Distribution of Hesitation Discourse Markers Used by Iranian EFL Learners during an Oral L 2 Test

Previous st udies o n hesi tation st rategies used by beginner or ad vanced L2 l earners revealed t hat beginners mostly leav e th eir hesitation p auses un filled which ca uses t heir sp eech t o s ound disfluent, a nd advanced learners tend to use various fillers in order to sound like native speakers. The present p aper reports o n a st udy w hich i nvestigated t he distribution of hesitation discourse markers including silen t pauses, silen t pauses and fillers, f illers, and n on-lexical words used b y Iran ian u niversity students in an oral (L2) test. The st udy e xamines t he l ocation of t he discourse m arkers of hesitation acr oss utterances produced by t he participants. The respondents were a group of students registered in the Tertiary English Language Program at a university in Kuala Lumpur, Malaysia. The aim was to identify the frequency of all h esitation strategies used in four locations of In itial, Middle, and Final position of the u tterances to find out the most frequent location of hesitation during an oral (L2) test.


Introduction
This p aper r eports on a study w hich aim s to d iscover the d istribution pattern of hesitation d iscourse markers across the utterances produced by Iranian university students who had studied English as a fo reign language in their home counry.This article begins with reviewing the literatu re of related studies carried ou t on h esitation strategies, and th en pr oceeds with an ov erview of t he research design, data collection proce dure, and data analysis organized in the frame of results.Finally, the last section of the paper develops the contributions of the present research to the domains of EFL and ELT in the conclusion part.

Hesitation and Hesitation Strategy
Hesitations are pauses of varying lengths, which are not usually left unfilled.They usually occur when a speaker finds himself/herself in a po sition where he/she lack s t he words to use or strug gles with cog nitive o r v erbal planning.Even native speakers fill h esitations when they speak and use fillers including non-lexical fillers like lengthening or stretching sounds, quasi-lexical fillers, repeating lexical items, and finally lexical fillers.(Rieger, 2003) Several researches have been carried out to i nvestigate the phenomenon of hesitation in Eng lish as a second or foreign language besides the remedies used by the speakers who cannot find their desired words or structures.This phenomenon app ears eith er in cogn itive or verbal p lanning stag es, and is categ orized as a typ e of communication st rategy being used t o help t he s peaker t o kee p t he ground of speech d uring hesi tation.A s Richards a nd Schmidth (2002) state, com munication st rategies help to compensate for weakn esses i n bu ilding and m aintaining ef ficient c ommunication, an d e nable t he s peaker t o m ake up or improve breakdowns i n communication efficiently.Little more (2003) fo und o ut th at learn ers of different c ognitive styles use d ifferent types of CSs, and Nakatani and Sayer (2005) discussed that the more communication strategies are focused, the more pervasive effect will be left on students' awareness to employ appropriate strategies in interviews.

Disfluencies
It has been frequently asked whether or not disfluencies are perceived.Although many students are quite good in basic listening and speaking skills, their oral communication skill seem stronger than the o thers.Those who are privileged with this c ompetence have been proved to be are more successful at sc hools and ot her areas of their lives.Studies have shown t he possibility of teaching oral communication strategies in L2 which can be furthe r practiced, improved, and applied by the learners successfully.
Disfluencies a re strate gic devices which si gnal the speaker's "under construction utterance".T hey have als o been ch aracterized as t he automatic effects of co gnitive burdens, particularly durin g sp eech pro duction management ( Nicholson et al,20 03).Spontaneous sp eech con tains all kinds of disfluency ph enomena su ch as silent p auses, hesitations, rep etitions, fillers, grammatical errors, m isselected lex ical i tems, self-co rrections, prolongations, false starts, slips of the tongue, etc., which occur because of disharmony between speech planning and exec ution stage.In fact , speech disfluencies are defined as phenomena interrupting t he fl ow of s peech without adding propositional content to an utterance.(Menyhárt, 2003)

Function of Disfluencies and Hesitation
Studies have shown that disfluencies have different functions.Some researchers have claimed that pause fillers (e.g.'uh' and 'um') serve the discourse function of turn taking.Clark and Fox Tree (2002) agree with this lexical role and discuss that "their communicative function will be lost if they are not recognized".(Lai.Gorman Yuan., & Liberman, 2008) Silent p auses have been known to facilitate breath ing, and en able the sp eaker t o harmonize h is/her sp eech processes, and at the same time allow the listeners to better comprehend and digest what they have heard.Other types of disfluency phenom ena are known as "errors", which ar e almost always dis tracting for the listener.Recently, a study on hesitations and disfluencies in speec h, showed there is an insta nce of disfluency for e very six words in spontaneous s peech, alth ough, in l onger monologues t he freq uency of e rror is for every 3.6 words.(this does not include silent pauses) (Menyhárt,2003).
According to Maclay & Osgood (1959) FPs can be used to control the c onversational "ball".They may be used to tak e th e lead of th e con versation away fro m th e coun terpart of t he i nteraction (Murata, 1994), and keep a conversational turn to ensure no one is taking over one's turn.
FPs have been mostly categorized as a ki nd of hesitation disfluncy including false starts, restarts, silent pauses, and filled pauses.(Maclay & Osgood,1959).
If a speaker "pauses long enough to receive the cue of his own silence, he will produce some kind of signal ([m, er]. ..) which says, in effect, 'I'm still in control-don't interrupt me!'" (Maclay & Osgood, 1959).
Beattie (1 977) stu died th is p henomenon in so me n aturally-occurring conv ersations, and d iscovered th at interruptions were occurring more frequently during silent pauses rather than in the presence of FPs.
Besides looking after the speech discourse structure, speakers need to signal the beginning and ending of their conversational turns to others too.It is commonly believed that interlocutors are freer to interrupt in interactive situations rat her t han th e public situ ations.Th erefore, it is en tirely im portant for t he speakers to main tain t he control of conv ersations in which FPs ar e of gr eat use pr oactively bu t at the sam e time coope ratively.(Rose, 2008) The classificatio n of un filled an d filled p auses un der cognitive/non-language-specific, an d semantic/language-specific groups recei ves some support from the way native s peakers pe rceive L2 learners.Studies have sh own th at the freq uency an d particularly th e d istribution of unfilled p auses m ight in fluence negatively on the way native speakers of English perceive and judge L2 learners' proficiency level (Dewaele, 1996;Lennon, 1990;Trofimovich& Baker, 2006), although using typical hesitation markers adds on proficiency in speech.(Schmid &, Fa¨gersten, 2010) Researches have shown that hesitation pause group is not really homogenous.In particular, silent pauses emerge in a real ly di fferent pattern from t hat of t he FPs' .Fo r e xample, si lent p auses (a nd ot her HP s excl uding F Ps) represent speaker's anxiety more than FPs (Goldman-Eisler, 1961;Kasl & Mahl, 1965;Krause & Pilisuk, 1961;Mahl, 1956;Ragsdale, 1976).It can be simply inferred that listeners who judge frequent hesitation negatively, build their negativity based on FPs, rather than the other HPs, or even the interaction between FPs and other HPs.(Rose, 2008) (1981) was one of the first few researchers who discovered that non-native speakers use more self-repairs compared to native speakers.Wiese (1984) found out that L1 and L2 productions entailed different processes in his study of self-repairs.Wiese also s howed that L2 speakers employed a larger number of self-repairs than L1 speakers did.He stated t hat L2 spe akers' e rror in s peech is mo re th an L1 s peakers', and L2 sp eakers ten d to correct their own errors more than L1 speakers do.He also proposed that L2 speakers required more time to plan their utterances du e to th eir in adequate knowledge of t heir L2, a nd they showed less aut omatization i n processing t heir seco nd l anguage w hen compared to what they did i n their L1.O 'Connor ( 1988) stu died the speech of beginner and advanced L2 learners and found out that beginners use fewer self-repairs than advanced learners.They tend to em ploy vari ous ki nds o f self-repair su ch as co rrective rep airs (rather th an anticip atory repairs (covert repairs), but advanced learners utilize more anticipatory s elf-repairs.Temple (1992) focused on self-repair in the speech of L1 and beginner L2 users.She analyzed speech and repair frequency in both groups and found out that native speakers seem to speak twice as fast when compared to non-native speakers because of the frequ ent an d sk illful app lication of fil lers.In con trast, n on-native speakers m ostly leav e th eir h esitation pauses unfilled, produce more false starts, and leave more errors uncorrected.(Rieger, 2003) Although disfluencies ca nnot be se parated from the spontaneous speech, listeners still discard repetitions a nd filled p auses easily.In fact, disfluencies are a cro ss li nguistic fact o f l ife, wh ich have n ot been affected b y inter-language differences in distributions.(Lai. Gorman.Yuan., & Liberman,2008) The find ings of so me co mparative st udies on d isfluency m arkers am ong m onolinguals and b ilinguals have shown that a bilingual's tas k is cognitively more complex; because, she/he accesses two linguistic system s and need to m anage bo th of th em si multaneously.Th is m ay cause an inc rease in the em ergence rate of CDMs i n speech particularly in the weaker language of the speaker.According to Grosjean (2001), the dominant language might be infl uenced too; because mostly all linguistic system s remain almost active in a bilingu al's mind at all time.

Hieke
As a result, it can be no longer surprising to find higher frequency of disfluency markers among bilinguals and a different distributional pattern from those by monolinguals.(Schmid & Fa¨gersten, 2010) 1.5 Previous Studies on the Distribution of Hesitation Discourse Markers Maclay and Osgood (1959) carried out a stu dy to find out the distribution pattern of hesitation markers across utterances w hithin different l exical cat egories; ho wever t hey di d not t ake spea kers' s ociolinguistic di fferences into account.The results showed th at filled and unfilled pauses might appear anywhere in the utterances, but in particular filled pauses are m ore likely to precede function words or at phrase boundaries, whereas unfilled pauses mostly occure before lexical words or within syntactic boundaries".(Purvis, 2008) In an other similar researc h, A ndersen (2001) focused on t he distribution of Eng lish d iscourse m arker: like.Andersen m ainly stud ied t he d istribution of th is lex ical un it in term s of its fun ction.Similar to Maclay an d Osgood, Andersen focused more on the syntactic environment rather than the sociological factors.(Purvis, 2008) On the other hand, Bailey and Ferreira (2003) carried out a psycholinguistic study focusing on the perception of short discourses.The findings of the study showed that listeners are more sensitive to short discourses.

The two following sentences are examples of what they presented to the respondents:
"Sandra bumped into the busboy and the uh uh waiter told her to be careful."and "Sandra bumped into the busboy and the waiter uh uh told her to be careful." They found out that in case filled pauses preceded the head noun (which is "the waiter" in the above example) of the second clause, the respondents tended to interpret the noun as a lex ical unit starting a n ew clause instead of the direct obj ect of the previous clause.Thus, it wa s discovered that, in case a filled pause appears in a place which might be perceived like a discourse boundary, then a listener would consider it like a discourse boundary too.(Rose, 2008)

Discourse Management
Conversation analysts co mmonly b elieve th at filled p auses m ostly e merge in two major lo cations with in a discourse structure: whether at the "discourse segment boundaries", or at the "beginning of conversational turns".Generally, discourse structure is perceived like a hierarchical structure consisting of some levels, and each level might contain one or more instances of the preceding level.For instance, Stenstrom (1994), having adapted from Sinclair and Coulthard's model (1975), defines five hierarchically arranged levels of spoken discourse structure: transaction, exchange, turn, move, and act.(Ro se, 2008) The cognitive representation of each level has relev ant features like the discourse purpose of each level besides a general sequence mapping of the le vels it contains .When a n i ndividual s peaks, he/ she pl ans t he discourse se gments bef ore cr ossing a ny particular discourse boundaries an d re gardless of t he hierarchy l evels.F urthermore, pl anning any discourse segm ent encom passes planning all hierarchically lower segments it includes.Therefore, beginning a new turn seems to involve greater planning rather t han st arting a ne w act .As a re sult, higher di scourse boundaries which em erge i n the hi gher levels of the hierarchy are expected to show greater language planning.This prediction was first suggested by Swerts (1998) who m ade a comparative a nalysis on the em ergence of FPs at "strong " and "w eak" discourse boundaries.He d iscovered th at th e i nitial p hrase being followed by a strong d iscourse bo undary co uld m ore likely contain an initial filled pause rather than a m iddle filled pause or i n some cases no filled pause at all.In contrast, those phrases followed by a weak discourse boundary have the lo west probability of containing initial filled pause.

Communication Strategies in EFL Context
For most people, the main purpose of learning a foreign language is gaining the ability to communicate.Even nowadays, being a ble t o c ommunicate eff ectively i s more im portant th an being ab le to read and write in a foreign language.Thus, communication strategies have turned out to be crucial components of EFL material and course books these days.(Ya-ni, 2007) According to Richards and Rodgers (2001) langua ge tea ching m ethods like the Gram mar Translation Method (GTM) a nd t he Audio-Lingual Method (ALM) have l ost t heir popularity because of t heir ina dequacy in preparing students for real communication out of the classroom.(Baleghizadfeh, 2010) According t o Wenden a nd Rubin ( 1987), an E nglish-speaking e nvironment sho uld be creat ed t o the l argest extent so that students can be exposed to natural conversation quite frequently, hear more of target language and learn to produce new utterances in L2.In this way, they have the opportunity to practice and evaluate themselves to see how much they have learned and also get strongly motivated to learn and use the target language correctly.Although many schools might have incorporated English in th eir language course curriculum but the fr equency and the extent of training might not be adequate.(Ya-ni, 2007) The reason why Iranians who speak English as their foreign language encounter communication problems in L2 can be attributed to the environment and educational system governing schools and universities.The only way to learn English in Iran is through formal instruction in classrooms where the language teachers are native speakers of Farsi.It is l ess likely to find an opp ortunity to in teract with native speakers of English unless they might be found as few tourists (Karimnia, & Salehi Zadeh, 2007).Persian learn ers of En glish mo stly find it d ifficult to communicate efficiently in the target language because they had not been t rained i n a way t o develop t heir communicative competence.Rossiter (2003) carried out a study to find out the e ffects of teachi ng communication strategies on the second language performance.The research focused on two classes of adult immigrants in Canada of which one class was assigned as the control group and another one received 12 hours of direct communication strategy training.The stud ents sat fo r two oral task s con sisting of picture sto ry narratives, an d object descr iptions, which w ere conducted in Week 1, Wee k 5, and Week 10.The post-test results showed a direct effect of the c ommunication strategy training on the variety of strategies used in the object description task.Interestingly, the variety of the strategies used in this task was considerably higher than those employed by the respondents in the narrative part of the test.(Lam, 2006)

Participants
In order to investigate the pragmatic markers Iranian EFL learners use during hesitation, the researcher selected a population of TEP (Tertiary Eng lish Program) students in a public university in Kuala Lumpur, Malaysia, wh o had already taken IELTS exam and recei ved a sc ore of 5.5.Thus, the language competence of the populatio n was alm ost th e sam e, bu t th eir lan guage background was ch ecked for a higher cong ruency scale.Th e instrumentation w hich was used at t his st age was a LB Q questionnaire (l anguage ba ckground questionnaire), which helped the researcher to identify the most congruent participants in terms of their language background.The respondents were six males and six females whose first language was Persian, and these respondents had learned th eir L2 (In th is case, th e English langu age) i nitially at a s chool, and th en co ntinued i n langu age institutes or private classes in their home country.
Each participant h ad to tak e part in an oral L2 test consisting of 4 part s: Introduction, Conversation (General) Questions, R etelling a pas sage, a nd Pi cture descri ption.Each of t hese t asks t ook ab out 10-12 m inutes.T he collected dat a y ielded ab out 14 0 m inutes of English i nteraction between t he student as t he s ubject and t he researcher played the part of an interviewer.

Research Instrument
The research instruments which were used in this study included a Language Background Questionnaire to find the most congruent subjects regarding their language background, twelve unseen passages(never seen before by the respondents) to give them to read and retell after a limited time, twelve unseen pictures(never seen before by the res pondents) t o s how t hem random ly f or a desc ription based o n t heir im agination, a nd fi nally 3 set s of general questions ex tracted from th e assessm ent d atabase of a lang uage institu te.(with th e prior perm ission granted from the institute).

Data Collection
For t he data co llection ph ase, th e researcher recorded the sessi ons a nd the n t ranscribed t he recordings of t he interview sessio ns.Th ese reco rdings ca ptured all the pauses a nd e ven in coherent so unds th e respondents produced.T he reco rdings were chec ked f or seve ral t imes i n or der not t o ski p eve n a short si lent pa use.T he researcher th en i dentified the h esitation strateg ies used by th e participants an d th en co ded th em a s d rawling, pauses, repeating wo rds, u sing h esitation filler words and producing incoherent vocals.No t all p ragmatic markers lik e "I t hink" ind icated a hesitation strategy, so the researcher had to e nsure that the a nalyzed data actually functioned as hesitation in utterance.
Figure 1.Procedures employed in the study

Method
The session s with the resp ondents were d igitally recorded.Th e findings of t he pilot study con tributed to th e reliability of the results by showing that the participants needed to read the passages in a timely manner that was around 1 -2 m inutes t o get a gi st of w hat t hey ha d t o read.Also t he s ubjects o f t he passages we re varied t o prevent the passing and sharing of the general idea of the texts among the respondents.The topics centered on social, historical, and environmental issues.
Each session began with s ome explanations about t he format of the test by the researcher, and proceeded with conversation qu estions.Th is later con tinued with retelling a p assage as th e th ird part, an d fin ally d escribing a picture was the fourth part of the test.

Result
This section presen ts th e an alysis o f th e t ranscribed data ab out the d istribution p attern of h esitation strateg ies employed by Iranian university students during an oral L2 test.For this purpose, each hesitation marker group was studied individually to find out where they mostly occur in a sentence: Initial, Middle, or Final position.

Repeated Words
According to the summary data table.1, the frequency of Repeating Words in the middle position is considerably higher than the initial and final locations, which implies that the participants tended to repeat words mainly in the middle of the sentences (30 times).The dat a table also shows that the s mallest number of re peated words was produced in the final position (only once).

Hesitation strategy initial middle final
Repeated Word 8 30 1

Hesitation Filler Words
Referring to summary table.2, it can be inferred that the participants of this study tended to use the Hesitation Filler Words like "I think", "You know" in the middle of their sentences more than the other positions.However, the difference between the first and the second highest groups does not seem considerable.

Hesitation Fillers as Vocals
Based on the results obtained from the analysis on th e tran scribed data, the participants of the study produced incoherent sounds during hesitation mostly in the middle of their sentences.The initial position was identified as the second mostly recorded location of hesitation.The summary of the data is shown in table.5.

Analysis on Each Hesitation Filler Vocal
It can be i nferred from dat a t able.6;t hat al l hesi tation i ncoherent v ocals pr oduced by t he pa rticipants of t his study have been id entified mainly in th e mid dle po sition of th e utterances (35 3 ti mes), a nd t he se cond m ost frequent place recorded for the initial location (67 tim es), and finally the l east frequent position has been found for the final location of the utterances.Based on the findings illustrated in figure.2, the distribution pattern of hesitation discourse markers produced by the participants of the study is mainly found in the middle of the sentences(550 times), while the initial position is recorded as the second most frequent position of hesitation emergence with the frequency of 101 times.
On the other hand, it is shown that the least possible position of producing hesitation discourse markers is at the end of the utterances (20 times).

Conclusion
The overall aim of this paper was to investigate the distribution pattern of hesitation discourse markers produced by EFL l earners in an oral L2 test.The results of t his paper show clear indications of the dominant occurring position of the identified hesitation markers in the middle of the sentences.
Referring to Barr (2001), Beattie (1979) who claimed:"Hesitation mostly occurs at the beginning of a sentence or phrase, probably as a result of the greater demand on planning processes at these junctures", the findings of the present research showed a different result.Based on the analysis of the transcribed data, the highest frequency of hesitation marker (550 times) belongs to the mid -position, which is con siderably higher than the initial position with t he rate of 101 ti mes.The b ig difference b etween the frequencies of m iddle an d in itial positions can represent t hat the I ranian EF L res pondents do not st ruggle m uch with planning p rocess at t he beginning of a sentence.In ot her w ords t he res pondents did not at tempt m uch at t he beginning of t heir speech t o pl an t he coming w ords.They t hink a nd speak si multaneously l eading t o a c onsiderable i ncrease of hesi tation i n t he middle of t heir sentences.Moreover, th e aim of studying the final position of the utterances as anothe r location of emerging hesitation was t o i nvestigate whether th e resp ondents su ddenly end t heir talk due to hesitation i n finding appr opriate w ords or structure to con vey their m essage.For in stance, so me respo ndents first hesitated about the coming utterance, and as they could not decide what to say, they preferred to end the conversation right away.This was identified as the hesitation at the final position of a se ntence.This also proves that Iranian EFL learners cannot follow an efficient planning process in L2 which can be the result of not being instructed well to use English communicatively.
f th e find ings is pro vided in th e following tab les for each hesitation strateg ies b esides a figure depicting the statistics clearly.
Figure 2. A comparative analysis on the distribution of hesitation strategies

Table 2 .
Distribution of hesitation filler wordsAccording to the summarized data table.3, it is discovered that drawling the words almost always occurred in the middle o f t he sentences rat her t han t he t wo other l ocations.However, t his st rategy sh owed a di fferent res ult regarding th e final position.Th e frequ ency o f drawlin g fo r t he final position stands in t he second highest location of emergence in cont rast with t he previous di stribution patterns of hesi tation di scourse markers identified in this study.Based on the findings, the respondents tended to keep silent while hesitating what to say next quite frequently in the middle of the sentences, which tops over the two other locations.The difference between the m ost frequent location and the two other positions is quite considerable which is shown in table.4.

Table 5 .
Distribution of hesitation fillers as vocals

Table 6 .
Distribution of each hesitation filler vocal