The relationship between task repetition and language proficiency

Introduction
Enormous growth of interest in task-based
language  learning  and  teaching  has  been
observed in recent years. There are several
reasons for this interest. First, a ‘task’ is a
construct  of  equal  importance  to  both
second  language  acquisition  researchers
and language teachers (Ellis, 2003, 2008).
Second,  task-based  pedagogy  is  capable
of  a  wide  range  of  interpretations;  that  is,
any single task, Ellis (2003) states, has the
potential  to  be  performed  in  a  number  of
ways,  depending  on  how  the  participants
orient  to  it.  This  perceived  flexibility  of
task-based  tradition  can  deflect  some  of
the  criticisms  leveled  against  it.  One  of
these criticisms is based on the claim that
performing  tasks  and  language  use  does
not necessarily lead to fluent and accurate
production  or  language  acquisition
(Reinders, 2009).  
 
From  the  vantage  point  of  information
processing  theories,  this  is  in  part  due  to
the fact that language learners’ attentional
or  processing  capacity  is  restricted,  and
hence,  they  cannot  process  ‘schematic’
and ‘systemic’ knowledge simultaneously
(see  Carroll,  2008;  Ellis,  1994,  2003,
2005; Randall, 2007; Skehan, 1998, 2007;
Skehan  &  Foster  1999,  2001).  This  being
so,  language  learners  tend  to  bypass
language  form  in  favor  of  meaning
drawing  on  their  wide  repertoire  of
communicative  strategies  to  which  they
have access (Skehan, 1998a).  
 
Researchers  have  proposed  several
implementation  and  task-based  variables
 
one  of  which  is  Task  Repetition.
However, there is not enough evidence as
to the relationship between task repetition
and  individual  difference  variables.  This
study aimed to compare the oral discourse
produced  on  two  encounters  of  the  same
task  and  the  way  in  which  task  repetition
correlates  with  language  proficiency
affecting  L2  oral  production  in  terms  of
accuracy, complexity and fluency.  
 
Background  
Task repetition
Task  repetition  is  a  very  important
concept  in  language  teaching  (Larsen-Freeman,  2012),  especially,  its  new
conceptualization  has  drawn  so  much
attention.  This  new  conceptualization  is
influenced by the view that our attentional
and  processing  capacity  during
communication  activities  is  inherently
both  limited  and  selective  (Anderson,
1995;  Schmidt,  2011).  As  a  result,  L2
learners  cannot  focus  on  both  meaning
and  form  simultaneously  (Van  Patten,
1990).  Because  tasks  are  essentially
meaning-centered,  when  it  comes  to
prioritizing  either  form  or  meaning,  it  is
likely  that  task  performers  choose
meaning. However, task repetition has the
potential  to  free  up  task  participants’
limited  attentional  resources  and  to  help
them  devote  much  of  their  cognitive
resources  to  the  formal  and  systemic
aspects  of  language  (Ahmadian  &
Tavakoli, 2011; Ellis, 2005). According to
Bygate  and  Samuda  (2005,  p.  45),  task
repetition  entails  two  phases:  a  first
performance  of  a  task,  in  which  task
performers organize the cognitive content,
select  the  useful  lexico-grammar,  process
it  online,  and  produce  “an  experientially
derived  multi-level  schema  to  support
subsequent linguistic work”; and a second
performance,  during  which  the  task
participant  can  build  upon  the  previous
one.  Several  studies  have  investigated  the
effects  of  task  repetition  on  L2
performance  and  in  this  section  some  of
the most relevant studies will be reviewed.  
 
Bygate’s studies
Martin  Bygate  is  perhaps  the  first  scholar
to study task  repetition in light of its new
conceptualization.  In  his  first  pioneering
study (1996), he  asked a language learner
to  perform  a  task  twice  with  a  three-day
time interval and without being told on the
first  occasion  that  the  task  would  be
repeated  three  days  later.  She  was  asked
to watch a Tom & Jerry video cartoon and
then to retell it.  Bygate  reported that this
form  of  repetition  resulted  in  some
striking  improvement  in  both  fluency  and
accuracy  (also  see  Bygate,  1999).
Accuracy,  too,  had  some  improvements
which  were  in  terms  of  vocabulary,
idiomaticity,  grammatical  markers  and
structure.  
 
Bygate  (2001)  drew  on  his  first
investigation  and  sought  to  compare  the
performances of 48 learners on a narrative
and  an  interview.  In  that  study,
participants  were  asked  to  perform  one
version  of  each  task  while  the  two
occasions  of  performance  were  10  weeks
apart.  Bygate  reported  that  over  the  10
week  interval  between  the  two
performances,  one  group  practiced
narrative  tasks  and  the  other  group
practiced  the  interview  tasks.  Overall,  he
attempted  to  address  three  things:  (a)  the
second performance of the same tasks that
they  had  performed  10  weeks  earlier;  (b)
performance  of  a  new  version  of  the  type
of task that participants had practiced over
the  10  weeks  and  the  one  they  had  not
practiced;  and  (c)  participants’  overall
performance  across  the  two  task  types.  It
was  found  that  task  type  had  no
meaningful  effect  on  learners’
 
performances.  However,  the  findings
revealed  that  task  repetition  had  a
significant  effect  on  fluency  and
complexity.  The  findings  of  this  study
were  similar  to  and  consistent  with
Bygate’s  (1996)  results  regarding  the
impacts of task repetition.  
 
Bygate  used  his  previous  dataset  in
Bygate  and  Samuda  (2005).  In  this  study
the  authors  attempted  to  test  whether  or
not  performing  a  communicative  task  for
the  second  time  may  assist  learners  to
combine  what  they  already  know  into
what  they  do.  Bygate  and  Samuda  (2005,
p.  45)  maintain  that  task  repetition  is  a
kind  of  planning  and  they  argue  that
repetition  has  “the  potential  to  lead  to
integration  of  knowledge  and
performance”  and  it  could  be  viewed  as
“facilitating  changes  particularly  in  the
conceptualization  and  formulation  phases
of  the  production  process”  (Bygate  &
Samuda,  2005,  p.45).  In  addition  to
confirming Bygate’s (2001) findings,  the
results  revealed  that  the  effects  of  task
repetition  “extends  well  beyond  the
domains  of  fluency,  accuracy,  and
complexity  and  into  aspects  of  language
use  which  involve  qualitative  issues  such
as […] in what ways speakers bring their
language  knowledge  into  action  to
generate  an  effective  piece  of  talk”
(Bygate & Samuda, 2005, 66).  
 
Lynch and McLean’s studies
Another  interesting  series  of  studies  on
task  repetition  has  been  carried  out  by
Lynch  and  McLean  (2000,  2001)  in  an
ESP context. They designed a special and
interesting task called ‘postal carousal’. In
this task students were required to read an
academic  article  and  prepare  a  poster
presentation  based  on  it.  Each  student
(poster  presenter)  had  six  visitors,  which
means  that  each  of  them  repeated  the
same  task  of  answering  to  the  same
question  posed  by  the  visitors  six  times.
Lynch  and  McLean  found  that  this
recycling  had  positive  impacts  on  both
accuracy  and  fluency  in  language
production.  Another  interesting
observation  was  that  while  highly
proficient  students  used  the  recycling
opportunity  to  improve  clarity  of  their
expressions,  low  proficient  students  made
use of these opportunities to improve their
accuracy and pronunciation.
 
Ahmadian’s studies
Ahmadian  and  Tavakoli  (2011)
investigated  the  effects  of  simultaneous
use  of  task  repetition  and  careful  online
planning (operationalized as the provision
of  ample  time  for  task  performance)  on
the  CAF  of  EFL  learners.  They  asked
intermediate  EFL  learners  to  repeat  an
oral narrative task with  an interval of one
week.  Results  of  their  study  revealed  that
task  repetition  positively  impacts
complexity and fluency. Moreover, it was
found that task repetition had the potential
to  compensate  for  the  dysfluency  which
resulted  from  engaging  in  careful  online
planning.  Overall,  the  findings  of  their
research confirmed Bygate and Samuda’s
(2005)  claim  that  task  repetition  could
complement  both  strategic  and  careful
online planning.  
 
In  another  study,  Ahmadian  (2011)
examined  the  effects  of  massed  task
repetition  on  the  CAF  triad  over  a  six-month  period  and  sought  to  see  if  these
effects transfer to a new task. The subjects
included  30  intermediate  EFL  learners
from two intact  classes  who were divided
into  two  groups.  Participants  in  the
experimental  group  were  asked  to  do  a
dialogic  narrative  task  on  occasions  1  to
11  and  an  interview  task  on  occasion  12,
each  occasion  was  two  weeks  apart.
 
However, participants in the control group
were  only  required  to  perform  the  oral
narrative task at time 1  and engage in the
interview  task  at  time  12.  The  results
revealed  that  massed  repetitions  of  the
same  task  assisted  subjects  in  the
experimental group to outperform those in
the  control  group  in  terms  of  complexity
and fluency, but not accuracy, to the effect
that  the  benefits  of  massed  repetitions  of
the  same  task  transfers  to  performance  of
a  new  task.  This  finding  is  important  in
that  it  demonstrates  that  task  repetition
could assist language learning.
 
Language proficiency
Language  proficiency  (LP)  could  be
defined as “a person’s overall competence
and  ability  to  perform  in  L2  [second
language]”  (Thomas,  1994,  p.  330,
footnote  1).  Most  of  the  investigative
attempts related to this construct pertain to
its  operationalization  (Hulstijn,  2011).
Theoretically,  there  are  grounds  to
hypothesize that the way second language
learners  make  use  of  task  repetition
opportunity  is  mediated  by  their
differential  LP.  This  hypothesis  is  based
on  two  interrelated  assumptions:  (a)  on
the  second  encounter  with  the  same  task,
language  learners  are  assumed  to  monitor
and  plan  their  speech  and  for  doing  so
they  would  need  to  fall  back  on  their
explicit  knowledge  (Ellis,  2005);  and  (b)
the  grammatical  knowledge  which  could
be  represented  as  explicit  knowledge
constitute  is  an  important  component  of
language  proficiency  (or  communicative
language  ability)  (Bachman  &  Palmer,
1996).
     
It  is  now  well  established  that  factors
external  to  the  classroom,  such  as  L2
learners’ age of first exposure to the target
language,  their  length  of  stay  in  an
environment  where  that  language  is
spoken,  and  their  percentage  of  weekly
use  of  the  target  language,  play  a
significant  role  in  determining  their  L2
proficiency  (see  Tremblay,  2011).  Hence,
even  in  an  instructed  environment,  L2
learners  show  considerable  variability  in
their  proficiency  due  to  these  factors
(Tremblay,  2011).  Because  proficiency
directly  influences  L2  learners’
performance  on  experiments,  it  seems
imperative  that  this  variable  be
characterized  as  precisely  and  accurately
as  possible  in  experimental  research.
Surprisingly,  most  of  the  studies  on  the
effects  of  task-based  implementation
variables  have  simply  been  controlled  for
the effects of L2 proficiency and therefore
no related studies exist to be reviewed. In
this study, this variable will play a pivotal
role.
 
Research question
In  light  of  the  above-mentioned
theoretical  and empirical backgrounds the
following  research  question  was
formulated:
  Is  there  any  relationship  between
language  proficiency  and  the
effects  of  task  repetition  on
complexity,  accuracy,  and  fluency
of L2 speech production?
 
Method
To  conduct  the  present  study,  a
correlational  design  was  applied  in  the
following  manner:  The  study  was  carried
out in three sessions. In the first session, a
cloze  test  was  given  to  the  participants  to
do.  In  the  second  and  third  sessions  the
participants  performed  the  oral  narrative
task.  
 
Participants  
Fifty Iranian EFL learners took part in this
investigation. There were 19 males and 31
females  and  their  average  age  was  19
 
ranging  from 17 to 21  years old. They all
signed  the  informed  consent  forms.  They
were  asked  several  questions  and  it
became  clear  that  they  had  studied
English  for  at  least  8  months.  None  of
them  had  had  any  opportunity  to  use
English  language  for  communicative
purposes  outside  the  classroom.  Since  in
this  study  language  proficiency  was  an
important  variable  they  were  selected
from among different levels of proficiency
and  totally  at  random  using  systematic
random sampling.
 
Task
In  line  with  the  previous  task  repetition
studies  (Ahmadian  &  Tavakoli,  2011;
Bygate, 1996, 2001), in the present study,
learners  did  an  oral  narrative  task  twice
with  a  one-week  interval.  The  video  was
silent  and  monologic.  This  action  is  good
to  make sure that L2 oral performance is
being  investigated  “as  an  individual
attribute” (De Jong, Groenhout, Schoonen
&  Hulstijn,  2013)  rather  than  the  product
of a process of co-constructing a message
which  is  typical  of  the  dialogic  mode  of
discourse  (De  Jong,  et  al.,  2013).  The
video  selected  for  this  study  was  One-Man Band (2005), the story of which was
found to be ‘organized’ and ‘structured’ in
a series of studies on task structure and L2
oral  performance.  The  video  tells  the
interesting  tale  of  a  peasant  girl  who
encounters  two  competing  street
performers who would prefer the coin find
its way into their tip jars. As the two one-man  bands’  rivalry  crescendos,  the  two
overly eager musicians vie to win the little
girl’s attention.  In none of the occasions
of  task  performance,  participants  were
allowed  to  either  take  notes  or  do
preparations  before  narrating  the  story  of
the  video.  Moreover,  they  were  not  told
that  they  were  going  to  repeat  the  same
task two weeks later.   
Language proficiency: Cloze test  
In  the  first  session  of  data  collection,  a
cloze  test  was  administered  to  the
participants.  The  use  of  cloze  tests  as  a
measure  of  proficiency  is  not  new.  The
cloze  procedure  was  first  introduced  by
Taylor  (1957)  and  has  been  the  object  of
much  testing  research  since  the  1970s
(e.g.,  Oller,  1972,  1973).  Some
researchers  have  argued  that  cloze  tests
assess  low-level  lexical  and  grammatical
competence  (e.g.,  Alderson,  1979),
whereas  others  have  proposed  that  cloze
tests  can  also  measure  higher  level
discourse  competence  (Oller,  1973).
Although no consensus has been achieved
as  to  what  aspects  of  linguistic
competence  cloze  tests  measure,  their
scores have been found to correlate highly
with  standardized  proficiency  scores  (see
Tremblay, 2011).  
 
The  validity,  reliability,  and
discriminability  of  cloze  tests  are
ultimately  a  function  of  the  extent  to
which  these  tests  are  tailored  for  the
targeted  population  of  L2  learners.  As
Messick  (1989)  suggested,  validity  is  not
a  property  of  a  test  but  rather  of  the
inferences made on the basis of the test. In
fact,  in  order  that    “such  inferences  be
accurate, the cloze test must be neither too
easy  nor  too  difficult  for  the  targeted  L2
learners;  otherwise,  the  test  may  not
reveal  much  about  these  learners’
proficiency  other  than  whether  it  meets  a
particular level” (Tremblay, 2011, p. 345).  
 
  To  sum  up,  as  Tremblay  (2011,  p.  346)
rightly points out: “more than just a useful
tool  by  testing  standards,  cloze  tests  are
also a practical tool for research purposes:
Unlike  standardized  proficiency  tests—
which at any rate are not always available
to  researchers—they  can  take  a  relatively
short  amount  of  time  to  complete  (e.g.,
 
15–35  min,  depending  on  the  difficulty
level  of  the  test).  Their  flexible  format
(e.g.,  choice  of  text,  length  of  text,  word-deletion  ratio,  scoring  method)  makes  it
possible  to  target  a  particular  range  of
proficiency  levels,  and  they  are  easy  to
create  and  score  if  clear  scoring  criteria
are  established.  These  advantages  are
certainly well known to some researchers,
as shown by a number of surveyed studies
that have already employed cloze tests (or
their  counterpart,  C-tests;  Raatz  &  Klein-Bradley,  1981)  as  a  “proficiency
assessment method.”
 
In the present study, a 300-word cloze test
was  designed  and  the  words  were  deleted
from  the  text  using  the  purposive  method
so  that  a  balanced  proportion  of  content
and function words could be elicited from
L2  learners.  The  test  had  an  open-ended
format  because  it  was  assumed  that  this
format  was  more  likely  to  provide  a
picture  of  the  production  skills  than  the
multiple-choice  format  did.  The  test  was
then  piloted  with  4  native  speakers  of
English.  The  test  was  revised  when  all  4
native  speakers  completed  it.  The  final
version of the test, provided in Appendix,
had  40  words  deleted,  of  which  about  20
were content words (i.e. open-class: nouns
adjectives,  main  verbs,  adverbs,  etc.)  and
about 20 were function words (i.e. closed-class:  determiners,  pronouns,  auxiliaries,
etc.).  Despite  the  occasional  difficulty
involved in determining what is and is not
an  acceptable  answer,  scoring  cloze  tests
on  the  basis  of  acceptable  answers  has
more  face  validity,  in  that  it  is  rarely  the
case that only one word is allowed in any
given  lexical,  morphosyntactic,  and
discourse  context  (Tremblay,  2011).  Two
native  speakers  of  English  were  asked  to
check  the  acceptable  responses  and  mark
those  which  were  not  acceptable  in
English. The cloze test scores could range
between 0 and 40.  
 
In  the  second  and  the  third  sessions  of
data  collection  the  participants  were
required to perform the oral narrative task
and  their  narrations  were  audio-recorded
for  further  analysis  and  coding.  The
transcribed  narrations  were  analyzed  in
terms  of  the  CAF  measures  described
below.  
 
Complexity, accuracy, and fluency
According to Ellis and Barkhuizen (2005),
one  way  to  make  studies  on  the  same
issue more comparable and, ultimately, to
help  reach  generalizations  is  to  use  the
same  measures  and  assessment  tools.  In
this  study,  attempts  were  made  to  use  the
same  measures  used  in  other  task
repetition  studies.  Following  Ahmadian
(2011), the following measures were used:
 
Complexity:  
  Syntactic  complexity  (amount  of
subordination):  the  ratio  of  clauses  to
AS-units  in  the  participants’
production.  The  rationale  behind
choosing  AS-unit  is  that  this  unit  is
essentially  a  syntactic  one  and
syntactic  units  are  genuine  units  of
planning  (Foster,  Tonkyn,  &     
Wigglesworth.,  2000)  which  might
make  them  good  units  for  analyzing
spoken language in this study.
 
AS-unit  is  defined  as  “…  a  single
speaker’s  utterance  consisting  of  an
independent  clause  or  sub-clausal  unit,
together  with  any  subordinate  clause(s)
associated with it” (Foster et al., 2000, p.
365).  
 
  Syntactic  variety:  the  total  number  of
different grammatical verb forms used
in  participants  performances.  Tense
   
(e.g.  simple  present,  simple  past,  past
continuous,  etc.)  and  modality  (e.g.
should,  must,  etc.)  were  taken  as
grammatical  verb  forms  used  for  the
analysis.
  Overall  complexity:  the  mean  length
of  AS-units  in  the  participants’  oral
performances  as  measured  by
calculating the mean number of words
per AS-unit.  
 
Accuracy:
  Error-free  clauses:  the  percentage  of
the clauses which were not erroneous.
All  syntactic,  morphological  and
lexical  errors  were  taken  into
consideration.  
  Correct  verb  forms:  the  percentage  of
all verbs which were used correctly in
terms  of  tense,  aspect,  modality,  and
subject-verb agreement.
 
Fluency:
  Rate  A  (number  of  syllables  produced
per minute of speech): the number of  
syllables  within  each  narrative,
divided  by  the  total  articulation  time
and multiplied by 60.
  Rate  B  (number  of  meaningful
syllables  per  minute  of  speech):  rate
A’s procedure was followed again, but
all  syllables,  words,  phrases  that  were
repeated,  reformulated,  or  replaced
excluded.
 
Given  the  nature  of  the  data,  Pearson
Correlation  Coefficient  was  used  to  test
the null hypotheses.
 
Results
This  study  aimed  to  investigate  the  way
task  repetition  correlated  with  language
proficiency to affect L2 oral production in
terms  of  accuracy,  complexity  and
fluency.  In  this  section,  the  results  of  the
study  will  be  reported  and  each  finding
will  be  interpreted  with  regards  to  the
relevant  theoretical  and  empirical
frameworks.    The  question  that
entertained  this  research  study  was
whether  or  not  language  proficiency
correlates  with  the  effects  of  task
repetition  on  the  CAF  triad.  The  three
variables, i.e., the CAF triad, were treated
separately.  
 
Tables 1 and 2 clearly indicate that on the
first  encounter  with  the  narrative  task  the
relationships  are  not  statistically
significant. This is an important finding in
that, as the results show, there are positive
relationships  between  LP  and  all  three
dimensions  of  L2  proficiency  (the  CAF
triad),  but  these  relationships  are  not
statistically significant

The  first  part  of  the  research  question
concerns  complexity.  A  comparison
between the first and second encounter of
the  task  performance  in  Tables  2  and  4
reveals  that  the  results  for  the  second
encounter  are  statistically  significant  (p  <
.04  for  syntactic  complexity  and  p  <  .003
for  syntactic  variety).  This  means  that
participants  with  higher  levels  of
proficiency have been more adept in using
the  task  repetition  opportunity  to  produce
more  complex  language.  Taking  into
account Cohen’s criterion for interpreting
effect  sizes,  the  former  correlation  (r  =
.41) is moderate and the latter is large (r =
.55). This is in line with the argument put
forth  by  many  researchers  regarding  the
important  role  of  language  proficiency  in
all  teaching  and  learning  techniques  (see
Trembly, 2011).  
      
The  second  and  the  third  part  of  the
research  question  concerns  accuracy  and
fluency.  Here,  again,  a  brief  look  at
Tables 3 and 4 reveals that  on the second
encounter  the  participants  with  higher
language  proficiency  have  been  able  to
produce  more  accurate  (p<  .001;  p<.001)
and fluent (p< .005; p<.001) language and
this  is  the  case  for  all  measures  which
have  been  utilized  to  assess  accuracy  and
fluency.  All  effect  size  magnitudes  for
accuracy  and  fluency  are  large.  A
theoretical  account  for  this  finding  would
be  that  our  attentional  resources  are
limited and selective (Schmidt, 2001) and
therefore  the  participants  with  higher  LP
might  be  more  able  to  allocate  their
attentional  resources  to  the  production  of
fluent and accurate language.

Conclusions
The  aim  of  this  study  was  to  see  if  LP
correlates  with  the  effects  of  task
repetition on CAF. The findings indicated
that  participants  with  higher  English
language  proficiency  are  more  capable  of
using  this  task-based  opportunity  to
produce  more  complex,  fluent,  and
accurate  language.  These  findings  have
some  implications  for  L2  pedagogy.
Anecdotal  evidence  has  it  that  in  the
   
Iranian  EFL  context,  explicit  teaching  of
grammar  is  of  prime  importance  to
language  learners.  Task-based
methodology, however, by its very nature,
puts premium on meaning. Given the fact
that  language  learners’  attentional  and
processing  capacity  is  limited,  they  might
be  induced  to  prioritize  meaning  over
form. Therefore, the explicit instruction of
grammar is in a way marginalized and the
use  of  strategic  competence,  contextual
support,  and  formulaic  chunks  might  be
encouraged.  Despite  all  this,
methodology,  Widdowson  (1990)  argues,
always  finds  some  room  for  maneuver.
TBLT  is  no  exception  in  this  particular
respect.  Task  repetition  is  an
implementation variable which has proved
to  be  useful  for  enhancing  the  CAF  of
language  learners.  Given  the  correlation
nature  of  this  study,  it  is  important  to
point  out  that  the  results  need  to  be
interpreted  with  due  caution.  In  fact,
correlation  is  not  equal  to  causation  and
therefore,  in  addition  to  language
proficiency,  one  needs  to  think  of  other
individual  variables  (such  as  willingness
to communicate, anxiety, etc.) when using
task repetition in the classroom.