Long-term Utilization of Interaction by Young EFL Learners: The Effects of Strategy Training

Introduction
Interaction has attracted a great deal of attention in studies of Second Language Acquisition
(SLA)  ever  since  the  traditional  enthusiasm  with  the  exaggerated  role  of  input  began  to  be
challenged. This trend was launched with a series of pioneering studies by Long (1981, 1983)
who analyzed the talk directed to L2 learners by native speakers (NSs) and the interaction in
which  they  engaged.  While  acknowledging  the  necessity  condition  for  input,  new  findings
started  to  contradict  the  sufficiency  condition  for  input  as  implied  by,  among  others,
Krashen’s input hypothesis (1982). Schmidt (1990, 1993), for instance, proposed that the
mere  exposure  to  input  may  not  be  beneficial  unless  “noticing”  occurs  and  input  is
transformed into “intake” or is “consciously registered” (p.130). Therefore noticing, defined
as “the mechanism by which learners, after sensitization to a particular structure, spot such
structure (or its absence) in subsequent natural input” (Fotos, 1993, p. 383) is speculated as a
pre-requisite  for  the  process  of  language  learning  and  a  proposition  to  challenge  the
sufficiency condition for input. Another trend which questioned the dominance of input in the
field of SLA in 1990s was Swain’s (1995) Output Hypothesis which was intended to shift the
focus  from  the  adequacy  of  input  to  the  benefits  of  output,  encompassing  a  learner’s
production. Swain (1995, 2000) highlighted the role of output as a supplement to input.
Finally,  building  on  the  arguments  put  forth  by  Schmidt  (1990,  1993),  Swain  (1995),
and  earlier  interactionist  research,  Long  (1996)  presented  an  updated  version  of  his
Interaction  Hypothesis  (Long,  1983),  forming  another  building  block  in  arguments
endeavouring to improve or redefine the notion of input in SLA. Among different accounts of
SLA  processes  that  have  been  put  forward  in  recent  decades,  the  Interaction  Hypothesis
(Long,  1996)  emphasizes  the  strong  connection  between  learners’  engagement  in
conversational interaction and second language acquisition based on negotiation of meaning.
According  to  Richards  and  Schmidt  (2002),  this  hypothesis  claims  that  “the  language
acquisition  requires  or  at  least  greatly  benefits  from  interaction,  communication  and
especially negotiation of meaning, which takes place when interlocutors attempt to overcome
problems in conveying their meaning, resulting in both additional input and useful feedback
on the learner’s own production” (p. 264). As Garcia Mayo and Alcon Soler (2013) remind
us,  the  process  of  acquisition  is  held  to  be  facilitated  by  a  learner’s  participation  in
meaningful conversational interaction with other learners or native speakers, featuring several
instances  of conversational  and  linguistic adjustments  where  “these  adjustments  were  not
unique  to  learner  discourse  but  were  significantly  more  numerous  in  instructional
 

environments” (p. 221).  Long (1996) had already suggested that those adjustments might
play  a  role  in  interlocutors’  provision  of  comprehensible  input.  In  fact,  enhancing
comprehensible input, encouraging modified output, and exposure to corrective feedback are
very often summarized as the main reasons why interaction and negotiation of meaning can
lead to learning (García Mayo & Lazaro Ibarrola, 2015). This is compatible with the claims
that  both  positive  and  negative  evidence  are  necessary  for  interlanguage  development,
particularly when the negative evidence is in form of corrective feedback provided within an
interactionist  framework  and  serving  meaning  and  meaningful  communication  (Leeman,
2007). The next section elaborates on these aspects, reports on the body of empirical works,
and  discusses  the  dearth  of  research  on  child  interaction  in  SLA  to  demonstrate  how  needy
this area of research is, especially in an EFL context.
 
Background
The field of SLA has been through a vivid history since the proposal of input hypothesis by
Krashen  in  1987.  The  input  hypothesis  along  with  Long’s  emphasis  on  the  role  of
engagement in conversation with native speakers in the acquisition of a new language (Long,
1983) opened a new pathway in second language research: the study of interaction. Varonis
and  Gass  (1985)  suggest  that  “the  modified  interactions  found  in  conversations  between
native speakers and non-native speakers are the sine qua non of second language acquisition”
(p.  71).  Long  (1981)  asserts  that  participation  in  conversation  with  native  speakers  made
possible  through  modification  of  interaction,  is  the  necessary  and  sufficient  condition  for
SLA.  
Since its introduction to the field of SLA, the study of interaction has developed a great
number of new dimensions and has evolved vastly  in its nature. Mackey  (2007) reminds us
that the questions addressed in interaction research today are now “qualitatively different”
since it has “carved out an initial area of inquiry” and “it has now come to the point where
researchers are contemplating the scope of the area” (Mackey, 2007, p. 1). As Mackey (2006)
suggests the study of interaction is no longer limited in scope to its beneficial effects since its
usefulness has been already established in the literature. In fact, the nature of the question has
shifted from “does interaction work” to “how it works”. Mackey goes one step further by
arguing that  Interaction Hypothesis is now closer in characteristics to a  theory rather than  a
hypothesis.
  
In  this  regard,  a  number  of  studies  have  been  conducted  to  delve  into  the  role  of
interaction and pinpoint the ways in which it benefits the learners (see Mackey, 2007, pp. 3-6
for a partial list). As a result of such research, interaction is now claimed to enhance learners’
interlanguage  development  based  on  a  number  of  arguments.  First,  engaging  in  interaction
facilitates the process of input enhancement by  giving and receiving feedback.  Long (1996)
believes  that  children  do  not  learn  the  grammar  first  and  subsequently  start  to  talk.  Rather,
children benefit from the interactive nature of communication and the feedback they receive
from more competent speakers (in this  case adults) when speech has  already begun. This is
argued  to  be  similarly  the  case  when  learners  engage  in  face  to  face  interaction,  especially
with  native  speakers,  in  which  learners  may  benefit  from  the  feedback  they  receive.  In  NS-NNS  (non-native  speaker)  interactions  learners  may  implicitly  attend  to  the  correct  form  of
the structure which they use erroneously. This type of feedback is known as “recast” (Lyster
&  Ranta,  1997).  By  receiving  this  feedback,  learners  benefit  from  enhanced  input  which
leads to noticing the intended target form and facilitates the process of acquisition. Second, it
enhances  the  saliency  and  frequency  of  targeted  forms.  According  to  Mackey  (1999),  when
learners  engage  in  interaction,  they  implicitly  and  naturally  face  numerous  instances  of  the
target  form.  This  is  where  “Input  Enhancement”  occurs  as  an  attempt  to  make  a  certain
linguistic  form  salient  to  L2  learners  by  manipulating  characteristics  of  input  (Sharwood
Smith, 1991, 1993) which in turn leads to the noticing of those features. By providing recasts
and consequently increasing the saliency of certain input features, learners ‘notice’ the target
form and SLA processes are facilitated.  
Swain’s (1985, 1995) Output Hypothesis provides yet another argument to support the
role  of  interaction  in  the  acquisition  of  a  second  language.  This  hypothesis  views  output  as
the  manifestation  of  learning  process  whose  benefits  to  learners  are  three-fold:  noticing/
consciousness  raising,  discovery  and  hypothesis  testing,  and  finally  reflection  as  a
metalinguistic function. The language provoked as a result of the aforementioned functions is
referred  to  as  pushed  output  and  helps  learners  revise  their  interlanguage  (Swain,  1995).
Output can also “serve a consciousness-raising function by helping learners to notice gaps in
their  interlanguage”  (Ellis,  2003,  p.  49).  Here,  the  role  of  interaction  is  to  facilitate  the
production of more and more output which can in turn lead to:
  noticing  the  gap  in  one’s  interlanguage  (the  difference  between  a  learner’s  present
competency and the target form),
  noticing the hole (what the learner doesn’t know at all),
 

  negotiation of meaning for more clarification as in LREs (language related episodes)
The role of interaction in SLA seems to have been well established in the literature by
turn  of  the  century.  A  series  of  seminal  studies  including  Lyster  and  Ranta  (1997),  Gass,
Mackey, and Pica (1998), Mackey (1999, 2006, 2007, 2012), Ellis, Basturkmen, and Loewen
(2001,  2002),  Leeman,  (2003),  Ellis,  Loewen,  and  Erlam  (2006)  and  many  others  have
already founded a strong argument in favor of interaction in language learning classes. As far
as  grammatical  development  (as  one  focus  of  the  present  study)  is  concerned,  Mackey
(1999),  for  instance,  asserted  how  important  active  participation  in  interaction  is  in  the
process  of  form  internalization.  Yet,  these  findings  had  to  be  tested  in  a  variety  of
environments  and  with  younger  learners  as  Polio  and  Gass  (1998),  Mackey  and  Oliver
(2002), Mackey and Silver (2005), and many others called for. Perhaps the most relevant to
the discussion of the present paper is the findings from a series of unique studies conducted
by Oliver (1998, 2002, 2009). Though restricted to ESL contexts, Oliver’s pioneering studies
are the quintessence of interactionist framework of research in child language learning.  
Oliver (1998) studied 96 pairs of young children ranging from 8-13 in age by engaging
them  in  conversational  tasks.  Interestingly  enough,  it  was  found  that  ESL  children  were
capable  of  deploying  interactional  strategies  of  meaning  negotiation  including  clarification
requests, repetitions, and confirmation checks, however in a different proportion from that of
their adult  counterparts.  Oliver (1998) concluded that engaging in interaction can lead to as
beneficent  outcomes  in  case  of  younger  learners  and  this  can  be  promoted  by  introducing
tasks  into  such  courses.  In  a  later  study,  Oliver  (2002)  investigated  whether  nativeness  and
proficiency level of interlocutors played a role in the utilization of interaction by young pairs.  
To this end, Oliver created three types of dyads by matching native speakers (NS) and
non-native speakers (NNS) of English and engaged them in conversational tasks. The results
revealed  that  NNS-NNS  pairs  created  the  highest  quantity  of  meaning  negotiation  in  their
interactions  while  the  NS-NS  pairs  engaged  in  the  least  amount  of  interaction.  Following  a
similar  pattern,  level  of  proficiency  significantly  affected  the  amount  of  negotiation  with
higher  proficiency  students  producing  smaller  amounts  of  interaction.  On  the  whole,  Oliver
(2002) concluded that the more native like the interlocutors, the less amount of negotiation of
meaning is produced and found an advantage for lower proficiency NNS-NNS pairs. Finally,
Oliver  (2009)  probed  whether  younger  children  aged  5-7  are  also  capable  of  meaning
negotiations  in  communicative  tasks.  The  results  indicated  that  not  only  did  these  children
negotiate for meaning but also they employed the same type of strategies that the older cohort
 
 
did and provided feedback to their interlocutors. This arsenal of findings on the interactional
behavior of ESL learners posits that adolescents and even younger children are both capable
of and benefit from engaging in interaction and strategic negotiation of meaning.  
Sharing a similar focus with the present study, Mackey and Silver (2005) investigated
the development  in  young  learners’  question  formation  ability  as  a  result  of engaging  in
interaction  with  adult  native  speakers.  For  this  purpose,  26  young  EFL  learners  aged  6-9  in
Singapore were assigned to two groups where the experimental group received feedback on
their question formation mistakes while interacting with adult native speakers and the control
group  engaged  in  mere  interaction  without  such  feedback.  The  results  indicated  that  the
group  receiving  feedback  during  their  interactions  significantly  outperformed  the  control
group and feedback was a determining element in benefitting from interactional exchanges.  
Philip,  Walter,  and  Basturkmen  (2010)  examined  whether  young  French  learners
noticed  the  form  while  engaging  in  peer  interaction  during  role-play  and  discussion  tasks.
Philip  et  al.  (2010)  concluded  that  a  number  of  task  and  social  considerations  affected  the
peers’ attention to forms and their willingness to interact. In  another  study,  Guillén  (2012)
investigated  the  role  of  interaction  in  developing  the  four  language  skills  in  case  of  young
learners. The  teaching  techniques  and  the learning  strategies  observed  in  young  learners’
classrooms  as  well  as  the  evidence-based  practices reported in Guillén’s study can be very
revealing. Findings from this study proved that engaging in conversational interactions with
teachers  actually  had  a  significant  effect  on  children’s  internalization  of  the  forms  and
mastering the four skills.  
Other  studies  have  taken  a  methodological  perspective  and  focused  on  the  mode  of
instruction  following  an  interactionist  framework.  For  instance,  Huang  (2011)  investigated
the  effect  of  content-based language instruction (CBLI) on young EFL learners’  attention,
engagement,  and  eager  volunteering,  as  well  as  classroom  verbal  interaction,  generally
termed  as  motivated  behaviors.  The  results  revealed  that  learners  showed  better  tendency
toward  interactive  and  subject  learning  classes  compared  to  traditional  language-input
classes. Sharing a similar instructional focus, Gunning and Oxford (2014) employed a mixed
method  design  to  check  whether  strategy  instruction  and  use  had  any  significant  effect  on
interactional  success  of  young  learners  of  French  performing  oral  tasks.  The  findings
revealed  that  instruction  was  effective  in  enhancing  strategy  awareness  and  use  in  case  of
young learners and this, in turn, led to more successful interaction and task performance.
 

Lazaro  and  Azpilicueta-Martinez  (2015)  set  out  to  identify  the  interactional  strategies
used  by  a  group  of  sixteen  7-8  year  old  Spanish  EFL  learners  who  engaged  in  a  guessing
game  task.  Utilizing  an  interactionist  framework,  the  study  concluded  that  these  young
children  do  negotiate  for  meaning,  however,  much  less  than  ESL  adult  and  child  learners.
The  findings  also  confirmed  those  of  Oliver  (1998,  2002)  in  that  the  type  and  frequency  of
conversational strategies employed by young EFL learners were different from ESL adult or
child learners. The main point was that children  were able to interact in  English in order to
accomplish  the  task  and  benefit  from  it  in  spite  of  having  a  very  limited  proficiency  in  the
target language.
In  another  study,  Garcia  Mayo  and  Lazarro  Ibarrola  (2015)  investigated  the  role  of
negotiation for meaning in task-based interactions among children in EFL and CLIL (Content
and  Language  Integrated  Learning)  settings.  To  fulfill  the  purpose,  eighty  8-11  year  old
children  participated  in  the  study  and  were  paired  to  form  40  age-  and-proficiency-matched
groups (20 EFL, 20 CLIL). In order to identify the different strategies they used to complete
the  task,  their  oral  production  was  analyzed.  Findings  proved  that  CLIL  learners  negotiated
more and resorted to L1 less frequently than EFL learners. On the other hand, older children
in both contexts showed a tendency to use their mother tongues in the class and to negotiate
less.  In  general,  the  beneficial  effects  of  interaction  reported  earlier  in  ESL  contexts  were
confirmed  for  an  EFL  context.  The  results  also  corroborated  the  general  trend  identified  by
Oliver (2002) in that the amount of interaction reduced with an increase in learners’ proficiency.
The  present  study  aimed  at  continuing  this  line  of  research  by  establishing  the
connection between strategy training and interactional feedback for young EFL learners, the
population  most  often  ignored  in  interaction  research.  In  fact,  although  as  already  asserted,
the  applicability  of  interaction  and  its  beneficial  effects  have  been  largely  supported  in  the
literature of SLA (for some recent evidence see García Mayo & Alcon Soler, 2013; Mackey,
2012; Mackey, Abbuhl, & Gass, 2011; Pica, 2013) it seems that there have been few studies
to investigate the utility of interaction in case of young learners (for some rare exceptions see
García  Mayo  &  Lazaro  Ibarrola,  2015;  Oliver,  1998,  2002,  2009).  As  Mackey  and  Silver
(2005) and Garcia Mayo and Lazaro Ibarrola (2015) remind us, little empirical studies have
been  conducted  to  investigate  the  role  of  interaction  among children while “SLA research
should  not  be  generalized  from  adult  learners  to  children  without  adequate  empirical
evidence” (Mackey & Silver, 2005, p. 243). More importantly, seminal studies such as Oliver
(1998, 2002, 2009) have focused on ESL settings and little evidence has been collected about
 
  
EFL  children  interactional  behavior.  Since  young  learners  are  believed  to  possess  more
limited  cognitive  abilities,  working  memory  capacity,  and  attention  span  (Cochran  et  al.,
1999;  Mackey  &  Silver,  2005;  Philip,  Oliver  &  Mackey,  2008),  it  is  probable  that  children
are  not  as  capable  as  adult  learners  in  providing  and  receiving  interactional  feedback  with
their  peers.  Thus,  this  study  sets  out  to  investigate  the  effects  of  interaction  on  young  EFL
learners and examine whether it is possible to train them in exchanging such feedback.  The
study  aims  to  point  out  if  engaging  young  EFL  learners  in  interaction  with  their  peer  NN
learners  of  English  has  any  significant  effect  on  the  grammatical  development  of  their
interlanguage  and  also  it  tries  to  find  out  if  there  is  any  significant  difference  between
instructed  and  uninstructed  interaction  in  terms  of  grammatical  development  in  the
interlanguage  of  young  learners.  That  is,  the  study  will  investigate  whether  training  young
learners  in  social  strategies  of  giving  and  receiving  corrective  feedback  can  significantly
benefit them in making the most of interaction with peers.
 
Research Question
Does  engaging  in  peer  interaction  with  or  without  prior  training  in  strategies  of  interactive
feedback influence young EFL learners’ ability to form accurate polar questions over time?
 
The Study
Participants
In order to answer the research question earlier put forth by the present study, 60 young adult
English learners in a national language school in Iran were designated as the participants. The
age  of  the  cohort  ranged  between  9  and  14  years  -  late  elementary  school  and  early
adolescence  ages  according  to  the  classification  by  Philp  et  al.  (2008)  -  with  an  average  of
12.13. The participants were all male as a result of a single-sex educational system. Since a
participant  randomization  was  inconceivable  due  to  administrative  constraints,  three  intact
classes  comprising  20  students  each  were  randomly  assigned  to  one  of  the  following
conditions: the Interaction with Training group (IT), the Interaction with No Training group
(IN),  and  the  No  Interaction  or  the  Control  group  (CG).  However,  to  make  up  for  this
shortcoming, all the classes were selected from the same level (the sixth one out of a thirteen-level  system)  representing  roughly  similar  language  proficiency  levels.  To  ensure  the
homogeneity  of  the  groups  prior  to  the  treatment,  one  way  ANOVA  was  used  to  compare
their achievement scores on the last term final exam as well as their performance on the pre-
 

test  of  the  present  study,  which  will  be  discussed  in  more  detail  later.  The  three  groups
displayed  no  significant  differences  in  terms  of  their  general  language  proficiency  as
measured  by  the  final  exam  of  the  previous  term  (F(2,57)=.38  ,  p=n.s.)  and  their  yes/no
question  formation  ability  as  indicated  by  the  results  of  the  pre-test  (F(2,57)=.19  ,  p=n.s.).
This built an argument to support the homogeneity of the three groups to begin with despite
the non-random distribution of the participants.  
 
Testing Instruments
The present study focused on the simple present yes/no question formation as an indicator of
interlanguage  development.  Due  to  the  unavailability  of  a  formerly  validated  test  of  this
particular grammatical ability for young English learners, the researchers were driven to write
their  own  test  with  30  items  encompassing  two  types  of  question  formation  tasks.  The  test
included  15  sentences  to  be  transformed  into  yes/no  questions  and  15  answers  for  the
students to write an appropriate polar question for. The test was initially piloted with a class
of  18  students  at  a  level  similar  to  the  main  participants  and  a  group  of  5  instructors  at  the
same school to locate any possible instances of ambiguity, ambivalence, or infelicity. Based
on  the  feedback  from  these  two  groups,  minor  modifications  were  applied  and  some  items
were  replaced  or  improved.  Particularly,  based  on  item  facility  and  difficulty  indices,  the
items  which  proved  to  be  invariably  easy  or  very  hard  were  either  eliminated  or  modified.
Consequently,  as  a  second  piloting,  the  modified  test  was  administered  to  a  group  of  39
students  from  two  different  proficiency  levels  (beginners  and  higher  level  students).
Comparisons of the results of the test across proficiency levels using a t-test revealed that the
test  was  adequately  functioning  in  discriminating  the  beginners  from  a  high  proficiency
group.
Once  the  first  version  of  the  test  was  ready,  the  items  were  slightly  changed  to  create
two more versions of the same test. For this purpose, the focus of each question was retained
while  the  insignificant  details  such  as  the  subject  or  the  object  of  the  sentence,  adverbs  of
time and place, and alike were substituted. The outcome was three versions of the same test
with 30 items each to be used as the pre-test, immediate post-test, and delayed post-test. To
ensure that all three versions of the test were parallel, they were administered to 58 students
at the same level in three different classes. Approximately one third of the students took one
form of the test while the other fractions  got the  other two versions.  Finally the results of a
one  way  ANOVA  indicated  that  there  was  no  significant  difference  between  the
 
    
performances  of  students  from  a  similar  proficiency  level  on  all  three  forms  of  the  test  and
they were statistically parallel (F(2,55)=.66 , p=n.s.). Furthermore, the reliability of the tests
based on Cronbach alpha estimate of internal consistency was computed to be α1=85, α2=87,
and α3=82 for the pre-test,  immediate  post  test,  and  delayed  post  test  versions  respectively.
Therefore,  all  three  measures  demonstrated  high  reliability  for  the  purpose  of  the  present
study.  These  three  forms  were  randomly  assigned  to  the  pre-test,  immediate  post-test,  and
delayed post-test situations.  
 
Procedure
The  treatment  was  devised  to  take  as  long  as  a  single  term  comprising  approximately  10
weeks during three months. Although the classes originally met two sessions every week, the
treatment was planned to be delivered for about twenty minutes once a week. This reflected
the  prescribed  syllabus  of  the  institute  which  demanded  explicit  focus  on  form  activities
every other session. Being taught by the same instructor (one of the authors) all three classes
followed  exactly  identical  lesson  plans.  To  begin  with,  one  of  the  three  test  versions  was
randomly  administered  as  the  pre-test  to  all  three  classes.  Following  the  pre-test,  the  three
classes were randomly assigned to the IT (interaction with training), IN (interaction with no
training),  and  CG  (control  group).  All  three  groups  were  similarly  presented  with  present
tense  structures  focusing  on  the  auxiliaries  be  and  can,  and  a  number  of  main  verbs.  Along
with  the  teaching  methodology  prevailing  at  this  school,  a  combination  of  inductive  and
deductive  grammar  presentation  techniques  was  employed  followed  by  ample  drilling  with
the newly taught grammar in the context of the previously practiced structures. However, the
IT and IN groups were given an opportunity to engage in peer interaction for twenty minutes
after  each  grammar  lesson  under  the  close  supervision  of  the  instructor  while  the  control
group engaged in other individual activities.  
To  further  distinguish  the  two  treatment  conditions,  the  IT  group  had  received  a  one
hour training session on giving and receiving interactional feedback on the first day of classes
following  a  categorization  of  these  strategies  provided  by  Long  (1983)  and  Lyster  (2004).  
The learners in this group were particularly introduced to recasts, elicitations, metalinguistic
clues,  prompts,  repetitions,  and  clarification  requests.  According  to  Lyster  (2004)
metalinguistic  clues  refer  to  comments,  information,  or  questions  related  to  the
wellformedness  of  the  student’s  utterance,  delivered  during  interactional  activities.
Elicitations include the direct questions asked in the process of interaction. By repetition, we
 

mean repeating the student’s wrong utterance with rising intonation and putting stress on the
erroneous  section.  Clarification  requests  involve  using  some  utterances  to  demonstrate  that
the student’s sentence has either been misunderstood or ill-formed.  The  teacher  made  sure
ample instances of each type of strategy was provided by interacting with individual students
and  strictly  monitoring  student  interactions  during  this  one  hour  workshop.  Figure  1
illustrates  some  examples  of  the  interactions  exchanged  between  the  teacher  and  students
during this training session. These interactions were prompted by using several cue cards and
encouraging the students to ask polar questions about the pictures. On the other hand, the IN
group did not receive such training prior to the study and merely engaged in interactions with
peer NNSs every week.

Upon  the  completion  of  the  term,  another  version  of  the  test  was  administered  to  all
three groups on the day of their end of the term exam as the immediate post-test. Following
that, the classes did not meet for approximately one month due to between-term and Iranian
New Year holidays. Accordingly, the delayed post test was administered after about a month
to  all  60  participants  which  were  by  this  time  assigned  to  new  classes.  Each  participant,
including the ones who were repeating the previous term after failing it, was located in their
new  classes  and  was  asked  to  answer  the  last  test.  The  results  from  all  these  three  test
administrations were submitted to IBM SPSS for statistical analysis.  
 
Analysis
Having administered the three parallel tests prior to, immediately after, and one month after
the treatment, the scores obtained by all three groups were analyzed using a mixed between

 
within  subjects  ANOVA  or  a  SPANOVA  (split  plot  ANOVA),  as  it  is  alternatively  called,
with a 3*3 design. The mixed ANOVA or the repeated measure factorial ANOVA allows for
the simultaneous analysis of variability both between and within subjects by mixing regular
ANOVAs  with  repeated  measure  ANOVAs  (Tabachnik  &  Fidell,  2013).  Furthermore,  the
advantage  of  this  design  is  the  ability  to  locate  any  possible  interaction  effects  between  the
independent variables (IVs), here being the type of interaction and time. Since each of these
two  independent  variables  or  factors  had  three  levels,  an  analysis  with  a  3*3  design  was
produced.  The  levels  of  the  between  subjects  IV  (type  of  interaction)  included  interaction
with training, interaction with no training, and no interaction. On the other hand, the levels
of  the  within  subjects  IV  included  the  three  time  intervals  (pre-test,  post-test,  and  delayed
post test). Finally, the dependant variable (DV) was the scores of the participants on the polar
question formation tests. A summary of the descriptive analysis of the group performances on
each of the testing occasions is provided in table 1 below.

As table 1 indicates, all three groups started the term with very similar means on their
question formation tests (𝑋̅= 8.80, 9.40, and 8.85) and showed some improvement during the
treatment time before the first post test. However, by the time of the immediate post test or
the end of the term, the three groups had maintained a distance while the mean performances
of  the  IT  and  IN  groups  were  ahead  of  the  control  group  (𝑋2̅̅̅̅𝐼𝑇=22.35,  𝑋2̅̅̅̅𝐼𝑁=20.00,
𝑋2̅̅̅̅𝐶𝐺=17.35).  The  performance  of  all  three  groups  proved  to  be  poorer  on  the  delayed  post

test  while  the  IT  group  demonstrated  the  mildest  drop  in  mean  (𝑋3̅̅̅̅𝐼𝑇=21.30,  𝑋3̅̅̅̅𝐼𝑁=14.55,
𝑋3̅̅̅̅𝐶𝐺=13.30). In order to establish the significance of the differences, these changes had to be  statistically analyzed.   
The  first  step  was  to  check  for  the  assumptions  of  the  ANOVA  in  general  and  those
pertaining to mixed model ANOVAs in particular. The normality of the scores on all testing
occasions  was  checked  using  the  explore  option  of  SPSS  and  examining  the  skewness  and
kurtosis  values.  Next,  Levene’s  Test  of  Equality  of  Error  Variances  and  Box’s  Test  of
Equality of Covariance Matrices were checked as two important assumptions of SPANOVA.
Since both tests were non-significant, neither of the two assumptions had been violated and
we  were  safe  to  proceed.  However,  as  a  common  occurrence  in  factorial  ANOVA
procedures, the assumption of sphericity was violated based on Mauchly’s Test of Sphericity.
Therefore,  instead  of  referring  to  the  univariate  statistics  results,  the  multivariate  statistics
table was consulted as it does not require the assumption of sphericity (Pallant, 2013). Both
main  effect  (the  individual  effects  of  the  IVs)  and  interaction  effect  (the  effect  of  the  IVs
combined)  were  checked  in  the  Multivariate  statistics  table,  a  summary  of  which  is
reproduced here as table 2.

a. Design: Intercept + group  
Within Subjects Design: Time
b. Exact statistic
c. The statistic is an upper bound on F that yields a lower bound on the significance level.

First,  the  interaction  effect  in  the  time*group  row  was  checked.  Although  all
multivariate  tests  produced  by  SPSS  syntax  tend  to  yield  very  similar  results,  Wilks’

Lambada  is  reported  here  as  it  is  common  practice.  According  to  table  2,  there  was  a
significant effect for the interaction between the two independent variables on the dependent
variable  with  a  large  effect  size;  F(4,112)=17.31,  p<.001,  𝜂2=.382.  This  indicates  that  our
grouping criteria or the treatment conditions did have an effect on participants’ polar question
formation  ability,  however,  through  the  levels  of  the  other  IV  which  was  time.  That  is,  the
question formation ability of the learners was differentially affected by the type of interaction
they  engaged  in  according  to  different  time  periods  and  the  change  for  the  three  different
groups  was  not  similar  over  time.  When  the  interaction  effect  is  significant  in  factorial
ANOVA,  the  interpretation  of  the  main  effects  proves  to  be  rather  tricky  and  should  be
treated  with  caution  as  it  will  be  less  than  revealing  (Pallant,  2013).  However,  as  table  2
indicates,  the  main  effect  of  the  IV  time  was  also  significant  with  a  very  large  effect  size;
F(2,56)=286.88, p<.001, 𝜂2=.911.  
As  suggested  by  Pallant  (2013),  since  the  interaction  effect  had  been  significant,  the
profile  plot  of  the  analysis  was  consulted  to  make  an  interpretation  of  the  results  possible.
This plot is reproduced below as Figure 2. 

As figure 2 and the descriptive statistics presented earlier in Table 1 suggest, while the
mean question formation ability of the three groups did not show any significant difference to
begin  with  (𝑋1̅̅̅̅𝐼𝑇=  8.80,  𝑋1̅̅̅̅𝐼𝑁=9.40,  and  𝑋1̅̅̅̅𝐶𝐺=8.85),  all  three  groups  demonstrated
considerable  changes  from  the  time  of  the  pre-test  to  the  first  post  test  (𝑋2̅̅̅̅𝐼𝑇=22.35,
𝑋2̅̅̅̅𝐼𝑁=20, and 𝑋2̅̅̅̅𝐶𝐺=17.35). In fact, all three groups dramatically improved in their ability to
make  polar  questions  after  the  treatment.  However,  after  a  one-month  interval,  their
 

performances  on  the  delayed  post  test  deteriorated.  This  decrease  was  obviously  more
substantial for the IN (interaction with no training) and CG (no interaction) groups and the IT
(interaction  with  training)  experienced  a  milder  decrease  (𝑋3̅̅̅̅𝐼𝑇=21.30,  𝑋3̅̅̅̅𝐼𝑁=14.55,  and
𝑋3̅̅̅̅𝐶𝐺=13.30). Although these patterns are observable from the profile plot in Figure 2 above,
post  hoc  tests  are  required  to  establish  the  statistical  significance  of  the  differences
demonstrated  by  the  means  of  the  three  groups.  For  this  purpose,  a  Bonferroni  post  hoc
analysis  was  requested  for  the  grouping  variable  in  the  SPSS.  The  results  are  presented  in
Table 3 below.

Based on observed means.
 The error term is Mean Square(Error) = 20.462.
*. The mean difference is significant at the .05 level.

As  Table  3  suggests,  Bonferroni  pos  hoc  analysis  indicated  that  the  only  significant
grouping  effect  occurred  between  the  two  groups  IT  and  CG  (p<.05).  Therefore,  it  can  be
concluded that the only group which performed significantly different from the control group
over  time  was  the  Interaction  with  Training  group.  The  implications  of  these  results  are
discussed more fully in the next section.
 
Discussion, Conclusion, and Implications
All  three  groups  of  the  study  showed  development  in  their  ability  to  form  yes/no  questions
immediately after the treatment period and differentially deteriorated in this regard after a one
month gap. This observation sounds rather plausible as long as the  immediate post test was
given  on  the  day  of  their  final  exam,  when  the  students  are  expected  to  show  the  greatest
amount  of  preparation.  Furthermore,  the  first  post  test  was  administered  immediately  after
 
three  months  of  class  practice  while  the  delayed  post  test  was  given  after  a  one  month
interval during which the students did not attend their classes and were not formally required
to  practice  English.  However,  in  order  to  make  claims  about  these  findings,  statistical
analyses were run as explained earlier in this text.
The  results  of  the  mixed  between-within  subjects  analysis  of  variance  (SPANOVA),
presented  in  the  previous  section  revealed  that  the  type  of  peer  interaction  that  young  EFL
learners  engaged  in  as  further  practice  after  a  grammar  lesson  had  an  effect  on  their  polar
question  formation  ability  over  time.  More  precisely,  as  suggested  by  the  analysis  of  the
profile  plot  and  post  hoc  tests,  although  all  groups  similarly  improved  in  their  question
formation ability during the treatment period and slightly regressed by the time of the delayed
post  test,  the  only  group  which  significantly  sustained  its  development  as  compared  to  the
other  groups  was  the  interaction  with  training  group.  In  fact,  while  engaging  in  interaction
with peers for both treatment groups did have an effect in attaining better scores on post tests,
only  the  students  who  were  trained  in  providing  and  receiving  interactional  feedback  could
significantly  outperform  the  control  group  on  the  immediate  post  test  and  maintain  their
interlanguage development over time until the delayed post test.  
These findings suggest that young learners may benefit from interaction with their peer
learners  only  if  it  is  accompanied  by  appropriate  training  in  dealing  with  interactional
feedback  and  close  mentoring  of  the  teacher.  This  is  particularly  true  when  it  comes  to  the
long term effects of peer interaction while the differences may not be of significant value in
the  short  run.  These  findings  tend  to  corroborate  the  results  from  a  number  of  studies
including  Philp  et  al.  (2010),  Mackey,  (1999),  and  Huang  (2011)  in  that  a  tendency  toward
interacting  in  the  class  among  learners  and  engaging  in  interactive  tasks  will  benefit  the
students and will also facilitate the learning process. Particularly, since teachers are expected
to be familiar with their learners’ cognitive capabilities, if this participation is accompanied
by the instructor’s close monitoring and feedback, the vividness of its effect will be more
dominant.  The  results  are  also  in  concordance with Mackey and Silver’s (2005) findings as
feedback  was  found  to  be  a  determinant  of  successful  grammatical  development  based  on
interactional  exchanges.  Also,  Gunning  and  Oxford’s  (2014)  claims  about  the  effects  of
strategy training on the quality of young learners’ peer interactions were corroborated and it
was  shown  how  important  such  training  can  be.  The  outcomes  of  the  present  study  also
confirmed Oliver’s (1998, 2002, 2009) findings about ESL learners in case of EFL children
in  that  they  were  found  to  be  benefitting  from  interactional  exchanges.  However,  it  also
 

challenged  these  earlier  results  by  indicating  that  long-term  improvements  in  child
interlanguage  can  occur  only  when  the  interlocutors  are  trained  and  well-rehearsed  in
strategies  for  negotiation  of  meaning  and  receiving  and  providing  feedback.  Therefore,  at
least  in  case  of  younger  learners,  sustained  grammatical  development  results  as  long  as
interaction is informed, meaningful, and orchestrated.
There are a number of pronounced implications for these results directly  applicable to
young  adult  classes  in  EFL  contexts.  The  implications  target  both  language  teachers  and
curriculum developers, especially in case of young learners’ language education. This study
showed  that  despite  the  doubts  cast  on  the  efficiency  of  NNS-NNS  interactions  in  EFL
classes, specifically in  case of  younger learners,  (Gunning & Oxford, 2014; Butler  &  Zeng,
2014;  Lazaro  &  Azpilicuteta-Martinez,  2015;  Garcia  Mayo  &  Lazarro  Ibarrola,  2015)
integration  of  such  interactive  activities  into  routine  classroom  procedures  is  worthwhile  as
long  as  the  interaction  is  guided  by  the  teacher.  In  order  to  maintain  the  interlanguage
development of young learners over time, it is recommendable to encourage meaningful peer
interaction  among  them  once  they  are  introduced  to  the  feedback  strategies  required  to
facilitate conversational exchanges and interlanguage development through such processes as
noticing.  This  is  especially  important  to  all  EFL  contexts  including  Iran  where  teachers  are
very often accused of running one-directional, teacher-fronted classes with little attention to
the  benefits  of  interaction  as  a  result  of  an  allegedly  lower  appreciation  for  individual
learning and interpersonal interaction in such non-western cultures (Mackey & Silver, 2005),
a  stereotype  which  needs  to  be  challenged  and  changed.  As  local  studies  including
Keivanpanah,  Alavi  and  Sepehrinia  (2012)  showed,  while  Iranian  EFL  teachers  insisted  to
perform most of the correction in the class, their learners preferred to maintain self and peer
correction  and  engage  in  interactional  activities  rather  than  being  corrected  by  their
instructors.  Therefore,  based  on  the  evidence  from  this  study  as  situated  within  the  body  of
literature, EFL teachers of young classes are suggested to trust the interactive capabilities of
their  students  and  give  a  chance  to  these  potentialities  to  burgeon  and  grow.  With  little
strategy  training  and  monitoring,  peer  interaction  can  significantly  enhance  learners’
internalization and long-term retention of language forms even in case of younger learners.
As to the limitations and delimitations of the present study, it must be reminded that the
development of learners’ interlanguage was limited to grammatical competence and narrowly
operationalized as polar question formation which can hardly represent all intricate aspects of
young learners’ SLA. Furthermore,  the  participants  of  the  present  study  were  all  male  and
 

came from intact classes of 20. Therefore, it is recommendable that in possible replications in
future,  larger  populations  of  both  male  and  female  learners,  randomly  assigned  to  various
treatment conditions, be targeted. It is also suggested that interactionist researchers interested
in the promising area of  young learners  research  continue this trend by focusing on a wider
range  of  grammatical  and  linguistic  features  including  phonological,  lexical,  and  pragmatic
exchanges. For this purpose, standardized tests now available for younger learners, including
TOEFL  Junior®  and  Cambridge  English  Placement  Test  for  Young  Learners,  can  be
employed to more fully grasp the development of their interlanguage.