disfluencies in spontaneous english dialogues poduced r by ... · english dialogues produced by...

53วารสารมนุษยศาสตร์ มหาวิทยาลัยนเรศวร ปีที ่13 ฉบบัที ่3 ประจ�าเดือนกนัยายน - ธนัวาคม 2559

Disfluencies in Spontaneous English Dialogues Produced by Thai Learners of English: A Pilot Study1

การพูดติดขัดในการสนทนาภาษาอังกฤษโดยไม่เตรียมการณ์ล่วงหน้า

ของผู้เรียนชาวไทย: การวิจัยน�าร่อง พัชรินทร์ ดวงศรี

Patcharin Duangsri เสาวภาคย์ กัลยาณมิตร

Saovapak Kallayanamit

Abstract This study aimed to quantitatively examine types of the disfluencies (DFs) in spontaneous English dialogues produced by Thai learners of English, their intentions, and native English-speaking hearers’ interpretations of each DF-type. The instruments were spontaneous dialogues and a checklist for intentions/interpretations of DFs. The findings revealed that the learners produced three main DF-types with their sub-types: pauses, repetitions and repairs. Through cognitive view, pauses and repetitions tended to reflect the learners’ difficulties in speech production process. Through sociolinguistic view, the hearers interpreted L2 DFs as the learners’ difficulties of speech production at the highest frequency, as the learners’ communication strategies serving interpersonal functions for social-interaction purpose at the second highest, while as discourse markers serving the textual functions for semantic coherence and relevance purpose as intended by the learners at the lowest frequency.Keywords: Disfluencies, Intentions, Interpretations, Spontaneous English Dialogues บทคัดย่อ

การวิจัยนี้มีวัตถุประสงค์เพื่อศึกษา การพูดติดขัด ขณะสนทนาภาษาอังกฤษโดยไม่เตรียมการณ์ล่วงหน้า

ของผู้เรียนชาวไทย เจตนารมณ์ของผู้พูด และการตีความการพูดติดขัดจากมุมมองของเจ้าของภาษา เครื่องมือ

ประกอบด้วยบทสนทนาภาษาอังกฤษ และแบบสอบถาม ผลการศึกษาพบว่าการพูดติดขัดของผู้เรียนประกอบด้วย

การหยุดชะงัก การพูดซ�้า และการแก้ไขถ้อยค�า เมื่อวิเคราะห์ตามแนวคิดด้านพุทธิปัญญาพบว่า การหยุดชะงัก

และการพูดซ�า้สะท้อนปัญหาในกระบวนการผลติถ้อยค�าของผูเ้รยีน เมือ่วิเคราะห์ข้อมลูตามแนวคดิด้านภาษาศาสตร์

สงัคมพบว่าผูฟั้งเจ้าของภาษามแีนวโน้มท่ีจะตคีวามการพูดตดิขัดในบทสนทนาของผูเ้รยีนว่าเป็นท้ังสญัญาณบ่งบอก

ถึงปัญหาในกระบวนการผลติถ้อยค�ามากทีส่ดุ และมแีนวโน้มจะตคีวามการพูดตดิขดัเป็นกลยุทธ์ในการสือ่สารเป็น

อันดับรอง และแนวโน้มตีความว่าเป็นดัชนีปริเฉทที่ท�าหน้าที่เชื่อมโยงเนื้อหาตามเจตนารมณ์ของผู้เรียนน้อยที่สุด

ค�าส�าคัญ: การพูดติดขัด การสนทนาภาษาอังกฤษแบบไม่เตรียมการณ์ล่วงหน้า เจตนารมณ์ การตีความ

1 This article is based on a Ph.D. dissertation titled Disfluencies in Spontaneous English Dialogues Produced by Thai Learners of English, Naresuan University (2015).

54 Journal of Humanities, Naresuan University Year 13 Volumn 3, September - December 2016

1. Introduction Speech communication is considered a fundamental ability of people. However, all related knowledge needed for forming the ability is rather complicated and requires a combination of a variety of knowledge and skills such as language knowledge, contextual knowledge, sociocultural knowledge, communication strategies knowledge, appropriate applicability knowledge, speaking skill, conversational skill, and interactional skill (e.g. Levelt, 1989; Levelt et al., 1999; Kormos, 2011; Swain, 1995; Schiffrin, 1987; Dornyei, 1995). Since speaking is a real-time phenomenon, all the required knowledge and skills are needed to be processed within a very limited time. The speaker’s speed in planning, formulating, and articulating an utterance is bound to the current time of communication (Levelt, 1989, Fulcher, 2003). Speakers’ speech production depends on their language proficiency, lexical range, ability to plan about what to say, ability to retrieve vocabulary, grammar, self-monitor for detecting errors in an utterance, hearer effect, and the nature of speaking topic (Levelt et al., 1999; Vural, 2008). Moreover, the interactional nature of speaking situation also requires speakers to make choices of language use to suit the social contexts and of interactional practices to conform to the conventions held by the native speakers (Fulcher, 2003; Levinson, 2003). The degree that the speakers can automatically control all the processes involved in speech production and appropriate speed in communication reflects speakers’ speaking fluency (Fulcher, 2003; Gass & Selinker, 2008). Due to these various sorts of knowledge and skill required for producing speech, it is common that there are some breaks occur in mid-utterance; especially, in the spontaneous speech made by second language speakers (Guara-Tavares, 2013) and they are considered indicators for the fluency (and the lack of fluency) in the speech through second language. Such breaks are known as disfluency phenomena (Johnson, 1961). “Disfluencies” (hereafter, DFs) is a technical term used to refer to normal breaks that disrupt the flow of speech without changing the meaning of the utterances (e.g Brutten, 1963; Jonhson, 1961; Wingate, 1984b; Fox Tree, 1995). They are seen common in spontaneous speech (Swerts et al., 1998). It was reported that, as a native norm of American English speakers’ conversational speech, DFs occur at the rate of 6 times per 100 words and in every 15 words (e.g. Bortfeld et al., 2001; Fox Tree, 2002; Shriberg, 1994).The forms of DFs that are widely recognized are uh, um, ah, er, slips of the tongue, and repetition or correction of words (e.g. Clark & Fox Tree, 2002; Oomen & Postma, 2001; Watanabe, Hirose, Den, & Minematsu, 2007). DFs that have been proposed by previous studies (e.g. Johnson, 1961; Schnadt, 2009; Allwood et al., 1990; Savova, 2002) can be broadly classified as pauses (filled pauses, unfilled pauses (silent pauses), repairs, prolongations, and repetitions. These DFs are found in the speech produced in both first language (L1) and second language (L2) (e.g Fehringer & Fry, 2007; Belz & Klapi, 2013; Hilton, 2007).Through the cognitive view, DFs can reflect speakers’ difficulties in processing speech production (Watanabe et al., 2007; Levelt, 1989; Levelt et al., 1999;


Harley, 2000). In contrast, through the sociolinguistic view, DFs serve as a conversational mechanism for facilitating interaction in communication (e.g. Hartsuiker & Notebaert, 2010; Shriberg, 1996; Clark & Fox Tree, 2002). For example, L2 Speakers tend to use repairs, repetitions, fillers and/or hesitation devices such as using filling words or gambits to fill pauses as their communication strategies to stalling or time-gaining strategies (Dornyei,1995; Canale & Swain, 1980; Fulcher, 2003).The occurrence of DFs is found associated with social and cultural factors: places, time, hearers, purposes, or speaking topics; that is, sociolinguistic competence or appropriateness is involved (e.g. Canale & Swain, 1980; Hymes, 1972; Luoma, 2004; Vural, 2008). When facing difficulties in speaking, L2 speakers tend to use communication strategies to manage their communication to be more efficient (Canale & Swain, 1980; Bachman, 1990; Fulcher, 2003). In spontaneous-speech condition, L2 speech is likely to be characterized with DFs and ungrammaticality while in planned speech condition, L2 speakers tend to demonstrate more fluency and complexity (Belz & Klapi, 2013; de Jong et al., 2013; Guara-Tavares, 2013). L2 disfluent speech is reflected through the high frequency of DFs and it is recognized as the signature of novice L2 speakers (Kasper, 2006). The findings of previous studies (e.g. Fehringer & Fry, 2007; Belz & Klapi, 2013; de Jong et al., 2013; Liu, 2013; Vural, 2008; Kim, 2010; Bila & Dzambova, 2011) revealed that non-native English speakers illustrate inappropriate segmentation of the discourse and overuse of pauses. The higher number of pauses in their oral productions is perceived and assessed as non-fluent revealing their deficiency of English language competence (Kasper, 2006; Bila & Dzambova, 2011) and as needs of more time for cognitive activities: planning, encoding, and articulating speech. The deviant segmentation tends to disturb the hearer and weakens hearer’s comprehension (Bila & Dzambova, 2011). Thus, DFs are signals of cognitive problems occurring during speech production processes and cannot be controlled by speakers. L2 DFs demonstrate the speakers’ difficulties, less automatically speech processing (e.g. pauses) than L1 speakers, and some are indicators of L2 acquisition (e.g. errors repairs, corrective repairs) (de Bot, 1992; Sajavara, 1987; Tang, 2015). L2 DFs are also viewed as communication strategies functioning as the stalling or time-gaining devices and they behave like discourse markers in doing textual and interpersonal functions such as turn taking (e.g Fuller, 2003; Wang, 2009; Zhao, 2013; Bu, 2013; Chen & Pu, 2002). However, these findings are mostly obtained through researchers’ observation. The evidence through the hearers’ and the speakers’ perceptions has been rarely found. With regard to the factors influencing the occurrence of DFs in L2 speech, based on the previous study results, they can be categorized into three categories which include (1) speaker’s speech production system: memory capacity and automaticity of speech processing (e.g. Fehringer & Fry, 2007; Belz & Klapi, 2013), (2) speaker’s language knowledge and ability: level of English language proficiency, linguistic skills (e.g. lexical retrieval speed, articulation speed, and sentence


building speed), language knowledge (e.g. vocabulary and grammar knowledge) and processing skills (Hilton, 2007; de Jong et al., 2013), and (3) the conditions as the speaker engaging in speaking context: familiarity with the addressee, the interlocutors’ mother tongue, the nature of speaking topic, and the speaking duration time (Belz & Klapi, 2013; de Jong et al., 2013; Vural, 2008). The above studies also reveal that types of DFs can (1) reflect the speakers’ difficulties in speech production processes such as lexical retrieval, grammaticality, pronunciation, and appropriateness, (2) reflect English language acquisition of non-native speakers, and (3) have communicative functions in speaking. They serve as indicators for evaluating speaking fluency in international institutions of English communication testing (e.g. International English Language Teaching System or IELTS). In Thai EFL context, the studies of DFs are rarely found. Most of them address the questions on how to promote Thai EFL learners’ fluency in speaking English in terms of teaching and learning. Some focused on investigating for factors affecting EFL speaking and examining pragmatic aspect. The studies by Chotirat and Sinwongsuwat (2011) and Sinwongsuwat (2012) have presented some pragmatic explanation on the DF phenomena in the speech by Thai EFL speakers. The findings demonstrated that repairs, repetitions, and fillers function as turn-holding devices and those speakers used repairs for the purposes of meaning clarification and fluency in speaking. The studies were analyzed in the dimension of speaking fluency in relation to communication strategies leaving the point of DFs as markers of speaking fluency unexplored although some DFs (repairs, repetitions, and fillers) were reported. The knowledge of DFs is a gap that needs to be fulfilled to complete the overall picture of speaking ability among Thai EFL speakers. In conclusion, the knowledge on the forms of the DFs produced by L2 speakers, as well as their communicative functions and interpretations through the native hearers’ perspective has been rarely found. The previous findings indicated an unclear point on the DF roles as communication strategies and as discourse markers facilitating interaction in a conversation as intended by speakers and/or as interpreted by hearers especially in EFL/ESL speech. Thus, the shortcomings indicate that the existing knowledge of DFs is largely through L1 studies. L2 DFs have been mainly studied focusing on the levels of fluency rather than on their interactional aspects in speech communication or their own characteristics and functions. As a result, investigation on L2 speakers’ intentions in using DFs and whether the hearers who are native speakers of English interpret these DFs should be conducted. In order to understand DFs from both cognitive and sociolinguistic views rather than analyzing through only one of them as previously studied, this study applied the framework from the two views. The speech production model of Levelt, (1989) in combination with Levelt et al. (1999) and Manyhart (2003) was used as the cognitive framework to examine the types of DFs in English speech produced by native Thai


learners of English and the speakers’ cognitive difficulties during speaking. The analytical framework of discourse marker functions adapted from Brinton (1996), Aijmer (2002), and Liu (2013) was applied to investigate the discourse-like functions which involve textual and interpersonal functions. The characteristics of L2 DFs were analyzed based on those found in L2 speech. The framework to identify the L2 DFs in doing the functions as communication strategies was adapted from Dornyei (1995) (cf. Dornyei, 1995; Dörnyei & Scott, 1997; and Fulcher, 2003).

2. ObjectivesThis research had 3 objectives as follows:1. To examine types of DFs in English speech produced by native Thai speakers2. To examine the speakers’ intentions during the occurrence of each DF-type 3. To examine native English-speaking hearers’ interpretations of each DF-type

3. Research Questions Based on the three objectives, these specific questions were raised. 1. What were the types of DFs and their frequency produced by Thai learners of English in their spontaneous English dialogues with native English speakers? 2. What were the Thai learners’ intentions when they produce each DF-type? 3. What were the native English-speaking hearers’ interpretations of each DF-type?

4. Research Method This study was designed to employ a quantitative research method to address the three research questions. To serve the study objectives, a survey-based method was applied for data collection.

4.1 Participants There were 9 participants purposively selected for this pilot study: 6 speakers and 3 hearers. The speakers were native Thai speakers who were the fourth-year undergraduate students of an English program at Loei Rajabhat University, a university under the context of English as a foreign language (EFL). They took the same English courses from the first to the fourth year. Thus, they were assumed to have similar experience with sufficient language knowledge and ability to communicate in English meaningfully and likely to produce information-rich data for the study. The 3 hearers were native English-speaking lecturers at Loei Rajabhat University with teaching experience to EFL Thai learners for 2-8 years, and live in Thailand for about 2-8 years. Their educational background was non-linguistics related and their knowledge of Thai language was considered at low-intermediate level. With these qualifications, their evaluation and their ability to articulate the problems and opinions were believed informative for the study.


4.2 Research instruments

4.2.1 Speaking task: the task was designed to create a spontaneous dialogue. The speaking topic involved general information about the speakers; such as, name, hometown, family, education, future work, and experience in traveling abroad.

4.2.1 Checklist for intention/interpretations of DFs: the checklist was developed based on both cognitive and sociolinguistic frameworks.

The cognitive framework was merged from the speech production models of Levelt, (1989), Levelt et al. (1999), and Manyhart (2003). The checklist contained the items related to speech production processes, language knowledge: lexical retrieval, grammaticality, and pronunciations. The items based on sociolinguistic framework involved the interpersonal functions of DFs as discourse markers (DMs) and as communication strategies (CSs). The functions of DMs were adapted from Brinton (1996), Aijmer (2002), and Liu (2013). They consisted of textual functions (coherence and relevance) and

59วารสารมนุษยศาสตร์ มหาวิทยาลัยนเรศวร ปีที ่13 ฉบบัที ่3 ประจ�าเดือนกนัยายน - ธนัวาคม 2559Ty

pes o

f DFs

Cha

ract

eris

tics

of D

FsC

ogni

tive

Vie

wSo

ciol

ingu

isti

c V

iew

Paus

es

Non

-lexi

caliz

ed

fi

lled

paus

es

Lex

icali

zed

filled

Pau

ses

U

nfille

d pa

uses

(Sile

nce

paus

es)

Susp

ensio

n in

mid

-utte

ranc

e w

ith o

r w

ithou

t fille

rsVo

calic

s e.g

. ah,

uh,

um

, eh,

mm

...an

d th

eir p

hono

logi

cal v

aria

nts &

so

und

leng

then

ing

(pro

long

atio

ns)

e.g. t

heee

…. t

huuu

uh,…

, too

ooo…

Non

-voc

alic fi

llers

e.g. w

ell, I

mea

n,

you

know

, and

like

Sus

pens

ion

with

sil

ence

Spee

ch p

rodu

ctio

n D

ifficu

lties

Text

ual F

unct

ions

Inte

rper

sona

l Fun

ctio

ns

-Mar

king

synt

actic

b

ound

ary

-Exp

ress

ing

cont

inui

ty-I

ntro

duci

ng p

ragm

atic

mea

ning

whi

le th

ey ar

e u

sed

with

a C

S -C

larif

ying

synt

actic

am

bigu

ity-H

oldi

ng sy

ntac

tic

com

mitm

ent t

o m

aint

ain

cont

inui

ty

Spea

kers

’ Int

enti

ons

Hea

rers

’ Int

erpr

etat

ions

-Hol

ding

tim

e-G

ivin

g tu

rn-E

xpre

ssin

g sp

eake

r’s

unc

erta

inty

-Hol

ding

turn

-Int

rodu

cing

pro

blem

s in

com

mun

icat

ion

-Hol

ding

tim

e-F

loor

hol

ding

-Pre

vent

ing

othe

rs fr

om in

terr

uptio

n-E

xpre

ssin

g sp

eake

r’s u

ncer

tain

ty

-Mar

kers

of d

iscou

rse

bou

ndar

y cu

eing

for

new

& d

ifficu

lt-to

- p

rodu

ce in

form

atio

n-C

lues

for p

redi

ctio

n of

n

ext u

ttera

nce

-Mea

ning

ful u

nit i

n an

utte

ranc

e in

volv

ing

coh

eren

t pas

sage

s to

the

rece

iver

s-C

ues f

or h

eare

rs to

i

nter

pret

the

inte

nded

m

eani

ng

Rep

etit

ions

S

elf-r

epet

ition

s

Oth

er-re

petit

ions

Repe

ated

wor

ds, p

hras

es, o

r se

nten

ces e

.g. t

o cl

ean…

to c

lean

Repe

atin

g on

e’s o

wn

wor

d or

str

ess o

f wor

ds im

med

iate

ly

afte

r the

y ar

e sa

idRe

peat

ing

som

ethi

ng

the

inte

rlocu

tors

said

Spee

ch p

rodu

ctio

n di

fficu

lties

-Mar

king

for

am

endm

ent o

f spe

ech

-Em

phas

izing

on

spea

ker’s

in

tend

ed m

essa

ge-H

oldi

ng ti

me

-Mak

ing

smoo

th sp

eech

afte

r lo

ng p

ause

-Res

umin

g flu

ency

afte

r in

terr

uptio

n-R

epre

sent

ing

a pr

elim

inar

y c

omm

itmen

t to

spea

king

(t

o av

oid

unne

cessa

ry si

lence

) -F

loor

hol

ding

-Pre

curs

or o

f rep

airs

-Stro

ng p

erce

ptua

l cue

to

the

repe

ated

lexi

cal i

tem

-Clu

es fo

r the

hea

rers

to

pre

dict

the

follo

win

g r

epai

r of u

ttera

nce

Rep

airs

Se

lf-in

itiat

ed

re

pairs

O

ther

-initi

ated

repa

irs

Amen

dmen

ts th

at sp

eake

rs m

ake

resu

lting

in re

form

ulat

ion

of th

e ar

ticul

atin

g ut

tera

nce

i.e. c

orre

ctin

g,

com

plet

ing,

revi

sing

wor

ds/p

hras

es/

sent

ence

s (re

phra

sing)

by

mea

ns o

f su

bstit

utio

ns/in

sert

ions

/ del

etio

nsRe

pairi

ng in

itiat

ed b

y spe

aker

s the

mse

lves

Repa

iring

as in

itiat

ed b

y the

inte

rlocu

tors

Indi

cato

rs o

f SLA

-M

aint

aini

ng se

man

tic

and

synt

actic

cohe

renc

e b

etw

een

inte

nded

spe

ech

plan

and

actu

al

spe

ech

prod

uctio

n -C

larif

ying

pre

viou

s s

tate

men

t -S

elf-c

orre

ctio

ns

-Exp

ress

ing

spea

ker’s

unc

erta

inty

-Tur

n-ta

king

dev

ices

(sp

eake

rs tr

y to

fini

sh t

heir

turn

)-M

anag

ing

spea

kers’

o

wn

perfo

rman

ce

pro

blem

s

No

inve

stiga

tion

foun

d

Tabl

e 1

Ana

lytic

al F

ram

ewor

k fo

r L2

Dis

fluen

cies

: Typ

es, C

hara

cter

istic

s, an

d Fu

nctio

ns


interpersonal functions (social functions). The functions of DFs as CSs were adopted from the Dornyei and Scott’s (1997) taxonomy of CSs which was based on both psychological and social interactional processes. See table 1 for the framework and Appendix A for the checklist for the intentions and the interpretations of DFs.

5. Data Collection The 6 speakers and 1 hearer participated in a speaking task: spontaneous English dialogue on a one-on-one basis. A screen was placed between the speaker and the hearer in order to avoid gestural interference. The length of the dialogue was about 5 minutes. The dialogues were recorded and, within the same day, the recorded verbal data were analyzed for categorizing the types of the DFs produced by the speakers. The next day, all the speakers and the hearer were invited to listen to their own sound records for checking the speakers’ intentions and the hearer’s interpretations of the DFs by using the provided checklist. The sound records were played for stimulating the participants’ recall of the DFs. Another two hearers did the checking individually. Prior to checking in the list, the participants were informed about the instructions and how to use the checklist. Then, they did it independently as the sound record was being played. The sound record was stopped or replayed each time a DF arose and the participants were asked to check the item(s) which was/were true to them.

6. Data Analysis The data consisted of (1) L2 DFs, 2) the speakers’ intentions and (3) the hearers’ interpretations of the DFs. To answer the research question 1, the DF-types obtained from the dialogues were categorized using the analytical framework as shown in table 1. The occurrence frequency of each DF-type was calculated to observe its statistical tendency through the mode value showing the most common category of the DFs. The research question 2 was answered by analyzing the data from the checklist (see Appendix A) where the speakers’ intentions as checked were counted and calculated for the frequency of each intention type indicated by the mode value. For answering to the research question 3, the data on the hearers’ interpretations of DFs were counted. The mode value was applied for statistical analysis.

7. Findings

7.1 Types of DFs and their frequencyThe answer to this question was obtained by using the L2 DFs categorization

criteria which was used as the framework to identify the DF types found in the verbal data from speaking task. The criteria were developed based on the types and the characteristics of L2 DFs revealed by the previous research findings. They contained description of each type of DFs as shown Table 1. The data were categorized by the researcher and another trained categorizer using the criteria. The mode value showed that the 6 Thai learners of English (hereafter referred to as “speaker(s)”) produced three main types of DFs: pauses, repetitions, and repairs. The details are shown in table 2.


7.1.1 Pauses The phenomena showed that the number of pauses was the highest and

much higher than the number of repetitions and repairs. Three sub-types were found: non-lexicalized filled pauses, unfilled pauses, and lexicalized filled pauses. Non-lexicalized filled pauses were found the highest. The markers ah, uh, um, eh, er, and mm…were produced by the learners. Unfilled pauses were found the second highest and marked by silences. Lexicalized filled pauses were the lowest. They were marked with like and and.

7.1.2 Repetitions The total occurrence frequency of the repetitions was at the second to the

pauses. Their sub-types included self-repetitions and other-repetitions. Self-repetitions were found the highest number among the sub-types. Other-repetitions were found the lowest and much lower than the first type.

7.1.3 Repairs Only one sub-type of repairs was found: self-initiated repairs. Its occurrence

frequency was found the lowest among all the three DF-types.

Table 2 Types and Frequency of DFs in English Speech by Thai Learners

7.2 Thai Learners’ Intentions of Producing DFsThe answer to the second question was obtained through the developed

Checklist for intentions/interpretations of DFs. The data were obtained through the speakers’ self-evaluation and analyzed using the mode value and the findings indicated that the Thai learners had intentions to use the DFs on three functions with different frequencies as in Table 3.

7.2.1 Cognitively Managing Speech Production DifficultiesThe frequency indicating the intentions for cognitively managing speech

production difficulties was the highest of 148 tokens in total. Based on the number,

DF-Types DF-Sub-Types Frequency

Pauses Non-Lexicalized Filled PausesUnfilled PausesLexicalized Filled Pauses

Total

108 28

4 140

Repetitions Self-RepetitionsOther-Repetitions

Total

40 3 43

Repairs Self-Initiated RepairsOther-initiated Repairs

Total

190

19Grand Total 202


pauses were used at the highest frequency of 92 for managing speech production difficulties, repetitions were used at the second highest frequency of 39 while repairs were used at the lowest frequency of 17.

Table 3 Frequency of Thai Speakers’ Intentions on Production of Each DF Type

7.2.2 Textual FunctionsComparing with the other two functions in using DFs as intended by the

learners, the frequency for textual functions were ranked the second highest at the total number of 77 tokens. Among them, the learners used pauses for textual functions at the highest frequency of 42. Repetitions were used for textural function purpose at the second highest of 20 while repairs were used at the lowest frequency of 15 for the same function.

7.2.3 Interpersonal FunctionsThe learners demonstrated their intentions in producing DFs for

interpersonal functions at the lowest frequency of 51 in comparison with the other two functions. The findings revealed that pauses were found serving interpersonal functions at the highest frequency of 28, repetitions were at the second highest frequency of 18, and the lowest was for repairs at the rate of 5 tokens.

7.3 Interpretations of DFs by Native English-Speaking HearersThe data for answering this question was also from the developed Checklist for

intentions/interpretations of DFs. The hearers’ interpretations of DF-type revealed that the learners’ DFs served three purposes as shown in table 4 and detailed below.

7.3.1 Cognitively Managing Speech Production DifficultiesThe data revealed that pauses, repetitions, and repairs were all interpreted

DF-Types Intentions Frequency(Tokens)

Pauses Speech Production DifficultiesTextual FunctionsInterpersonal Functions

Total

92 42

28 162

Repetitions Speech Production DifficultiesTextual FunctionsInterpersonal Functions

Total

39 20

18 77

Repairs Speech Production DifficultiesTextual FunctionsInterpersonal Functions

Total

17155

37


as the learners’ cognitively managing speech production difficulties at the highest frequency in comparison with other aspects of interpretations.

7.3.2 Textual FunctionsIt was found that pauses and repetitions were interpreted as being used

for textual functions at the lowest frequency (93 and 45, respectively) in comparison with other aspects of interpretations for pauses and repetitions. In contrast, repairs were interpreted as being used for textual functions and for cognitively managing speech production difficulties at the same highest frequency of 39.

7.3.3 Interpersonal FunctionsFor interpersonal functions, pauses and repetitions were interpreted as

having interpersonal functions at the second highest frequency of 126 and 53, respectively (the highest was for cognitively managing speech production difficulties). In comparison with pauses and repetitions, repairs were interpreted as being used for interpersonal functions at the lowest frequency of 4.

Table 4 Frequency of Interpretations of DFs by Native English-Speaking Hearers

8. Discussion

The findings revealed that through cognitive view, the DF-types produced by the speakers reflected both their speech processing difficulties and their L2 acquisition. Through sociolinguistic view, the speakers and the hearers indicated that the DFs were mainly used as a means for managing the speakers’ speech production difficulties. However, the DFs were sometimes used as a textual coherence device by the speakers while the hearers perceived that the DFs were likely to be used as the speakers’ communication strategies. More details were discussed below.

DF-Types Intentions Frequency(Tokens)

Pauses Speech Production DifficultiesTextual FunctionsInterpersonal Functions

Total

185 93

126 404

Repetitions Speech Production DifficultiesTextual FunctionsInterpersonal Functions

Total

86 45

53 184

Repairs Speech Production DifficultiesTextual FunctionsInterpersonal Functions

Total

39394

82


8.1 DF-TypesPauses: Thai learners tended to produce pauses at the highest frequency.

This phenomenon confirmed the assumption by the previous studies (Levelt, 1989; Levelt et al, 1999) that pauses occurred at the formulating level rather than the articulating level. The DFs that occurred at the formulating level were resulted from the speakers’ difficulties in memory capacity and automaticity of speech processing. That is, the speakers tended to process words by encoding the preverbal message with syntactical, morphological, phonological, and phonetic features. They paused both with and without fillers and self-monitored to detect errors in their speech production process before articulating the encoded preverbal message. The pausing was a reflection of the internal checking process that the speakers were performing in their minds (Levelt, 1983, 1989; Kormos, 2006). It can be concluded that the checking process involved self-monitoring of covert speech (cf. Menyhárt, 2003; Levelt et al., 1999; Levelt, 1989), an utterance had not yet been articulated; the DFs were then produced in the forms of pauses. In addition, since speech production process is feed-forward on a step-by-step basis (Levelt, 1989; Levelt et al., 1999), the execution of the next process will not be possible if the current process is not completed. As a result, the arising pauses were the DFs that occurred before the speakers articulated an utterance; in other words, the speakers paused while they were formulating their preverbal message. The findings were in line with the previous studies that pauses produced by L2 speakers reflected their low automaticity in producing an L2 utterance.

In terms of the sub-types, the difficulties were marked with non-lexicalized filled pauses such as ah, uh, lexicalized filled pauses such as like and and, or unfilled pauses (silences). According to the studies, the use of lexicalized filled pauses such as like, you know, well, and ok were indicators of L2 acquisition. The current data informed that there were two lexicalized fillers: like and and produced by a speaker with high level of language proficiency. The present finding agreed with the results from Bu (2013) and Zhao (2013) that this type of fillers tended to be markers for L2 acquisition. Although more information on the interaction between the speakers’ language proficiency level and the use of this DF-type is required for valid statistical analysis, the DFs can be considered a positive sign of the speakers’ potential to develop their English language since the speakers’ efforts to produce English utterance were obviously observed.

In contrast, it can be noticed that the numbers of non-lexicalized filled pauses and unfilled pauses were found the highest in the speech made by the lower L2 proficiency group. It was found that the weakest speaker of the group produced the highest number of pauses in total of 45 tokens in a 5-minute dialogue. The frequency of non-lexicalized filled pauses was the highest and that of the unfilled pauses was the second (38 and 7 tokens, respectively). Although the result confirmed the findings in the previous studies that the overuse of pauses in L2 speech revealed the speakers’ weak L2 competence, there was an interesting finding in the current study that the strongest


speaker of the advanced group also produced high number of pauses, 21 tokens in total, consisting of non-lexicalized filled pauses at the highest number of 15, 2 unfilled pauses, and 4 lexicalized filled pauses. However, the difference between the two cases was that the non-lexicalized filled pauses produced by the advanced speaker were mostly found at the initial position of the utterance (before starting to answer the questions asked by the interlocutor) while the occurrence pattern of the fillers produced by the lower proficiency speaker was unpredictable and he did not produce any lexicalized filled pauses. The findings agreed with Shriberg (1994), Fox Tree (2001), and Bailey & Ferreira (2003) that the fillers occurring at the beginning of an utterance were associated with macro-planning of the upcoming speech whereas the fillers within an utterance were related to micro-planning process. That is, the pauses produced by the advanced speaker tended to deal with difficulties concerning idea conceptualization rather than grammatical aspects as did among the low proficiency speakers. It can be concluded that pauses tended to reflect L2 speakers’ speech production difficulties occurring at the formulating level. The sub-types of pauses and the occurrence position of non-lexicalized filled pauses can be the markers of L2 acquisition since the speakers’ language proficiency level tended to interact with the types and the behaviors of the DFs that they produced. However, future research on the interaction between L2 speakers’ language proficiency and the use of lexicalized filled pauses should be conducted quantitatively.

Repetitions: based on the data, the two sub-types of repetitions found in the previous studies also appeared in the speech produced by the speakers in the current study. Although the frequency of repetitions was much lower than that of the pauses, the positions of their occurrence in an utterance clearly indicated the speakers’ cognitive difficulties in processing their speech. It was found that they tended to co-occur with other DF types. They were found both before and after pauses (Ex.: ...wave people, wave people …uh…people, people wave). This phenomenon can be explained that repetitions were markers of cognitive difficulties. This is because when they preceded pauses. It meant that the speakers found an error in their articulated utterance before the pause and tried to re-encode the message while repeating the message. However, they failed to do it or were uncertain to change their utterance. Their uncertainty could be about conceptualizing the ideas or encoding linguistic features. As a result, they stopped their repetition and paused for thinking. On the other hand, when repetitions followed pauses, it suggested that after the speakers’ first checking had been completed, they decided to produce an utterance but it was found not as intended. Thus, they made repetition of the entire or a part of the utterance trying to detect the error in the utterance articulated. Thus, repetitions served as hesitation devices and reflected low automatically speech processing which were out of the speakers’ control. Other than pauses, repetitions also co-occur with repairs. They were mostly found preceding repairs. The findings suggested that repetitions tended to be markers for repairs. The findings supported the previous results studied by Chotirat & Sinwongsuwat (2011) that


repetitions in L2 speech occurred prior to repairing the utterance. It can be explained through cognitive view that repetitions were markers of low automaticity in speech production of L2 speakers as evidenced by the repairs that they made after repetitions. This result supported the above findings of repetitions when they co-occurred with pauses. The mode value showed the highest frequency of self-repetitions in the advanced group. The frequency was considered a positive sign of L2 acquisition that the speakers were able to self-monitor, self-detect errors, and self-initiate to correct the errors in their own utterance. In contrast, the lower proficiency group produced the highest frequency of other-repetitions (2 tokens) while it was found only 1 in the advanced group. It can be explained that the higher frequency in using self-repetitions in the advanced speaker group suggested their higher automaticity in speech production than the weak group since they were able to self-notice the errors in their own utterances. However, they may not have sufficient required knowledge for amending the utterance. Thus, they felt uncertain and made repetition of the utterance. On the other hand, the lower language proficiency group showed higher number of other-repetitions than the advanced group. That means they were unable to identify their own errors unless there was a signal from the interlocutor. Thus, they made repetition of the interlocutor’s utterance while thinking to amend the errors. The findings about the phenomena of repetitions confirmed the conclusion that repetitions indicated L2 speakers’ speech production difficulties, and the L2 speakers’ language proficiency level tended to interact with the sub-types of repetitions: self-repetitions tended to be used by the advanced speaker group while other-repetitions tended to be used by the lower one. The repetitions that were found both before and after pauses suggested the speakers’ engagement in the self-monitoring of both covert and overt speeches to detect errors in their utterances. The processes occurred in sequence. This is a piece of evidence confirming that speech production process goes in only one direction (see also Levelt, 1989, Levelt et al., 1999). The articulation of message will not be possible if the message formulating is not finished (regardless of its mistakes). As a consequence, repetitions were found preceding and following pauses.

Repairs: the findings showed two sub-types of repairs: self-initiated repairs and other-initiated repairs as previously analyzed in Liyanage & Gardner (2013), Chotirat & Sinwongsuwat (2011), van Hest (1996), and Kormor (2002). Self-initiated repairs were the DFs that occurred when the speakers found errors in their speech through self-monitoring process and decided to repair the utterance. Doing the repair, the L2 speakers were required to possess more communicative ability which involved language knowledge, contextual knowledge, socio-cultural knowledge, communication strategies knowledge, and appropriate applicability of all the knowledge while other-initiated repairs did not because the repairs were initiated by others. Although self-initiated repairs were produced by the learners, the frequency was found the lowest among the three DF-types. It can be explained that due to the highly cognitive demanding for


making an utterance repaired, the L2 speakers tended to produce the DFs at a lower rate. Thus, they tended to produce pauses rather than this type of DFs. However, the findings revealed very close frequencies of self-initiated repairs produced by the speakers with different levels of language proficiency and also the slightly difference in nature of their self-initiated repairs. It was also found that speakers in each level produced the same frequency of 5 for repairing by insertion and 3 by revision. The data showed that the lower proficiency speakers repaired their speech by deletion at the higher number than the advanced speakers did. This can be explained that as repairing did not involve only the well-formedness rule but also the pragmatic rule, the speakers were required to do the two processes simultaneously (as discussed in van Hest, 1996 and Kormor, 2002). Thus the speakers with low language proficiency level may not have sufficient knowledge to make a repair by insertion or revision which required more communicative ability than repairing by deleting information. That is, the nature of repairs could reflect the speakers’ L2 acquisition and the frequency of repairs could not be claimed as markers of the speakers’ non-fluency in L2 speaking. It was likely that the speakers’ level of language proficiency did not interact with the frequency of repairs but it tended to deal with the nature of repairs. The findings were supportive to previous studies (e.g. Liu, 2009) that the qualitative characteristics of DFs rather than the frequency reflected the speakers’ L2 acquisition. However, the qualitative characteristics of repairs in the current study were based on communication strategies (insertion, revision, and deletion) rather than semantic aspects (sentential and discourse levels). Thus, the sociolinguistic -based interpretation of repairs by the current study contributes to the more understanding of repairs as a type of DFs that the Thai learners used for managing their difficulties in communicating in English.

8.2 Thai Speakers’ Intentions when Producing a Type of DFBased on the speakers’ intentions, the phenomena showed the same direction

of how they made use of all the DF-types. All DF-types: pauses, repetitions, and repairs, were found serving the functions to cognitively manage speech production difficulties at the highest frequency, having textual functions for semantic coherence and relevance purposes at the second highest frequency, and doing interpersonal functions for social-interaction purposes at the lowest frequency. This can be explained that the DFs produced were likely to be used as signals for cognitively managing speech difficulties and as discourse markers for textual functions rather than as communication strategies. Thus, the DFs produced reflected that the problems that the speakers mainly concerned were in the domain of language knowledge rather than contextual or sociocultural knowledge. Thus, their DFs were considered hesitation devices signaling their English language deficiency with low automaticity in processing speech production. In other words, through the Thai learners’ lenses, their DF phenomena demonstrated their non-fluency state and their degree of English language acquisition.


8.3 Native English-Speaking Hearers’ Interpretations of the DFsThrough the hearers’ interpretations, the DFs served for cognitively managing

speech production difficulties at the highest, for interpersonal functions at the second highest and as textual functions at the lowest. The findings indicated that, mainly, the hearers viewed all types of the DFs as signals of speech production difficulties. Such view corresponded to the speakers’ intentions when producing each DF. The difference is that the hearers’ found that the DFs served interpersonal functions at the second highest while the speakers viewed them as signals of language deficiency. This can be explained that the hearers tended to interpret the DF markers as signals for communication strategies which were the speakers’ efforts to achieve the communication goal. Thus, for the hearers, although the DFs were perceived as hesitation devices signaling the speakers’ low language proficiency, the hearers tended to understand the DFs as the speakers’ means to socially interact with the interlocutor on a purpose. That is, through the hearers’ interpretations, the DFs like pauses and repetitions tended to be perceived as the speakers’ non-fluency but they were likely to be used as communication strategies while repairs were indicators of speakers’ uncertainty and functioned like a discourse marker for textual function in clarifying the previous statement or as self-corrections (Tsychiya and Handford (2014). Thus, the current findings do not agree with Dörnyei and Scott (1997) and Faerch and Kasper (1983) that repairs are classified as communication strategies.

9. Conclusion

The Thai learners of English produced three types of DFs: pauses, repetitions, and repairs. Through cognitive view, the DF-types reflected the leaners’ speech production difficulties and the degree of their English language acquisition. Their language proficiency levels tended to interact with the types rather than the occurrence frequency of the DFs. Through sociolinguistic view, the most- likely-to-occur functions as intended by the speakers and as interpreted by the hearers was DFs as markers signaling the speakers’ less automaticity in producing speech and DFs as discourse markers serving textual functions in linking the messages in speech to create meaning relationship within utterances rather than DFs as communication strategies and discourse markers for social interaction serving interpersonal functions during conversation. The findings of the current study offer a body of knowledge for the researchers in the fields of second language acquisition, linguistics, sociolinguistics, applied linguistics, and other related fields. Since DFs serve as indicators for evaluating speaking fluency in international standardized tests such as International English Language Teaching System or IELTS, the speakers’ realization of the hearers’ interpretations can make them pay more attention to the DFs to control the frequency of the DF occurrence in speech, and make use of the DFs for socially interact with the interlocutor on a purpose. The findings cannot only help promoting EFL/ESL teachers’ and learners’ realization of the native English-speaking hearers’ interpretations of L2 DFs but also the interaction between


EFL/ESL learners’ levels of English proficiency and the types of DFs that they produce. The EFL/ESL teachers may apply the knowledge for creating an appropriate lesson plan to promote EFL/ESL learners’ speaking ability and to void overusing DFs. In addition, the findings can be of benefits to the use of English speech communication in other contexts such as in business negotiation where DFs can be useful if the speakers know how the hearers’ interpretation of the signals is likely to be.


References

Aijmer, K. (2002). English discourse particles: Evidence from a corpus.(Topics in English. Linguistics). Amsterdam: John Benjamins.Allwood, J., Nivre, J. & Ahlsén, E. (1990). Speech management - On the non-written life of speech. Nordic Journal of Linguistics 13, 3-48.Bailey, K.G.D. & Ferreira, F. (2003). Disfluencies affect the parsing of garden-path Sentences. Journal of Memory and Language, 49, 183-200.Belz, M. & M. Klapi (2013) Pauses following fillers in L1 and L2 German map task dialogues. In R. Eklund, (Ed.), In Proceedings of DiSS 2013: The 6th Workshop on Disfluency in Spontaneous Speech, 9–12 Brinton, L. (1996). Pragmatic markers in English. Grammaticalization and discourse functions. Berlin:de Gruyter.Brutten, E. (1963). Palmar sweat investigation of disfluency and expectancy adaptation. Journal of Speech and Hearing Research, 6, 40-48.Canale, M. and Swain, M. (1980). Theoretical bases of communicative approaches to second language teaching and testing. Applied Linguistics, 1(1), 1–47.Chotirat, S. & Sinwongsuwat, K. (2011). Effects of scripted and non-scripted role play Activities on oral performance: A case study of repair organization in conversation in conversation of Thai college students. In Proceedings of the 3rd International Conference on Humanities and Social Sciences, Faculty of Liberal Arts, Prince of Songkhla University, Thailand.Clark, H.H. & Fox Tree, J. (2002). Using uh and um in spontaneous speaking. Cognition 84 (1),73 - 111.de Jong, N., Groenhout, R., Schoonen, R. & Hulstijn, J.H. (2013). Second language fluency:Speaking style or proficiency? Correcting measures of second language for fluency first language behavior. Applied Psycholinguistics, 36(2), 223 - 243.Dörnyei, Z. (1995). On the teachability of communication strategies. TESOL Quarterly, 29(1),55-85.Dörnyei, Z. & Scott, M.L. (1997). Communication strategies in a second language: Definitions and taxonomies. Language Learning, 47(1), 173-210.Dörnyei, Z., & Kormos, J. (1998). Problem-solving mechanisms in L2 communication: A psycholinguistic perspective. Studies in Second Language Acquisition, 20, 349–385.Fox Tree, J. E. (1995) The effects of false starts and repetitions on the processing of subsequent words in spontaneous speech. Journal of Memory and Language, 34, 709-738.Fox Tree, J. E. (2002). Interpreting pauses and ums at turn exchanges., Discourse Processes, 34 37-55.Fulcher, G. (2003). Testing Second Language Speaking. London: Pearson Education.


Hilton, H.E. (2007). What pauses reveal about language processing in spoken L2 production. 6th International Symposium on Bilingualism, Hamburg.Hilton, H. E. (2008). The link between vocabulary knowledge and spoken L2 fluency. Language Learning Journal, 36, 153-166. Kormos, J. (2006). The structure of L2 self-repairs in the speech of Hungarian learners of English. Acta Linguistica Hungarica, 53(1), 53-76.Kormos, J. (2011). Speech production and the cognition hypothesis. In P. Robinson (ED.),Second language task complexity. Researching the cognition hypothesis of Language learning and performance, 39-60. Amsterdam: John Benjamins.Levelt, W., Roelofs, A., and Meyer, A.S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22 (1), 1-75.Liu, B. (2013). Effect of first language on the use of English discourse markers by L1 Chinese speakers of English. Journal of Pragmatics, 45, 149-172.Luoma, S. (2004). Assessing speaking. Cambrideg: Cambridge University Press.Menyhárt, K. (2003). Age-dependent types and frequency of disfluencies. DiSS'03, 45-48.Schiffrin, D. (1987). Discourse Markers. New York: Cambridge University Press.Schnadt, M.J.(2009). Lexical influences on disfluency production (Doctoral dissertation). Retrieved from ProQuest Dissertation and Theses Database. (AAT U572820)Shriberg, E. (1996). Disfluencies in switchboard. Paper presented at the International Conference on Spoken Language Processing (ICSLP '96), Philedephia, PA.Swerts, M. (1998). Filled pauses as markers of discourse structure. Journal of Pragmatics, 30 (40), 485-496.Tang, C. (2015). Applications of stalling mechanisms in Chinese-English bilinguals’ L1 and L2 spoken discourse. International Journal of Bilingualism, 19(1) 92–114.van Hest, E. (1996). Self-repair in L1 and L2 production. Tilburg, the Netherlands: Tilburg University Press.Vural, E. (2008). Disfluency in second language: A study of Turkish speaker of English (Master’s Thesis) Retrieved November 24, 2014, from http://etd.lib.metu.edu. tr/upload/126100/index.pdf.

disfluencies in spontaneous english dialogues poduced r by ... · english dialogues produced by...

Documents