(LUẬN VĂN THẠC SĨ) A comparative study on how hesitation and reservedness is expressed via prosodic means in English and the equivalent expressions in Vietnamese

NGUYỄN THỊ HIỀN A COMPARATIVE STUDY ON HOW HESITATION AND RESERVEDNESS IS EXPRESSED VIA PROSODIC MEANS IN ENGLISH AND THE EQUIVALENT EXPRESSIONS IN VIETNAMESE Nghiên cứu so sánh sự do

NGUYỄN THỊ HIỀN

A COMPARATIVE STUDY ON HOW HESITATION AND

RESERVEDNESS IS EXPRESSED VIA PROSODIC MEANS IN ENGLISH AND THE EQUIVALENT EXPRESSIONS IN

VIETNAMESE

(Nghiên cứu so sánh sự do dự và dè dặt được thể hiện thông qua phương tiện ngôn điệu trong tiếng Anh và các hình thức diễn đạt tương đương

trong Tiếng Việt)

M.A Minor Program Thesis

Field: English Linguistics Code: 60 22 15

HANOI - 2010

TIEU LUAN MOI download : skknchat@gmail.com

NGUYỄN THỊ HIỀN

A COMPARATIVE STUDY ON HOW HESITATION AND

RESERVEDNESS IS EXPRESSED VIA PROSODIC MEANS IN ENGLISH AND THE EQUIVALENT EXPRESSIONS IN

VIETNAMESE

(Nghiên cứu so sánh sự do dự và dè dặt được thể hiện thông qua phương tiện ngôn điệu trong tiếng Anh và các hình thức diễn đạt tương đương

trong Tiếng Việt)

M.A Minor Program Thesis

Field: English Linguistics Code: 60 22 15

Supervisor: Nguyễn Hương Giang, M.A

HANOI - 2010

TIEU LUAN MOI download : skknchat@gmail.com

ACKNOWLEDGEMENTS ii

PART A: INTRODUCTION

1.3.2.1 Pitch (or fundamental frequency) 11

Chapter 2: HESITATION AND RESERVEDNESS EXPRESSED VIA

PROSODIC FEATURES IN ENGLISH AND VIETNAMESE

2.3 Comparison: prosodic cues for hesitation and reservedness in English

Chapter 3: IMPLICATIONS FOR TEACHING ENGLISH LANGUAGE SPEAKING

30

3.1 Hesitation phenomena in English Language Teaching classroom (ELT) 30 3.2 Using filled pauses to gain processing time 32 3.3 Implications for improving students‟ speaking fluency 33

Chapter 4: SOME KEY FINDINGS

4.1 Hesitation and reservedness in English and Vietnamese speech 37

LIST OF FIGURES Figure 1: Praat Editor showing waveform, spectrogram and TextGrid 16

Figure 2: Illustration of pitch contour of “No” on Praat Screen 17

Figure 3: Pitch contour of the sentence “Um…u…no I don‟t think so I can‟t

PART A: INTRODUCTION

1 Rationale of the study

For many people, learning English is uneasy and mastering it is more difficult Some people underestimate that a learner only needs to have a treasury of vocabulary or good knowledge of grammar, he can speak English fluently It is a concerning problem in teaching and learning a foreign language at schools, universities and centers because it causes learners to waste much more time and effort while the achievements are not up to their expectations At English lessons, there exist many situations in which teachers ask students to answer the questions or read a text aloud, the students feel embarrassed and shy even they only response in whisper or murmur It is not because they do not have any vocabulary in their mind or they do not know how to arrange words in a correct grammar But the fact that students are afraid of pronouncing words wrongly or making mistakes These phenomena directly or indirectly hinder the effectiveness and motivation of learning

a foreign language

It is common knowledge that among phonetic aspects, prosody is considered a crucial factor which helps learners master a communicative skill People usually talk much about intonation, stress or rhythm in language while prosody is rarely mentioned However, in linguistics, intonation, stress and rhythm mainly refer to prosodic components

The title “a comparative study of how hesitation and reservedness is expressed via

prosodic features and its equivalents in Vietnamese” was chosen for my M.A thesis

because of some following reasons First of all, prosody is a new area and not many studies

of students have been conducted on the basis of this theory because it is difficult and challenging Meanwhile, prosody plays an important role in the comprehension of spoken language, it helps human in recognition of spoken words, in resolving global and local ambiguities and in processing discourse structure Therefore, it will be worth conducting a study on this area Secondly, in the world of English language teaching (ELT), however, the communicative value of hesitation in speech has been largely ignored The prevailing view of hesitation seems to be that they are evidence of disfluency and should therefore be discouraged This view is inadequate when it assumes that fluency is directly related to communicative ability while disfuency is inversely related The present study examines the

evidence that speech hesitation sometimes supports and enhances communication and suggests the ways they may be dealt with in the ELT classroom Finally, there are few studies of hesitation across languages from the perspectives of prosody and using English

I do hope that my work can give more insight into the similarities and differences of hesitation via prosodic means between two languages in which English is a stress-timed language and Vietnamese is a tonal language

2 Aims of the study

The main aims of this study are:

- To explore prosodic features in English which express hesitation and reservedness,

- To provide a brief account of similarities and differences between hesitation and reservedness expressed via prosodic features in English and Vietnamese,

- To give some proposals for further study and suggestions for improving speaking skill

To fully achieve the stated aims, the study should answer the following basic questions:

- What are the prosodic features used to express hesitation and reservedness in English spontaneous speech?

- What are the similarities and differences in the expression of hesitation and reservedness by prosody in English and Vietnamese?

- What tips are utilized to improve speaking fluency?

3 Scope of the study

Many fields relating to hesitation phenomena and prosodic features need to be explored However, due to the limited time and available facilities, this thesis only focuses

on the following aspects:

- Hesitation and reservedness in spontaneous speech;

- Typical types of hesitation and reservedness in English and their equivalences in Vietnamese including silent pauses, filled pauses, repetitions, syllable lengthening

- Only prosodic features of pitch, length, loudness and tempo are explored in the expression of hesitation and reservedness

- Only British English is chosen for the standard sound and the Northern dialect in Vietnamese is utilized as the standard sound in this study

- English non- major students at Ha Tinh Education and Information Center are chosen in the survey to find out how frequent hesitation phenomena occur in English language teaching classroom Thereby, some tips are suggested to improve speaking fluency with hesitation strategies

4 Methods of the study

In order to explore the differences and similarities in expressing hesitation and reservedness via prosodic features, comparative study (CS) is utilized as the key method of the study Here, English is employed as the instrumental language Besides, systemization and generalization from previous studies are also integrated as reference for this thesis

To analyze exactly the expression of hesitation and reservedness via prosody, computer software such as PRAAT is applied despite the fact that transcribing prosodic features is really challenging and costly for a M.A student The information sources for analysis come from interviews at different situations recorded from Vietnamese channels such as VTV1 and textbooks such as TOEFL, NEW HEADWAY Other sources are also utilized in this study

A survey is also conducted to find out how students express their hesitance and how they adapt to hesitation strategies to gain speaking fluency Then, the data are collected, analyzed and synthesized

5 Design of the study

This study is completed on the basis of three separating parts: introduction, development and conclusion

Part A is “INTRODUCTION” which gives the readers an overview of the reasons for

choosing the topic, the aims, the scope, the methods applied and the design of the study

Part B entitled “DEVELOPMENT” which plays the most crucial role in the whole

study It is considered as the backbone of the study This part consists of three main chapters Chapter 1 shows the theoretical background of hesitation, reservedness and prosody Chapter 2 explores the similarities and differences of hesitation and reservedness

expressed via prosodic features in English and Vietnamese equivalent expressions Chapter

3 suggests some tips for the teacher to improve the students‟ English speaking fluency Part C is “CONCLUSION” in which the author will give the readers some concluding remarks as well as suggestions for further study

PART B: DEVELOPMENT Chapter 1: THEORETICAL BACKGROUND

In order to create the basis for analyzing and synthesizing the data in the main part of the study, it is necessary for the author to have a comprehensive understanding of theoretical background In this part, the author will help the readers understand more about the history of hesitation phenomena research which is also the basis for the author to conduct the study Besides, the nature of hesitation and reservedness in spontaneous speech

is revealed with the provision of definitions and types Prosody is another core point which the study focuses on so the author tries to clarify its concepts and features as well Here, prosody is simultaneously clarified in both English and Vietnamese language This enables the author to have a good approach in analyzing the similarities and differences of both languages in the later chapter

1.1 Literature review

The existence of hesitation and reservedness phenomena is a universal characteristic of spontaneous speech in any language Hence, this phenomenon has really attracted the attention of many researchers all over the world who inspire to find out what features contribute to the impression of hesitant speech

Much contemporary research on hesitation phenomena in speech was derived from the work of Goldman-Eisler (1968: 48), who argues that the analysis of speech pauses provides an external window upon the internal constructive processes of speech selection and organization Following Goldman- Eisler, a fair amount of work has been done on hesitations in spontaneous dialogues and monologues For example, Beattie (1979: 61-78) suggests that hesitation phenomena can be useful in studying various psycholinguistic processes

In this study, I am interested in exploring prosodic features to express hesitations and reservedness; therefore, the studies of Eklund (2004) and Lovgren & Van Doorn (2005) provide a useful background when they have shown that pauses and retardations are among the acoustic correlates of hesitations Moreover, it is worth mentioning the studies ofRolf Carlson, Kjell Gustafson, and Eva Strangert (2006:1) when they prove that the total

duration increase, the combination of both pause duration and final lengthening are potential cues to hesitation

The results of mentioned-above English studies will be extremely valuable for the present study However, in Vietnamese language there have not yet been any works about the expressions of hesitation via prosody for reference Only some studies related to

Vietnamese tones are considered such as Tiếng Việt mấy vấn đề Ngữ âm - Ngữ pháp - Ngữ

nghĩa by Cao Xuân Hạo (1998) and Ngữ âm Tiếng Việt by Đoàn Thiện Thuật (2003) On

the basis of available materials, this study will give one more insight into hesitation expressed by prosodic means in both English and Vietnamese

1.2 Hesitation and reservedness 1.2.1 Definition of hesitation

The term “hesitation” is defined from many different perspectives First of all, from the perspective of psychology, the Oxford Advanced Learner‟s Dictionary 5th Edition

(1995:559) gives a clear concept: “Hesitation is the status of being slow to speak or act

because one feels uncertain or unwilling, to PAUSE in doubt or being worried about or shy of doing something” From the similar point of view, the Macmillan Dictionary for

Advanced Learners 2nd Edition (2002), a popular online dictionary for English learners, also explains that hesitation is a pause before doing something, or a feeling that you should not do it, especially because you are nervous, embarrassed, or worried Synonyms or related words for this meaning of hesitation can be found in words such as “uncertainty, doubt, reservation, question, reserve”

With regard to hesitation in spontaneous speech, a lot of definitions are given by linguists but it is uneasy to have a common definition Firstly, Fox Tree and Clark

(1997:152) defined hesitation as a phenomenon which occurs when “the speaker does not

immediately find an adequate option for language production and is compelled to temporarily delay the output to solve his or her difficulties” Later, Rolf Carlson, Kjell

Gustafson, and Eva Strangert in another study gave similar concept about hesitation

(2006: 21–24): “hesitation is the phenomenon when you are uncertain to what to say or

you have problems in lexical access or in the structuring of utterances or in searching feedback from a listener”

Obviously, Fox Tree and Clark and Rolf Carlson gave out the same explanations when hesitation occurs in spontaneous speech They focus more on the causes of hesitation rather than their expressions In this study, the expressions of hesitation are more considered

about whether something is good or right (The Macmillan Dictionary for Advanced

be understood similarly as hesitation when both of them refer to doubt of doing something

or uncertainty to give an utterance However, in the light of linguistics, the term

“reservedness” is seldom mentioned because it is considered a synonym to hesitation

1.2.3 Types of hesitation and reservedness in spontaneous speech

In spontaneous speech, when being hesitant, the speaker can use various ways to convey his message For example, while studying verbal planning in children‟s speech, Brian MacWhinney & Harry Osser (1977: 980) discovered 9 types of hesitation appearing

in speech which can be recognized by the hearers They consist of silent pause, filled pause, drawls, initial segment phonological repetitions; word- included phonological repetitions, word repetitions, sentence incompletion, false starts and phonological corrections Actually, these nine types of hesitation can be grouped into five: silent pause, filled pause, drawls, repetitions and false starts Similarly, Hieke (1981: 157-158) offered one possible classification of hesitation phenomena including silent pauses, filled pauses, prospective repeats and syllabic prolongations, false starts Heike links hesitations to quality control in the speech production process

Due to limited time and the context of this study, I only focus on the most common aspects of hesitation phenomena from the prosodic point of view Therefore, the following types of hesitation will be studied into detail including silent pause, filled pause, repetitions and syllable prolongation (lengthening)

1.2.3.1 Silent pause

In the studies of hesitation phenomena, Lounsburry (1954: 99) shows that “hesitation pauses are indicative of the strength of association between sequential linguistic elements” Silence has its own communicative value It is possible that the speaker deliberately put pauses into his speech to make the listener‟s job easier, or to aid them to segment speech or

to give them time to parse the speech We have pauses at the end of syntactical boundaries, breathing pauses and hesitation pauses In order to differentiate among these types of pauses, we can look at the below example in a conversation taken from the study of Davidson in which silence is inserted into

A: Well did you want me to just pick you- get into Robinson‟s so you could buy a little pair of slippers?

(silence)

A: I mean or can I get you something?

(based on Davidson, 1984: 104)Obviously, the silence follows the proposal or request which the speaker offers to the hearer However, the appearance of silence implies that the speaker understands their hearer to be reluctant, not hearing, or for some other reason slow to respond Hence, the speaker‟s silence is an intentional signal or it is a hesitation pause

1.2.3.2 Filled pause Filled pauses are hesitation sounds that speakers employ to indicate uncertainty or to maintain control of a conversation while thinking of what to say next Filled pauses do not add any new information to the conversation and they do not alter the meaning of what is uttered They are called fillers which are regarded as extra linguistic noise In English, the set of filled pauses includes /ah, eh, er, uh, um, erm, hm/ Among them, a nasalized “um” and an oral “ah” are the most common fillers Other sounds or non-lexemes can occasionally be used as a filled pause, and some speakers may adopt the words "well", “I mean” and "you know" In Vietnamese, the assertion of /ừm/, /à/, /ờ/ is the distinguishing fillers The examples of (1) and (2) illustrate the most prevalent forms of filled pauses in English and Vietnamese

(1) A: Tomorrow will you go to the cinema with me?

B: I… uh … am busy

(2) A: Em còn đứng đấy làm gì nữa?

B: Em nghĩ … ừm… là chúng ta có thể giúp anh ta

It can be seen that /uh/ in English and /ừm/ in Vietnamese are not uttered randomly by the speakers In both different situations, the speakers tend to produce more /uh/ and /ừm/ before answers because they feel hesitant Below is the summary of filled pauses easily found in English and the equivalences in Vietnamese

Table 1: Summary of vocalized fillers in English and the equivalences in Vietnamese Vocalized fillers in

English

Equivalences inVietnamese

Examples

….er…England (London là thủ đô của… à ờ…Anh) Hmm ừm, hừm Hmm I am not so sure

(Ừm Tôi không chắc lắm.)

(Ừm Tôi không biết dùng nó thế nào cả)

Um, umn ừm "75 divided by 5 is um 15."

(75 chia 5 bằng ừm 15)

Ah a, à "Ah well, I will try."

(À, tôi sẽ thử)

1.2.3.3 Repetitions Repetition is a common phenomenon in spontaneous speech It is often the case that repetition to a spoken dialogue system occurs when the users fail to make themselves understood Repetition can occur when a unit of speech, such as a sound, syllable, word, or phrase is repeated, e.g: "to-to-to-tomorrow"

Repetitions in spontaneous speech in most cases involve a first instance of the repeated word (R1), a possible silent pause (SIL), a second instance of the repeated word (R2), and continuation of the utterance An example is given below:

(a) I might (R1) might (R2) have to go to the cla- class

(b) I might (R1) (SIL) might (R2) have to go to the cla- class

In Vietnamese, repetition often occurs in dialogues as a way of expressing the speaker‟s attitude Like English language, Vietnamese people can make use of repeating a

word; e.g: “Tôi tôi tôi xin lỗi em Có lẽ không cần nữa "(From the short story Con

bé và Gã lang thang: 15 by Chiêu Hoàng)

1.2.3.4 Syllable lengthening Apart from hesitation markers like filled pauses, syllable lengthening is quite common

in spontaneous speech The speaker can lengthen the syllables or words, e.g: English speakers often lengthen the words “a:nd” and “we:ll” in their utterance as in the following example: “Yesterday he came a:nd asked about your book” Moreover, the most common instance of lengthening occurs when the particle “the” is pronounced as “thee” in and the ending vowel sound is drawn out past its usually enunciated duration

In Vietnamese language, syllable lengthening is quite popular in daily conversations as

in the below example:

Teacher: Tại sao em không làm bài tập?

Student: E:m…xin lỗi thầy nhưng mà:…

The student is quite embarrassed when the teacher asks him the reason why he has not finished homework In order to avoid giving a complete answer to the teacher, the student lengthens the words “em” and “mà” in the same utterance

1.3 Prosody 1.3.1 Definition of prosody

Prosody is an important component of language and speech so it is necessary for the readers to catch possible definitions and concepts of this term from different views In linguistics, prosody can be defined as the features that do not determine what people are saying, but rather how they are saying it We can use the term "prosody" broadly, meaning

“a time series of speech-related information that is not predictable from a reasonable

window (i.e word-sized or sentence-sized) applied to the phoneme sequence” (Cutler, A.,

Dahan, D., van Donselaar, 1997: 141-201) Clearly, prosody is a parallel channel for communication, carrying some information that cannot be simply deduced from the lexical channel With this understanding, hand gestures, eyebrow and face motions, can be

considered prosody, because they carry information that modifies and can even reverse the meaning of the lexical channel

In the aspect of speech, prosody only comprises suprasegmental features of fundamental frequency, duration and intensity that contribute to the melody of speech production Most linguists agree that these features are the main components of prosody in English However, the terms used may differ, depending upon whether they are being considered from the speaker‟s point of view (physiological/production), the listener‟s point

of view (perception), or as an acoustic manifestation (measurement)

1.3.2 Prosodic features

Learners who have mastered the pronunciation of the individual sounds of English and the way sounds combine in syllables have gone a long way toward mastering the sound system It is important to learners‟ ability to make themselves understood as any of the other features of the sound system What helps people understand is the characteristic “melody” – the prosodic features of the language –includes variations in pitch, loudness, tempo and length

1.3.2.1 Pitch (or Fundamental Frequency)

Pitch is an important component to establish intonation of utterances, especially for

English It is defined as the frequency of vibration of vocal cords and the relative height of speech sounds as perceived by a listener Pitch represents the fundamental frequency (F0)

of its signal which is calculated as the number of repetitions, or cycles, of its waveform per

second, and given in Hertz (Hz.) Pitch varies over an entire phrase or sentence, which is

manifested by different pitch curves In English, three main levels: fall, rise and level can combine to formulate patterns of pitch including fall, rise, fall- rise, rise – fall, level Each pattern of pitch carries its own function, e.g: fall (the impression of completeness and finality), rise (a certain degree of doubt/ uncertainty), rise – fall (strong feeling of approval

or certainty), fall – rise (limited agreement or response with reservations or hesitation), level (a feeling of saying something routine, uninteresting or boring) For example:

- Today we learn English phonetics (With falling tone, the speaker wants to give a statement)

- Today we learn English phonetics (With rising tone, the speaker puts a question to the hearer)

Different from English language, Vietnamese is a tonal language in which there exist

six tones (Ngo Nhu Binh, 2001: 12-14) Tones have distinctive pitch contours: Ngang has

an almost level contour (e.g: ma); Sắc has high rising contour (e.g: má); Ngã also has an overall rising pattern, but interrupted by a glottalization in the middle (e.g: mã ); Huyền has

a falling tone (e.g: mà); Nặng has a drop tone interrupted by a glottalisation (e.g: mạ) and

Hỏi is gradually falling then rising in the last third back to the original level (e.g: mả) A

change of pitch in Vietnamese can make words change their meaning but it can not change the meaning of the whole utterance like that in English For example: "Đấy là một nhà thơ lớn" (that is a famous poet) versus "Đấy là một nhà thờ lớn" (that is a big church) In the first example a level tone is indicated for the underlined word, whereas in the second example, the underlined word has the falling tone.)

1.3.2.2 Loudness (or intensity)

According to Peter Roach, most people seem to feel that stressed syllables are louder than unstressed; in other words, loudness is a component of prominence (1983:72-73) If in

a sequence of identical syllables, one is uttered with the intensive loudness, it would be heard as stressed syllable Nevertheless, only with changing the loudness of one syllable, the speaker will encounter some difficulties and the perceptual effect of stress is not strong

Loudness is a perceptual response to the physical property of intensity

Unlike English language, Vietnamese is a syllable-timed language in which the rhythm

appears to be fairly even, with each syllable giving the impression of having about the same duration and force as any other; therefore, Vietnamese try to stress syllables by increasing only the volume Loudness in Vietnamese appears in the whole word if the volume changes

1.3.2.3 Tempo

Every speaker knows how to speak at different rates, and this linguistic use of speaking rate is frequently called tempo In the study of speech rate, it is usual to measure either syllables per second or phonemes per second Most speakers seem to produce speech at a rate of five or six syllables per second, or ten to twelve phonemes per second

Many studies have shown that the speech rate of stress- timed languages (e.g English) is generally lower than that of syllable-timed languages (e.g Vietnamese) owing to the fact that the syllable-timed languages tend to have a relatively simple syllable structure

E.g: ba /ba/ father /fð/

mẹ / me/ mother /mð/

1.3.2.4 Length (or duration)

The amount of time that a sound lasts for is a very important feature of that sound In the study of speech, it is usual to use the term “length” for the listener's impression of how long a sounds lasts for, and duration for the physical, objectively measurable time For example, I might listen to a recording of the following syllables and judge that the first two contained short vowels while the vowels in the second two are long: / bit bet bi:t bæt /; that is a judgment of length But if I use a laboratory instrument to measure those recordings and find that the vowels last for 100, 110, 170 and 180 milliseconds respectively, I have made a measurement of duration

Chapter 2: HESITATION AND RESERVEDNESS VIA PROSODIC FEATURES IN

ENGLISH AND VIETNAMESE

In this main part of the study, what I would want to do is to apply the comparative study approach in analyzing English and Vietnamese samples The analysis is to uncover what prosodic features contribute to the expression of hesitation and reservedness in English and Vietnamese spontaneous speech Here, English is used as an instrumental language Prosodic features including pitch, duration, loudness and tempo are tested on PRAAT software to determine whether they have any influence on hesitant speech On the basis of data collection and analysis, the author draws out the similarities and differences between English and Vietnamese prosodic features in showing hesitation and reservedness

2.1 Procedures 2.1.1 Collecting samples of spontaneous speech

In this study, the following data are utilized:

 In English: three job interviews are extracted from online TOEFL tests Each individual short sequence lasts for about 5-7 minutes The applicant in the first interview is male; two applicants in the other interviews are female Each applicant is in a different mood (overconfident, shy and technical) However, the study only focuses on the prosodic cues for hesitation; therefore, psychological factors are omitted

 In Vietnamese: three interviews are recorded from the program “Gõ cửa ngày mới” on VTV1 (a national television channel) Each interview lasts for 4- 5 minutes and famous people are invited for the interview to talk about their career and life The interviewees

who appear in the first and the last interview are male The other interview is female

Most selected samples in both English and Vietnamese are at the scenario of interviews The reason is because the nature of an interview is that one asks and the other answers Therefore, the speakers are not always equally confident about or committed to what they are saying When asked a question, for instance, they can be certain or rather doubtful about the correctness of their answer, and they may be unable to respond at all, even though in some cases it might feel as if the answer lies on the tip of the tongue These

characteristics are appropriate for analyzing hesitation phenomena in spontaneous speech

2.1.2 Methods

After careful listening, the author analyzed the records in both English and Vietnamese

by using automatic speech digitalization technique PRAAT, a programme especially

developed and designed for speech analysis computer by P Boersma and D.Weenink at the Phonetic Sciences Department of the University of Amsterdam was applied Duration, fundamental frequency, loudness and tempo were analyzed, whereby the average F0 was measured (expressed in Hz), and its duration (in milliseconds) Read more about PRAAT software program at http://praat.org

In order to mark hesitation phenomena in each extract, the transcriptions were also used The following conventions of transcription were defined during the analysis All texts are written in small letters “ : ” indicates lengthening of syllable {.} refers to silent pause, “* *” refers to a inserted filled pause, “{number}” marks a pause with a duration

of milliseconds (e.g: {102} means a pause with a duration of 102 milliseconds), “//” indicates a tone unit boundary marker

2.1.3 Methodological difficulties

A total of 19 minutes 29 second in English speech and 13 minutes 32 second in Vietnamese speech was collected from six subjects However, many issues arose during recording that are worth mentioning here Firstly, the processing of audio and video data required very powerful computer resources while the author finds it really unfamiliar Even, this is the first time the author has the chance to access to phonetics analysis by computer software Besides, transcribing and analyzing audio and video data is extremely time- consuming Especially, the author had to calculate the speaking rate Another difficulty is that purely automatized speech processing is error-sensitive because it requires the subjectivity Finally, the subjects were downloaded again from the available resources but not directly recorded; therefore, the quality of sound was not good and some errors during playing the records could happen

2.2 Data analysis 2.2.1 Prosodic feature analysis of hesitation and reservedness in the English samples

In order to characterize the speech at the prosodic level, prosodic values such as duration, F0 slopes, speaking tempo and loudness were measured by Praat version 5.1.3.7

All the samples were recorded and saved in a single wav file for each block for each

participant Not all the sentences in three English samples were analyzed The author only refined 20 typical sentences which contain as much hesitation as possible after listening carefully to each corpus These sentences were digitized and then transcribed in Praat using TextGrids Figure 1 illustrates the sentence which was entered into Praat using Textgrid for analysis In this figure, blue line, grayish image, red speckles and yellow or green line show pitch contour, spectrum, formant and pulses respectively We can get the values of fundamental frequency in Hz, duration in seconds and speaking tempo in syllable/second Besides, pauses are represented as blank lines (see Praat screen on Figure 1)

Figure 1: Praat Editor showing waveform, spectrogram and TextGrid

2.2.1.1 Pitch contour

An important feature to take into account when modeling pitch in hesitation phenomena is F0 slope For example, syllables at the end of a sentence are mainly pronounced with a descending pitch slope, but an interrogative sentence ends with a rising pitch Therefore, it is reasonable to investigate whether there is a standard pitch slope for hesitations or whether, on the other hand, it depends on other aspects, such as semantics or syntax

We will check pitch contour in a typical example extracted from different samples Firstly, we begin with the second sample in which the applicant appears to be unconfident just at the beginning of the interview Look at the following dialogue:

(1) PP: Welcome Now I know you live locally, so I trust you didn‟t have to travel too far Candidate: No

(Interview 2, line 1-4)

The candidate responds to the question by an answer “no” which consists of a fall pitch from rather high to low and then rise to about the middle of the voice In this context, the candidate is quite nervous to begin answering the question given by the interviewer So, she is reluctant to agree what the interviewer says or in other words, she responds with reservation On Praat screen, this is represented in a fall- rise pitch contour (see black curve in figure 2)

Figure 2: Illustration of pitch contour of “No” on Praat screen

In another example, hesitation is expressed more clearly when filled pauses and pauses appear at the same time Consider the following extract:

(2) PP: Finally, are there any questions you‟d like to ask us?

[She thinks for a while]

C: Um….u…no I don‟t think so I can‟t think of anything

(Interview 2, line 78-80)

Figure 3: Pitch contour of the sentence “Um…u…no

I don’t think so I can’t think of anything”

On Praat screen, the answer “no” also is represented in a fall- rise pitch contour Obviously, the use of the fall- rise tune in this case shows reservation on the part of the speaker when it is preceded by filled pauses “um” and “u” and long silent pauses Also, we can use fundamental frequency (F0) to demonstrate the change of pitch contour F0 slope

is manipulated in Hz and it ranges from 75-500 Hz The word “no” /nəu/ is a single syllable which has an onset The F0 value of onset (F0 start value) and the F0 value of the syllable- ending (F0 end value) are measured in Hz From automatic analysis using Praat TextGrid, we get F0 start value at 228.1 Hz, then decreasing to 181.3 Hz and rising again

at F0 end value (200.1 Hz) This means that pitch descends and then rises again

In conclusion, when the speaker wants to implicate a limited agreement or respond with reservations, he can use a fall-rise pitch to convey his intention

2.2.1.2 Duration Duration is the most important cue to the impression of hesitation Firstly, we pay attention to the duration of the pauses and determine how it changes on the part of hesitant speech As far as duration is concerned, the first parameter we test is the length of pauses which is measured in millisecond (ms) Length is scaled automatically using the Praat version 5.1.3.7 duration tier manipulation standard settings Consider the following examples in which pauses are represented in milliseconds between brackets:

(3) // Yeah, no problem.{120} I left loads of time „cos you know what the trains are like nowadays {240}// And I wasn‟t really sure where the Heppleworth site was, but

er…{333} the directions you *er* sent me were crystal clear.// (Interview 1, line 6- 11)

(4)// Um…{5005} u {1831} no {1221}// I don‟t think so {2099} // I can‟t think of

anything.// (Interview 2, line 80)

(5) //…{2332} no, sorry {598} that‟s the molecular ion.{2710} The base peak is….{2881} the most intense peak {1086} // All the other peaks are relative to this {927}// I can‟t

believe I got them mixed up.// (Interview 3, line 29-30)

In example 3, there are very short silent pauses (120 ms, 240 ms) which appear when the applicant has apparently figured out what he wants to convey (subsequent delivery is fairly fluent), but is just trying to stop for breathing In other words, these pauses play the demarcative role between different syntactic components The pause after the filler /er/ seems to last longer (333 ms) when the applicant is trying to find next words However, the duration increase in this case is not significant In contrast, there are significant pauses of more than one second in example 4 producing the delay device of „no immediately forthcoming talk” Even, the pause after the “um” lasts within five seconds and occurs when the speaker is struggling with her hesitant response Similarly, in example 5 the applicant is finding it difficult to express her idea so she is trying to delay the time for each next act Minimum duration of pauses in example 5 is 598 ms, which is longer than the maximum duration in example 3 (+265 ms) In terms of silent pauses, it can be concluded that the longer the pause is, the more certain the impression of hesitance exists

We continue to test in terms of filled pauses and syllable lengthening whether the duration has any change In the examples (3), (4) and (5), there exist three filled pauses

“er”, “um” and “u” and one syllable lengthening “no” The average duration of filled pauses is 510 ms and the average duration of syllable lengthening is 600 ms Whereas, other normal words in the same utterance only get average duration of 100 ms or less It means that there is the total duration increase of filled pauses and syllable lengthening in case of hesitation

2.2.1.3 Speaking tempo

Speaking tempo (or speaking rate) has also been used to demonstrate its influence on hesitant speech Phonetic speaking rate is usually expressed in a syllable- per- second scale

or average syllable duration We will test the expression of hesitation phenomena by calculating speaking rate before and after a pause Normally, a long silent pause or a filled pause shows that the speaker is being hesitant to utter his speech We do not measure the tempo at pause points Instead, the tempo before and after a pause is tested to determine whether speaking rate changes when hesitation occurs Three most typical examples are extracted from different samples in which the pauses are demonstrated to reflect hesitation

mixed up.// (Interview 3, line 29-30)

From the above examples, we can have the total of 12 pauses (pauses are represented in period between brackets) Pauses are marked by ordinal number to identify the order of each pause From the automatic analysis of Praat software, we get the speaking rate before and after a pause as shown in the below table

Pause in order

Speaking rate (syllables/second)

(9) 0.913 1.260

Table 2: Speaking rate (syllables/second) before and after each pause

If we set p1 as average syllable duration before pause and p2 as average syllable duration after pause, we get p1<p2 (p1=0.908, p2= 0.950) The rate of speech of the speaker is slower before the pause than after it This means that the speaker has a lower rate of speech while looking for her words or ideas, then pauses, and has a quicker rate of speech once she has found her words

2.2.1.4 Loudness (or intensity)

Hesitation phenomena are manipulated most clearly in the English corpora through filled pauses Some people perceive them as non-linguistic signals (pauses filled with sound) and other perceive them as linguistic signals – words of English To be neutral and yet retain a bit of some views, we call them fillers in the thesis Fillers are characterized by

“um”, “erm”, “er”, “em”, “uh” in the given samples Below are some examples in which the fillers occur quite frequently

(9) I think, *erm* it was probably problem-solving *Erm* I mean the course, the

apparatus we used, *erm* weren‟t really explained within the course.// (Interview 1, line

81-82)

(10) *Um*…*u*…no I don‟t think so I can‟t think of anything (Interview 2, line 80)

Fillers in the above sentences are represented in unstressed syllables Although the speakers try to lengthen them with longer duration but they are heard much quieter and softer These fillers appear as reduced vowels or as schwas If we listen carefully to the sentences in the records, we will see these fillers pronounced with lower intensity in comparison with other syllables in the same utterance

2.2.1.5 Summary

In the English corpus, silent pauses and filled pauses are the most powerful signals for hesitation than other factors Repetitions and final lengthening also play a quite important role in marking hesitation in the utterances; however, their appearance is not significant in

the given English samples We have shown what is specific in terms of prosody to contribute to the perception of hesitant speech on surface level The results support the conclusion that the change of pitch contour, duration and speaking tempo are the valid cues for hesitation These cues are summarized in the following table:

Pitch contour Fall-rise Duration Duration increase or the pause is longer Speaking tempo Speaking rate is slower before than after the

pause Intensity Lower intensity of filled pauses

Table 3: Summary of prosodic features which contribute to the expression of

hesitation

What is shown in this table is that hesitation is marked by fall-rise pitch contour and total duration increase Moreover, the slower speaking rate before than after the pause also expresses hesitant speech in a clear way Lower intensity of filled pauses also contribute to the expression of hesitation in the speech

2.2.2 Prosodic feature analysis of hesitation and reservednes in the Vietnamese samples

We apply the same methods as in the English samples to analyzing and synthesizing hesitation phenomena in the Vietnamese corpus We have analyzed the presence of four types of hesitation including silent pauses, filled pauses, repetitions and final lengthening

in three different samples All of the samples are spontaneous speech in which the participants involve into the interviews The corpus here was a selection of sentences from Vietnamese samples under Praat using Textgrid 24 sentences were selected from three different speakers (two male and one female) to contain as much hesitations as possible: 34 silent pauses, 9 filled pauses and 5 repetitions Phonetic segmentation was performed automatically and corrected manually The prosodic features were analyzed to test their effect on hesitant speech

2.2.2.1 Pitch contour

In the selected Vietnamese corpus, hesitation points fall on the individual words; thus, pitch is also an inherent part of the individual words Hesitation point is a flat contour if it exists on a word with “ngang” tone However, pitch contour will change at hesitation point

if that word carries “huyền” tone (falling tone) Look at the following example:

(11) Em đang là thợ cắt tóc ở dưới tỉnh Hồi trước em… (M) em có tham gia câu lạc bộ người mẫu, nhưng mà…(E)

In the context of asking for a help, the speaker is hesitant to give out her intention to the listener Her hesitation is expressed by pausing at the words “em” and “nhưng mà” Pitch contour will be tested at points in which a hesitation is placed either in the middle part (M)

or at the ending (E) of the utterance We get a flat contour at the point M and a falling contour at the point E because they are associated with individual syllables “em” and

Figure 4: Illustration of pitch contour of two hesitation points M and E

It can be easily seen that two hesitation points M and E have different pitch contours: falling and rising respectively Pitch contour does not change on the part of each syllable which hesitation exists However, when listening carefully to the above sentences in the records and watching the gestures shown by the speaker, it appears that the speaker is turning down her volume at hesitation points and trying to flatten her voice

In conclusion, pitch contour does not convey purposes of the speaker in the same Vietnamese utterance In this case, a question is raised which prosodic factors contribute to the expression of hesitation in Vietnamese speech At hesitation points, there exist the pauses followed and preceded by filled pauses or repetitions A hypothesis is given out that the duration and volume of these filled pauses and repetitions changes

2.2.2.2 Duration

Normally, Vietnamese speakers may have difficulty in producing the longer words of stress-timed languages like English because each syllable of Vietnamese utterance will receive equal weight Is this true for the case in which hesitation points exist? In example

11 and 12, The words “em”, “nhưng mà”, “ừm”, “khó” exist before hesitation points, which are followed by pauses From automatically extracted Praat software, the average duration of these words is 310 ms; meanwhile the average duration of other normal words

is 305 ms The difference between the average duration of the words before hesitation points and that of other words is not significant although there is an increase in total duration Besides, it is very difficult for the author to extract the average duration of each Vietnamese syllable because the speaker maintains moderate speaking speed between words Look at another example:

(13) PV: Chị có thấy sự đổi khác không khi mà chị nhập vai?

HX: Tôi nghĩ là có.(1) Bởi vì tất cả những vai diễn trước đây của tôi cũng có những vai

mà đều là có gia đình rồi và có conthế nhưng mà…(3) thực sự nó (2) nónó không đời thường như vai diễn “Nhà có nhiều cửa sổ”

Hesitation points are marked by ordinal number 1, 2, 3 At these points, the preceding words receive the average duration of 318 ms, 317 ms and 320 ms respectively Other normal words get the average duration of 315 ms Obviously, syllables in Vietnamese utterance receive relatively equal weight regardless of their occurrence before hesitation points