Professional Documents
Culture Documents
Derwing 2008
Derwing 2008
doi:10.1093/applin/amm041
COMPREHENSIBILITY
We define comprehensibility as the ease or difficulty with which a listener
understands L2 accented speech. Numerous studies have shown that
FLUENCY
Oral fluency, interpreted here as an automatic procedural skill on the part of
the speaker (Schmidt 1992) and a perceptual phenomenon in the listener,
has been investigated from a number of perspectives: the characteristics
of fluent and dysfluent L2 speech (Lennon 1990, 2000; Riggenbach 1991,
2000; Wennerstrom 2001); the effects of planning (e.g. Foster and Skehan
1996; Wigglesworth 1997; Ortega 1999; Yuan and Ellis 2003); self-
monitoring (Kormos 1999); neurobiological factors (Dewaele 2002); and
task type (Bygate 1996; Skehan and Foster 1999; Derwing et al. 2004). In
addition, there has been considerable research on the effects of a study-
abroad experience on fluency (Lennon 1990; Freed 1995; Towell et al. 1996;
Segalowitz and Freed 2004; Towell and Dewaele 2005). Generally, fluency
improves with greater opportunities to interact in the L2 outside of the
classroom. However, most study-abroad research has involved a relatively
narrow population comprising young university students who have self-
selected to take language courses. Like comprehensibility, fluency, or
automaticity of production, is a vital aspect of successful communication.
Filled pauses, excessive pausing, pausing in inappropriate places, false starts,
and a slow speaking rate can all affect the listener negatively (Derwing and
Munro 2001; Munro and Derwing 2001). Fluency problems can also be
exasperating for L2 speakers; it is thus important for L2 learners to develop
oral fluency skills early in L2 acquisition.
WILLINGNESS TO COMMUNICATE
As important as comprehensibility and fluency are, L2 learners’ willingness
to communicate (WTC) is also an essential prerequisite for successful
interactions, particularly in the case of people who live and work in the L2
environment (MacIntyre et al. 1998). Originally conceptualized for the L1,
WTC refers to the ‘probability of engaging in communication when free to
choose to do so’ (p. 546). WTC in the L2 is highly complex as shown in the
TRACEY M. DERWING, MURRAY J. MUNRO, and RON I. THOMSON 361
RESEARCH QUESTIONS
METHOD
Our mixed-methods longitudinal design comprised quantitative evaluations
of the L2 learners’ comprehensibility and fluency, and qualitative interviews
to probe learners’ English learning experiences.
English L2 learners
The learners were sixteen high beginner MAS ESL students (six men and ten
women) from 27 to 42 years of age (M ¼ 34.4years) and sixteen high
beginner SLS ESL students (eight men and eight women), from 19 to 49
years of age (M ¼ 38.6years). The SLS group consisted of fifteen Russian
speakers and one speaker of Ukrainian. These two languages are very similar
in terms of grammar, phonology, and lexis, and are mutually intelligible.
Furthermore, Russians and Ukrainians shared many experiences under the
Soviet Union.
Once they arrived in Canada, the students’ proficiency was tested with the
Canadian Language Benchmarks assessment tool (Centre for Canadian
Language Benchmarks 2006). All were in Stage 1 of the Canadian Language
Benchmarks, and all were enrolled in full-time beginner ESL classes at the
outset of the study. In total, data were collected seven times over the course
of 2 years. The first six data collection points were at approximately 2-month
intervals, while the seventh and final collection date was a year after the
sixth. For the purposes of this paper, we chose to use the data collection
points from Time 2 (T2), Time 6 (T6) 8 months later, and Time 7 (T7), a year
after T6. The picture narratives given to the speakers at these three times
were identical; furthermore, since the same narrative had been used at Time
1, the speakers were familiar with the task. We realize that this raises the
potential problem of practice effects in our data; however, the drawbacks of
using the same task repeatedly were outweighed by the importance of
obtaining comparable speech samples at each testing session. In an earlier
study, comparisons were made of a larger set of these same speakers at Times
1 and 2 (Derwing et al. 2006). There were no significant between-group
differences in fluency at T2 (although the MAS appeared to have a fluency
advantage at Time 1); for this reason, T2 seemed to be the best starting point
for the current comparisons. At that point, the students had completed
between 3 and 5 months of ESL study. Some of the participants had been in
Canada for up to 3 months before beginning ESL classes, but reported little or
no contact with native speakers during that period. The MAS indicated that
they had studied English as a foreign language (EFL) in China for between 1
364 ESL LEARNERS’ FLUENCY AND COMPREHENSIBILITY DEVELOPMENT
and 20 years (M ¼ 6.94years). The SLS had had EFL instruction for
between 2 and 13 years (M ¼ 5.7years); however, in all cases, members of
both groups had very little experience with listening and speaking in
Speaking tasks
The ESL participants narrated the same story in English from a set of pictures
at each of the three data collection points. The pictures were of a man and a
woman who collided on a busy street corner in a large city. They fell to the
ground, dropping their identical green suitcases in the process. After standing
up and apologizing, they each picked up a suitcase and went on their
respective ways. Finally, upon opening their luggage, they discovered that
they had mistakenly switched bags. In each session the speakers were given a
few minutes to familiarize themselves with the pictures and to ask questions.
Recordings of the learners’ narratives were made using high quality digital
recording equipment in a quiet room.
Stimulus preparation
For presentation to the listeners, 20-second excerpts were taken from the
beginning of each production, eliminating sample-initial false starts or
hesitations. Speakers sometimes employ dysfluencies in turn beginnings as
interactional strategies (Carroll 2004); however, our listeners could not see
the speakers face-to-face and heard these narratives as a single turn, and thus
might have been unduly influenced by initial dysfluencies. Samples from the
beginning of the narrative ensured that content was held relatively constant
across speakers and times. The ninety-six 20-second samples (32 speakers 3
times) were randomized. An additional three native speaker samples were
added to the stimulus set to verify that raters stayed in step during the rating
task. Three different randomizations were recorded onto separate CDs for
presentation to the listeners.
Raters
A total of thirty-three native-speaking English listeners (six males and
twenty-seven females; age range 20–48 years, M ¼ 30 years) were recruited
to evaluate the fluency and comprehensibility of the L2 speakers using
7-point Likert scales. Fourteen raters were assigned to randomization 1, ten
to randomization 2, and nine to randomization 3. All of the raters were
students enrolled in courses in the Faculty of Education at the University of
Alberta. None had studied Mandarin, but one person reported having studied
Ukrainian, although she indicated that she was not at all proficient. Twelve
listeners said that they had some familiarity with Slavic or Mandarin accents,
although none of the participants declared a high degree of familiarity.
All the listeners reported normal hearing.
Rating task
The listeners completed the rating task in small groups. Two 7-point scales
were used, one for fluency (1 ¼ extremely fluent, 7 ¼ extremely dysfluent) and
one for comprehensibility (1 ¼ very easy to understand, 7 ¼ extremely difficult to
understand). Before the task began, we briefly explained that fluency
judgements should be based on factors such as speech rate, filled pauses
such as ums and uhs, self-corrections, and silent pauses, as well as the overall
flow of speech, and that grammar and vocabulary should not be taken into
consideration. This explanation was made to ensure that raters did not
confuse ‘fluency’ with ‘proficiency’, a common connotation of this term
366 ESL LEARNERS’ FLUENCY AND COMPREHENSIBILITY DEVELOPMENT
RESULTS
The ratings of the NS samples, which consistently showed high
comprehensibility and a high level of fluency, indicated that none of the
listeners lost their place during the task; these ratings were subsequently
removed from the data set. Intraclass Correlation Coefficients for the NNS
were .95 for comprehensibility and .97 for the fluency ratings.
Having established strong inter-rater reliability, we pooled mean
comprehensibility and fluency ratings over all raters to derive means for
each learner as well as for each group of learners. As can be seen from
Figures 2 and 3, the MAS and SLS groups’ mean scores at 2 months were
nearly identical on both scales. However, over time, the two groups appear to
diverge.
Mean ratings for each learner on each scale were submitted to two
separate mixed-design ANOVAs, one to evaluate changes in comprehensi-
bility ratings over time and another to assess changes in fluency ratings.
7
Mandarin
7=Extremely difficult to understand
6 Slavic
1=Very easy to understand
1
Two months Ten months Two years
7
Mandarin
6 Slavic
1
Two months Ten months Two years
First language (L1) was the between-groups factor and Time (3 levels) was
the within-groups factor.
For comprehensibility ratings, the effect of L1 was significant,
F(1, 30) ¼ 4.69, p ¼ .038, partial 2 ¼ .135, as was the effect of Time
F(2, 60) ¼ 13.02, p 5 .001, partial 2 ¼ .303. and the L1 Time interaction,
F(2, 60) ¼ 6.19, p 5 .001, partial 2 ¼ .171.
The ANOVA for mean fluency ratings also revealed significant effects of L1,
F(1, 30) ¼ 5.29, p ¼ .029, partial 2 ¼ .15, Time F(2, 60) ¼ 13.76, p 5 .001,
partial 2 ¼ .314, and the L1 Time interaction, F(2, 60) ¼ 3.98, p ¼ .024,
partial 2 ¼ .171.
Because of significant interaction effects, Bonferroni-adjusted t-tests were
used to evaluate within-group performance. There was no significant change
over time in comprehensibility for the MAS. However, for the SLS,
significant differences were found in mean comprehensibility ratings from T2
to T6 (Ms ¼ 4.19, 3.39; SDs ¼ .83, .60), t(15) ¼ 4.77, p 5 .001, and from T2 to
T7 t(15) ¼ 5.24, p 5 .001, but not from T6 to T7. No significant differences
were found in the fluency judgements of the MAS productions. Significant
differences were found in the fluency ratings of the SLS between T2 and
T6 (Ms ¼ 4.8, 4.15; SDs ¼ .96, .77), t(15) ¼ 3.19, p ¼ .006, and from T2 to
T7 (M ¼ 3.75; SD ¼ .81), t(15) ¼ 5.09, p 5 .001, but not from T6 to T7.
In summary, there was no indication of improvement over time for the MAS
in either comprehensibility or fluency. The SLS, however, improved
somewhat on both dimensions.
96
88
Number of responses (n=96) 80
72
1-3
times/week
4-6
times/week
once a day
several
times/day
never
1-3
times/week
4-6
times/week
once a day
several
times/day
Mandarin Slavic
and non-native speakers on the following scale: 1 ¼ never, 2 ¼ 1–3 times per
week, 3 ¼ 4–6 times per week, 4 ¼ once a day, and 5 ¼ several times a day.
These scores were pooled across times and submitted to non-parametric
Mann–Whitney U tests because of uneven distributions. They revealed that
the SLS interacted significantly more often than did the MAS with both
native speakers of English (U ¼ 691, p 5 .001) and non-native English
speakers (U ¼ 714, p ¼ .001). Because of the parallel findings for both kinds
of interlocutors, we pooled the responses regarding conversations with native
speaker and non-native speaker interlocutors in the histogram in Figure 4
to illustrate between-group differences.
At each recording session, the L2 learners reported the time they spent
listening to English talk radio and watching English television daily using a
scale of 1–5 (1 ¼ less than 1 hour, 2 ¼ 1 hour, 3 ¼ 2 hours, 4 ¼ 3 hours,
5 ¼ more than 3 hours). The MAS’s radio scores ranged from 1 to 5 across the
three times, while the SLS’s scores ranged from 1 to 4. Both groups’ scores
for television viewing ranged from 1 to 5. As can be seen in Figure 5, the SLS
listened to the radio significantly more often than the MAS (U ¼ 887,
p ¼ .027), but no significant between-group difference was found for time
watching television (U ¼ 901, p 4 .05).
40
24
16
0
< 1 2 3 >3 < 1 2 3 >3
1 hour hour hours hours hours 1 hour hour hours hours hours
Mandarin Slavic
p 5 .01, 2-tailed.
Comprehensibility scale: 1 ¼ very easy to understand; 7 ¼ very difficult to understand.
Fluency scale: 1 ¼ extremely fluent; 7 ¼ extremely dysfluent.
Affective-cognitive context
The affective-cognitive context encompasses intergroup attitudes, social
situation, and L2 communicative competence. The latter clearly affects
success in L2 interaction. The difference in the two groups’ ratings
could be attributable to the greater typological distance of Mandarin
TRACEY M. DERWING, MURRAY J. MUNRO, and RON I. THOMSON 371
from English. ‘I don’t have enough words, so I cannot read the news-
paper and watch TV. And sometimes for me, the pronunciation . . .
I cannot understand’ (MAS #15). While participants from both groups
Motivational propensities
One of the categories in the motivation layer of the WTC model is L2 self-
confidence. Again we noted some differences between the two groups, where
DISCUSSION
We compared listeners’ assessments of oral comprehensibility and fluency in
this longitudinal study of speakers of Mandarin and East Slavic languages
TRACEY M. DERWING, MURRAY J. MUNRO, and RON I. THOMSON 373
who were learning English. Our first research question asked how these two
features of L2 speech changed over a period of 2 years. For both
comprehensibility and fluency, the SLS improved from T2 to T6 and T7,
(1–3 times a week). The SLS also reported significantly more exposure
to English talk radio than the MAS. We recognize the limitations of scalar
self-report as a means of assessing exposure, but the specific instances in the
RECOMMENDATIONS
In this section we make recommendations for language programmes to
improve L2 learners’ comprehensibility and fluency development. Of course,
we realize that adult immigrant students must make many life-changing
adjustments to fulfil their potential in their new country, and we do not
advocate a unidirectional perspective on adjustment, even from a linguistic
standpoint. There is often a misapprehension that, in interactions between L2
speakers and native speakers, the L2 individual is entirely responsible for
communicative success. In fact, native speakers sometimes abdicate their
own responsibilities in such exchanges (Terrell 1990; Varonis and Gass 1982).
However, as Derwing et al. (2002) have shown, some native speakers, at
least, can be helped to see that they can play a greater role in ensuring
successful interactions. Post-secondary institutions and workplaces with
diverse populations should consider implementing not only anti-racism and
cross-cultural awareness programmes, but practical guidelines for enhancing
NS interlocutors’ ability to converse (and understand) second language users.
Current severe labour shortages, a low birth rate, and the looming
retirement of baby boomers in Canada ensure high rates of immigration long
into the future. The federal government views language instruction for adult
newcomers as a crucial component of their integration into Canadian society.
However, as this study has demonstrated, the instruction the participants
received appeared to do very little to enhance their oral fluency and
comprehensibility in ways that could be recognized by naive native speaking
listeners. There appears to be a relationship between the SLS’s greater
(although limited) success and their exposure to English outside the
classroom. Although we have not established a direct causal link in this
study, our findings support MacIntyre et al.’s (1998) suggestion that the
376 ESL LEARNERS’ FLUENCY AND COMPREHENSIBILITY DEVELOPMENT
ACKNOWLEDGEMENTS
We thank the participants of the study for their willingness to meet with us seven times; they
were very generous and candid in sharing their experiences. We appreciate the comments of
Marian Rossiter, Kim Noels, and the anonymous reviewers. We thank NorQuest and Metro
Colleges for their cooperation. The Social Sciences and Humanities Council of Canada funded
this research.
1 What has been the most difficult thing about adjusting to life in Canada?
2 Describe how you use English during a normal day.
3 What have been the most difficult things about learning English?
4 What are the easiest things about learning English?
5 Is it difficult to make small talk? (Why?, why not?)
6 What did you find most useful about your language classes? What was
the least useful?
7 Is there anything you think you missed in language classes?
NOTE
1 ‘Visible minority’ is a legal term in Canada referring to ‘persons, other than
aboriginal peoples, who are non-Caucasian in race or non-white in colour.’
Employment Equity Act, 1995, c 44, retrieved 3 April 2006 from http://
laws.justice.gc.ca/en/E-5.401/text.html
REFERENCES
Bradac, J. 1990. ‘Language attitudes and impression (eds): Handbook of Language and Social Psychology.
formation’ in H. Giles and W. P. Robinson Chichester: John Wiley & Sons, Ltd, pp. 387–412.
378 ESL LEARNERS’ FLUENCY AND COMPREHENSIBILITY DEVELOPMENT
Lennon, P. 2000. ‘The lexical element in spoken Ortega, L. and G. Iberri-Shea. 2005. ‘Longitu-
second language fluency’ in H. Riggenbach (ed.): dinal research in second language acquisition:
Perspectives on Fluency. Ann Arbor: University of Recent trends and future directions,’ Annual
Michigan Press. Review of Applied Linguistics 25: 26–45.