan Mandarin subjects have a clear preference to label the tone with Mandarin Tone 1, Putonghua subjects also label the tone with Mandarin Tone 2 and Mandarin Tone 3 at times.. 31.

(41) 17 . 18 16 14 12 10 . 10 Putonghua . 8 . 6 . 6 . T Mandarin . 4 1 . 2 . 2 0 . 0 M Tone 1 . Figure 9.. 16 . M Tone 2 . 0 0 . M Tone 3 . M Tone 4 . Labeling of Cantonese non-E Tone 1 by both subject groups. 15 . 14 12 10 8 . 8 . 7 . Putonghua T Mandarin . 6 4 . 3 . 2 . 2 . 0 . 0 M Tone 1 . M Tone 2 . 1 . 0 . M Tone 3 . M Tone 4 . Figure 10. Labeling of Cantonese E Tone 1 by both subject groups When presented with Entering Tone 1, a drift in choice towards Mandarin Tone 4 can be observed, particularly for Putonghua subjects. Tone 2 (HR). Both groups of subjects apply Mandarin Tone 2 and Mandarin Tone 3 labels for the high rising Tone 2. However, the two groups have a different preference, where Putonghua subjects prefer Mandarin Tone 3 over Mandarin Tone 2 and Taiwan. 32.

(42) Mandarin subjects prefer Mandarin Tone 2 most of the time. Neither Mandarin Tone 1 or Mandarin Tone 4 are chosen here.. 16 . 14 . 14 12 . 11 . 10 8 . 7 . Putonghua T Mandarin . 6 . 4 . 4 2 0 . 0 0 M Tone 1 . 0 0 M Tone 2 . M Tone 3 . M Tone 4 . Figure 11. Labeling of Cantonese Tone 2 by both subject groups Tone 3 (ML). Similar to the case in Tone 1, both groups tend to label the mid-level Tone 3 as Mandarin Tone 1. It should be noted that Putonghua subjects apply the label even more often for Tone 3 (and Entering Tone 3) than for Tone 1.. 18 . 16 . 16 14 . 13 . 12 10 . Putonghua . 8 . T Mandarin . 6 . 4 . 4 . 1 . 2 . 1 1 . 0 0 . 0 M Tone 1 . M Tone 2 . M Tone 3 . M Tone 4 . Figure 12. Labeling of Cantonese non-E Tone 3 by both subject groups 33.

(43) 16 . 15 15 . 14 12 10 8 . Putonghua . 6 . T Mandarin . 4 . 3 . 2 . 1 . 0 0 . 0 M Tone 1 . 2 . M Tone 2 . 0 . M Tone 3 . M Tone 4 . Figure 13. Labeling of Cantonese E Tone 3 by both subject groups Unlike the other two entering tones, it is relatively rare for Entering Tone 3 to be labeled with Mandarin Tone 4. Tone 4 (LF). While Taiwan Mandarin subjects show a strong preference for Mandarin Tone 3 labels, Putonghua subjects use labels of four Mandarin Tones evenly. The low falling tonal feature assigned to Mandarin Tone 3 by Taiwan Mandarin subjects also prevent the use of high level Mandarin Tone 1, since a more apt label is available to this group of subjects now.. 17 . 18 16 14 12 10 . 6 . 0 . T Mandarin . 5 . 4 . 4 2 . Putonghua . 7 . 8 . 2 0 M Tone 1 . 1 . M Tone 2 . 0 M Tone 3 . 34. M Tone 4 .

(44) Figure 14. Labeling of Cantonese Tone 4 by both subject groups Tone 5 (LR). The choice of labels is similar to that for Tone 2: Both Mandarin Tone 2 and Mandarin Tone 3 labels are applied, while Putonghua subjects prefer Mandarin Tone 3 and Taiwan Mandarin subjects prefer Mandarin Tone 2.. 18 . 16 . 16 14 12 . 10 . 10 . Putonghua . 7 . 8 . T Mandarin . 6 4 2 . 1 1 . 1 . 0 0 . 0 M Tone 1 . M Tone 2 . M Tone 3 . M Tone 4 . Figure 15. Labeling of Cantonese Tone 5 by both subject groups Tone 6 (LL). For the low level Tone 6, Putonghua subjects mostly apply the Mandarin Tone 1 label, reasonably owing to the level contour of the tone. Taiwan Mandarin subjects, on the other hand, equally apply Mandarin Tone 1 and Mandarin Tone 3 labels.. 35.

(45) 14 . 13 . 12 10 8 . 8 . 8 Putonghua . 6 . T Mandarin . 4 . 3 2 . 2 . 1 . 1 0 . 0 M Tone 1 . M Tone 2 . M Tone 3 . M Tone 4 . Figure 16. Labeling of Cantonese non-E Tone 6 by both subject groups The drift towards Mandarin Tone 4 for entering tones can also be found here, which, combined with the preference for Mandarin Tone 1 and Mandarin Tone 3 labels, made the categorization of Tone 6 for Taiwan Mandarin subjects rather ambiguous.. 8 7 6 5 . 7 6 5 . 5 . 5 . 4 . 4 . Putonghua . 3 . 3 2 . T Mandarin 1 . 1 0 M Tone 1 . M Tone 2 . M Tone 3 . M Tone 4 . Figure 17. Labeling of Cantonese E Tone 6 by both subject groups In summary, this section looks into how Mandarin-speaking subjects in Putonghua and Taiwan Mandarin groups perceive of the non-native tones of Cantonese. Based on these findings, the study in turn reveals how subjects think of their native tonal categories of by examining how non-native tones are grouped and related with their 36.

(46) native tones. Influences of segmental factors, such as vowels and syllabic structures, are also investigated.. 4.2. Identification. 4.2.1 Overview The Identification task posed challenges to subjects in both groups. The overall performance is poor, which is foreseeable given that the task is conducted after a short priming on Cantonese tones, which all subjects are foreign to. Still, a few observations can be drawn from the results. The accuracy drops among tones with lower f0 heights (Tones 4-6) and in particular Tone 6. One possible account is memory limitation—subjects are confronted with a foreign tone system that is more complicated than their native one, and they can at best memorize first three tones, numbered one to three, and the rest is in turmoil. While this may hold true when describing how subjects attempt to remember the tones throughout the tasks in general, it should be pointed out that a tone graph is provided every time during this task before subjects made their decisions, and they do not have to solely rely on their remaining impression of Cantonese tones after the priming process. As a result, it could be inferred from the results that distinguishing tones with lower f0 is somehow challenging to subjects. The statement does not imply that the subjects have deficiencies in detecting low frequencies, rather, it proposes that subjects have trouble relating what they heard with new categories in relatively low f0 and thus identify most of their stimuli as higher tones.. 37.

(47) 100% 90% 80% 70% 60% 50% . 0.47 . 0.44 . 0.46 0.36 . 40% . 0.36 . 30% 0.14 . 20% 10% 0% Tone 1 . Tone 2 . Tone 3 . Tone 4 . Tone 5 . Tone 6 . Figure 18. Overall accuracy in Identification task The separation of entering tones from non-entering ones does not show major differences in accuracy, though the accuracy for entering tones are slightly higher than their non-entering counterparts.. 100% 90% 80% 70% 60% 50% 40% . 0.47 0.47 0.42 . 0.50 0.42 . 0.36 . non‐E Tones . 0.36 . E Tones . 30% 0.17 0.11 . 20% 10% 0% . Tone 1 Tone 2 Tone 3 Tone 4 Tone 5 Tone 6 . Figure 19. Overall accuracy in Identification task (non-E tones and E tones). 38.

(48) 4.2.2 Subject groups 4.2.2.1 Accuracy In general, Taiwan Mandarin subjects have better ID performance than Putonghua subjects for all Cantonese tones.. 100% 90% 80%. 0.72 . 70%. 0.58 . 60%. 0.5 0.44 . 50% 40%. 0.5 0.33 . 30% 20%. 0.22 . 0.17 . 0.44 . Putonghua T Mandarin. 0.28 0.19 0.08 . 10% 0% Tone 1. Tone 2. Tone 3. Tone 4. Tone 5. Tone 6. Figure 20. ID accuracy between subject groups For non-entering tones, Putonghua subjects have the uppermost performance identifying Tone 2 and are most inept for the identification of Tone 6. Taiwan Mandarin subjects, however, show relatively equal performance for all Cantonese tones, except for Tone 6.. 39.

(49) 100% 90% 80% 70% . 0.61 . 60% . 0.56 . 0.50 0.44 . 50% 40% 30% . 0.28 . 0.22 . 0.50 . 0.22 . 0.44 . Putonghua T Mandarin . 0.28 0.17 . 20% . 0.06 . 10% 0% Tone 1 . Tone 2 . Tone 3 . Tone 4 . Tone 5 . Tone 6 . Figure 21. ID accuracy of non-entering tones between subject groups For entering tones, Putonghua subjects are particularly good at identifying Entering Tone 3, while for Taiwan Mandarin subjects, the identification of entering tones deteriorates with f0 heights.. 100% 90% . 0.83 . 80% 70% . 0.61 . 60% 50% . Putonghua . 0.39 . 40% . T Mandarin . 30% 20% . 0.22 0.11 . 0.11 . 10% 0% Tone 1 Tone 2 Tone 3 Tone 4 Tone 5 Tone 6 . Figure 22. ID accuracy of entering tones between subject groups. 4.2.2.2 Confusions This section provides detailed discussion concerning subjects' response to the stimuli of each Cantonese tone. 40.

(50) Tone 1 (HL). For non-entering Tone 1 and Entering Tone 1, answers from Putonghua subjects are rather diverse and sporadic, with a preference to identify Tone 1 stimuli as Tone 3.. 15 12 9 5 . 6 4 3. 2 . 6 . E Tone 1. 4 . 3 . non-E Tone 1. 6 3 . 2 . 1 . 0 0 . 0 Tone 1. Tone 2. Tone 3. Tone 4. Tone 5. Tone 6. Figure 23. Answers for Tone 1 stimuli from Putonghua subjects Taiwan Mandarin subjects, on the other hand, perform rather well when identifying Tone 1 and especially Entering Tone 1.. 16 . 15 . 14 12 . 11 . 10 8 . non‐E Tone 1 . 6 . E Tone 1 . 4 2 0 . 3 . 2 0 . 0 . 0 0 . 1 . 2 . 1 1 . Tone 1 Tone 2 Tone 3 Tone 4 Tone 5 Tone 6 . Figure 24. Answers for Tone 1 stimuli from T Mandarin subjects. 41.