Available online at www.sciencedirect.com

System 36 (2008) 279–294

www.elsevier.com/locate/system

SYSTEM

The effects of backchannels on fluency in L2 oraltask production

James P. Wolf *

Department of Tourism, Faculty of Regional Development Studies, Toyo University Itakura, Japan

Received 21 January 2007; received in revised form 3 July 2007; accepted 6 November 2007

Abstract

This article reports on an experimental study that investigated the effect of different conditions oflistener backchannels on the fluency of L2 speakers. Participants were 14 non-advanced Japaneselearners of English who each performed three oral tasks in three different backchannel conditions:(1) verbal/nonverbal (V/NV), (2) nonverbal-only (NV), and (3) no backchannels (NB). Verbal back-channels included ‘mm-hm’ and ‘uh-huh’ while the nonverbal backchannels involved head nodding.Fluency was assessed via five temporal measures. As hypothesized, the results showed that the 14Japanese participants were, on average, most fluent in the V/NV condition, less fluent in the NV con-dition and least fluent in the NB condition. The differences obtained in fluency between the V/NVand NB conditions were found to be significant. These results lend support to the ‘backchannel out-put hypothesis’ which suggests that backchannels may facilitate the fluency of non-advanced learnersof English during oral tasks depending on the nature of backchannel use in their L1 and sociocul-tural environments.� 2008 Elsevier Ltd. All rights reserved.

Keywords: Second language learning; Backchannels; Language teaching; Second language fluency

1. Introduction

Second language (L2) fluency is a complex, multifaceted construct that has resistedclear-cut definitions as noted by a number of researchers (e.g. Chambers, 1997; Lennon,

0346-251X/$ - see front matter � 2008 Elsevier Ltd. All rights reserved.

doi:10.1016/j.system.2007.11.007

* Home address: 2-9-11 Shinkoiwa, Katsushika-Ku, Tokyo, Japan. Tel./fax: +81 03 5662 7673.E-mail address: wolves626@aqua.plala.or.jp

280 J.P. Wolf / System 36 (2008) 279–294

1990, 2000; Wood, 2001). For the purposes of this paper, L2 fluency will be defined as anautomatic procedural skill (Schmidt, 1992) and thus a temporal phenomenon. In recentyears, several researchers have attempted to elucidate factors that influence L2 speakers’fluency during oral tasks. Some have investigated the effects of task characteristics them-selves, such as task types (Skehan and Foster, 1999; Derwing et al., 2004). Others haveexamined how task conditions, such as preplanning and online planning, affect fluency(e.g. Crookes, 1989; Foster and Skehan, 1996; Mehnert, 1998; Ortega, 1999; Wiggles-worth, 1997; Yuan and Ellis, 2003). Aside from tasks, some researchers have investigatedthe effects of L2 speaker performance and speech variables, such as self-monitoring (Kor-mos, 1999) and intonation (Wennerstrom, 2000), while still others have studied how timespent in an L2 learning context may affect fluency (Freed, 1995; Freed et al., 2004; Lennon,1990). In sum, fluency has been investigated with respect to three different variables: (1)task characteristics, (2) L2 speaker characteristics, and (3) learning context.

It appears then that at least one important variable that may affect L2 speakers’ fluencyduring oral tasks has eluded consideration; that is, the listener behaviors of their interloc-utors. When L2 speakers perform oral tasks, teachers or testers are often present andrespond to their production with a variety of verbal and nonverbal messages. Verbal mes-sages might include, for example, ‘uh-huh’, ‘mm-hm’, and ‘yeah’, while nonverbal mes-sages might include head nods or smiles. In the literature, such messages have beenvariously described as ‘signals of attention’ (Fries, 1952), ‘accompaniment signals’ (Ken-don, 1967), ‘listener responses’ (Dittmann and Llewellyn, 1968) and ‘backchannels’(Yngve, 1970), among several other descriptors. In this paper, I shall refer to them inter-changeably as backchannels or backchannel cues. What effects then might backchannelcues have on L2 speakers’ fluency during oral tasks? Do they facilitate fluency? If so,are they equally facilitative for all L2 speakers regardless of L1 and cultural background?

Initially, most backchannel research was conducted by American linguists who soughtto identify the types and functions of backchannels in American English (e.g. Duncan,1974; Duncan and Fiske, 1977; Fries, 1952; Goodwin, 1986; Jefferson, 1984; Schegloff,1982; Yngve, 1970). More recently, much research has been done to compare and contrastthe use of backchannels across cultural and linguistic groups (e.g. Clancy et al., 1996; Cut-rone, 2005; Heinz, 2003; LoCastro, 1987, 1999; Maynard, 1986, 1990, 1997; Tao andThompson, 1991; White, 1989, 1997). This line of research has revealed differencesbetween these groups concerning the frequency, placement, functions, and types of back-channels. Due to the nature of this research, contrastive conversational analysis has tradi-tionally been the research methodology of choice. As a result, no L2 experimentalresearch, to my knowledge, has been done to investigate the effects of backchannels on flu-ency during oral tasks.

Experimental research has been done, however, in the psychotherapy field in an EnglishL1 context to assess how interviewer backchannel cues affect interviewees’ verbal produc-tivity. These studies aimed to show that patients could be induced to speak more duringpsychotherapy by responding to them with backchannel cues. Some studies (Kanfer andMcBrearty, 1962; Matarazzo et al., 1964a, 1964b) found verbal and nonverbal backchan-nels to facilitate verbal productivity while others did not (Siegman, 1976). At any rate,these results cannot be used to infer how L2 speakers might react to backchannel cuessince they involved L1 speakers in interviews rather than L2 speakers in tasks. Under-standing how backchannels might affect L2 learners’ fluency is especially important forteachers who interact and negotiate meaning with them in the target language, as well

J.P. Wolf / System 36 (2008) 279–294 281

as for language testers who evaluate their oral production. Hence, this experimental studytakes place in a Japanese setting and aims to determine how English backchannel cuesinfluence Japanese EFL learners’ fluency during oral tasks.1

2. Literature review

2.1. Backchannel research

Fries (1952) was perhaps the first to lump backchannel cues together based on his anal-ysis of telephone conversations. He viewed them as ‘signals of attention’ that did not inter-rupt the speaker’s talk. Yngve (1970) coined the term ‘backchannels’ to describe them andalso considered them to support the primary speaker’s turn.

1 No

When two people are engaged in conversation, they generally take turns. . .. In factboth the person who has the turn and his partner are simultaneously engaged in bothspeaking and listening. This is because of the existence of what I call the back-channel, over which the person who has the turn receives short messages such asyes and uh-huh without relinquishing the turn (Yngve, p. 568).

In addition to short messages that indicate interest and attention, Yngve also regardedshort questions as backchannels. Following this, Duncan (1974) and Duncan and Fiske(1977) expanded the scope of backchannels by incorporating (1) requests for clarification,(2) sentence completions, (3) brief restatements, and (4) nonverbal messages such as headnods. Subsequently, Brunner (1979) claimed that smiles can also function as backchannels.

From here, however, a major shift in perspective occurred owing to Schegloff (1982)cogent argument that, rather than lumping backchannels together into discrete categories,they must instead be analyzed in view of their interactive functions within discourse.Whereas Fries (1952) and Yngve (1970) viewed backchannels as having a supportive func-tion in that the non-primary speaker uses them to signal agreement and understanding tothe primary speaker, Schegloff (1982) considered them to be ‘continuers’, meaning thatthey serve to pass an opportunity to produce a full turn, and thus viewed them as havinga regulative function. Schegloff’s work led to a spate of research aimed at identifying otherinteractive backchannel functions (e.g. Goodwin, 1986; Goodwin and Goodwin, 1987; Jef-ferson, 1984; Drummond and Hopper, 1993a, 1993b).

2.2. Comparing backchannel use between Japanese and American English speakers

According to Heinz (2003), backchannels are a universal human behavior since all cul-tural and linguistic groups use some form of backchannel response, although specificbackchannel behaviors are particular to language and culture. How does backchanneluse differ between Japanese and American English speakers? To address this question, Iwill compare the (1) frequency, (2) placement, (3) functions, and (4) types of backchannelsused between these two cultural and linguistic groups.

Several researchers have noted that Japanese speakers make much greater use of back-channels compared to American English speakers (e.g. Clancy et al., 1996; LoCastro,1987, 1999; Maynard, 1990, 1997; White, 1989). The Japanese participants in White’s

te that the influence of American English backchannels in particular is examined.

282 J.P. Wolf / System 36 (2008) 279–294

(1989) study used one backchannel for every 14 words while speaking English with Amer-icans who used one for every 37 words. White observed that this 3:1 ratio holds when Jap-anese and Americans speak with fellow native speakers in their respective languages. Whydo Japanese speakers use backchannel cues so pervasively? Several researchers have citedJapanese cultural influences, linking their relatively frequent use to social values such asconsideration of others, cooperation, and the desire for harmonious communication(e.g. LoCastro, 1987, 1999; Maynard, 1997; White, 1989). Some have taken a linguisticperspective, suggesting that discourse structures inherent in Japanese naturally elicit back-channel responses (Maynard, 1997; Miller, 1988; White, 1989). In contrast, LoCastro(1999) proposed a functional perspective, claiming that the language code has evolveddue to cognitive and sociocultural pressures to allow more frequent backchannel cues.

Secondly, Japanese and American English speakers also differ with respect to wherethey place backchannels. American English speakers consistently tend to provide back-channels at points of grammatical completion and therefore do not usually overlap theirinterlocutors’ talk with them (Clancy et al., 1996; LoCastro, 1987; Maynard, 1990, 1997).Conversely, the discourse contexts of Japanese speakers’ backchannels are varied. May-nard (1990) reported that her Japanese participants’ backchannels occurred at grammat-ical junctures, during unfilled pauses, and were also prompted by cues from the primaryspeaker such as sentence final particles and head nods. Moreover, Japanese speakers aremore likely to send backchannels during the primary speaker’s turn thus producing simul-taneous talk (e.g. Clancy et al., 1996; LoCastro, 1987; Maynard, 1990, 1997).

Thirdly, functional differences also exist between Japanese and American Englishspeakers in their backchannel use. These differences, however, are not as stark as thoseconcerning the frequency and placement of backchannel cues already discussed. In fact,based on her analysis, Maynard (1997, p. 46) reported that both Japanese and AmericanEnglish speakers use backchannels for the same purposes: (1) continuer (Schegloff, 1982);(2) display of understanding of content; (3) support toward the speaker’s judgment; (4)agreement; (5) strong emotional response; (6) minor addition, correction, or request forinformation. While these functions can overlap and are not mutually exclusive, Maynardobserved that Japanese favor the display of understanding of content function to expressmoral support for the speaker, while Americans favor the ‘continuer’ function (i.e., tosend a backchannel to pass up an opportunity for a speaking turn at a grammaticaljuncture).

Lastly, somewhat more similarity also exists regarding the types of backchannels used.Maynard’s (1990) contrastive analysis of 40 (20 American and 20 Japanese) dyadic conver-sations found brief utterances (i.e., ‘uh-huh’, ‘mm-hm’, etc.) the most frequent type usedby both Japanese and American English speakers (70.49% and 50.23%, respectively) andhead movement (i.e., head nods) the second most frequent (18.83% and 35.05%, respec-tively). Concerning brief utterances, however, LoCastro (1987, 1999) noted that thereappears to be a greater variety of these in Japanese than in American English. Overall,Maynard (1990) declared the biggest differences between Japanese and American Englishbackchannels to be in their frequency and discourse contexts, as discussed earlier.

2.3. Second language (L2) fluency

Defining L2 fluency has been a difficult task for researchers due to conflicting meaningsassociated with the term. In an effort to disambiguate the meaning of fluency in foreign

J.P. Wolf / System 36 (2008) 279–294 283

language learning, Lennon (1990) distinguished between a ‘‘broad” sense and a ‘‘narrow”

sense of fluency. According to the broad sense, he noted that fluency serves as a cover termfor oral proficiency, representing ‘‘the highest point on a scale that measures spoken com-mand of a foreign language” (p. 389). On the other hand, Lennon observed that, in its nar-row sense, fluency in EFL pertains to one, isolatable component of oral proficiencydescribing learners who are ‘‘fluent but grammatically inaccurate” or ‘‘fluent but [lack]a wide and varied vocabulary.” Furthermore, Lennon (1990) mentioned that the overallemphasis on fluency in the narrow sense seems to be on native-like rapidity.

A focus on native-like rapidity dovetails well with Schmidt’s (1992) conception of L2fluency. Schmidt defines second language fluency as a ‘‘procedural skill,” and states thatan extreme example of such a fluent speaker ‘‘would be the speaker of a pidginized inter-language, largely unanalyzed and agrammatic with respect to the target language, whospeaks that variety in a fluid rather than a halting manner” (p. 358). Researchers sharingsuch a view of L2 fluency operationalize and measure it in terms of various temporal anddysfluency markers such as speech rate, the length of pauses, the length of speech runsbetween pauses, and the number of repetitions and reformulations (e.g. Derwing et al.,2004; Freed, 1995, 2000; Kormos and Denes, 2004; Lennon, 1990; Riggenbach, 1991;Towell et al., 1996). While some researchers in recent years have studied how (1) task char-acteristics, (2) L2 speaker characteristics, and (3) learning context affect fluency as men-tioned earlier, others have also attempted to understand which linguistic and temporalspeech variables tend to influence listeners’ perceptions of fluency. One variable in partic-ular, speech rate, has emerged as a core component of fluency consistently cited acrossstudies as affecting perceptions of fluency (e.g. Ejzenberg, 2000; Freed, 1995, 2000; Len-non, 1990; Kormos and Denes, 2004; Riggenbach, 1991).

3. Theoretical perspective

3.1. Backchannel findings informing the perspective

Tao and Thompson (1991) studied two native Mandarin speakers for whom AmericanEnglish had become their dominant language after living in the United States for manyyears. They were found to transfer their L2 American English backchannel behavior totheir L1 Mandarin conversations. Likewise, the native German speakers studied by Heinz(2003) were found to transfer their L2 American English backchannel behavior to their L1German conversations after becoming highly proficient in American English. These stud-ies show that, as English learners attain advanced levels of oral proficiency, they graduallydevelop implicit knowledge of English backchannel cues.

On the other hand, non-advanced English learners seem likely to transfer their L1 back-channel behavior to their L2 conversations. Both White (1989) and Maynard (1997) foundthe Japanese L2 learners of English in their respective studies to transfer their L1 back-channel behavior to their L2 English conversations. Maynard concluded that Japaneselearners of English use backchannels in intercultural discourse as they do within theirown cultural context. Rather than being restricted to Japanese learners of English, it ismy view that non-advanced learners of English in general, due to their lack of implicitknowledge of English backchannel behavior, will tend to have an implicit preference forproducing and receiving backchannels in a manner which reflects the backchannel conven-tions of their respective L1 and sociocultural environments.

284 J.P. Wolf / System 36 (2008) 279–294

3.2. The backchannel output hypothesis

Intimately intertwined with an individual’s implicit preference for how an interlocutorshould respond to their talk with verbal and nonverbal backchannel cues is the value thatis attached to such listener behaviors. It seems likely that non-advanced English learnersfrom different linguistic and sociocultural backgrounds will react to English backchannelcues in ways that reflect different appraisals of their value. Therefore, I would like to pro-pose the backchannel output hypothesis, which suggests that non-advanced learners ofEnglish from various L1 and sociocultural contexts will tend to evaluate verbal and non-verbal English backchannel behavior differently based on their implicit preferences forreceiving backchannels, and these evaluations will determine whether or not English back-channel cues facilitate their fluency during oral tasks.

4. Research questions and hypotheses

How then might the fluency of non-advanced Japanese EFL learners be influenced bybackchannel cues during oral tasks? To address this broad question, three specific researchquestions and hypotheses will be posed.

(1) How is non-advanced Japanese EFL learners’ fluency affected by verbal and nonver-bal backchannels (head nods) during an oral narrative task? Although AmericanEnglish backchannel behavior differs from that of native Japanese speakers, it ishypothesized that the Japanese participants will be most fluent in this conditiondue to the high value placed on verbal and nonverbal backchannel cues in Japanese.

(2) How is non-advanced Japanese EFL learners’ fluency affected by only nonverbalbackchannels (head nods) during an oral narrative task? Maynard (1990) found headmovement which did not accompany verbal backchannel cues somewhat infrequentfor both her American and Japanese participants (35.05% and 18.83%, respectively).Therefore, it is hypothesized that the Japanese participants in this study will be lessfluent in this condition compared to the verbal and nonverbal backchannelcondition.

(3) How is non-advanced Japanese EFL learners’ fluency affected by the absence of ver-bal and nonverbal backchannels during an oral narrative task? Again, due to thevalue attached to backchannels in Japanese, it is hypothesized that the Japanese par-ticipants will be least fluent in this condition.

5. Method

5.1. Design

This study is a single-factor within-participants design with three levels of backchannelconditions (verbal/nonverbal, nonverbal-only, and no backchannels). Participants with alower to intermediate level of oral proficiency were sought for the study. Fourteen volun-teered to participate. They were introduced to the researcher by colleagues and acquain-tances. All participants were told the purpose of the study was to examine their oraltask production; however, no information was given about the independent variable,

J.P. Wolf / System 36 (2008) 279–294 285

backchannels. Consent forms for participation in the research were signed by all partici-pants. Three oral narrative tasks were performed by each participant and were elicitedby means of a series of three picture stories in each task condition.

5.2. Participants

Fourteen Japanese EFL speakers participated in the study. Table 1 shows there were 3males and 11 females ranging in age from 19 to 55. Despite age differences, participantswere judged fairly homogenous in terms of both their experiences using English and theiroral proficiency. Most had been speaking English 5–10 years. Only four participants hadlived in an English-speaking country for a limited period of 5–7 months. Two steps weretaken to ensure the participants were sufficiently similar in oral proficiency. Firstly, in gen-eral terms, potential participants were asked before the study not to volunteer if theyjudged themselves to be either complete beginners or advanced speakers of English. Sec-ondly, the researcher, an American male who has taught English in Japan for over10 years, interviewed each participant before the study to elicit biographical data andinformation regarding their language learning experience. On this basis, the researcherthen selected the best descriptor of their communicative ability by referring to the ACTFLguidelines (1986). All participants were judged to range in proficiency from intermediate-mid to intermediate-high, meaning they could function competently in many routine com-municative tasks and social situations.

5.3. Task materials

Participants were required to orally narrate a series of three picture stories during eachof the three tasks. Thus, series 1 (S1), series 2 (S2), and series 3 (S3) each contained threesuch stories. Eight were from Papajohn (1998) and one was developed for the study. Eachhad a sequence of six pictures and typically involved a problem; for example, one involveda man who rented a bicycle, but had to return it due to a flat tire. As shown in Table 2, S1,S2, and S3 were restrictively randomized across the tasks for each participant.

Table 1Participant information

Gender Age L1 Time speaking English Time living in English-speaking country

Speaker 1 Female 19 yrs Japanese 6–7 yrs NoneSpeaker 2 Female 19 yrs Japanese 6–7 yrs NoneSpeaker 3 Male Late 20s Japanese 4–5 yrs NoneSpeaker 4 Female Late 20s Japanese 5–6 yrs NoneSpeaker 5 Female Late 20s Japanese 7–8 yrs NoneSpeaker 6 Male Early 30s Japanese 7–8 yrs NoneSpeaker 7 Female Early 30s Japanese 4–5 yrs 5 months (Singapore)Speaker 8 Female Late 30s Japanese 4–5 yrs 4 months (USA)Speaker 9 Female Late 30s Japanese 4–5 yrs 6 months (USA)Speaker 10 Female Late 30s Japanese 5–6 yrs NoneSpeaker 11 Female Late 30s Japanese 10–11 yrs NoneSpeaker 12 Female Early 40s Japanese 3–4 yrs 7 months (Canada)Speaker 13 Female Early 50s Japanese 10–11 yrs NoneSpeaker 14 Male Mid 50s Japanese 4–5 yrs None

Table 2Restrictive randomization of the three series of picture stories

Participant Task 1 Task 2 Task 3

1 S3 S2 S12 S1 S3 S23 S2 S1 S34 S2 S1 S35 S3 S2 S16 S1 S3 S27 S1 S3 S28 S2 S1 S39 S3 S2 S110 S3 S2 S111 S1 S3 S212 S2 S1 S313 S2 S1 S314 S3 S2 S1

S1 = picture series 1, S2 = picture series 2, S3 = picture series 3.

286 J.P. Wolf / System 36 (2008) 279–294

5.4. Task procedures

Participants were allotted four-minutes to plan before each task.2 To finish each task, afour-minute time limit was given during which participants had to, if possible, narrate allthree picture stories. During each task, a sheet with the three stories was held up by theresearcher, while sitting across from the participants, so that any head nodding wouldbe noticed. Backchannels were given in response to the participants’ production by theresearcher according to the task conditions. The time spent on task was recorded by theresearcher after each performance. All three tasks were performed consecutively in indi-vidual sessions with the researcher and audio-recorded for later transcription. Each sessionlasted 25–30 min. Instructions were given in Japanese before starting by the researcher,who has advanced Japanese oral proficiency certification.

5.5. Task conditions: independent variable

Backchannels were operationalized at three levels: (a) verbal/nonverbal (V/NV), (b)nonverbal-only (NV), and (c) no backchannels (NB). To ensure internal validity, thesethree conditions were restrictively randomized in a counterbalanced design in which exper-imental control was achieved by entering all participants into all treatments as shown inTable 3.3

2 A reviewer suggested four minutes for planning was perhaps too long, thus possibly reducing the effect ofbackchannels since participants would not need to produce language spontaneously. It was felt, however, that thefour-minute time limit for task performance would help to offset the planning time benefits since limited timewould be available for on-line planning. Still, the interaction of planning time and backchannel effects is aninteresting topic for future research.

3 See Campbell and Stanley (1963, p. 50) for a discussion of counterbalanced research designs.

Table 3Counterbalanced design for the three backchannel conditions

Participant Task 1 Task 2 Task 3

1 NB (S3) NV (S2) V/NV (S1)2 NV (S1) V/NV (S3) NB (S2)3 V/NV (S2) NB (S1) NV (S3)4 NB (S2) NV (S1) V/NV (S3)5 NV (S3) V/NV (S2) NB (S1)6 V/NV (S1) NB (S3) NV (S2)7 NB (S1) NV (S3) V/NV (S2)8 NV (S2) V/NV (S1) NB (S3)9 V/NV (S3) NB (S2) NV (S1)10 NB (S3) NV (S2) V/NV (S1)11 NV (S1) V/NV (S3) NB (S2)12 V/NV(S2) NB (S1) NV (S3)13 NB (S2) NV (S1) V/NV (S3)14 NV (S3) V/NV (S2) NB (S1)

Note: V/NV = verbal/nonverbal, NV = nonverbal-only, and NB = no backchannels.S1 = picture series 1, S2 = picture series 2, S3 = picture series 3.

J.P. Wolf / System 36 (2008) 279–294 287

1. Verbal/nonverbal (V/NV): Both verbal and nonverbal backchannels were given. Theverbal backchannels included ‘mm-hm’ and ‘uh-huh’ and the nonverbal backchannelsinvolved head nodding. No attempt was made to control how the backchannels weregiven; rather, it was desired they be given in a manner that felt as natural as possibleto the researcher. In general, however, backchannels were given at points of grammat-ical completion without overlapping participants’ speech, which is the norm for Amer-ican English speakers.

2. Nonverbal-only (NV): Only nonverbal backchannels (head nodding) were given. Againno attempt was made to control how they were given.

3. No backchannels (NB): Neither verbal nor nonverbal backchannels were given. Whileholding up a sheet with a series of picture stories, the researcher unobtrusively directedhis gaze toward the sheet, rather than the participants.

5.6. Fluency measures: dependent variables

Fluency was measured in terms of (a) the number of syllables per minute (Rate A),(b) the number of meaningful syllables per minute (Rate B), (c) the mean length of run

(MLR), (d) the mean pause time (MPT), and (e) the total pause time (TPT). In addition tofluency, a separate measure was used to assess whether the length of time on task wasinfluenced by the backchannel conditions although no a priori hypothesis was establishedfor this measure.

1. Rate A (the number of syllables per minute): the number of syllables within each nar-rative, divided by the number of seconds used to finish the task and multiplied by 60.

2. Rate B (the number of meaningful syllables per minute): same as Rate A, but excludingall syllables, words, phrases that were repeated, reformulated, or replaced.

3. MLR (mean number of syllables between all pauses): obtained by counting the totalnumber of syllables between all filled (e.g. hm, um, and uh) and unfilled (i.e. silent)pauses and calculating the mean number of syllables.

Table 4Time on task according to the backchannel conditions

Backchannel Conditions F-value Significance

V/NV NV NB

M SD M SD M SD

Length of time (s) 184.43 49.60 200.64 53.79 204.07 48.48 2.739 .083

Note: V/NV = verbal/nonverbal, NV = nonverbal-only, and NB = no backchannels.

288 J.P. Wolf / System 36 (2008) 279–294

4. MPT (mean number of seconds for all filled and unfilled pauses): derived by countingthe total number of seconds for all filled and unfilled pauses and calculating the meannumber of pause seconds.

5. TPT (total pause time [filled and unfilled]): expressed as a percentage of the total timeon task.

5.6.1. Length of time on task: backchannel conditions

1. Time on task: the total number of seconds on task was counted for participants in eachbackchannel condition.

6. Results

6.1. Inter-rater reliability and data analysis

After transcription, two raters worked independently to analyze a third of the data.Inter-rater reliability was over 89% on all five measures.4 Pause time and temporal vari-ables were measured manually. Following Mehnert (1998), pauses of 1 second or longerwere identified, the length of each pause was measured twice via a stop watch, the meanwas recorded, and the location and length of each pause was noted in the transcripts. Aseries of one-way within-subjects ANOVAs were performed on all the measures followedby post hoc t-tests. The alpha for achieving statistical significance was set at .05 for theANOVAs and at .017 for the post hoc tests.

6.2. Time on task according to backchannel conditions

The means for the length of time taken for task completion in each condition are shownin Table 4. They reveal that the time spent for task completion in the V/NV condition(mean = 184.43 s) was noticeably less than in both the NV (mean = 200.64 s) and NB(mean = 204.07 s) conditions. In contrast, the time spent for task completion in the NVand NB conditions was quite similar. However, a one-way within-subjects ANOVA shows

4 One of the three audio transcripts for each participant was randomly selected for the inter-rater reliabilityanalysis. Thus, this involved 33% of the data (14 of the 42 transcripts), which exceeds the 10% criterionrecommended by Brown (2001, p. 238–239)) for inter-rater reliability analyses.

Table 5Fluency results for the 14 Japanese participants in the backchannel conditions

Conditions F-value Significance Locations of significance

V/NV NV NB V/NV-NV V/NV-NB NV-NB

M SD M SD M SD

Rate A 76.46 18.74 70.03 18.09 67.39 17.44 4.387 .023* .021 .008** .217Rate B 65.73 18.33 59.34 15.93 56.76 16.07 3.982 .031* .026 .013** .223MLR 5.53 1.76 5.12 1.58 5.00 1.02 1.710 .201 .079 .066 .338MPT 1.98 0.44 2.19 0.47 2.42 0.63 5.199 .013* .061 .010** .031TPT 46.96 11.07 50.61 10.21 53.13 10.36 3.023 .066 .061 .030 .149

Comparisons for locations of significance were tested at the .017 level and the p-values derived from a one-tailedtest.* p < .05.** p < .017.

J.P. Wolf / System 36 (2008) 279–294 289

that the difference in time for task completion across the conditions is not statistically sig-nificant (p = .083).

6.3. Fluency results

Hypothesis 1 predicted that participants would be most fluent in the V/NV condition.Table 5 shows the overall scores support this prediction for all five measures. Of these, sig-nificant differences were obtained for three measures: Rate A (p = .023); Rate B (p = .031);and the Mean Pause Time (p = .013).5 Post hoc t-tests show the differences between the V/NV and NB conditions to be statistically significant for Rate A (p < .008), for Rate B(p < .013), and for the Mean Pause Time (p < .010). Thus, the participants uttered a largernumber of overall syllables (Rate A) and meaningful syllables (Rate B) per minute in theV/NV compared to the NB condition, while using less time per pause (Mean Pause Time).Hypothesis 2, which predicted less fluent results in the NV condition compared to the V/NV condition, was also supported by the raw scores for all measures, but no statisticallysignificant differences were found between the V/NV and NV conditions. Hypothesis 3,which predicted the least fluent results in the NB condition, was also supported by theraw scores for all measures. However, despite the significant differences between the V/NV and NB conditions for the Rate A, Rate B and Mean Pause Time measures, no sig-nificant differences were found between the NV and NB conditions.

6.4. Practice effects for fluency

To investigate whether practice effects accounted for differences in fluency, a one-waywithin-subjects ANOVA was performed to assess how participants’ fluency developedsequentially across the three tasks irrespective of backchannel conditions. The means

5 Table 5 shows the Mean Length of Run measure furthest from achieving significance (p = .201). Kormos andDenes (2004) noted that several studies have found the MLR a good predictor of fluency. However, most studieshave tended to investigate fluency longitudinally (e.g. Lennon, 1990) or by comparing fluent and non-fluentspeakers (e.g. Riggenbach, 1991). Thus, less variation in this measure should perhaps be expected whenparticipants act as their own controls in a within-participants design.

Table 6Results for possible practice effects for fluency

Task sequence F-value Significance

Task 1 Task 2 Task 3

M SD M SD M SD

Rate A 70.55 16.78 70.77 22.74 71.03 16.41 0.009 .991Rate B 59.53 16.52 60.15 20.44 60.44 15.84 0.031 .970MLR 5.13 1.27 5.37 1.88 5.12 1.28 0.401 .674MPT 2.24 0.59 2.27 0.69 2.19 0.48 0.128 .881TPT 50.98 10.32 50.09 13.70 50.42 8.36 0.049 .952

Table 7Factor analysis results for the five temporal measures

Fluency measures Factor Communality

Speech rate A .97 .94Speech rate B .95 .90Mean length of run .88 .78Mean pause time �.85 .72Total pause time �.92 .85

290 J.P. Wolf / System 36 (2008) 279–294

for the fluency measures according to task sequence are shown in Table 6. They suggestextremely limited, if any, practice effects as participants’ fluency changed little over thethree tasks and statistical significance for the effect of task sequence on fluency was notattained for any measure.

6.5. Fluency construct

To verify that the five temporal measures used assessed the same construct, the data forall 14 participants were subjected to a factor analysis. Table 7 shows high loadings on onefactor indicating the existence of a general fluency factor, affecting all five temporalmeasures.

7. Discussion

7.1. Interpreting fluency vis-a-vis the backchannel conditions

The participants were most fluent in the V/NV condition as predicted by Hypothesis 1.The rationale for Hypothesis 1 was that non-advanced Japanese EFL learners would bemost fluent when receiving verbal and nonverbal English backchannel cues due to the highvalue attached to such listener behaviors in Japanese. Thus, the result obtained in the V/NV condition supports this line of reasoning.

Furthermore, the overall scores show the participants were less fluent in the NV condi-tion compared to the V/NV condition as predicted by Hypothesis 2. The rationale forHypothesis 2 was that non-advanced Japanese EFL learners would be less fluent whenreceiving only nonverbal backchannels (head nods) since Maynard (1990) found that head

J.P. Wolf / System 36 (2008) 279–294 291

movement without verbal backchannel cues occurred infrequently in her Japanese data.Table 5 showed that the differences obtained between the V/NV and NV conditions forthe Rate A (p = .021) and Rate B (p = .026) measures just missed achieving significancealthough the sample size of the present study was somewhat limited. Therefore, ratherthan indicating equality between the V/NV and NV conditions, the non-significant resultlikely reflects a lack of statistical power due to the number of participants. Moreover, theless fluent result obtained in the NV condition makes sense in light of Maynard (1990)findings concerning the relative infrequency of listener head movement without verbalbackchannels in Japanese.

Lastly, the overall scores show participants were least fluent in the NB condition as pre-dicted by Hypothesis 3. The rationale for Hypothesis 3 was that non-advanced JapaneseEFL learners would be least fluent when receiving no backchannels at all due to the valueplaced on backchannel behavior in Japanese. Despite the significant fluency differencesfound between the V/NV and NB conditions, no significant differences were detectedbetween the NV and NB conditions. However, since listener head movement unaccompa-nied by verbal backchannel cues occurs somewhat infrequently in Japanese, the Japaneseparticipants may have perceived little difference between receiving only head nods in theNV condition and not receiving any backchannels at all in the NB condition. This wouldthus help account for the similar results obtained in the NV and NB conditions.

7.2. Preferences for backchannels and somatic value

The major factor influencing whether backchannels can facilitate non-advanced L2speakers’ fluency in oral tasks was suggested to be the nature of backchannel use in theirL1 and sociocultural environments from which they develop a preference. A better under-standing of the term ‘preference’ can be gained from Schumann’s (1997, Schumann andWood, 2004) notion of somatic value. Schumann and Wood (2004) explain that somaticvalue ‘‘involves preferences and aversions acquired in the lifetime of the individualthrough experience, socialization, enculturation, and education” (p. 24). Therefore, pref-erences for producing and receiving backchannels can be viewed as a form of somaticvalue. Schumann (1997) further explains that individuals evaluate stimuli along severalpsychological dimensions, identified by Scherer (1984), such as ‘pleasantness’ and ‘rele-vance to their goals and needs’, to determine their emotional and motivational signifi-cance. From this perspective, backchannels might be considered emotional stimuli thatL2 speakers evaluate via their somatic value systems and that may, depending on theirpreferences, facilitate their fluency.

8. Possible Implications

This study provides some support for the backchannel output hypothesis, which sug-gests that non-advanced learners of English will tend to evaluate verbal and nonverbalEnglish backchannel behavior based on their implicit preferences for receiving backchan-nels, and these evaluations will, in turn, determine whether or not English backchannelcues facilitate their fluency during oral tasks. For non-advanced Japanese EFL learners,the present study suggests that verbal and nonverbal English backchannel cues can facil-itate their fluency during tasks. However, non-advanced learners of English from other L1and sociocultural contexts, who possess different implicit preferences for receiving back-

292 J.P. Wolf / System 36 (2008) 279–294

channels, may not show improved fluency in response to English backchannel cues. Forexample, Tao and Thompson (1991) noted that Mandarin speakers use backchannelsmuch less frequently than American English speakers do, while Clancy et al. (1996)reported that Mandarin speakers avoid sending backchannel cues during their interlocu-tors’ speaking turn out of respect for their right to produce their talk undisturbed. There-fore, one might theorize that non-advanced Chinese learners of English will not showimproved fluency during tasks in response to English backchannel cues.

What are the possible implications for language teachers and testers? This study sug-gests that teachers’ verbal and nonverbal listening behaviors may be interpreted differentlyby non-advanced learners of English from various L1 and sociocultural backgrounds.Non-advanced Japanese learners of English appear likely to value and show enhanced flu-ency in response to their teachers’ backchannel behavior. Conversely, other non-advancedlearners of English may assign less value to English backchannel cues and therefore be ableto maintain the same level of fluency regardless of the types or amounts of backchannelsreceived from their teachers. Furthermore, in oral testing situations, those who attach highvalue to receiving backchannels, such as non-advanced Japanese learners of English, maydisplay greater or lesser fluency depending on their availability.

9. Limitations and future research

One limitation of the present study is the sample size, which included 14 participants.Therefore, it would be useful to replicate this study with a larger sample of non-advancedJapanese EFL learners in the future. For future research, it would also be worth examininghow participants react to backchannels when given less time for pre-task planning.

Additionally, although production was elicited via oral narrative tasks in the presentstudy, future studies might examine how backchannels influence fluency during moreopen-ended tasks or perhaps even interviews. Lastly, to test the backchannel outputhypothesis, it is necessary to investigate the fluency of non-advanced learners of Englishfrom different L1 and sociocultural groups in response to different types and amountsof English backchannel cues during tasks. Through such research endeavors, the meritsof the backchannel output hypothesis can be more fully understood.

Acknowledgements

I would like to thank Dr. Mark Sawyer and two anonymous reviewers for their helpfulcomments on earlier drafts of this paper.

References

ACTFL oral proficiency guidelines by the American Council for the Teaching of Foreign Languages, 1986.

Available from <http://www.sil.org/lingualinks/LANGUAGELEARNING/OtherResources/ACTFLProfi-

ciencyGuidelines/TheACTFLGuidelines.htm> (retrieved 31.08.06).

Brown, J.D., 2001. Using Surveys in Language Programs. Cambridge University Press, Cambridge.

Brunner, L., 1979. Smiles can be back channels. Journal of Personality and Social Psychology 37 (5), 728–734.

Campbell, D.T., Stanley, J.C., 1963. Experimental and Quasi-Experimental Designs for Research. Houghton

Mifflin Company, Boston.

Chambers, F., 1997. What do we mean by fluency? System 25 (4), 535–544.

Clancy, P., Thompson, S., Suzuki, R., Tao, H., 1996. The conversational use of reactive tokens in English,

Japanese, and Mandarin. Journal of Pragmatics 26, 355–387.

J.P. Wolf / System 36 (2008) 279–294 293

Crookes, G., 1989. Planning and interlanguage variation. Studies in Second Language Acquisition 11, 367–383.

Cutrone, P., 2005. A case study examining backchannels in conversations between Japanese–British dyads.

Multilingua 24 (3), 237–274.

Derwing, T.M., Rossiter, M.J., Munro, M.J., Thomson, R.I., 2004. Second language fluency: judgments on

different tasks. Language Learning 54 (4), 655–679.

Dittmann, A.T., Llewellyn, L.G., 1968. Relationship between vocalizations and head nods as listener responses.

Journal of Personality and Social Psychology 9 (1), 79–84.

Drummond, K., Hopper, R., 1993a. Back channels revisited: acknowledgement tokens and speakership

incipiency. Research on Language and Social Interaction 26, 157–177.

Drummond, K., Hopper, R., 1993b. Some uses of yeah. Research on Language and Social Interaction 26, 203–

212.

Duncan, S., Fiske, D., 1977. Face-to-Face Interaction: Research Methods, and Theory. Erlbaum, Hillsdale, NJ.

Duncan, S., 1974. On the structure of speaker–auditor interaction during speaking turns. Language in Society 2,

161–180.

Ejzenberg, R., 2000. The juggling act of oral fluency: a psycho-sociolinguistic metaphor. In: Riggenbach, H.

(Ed.), Perspectives on Fluency. Michigan University Press, Ann Arbor, pp. 287–313.

Foster, P., Skehan, P., 1996. The influence of planning and task type on second language performance. Studies in

Second Language Acquisition 18, 299–323.

Freed, B.F., Segalowitz, N., Dewey, D.P., 2004. Context of learning and second language fluency in French:

comparing regular classroom, study abroad, and intensive domestic immersion programs. Studies in Second

Language Acquisition 26, 275–301.

Freed, B.F., 1995. What makes us think that students who study abroad become fluent? In: Freed, B.F. (Ed.),

Second Language Acquisition in a Study Abroad Context. Benjamins, Amsterdam, pp. 123–148.

Freed, B.F., 2000. Is fluency, like beauty, in the eyes (and ears) of the beholder? In: Riggenbach, H. (Ed.),

Perspectives on Fluency. Michigan University Press, Ann Arbor, pp. 243–265.

Fries, C., 1952. The Structure of English. Harcourt Brace, New York.

Goodwin, C., Goodwin, M., 1987. Concurrent operations on talk: notes on the interactive organization of

assessments. IPRA Papers in Pragmatics 1 (1), 1–54.

Goodwin, C., 1986. Between and within: alternative sequential treatments of continuers and assessments. Human

Studies 9, 205–217.

Heinz, B., 2003. Backchannel responses as strategic responses in bilingual speakers’ conversations. Journal of

Pragmatics 35, 1113–1142.

Jefferson, G., 1984. Notes on a systematic deployment of the acknowledgement tokens ‘‘yeah” and ‘‘mm hm”.

Papers in Linguistics 17 (2), 197–216.

Kanfer, F., McBrearty, J., 1962. Minimal social reinforcement and interview content. Journal of Clinical

Psychology 18, 210–215.

Kendon, A., 1967. Some functions of gaze direction in social interaction. Acta Psychologica 26 (1), 22–63.

Kormos, J., Denes, M., 2004. Exploring measures and perceptions of fluency in the speech of second language

learners. System 32, 145–164.

Kormos, J., 1999. The effect of speaker variables on the self-correction behavior of L2 learners. System 27, 207–

221.

Lennon, P., 1990. Investigating fluency in EFL: a quantitative approach. Language Learning 40, 387–417.

Lennon, P., 2000. The lexical element in spoken second language fluency. In: Riggenbach, H. (Ed.), Perspectives

on Fluency. Michigan University Press, Ann Arbor, pp. 25–42.

LoCastro, V., 1987. Aizuchi: a Japanese conversational routine. In: Smith, L. (Ed.), Discourse Across Cultures.

Prentice Hall, New York, pp. 101–113.

LoCastro, V., 1999. A sociocultural functional analysis of fragmentation in Japanese. Multilingua 18 (44), 369–

389.

Matarazzo, J., Saslow, G., Wiens, A., Weitman, M., Allen, B., 1964a. Interviewer head nodding and interviewee

speech durations. Psychotherapy: Therapy, Research, and Practice 1, 54–63.

Matarazzo, J., Wiens, A., Saslow, G., Allen, B., Weitman, M., 1964b. Interviewer mm-hmm and interviewee

speech durations. Psychotherapy: Theory, Research, and Practice 1, 109–114.

Maynard, S., 1986. On back-channel behavior in Japanese and English casual conversation. Linguistics 24, 1079–

1108.

Maynard, S., 1990. Conversation management in contrast: Listener response in Japanese and American English.

Journal of Pragmatics 14, 397–412.

294 J.P. Wolf / System 36 (2008) 279–294

Maynard, S., 1997. Analyzing interactional management in native/non-native English conversation: a case of

listener response. International Review of Applied Linguistics in Language Teaching 35 (1), 37–60.

Mehnert, U., 1998. The effects of different lengths of time for planning on second language performance. Studies

in Second Language Acquisition 20, 83–108.

Miller, L., 1988. Listening behavior in conversations between Japanese and Americans. In: Blommaert, J.,

Verschueren, J. (Eds.), Intercultural and International Communication. Benjamins, Amsterdam, pp. 111–130.

Ortega, L., 1999. Planning and focus on form in L2 oral performance. Studies in Second Language Acquisition

21, 109–148.

Papajohn, D., 1998. Toward Speaking Excellence: The Michigan Guide to Maximizing your Performance on the

TSE Test & Speak Test. Michigan University Press, Ann Arbor.

Riggenbach, H., 1991. Toward an understanding of fluency: a microanalysis of nonnative speaker conversations.

Discourse Processes 14, 423–441.

Schegloff, E., 1982. Discourse as an interactional achievement: Some uses of ‘‘uh huh” and other things that come

between sentences. In: Tannen, D. (Ed.), Analyzing Discourse Text and Talk. Georgetown University Press,

Washington, DC, pp. 71–93.

Scherer, K.R., 1984. Emotion as a multicomponent process: a model and some cross-cultural data. In: Shaver, P.

(Ed.), Review of Personality and Social Psychology, Emotions, Relationships and Health, vol. 5. Sage,

Beverly Hills, CA, pp. 37–63.

Schmidt, R., 1992. Psychological mechanisms underlying second language fluency. Studies in Second Language

Acquisition 14, 357–385.

Schumann, J.H., Wood, L.A., 2004. The neurobiology of motivation. The Neurobiology of Learning:

Perspectives from Second Language Acquisition. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 23–42.

Schumann, J.H., 1997. The Neurobiology of Affect in Language. Blackwell, Boston.

Siegman, A., 1976. Do noncontingent interviewer mm-hmms facilitate interviewee productivity? Journal of

Consulting and Clinical Psychology 44 (2), 171–182.

Skehan, P., Foster, P., 1999. The influence of task structure and processing conditions on narrative retellings.

Language Learning 49 (1), 93–120.

Tao, H., Thompson, S., 1991. English backchannels in Mandarin conversations: a case study of superstratum

pragmatic interference. Journal of Pragmatics 16, 209–223.

Towell, R., Hawkins, R., Bazergui, N., 1996. The development of fluency in advanced learners of French. Applied

Linguistics 17, 84–119.

Wennerstrom, A., 2000. The role of intonation in second language fluency. In: Riggenbach, H. (Ed.), Perspectives

on Fluency. Michigan University Press, Ann Arbor, pp. 102–127.

White, S., 1989. Backchannels across cultures: a study of Americans and Japanese. Language in Society 18, 59–

76.

White, R., 1997. Back channelling, repair, pausing, and private speech. Applied Linguistics 18 (3), 314–344.

Wigglesworth, G., 1997. An investigation of planning time and proficiency level on oral test discourse. Language

Testing 14, 85–106.

Wood, D., 2001. In search of fluency: what is it and how can we teach it? Canadian Modern Language Review 57,

573–589.

Yngve, V., 1970. On getting a word in edgewise. In: Papers from the Sixth Regional Meeting of the Chicago

Linguistic Society, pp. 567–578.

Yuan, F., Ellis, R., 2003. The effects of pre-task planning and on-line planning on fluency, complexity, and

accuracy in L2 monologic oral production. Applied Linguistics 24, 1–27.