Learning of relative clauses by L3 learners of English

Introduction
The  spread  of  English  as  a  lingua  franca
throughout  the  world  has  promoted  the
acquisition of the English language not only
as  a  second  language,  but  also  as  a  third
language  (Cenoz  &  Jessner,  2000).  For
instance,  English  is  the  third  language  for  a
large number of speakers in Asia. Therefore,
third  language  acquisition  (TLA)  is  a  very
common  phenomenon  all  over  the  world
(Cenoz,  2008).  Generally  speaking,  TLA
means  "the  acquisition  of  a  non-native
language  by  learners  who  have  previously
acquired  or  are  acquiring  two  other
languages  (Cenoz,  2003,  p.71).  TLA  is  a
relatively new area of research in the field of
second  language  acquisition  (SLA),  yet  it
has recently become the focal point of many
studies  (Molnár,  2008)  as  research  on  TLA
is  scarce    in  comparison  to  that  already
carried  out  on  SLA,  and  first  language
acquisition (FLA). However, the young field
of  TLA  research  has  experienced  a  number
of  important  developments  in  recent  years
(e.g. Ringbom 1987; Edwards, 1994; Cenoz
&  Genesee,  1998;  Cenoz  &  Jessner,  2000;
Cenoz,  2000,  2009).  Results  have  indicated
that,  despite  sharing  many  characteristics,
 
the  acquisition  of  an  L3  is  qualitatively
different  from  that  of  an  L1  and  L2,  since
the  L3  learner  has  already  acquired  one  L2
(up to a certain  level) in addition to  an  L1,
and  this  knowledge  plays  a  role  in  the
acquisition  of  additional  languages  (Cenoz
& Jessner, 2000). Furthermore, according to
researchers,  it  is  believed  that  different
variables  present  greater  diversity  and
complexity  when  more  than  two  languages
are involved. This complexity is the result of
the  interaction  between  various  factors:
linguistic,  social,  and  individual  (Cenoz  &
Jessner,  2000).  The  following  section
explains  the  basic  differences  between  SLA
and TLA.  
 
Basic differences between SLA and TLA
According to researchers who actively work
on  TLA  (Cenoz,  2001),  important
differences  exist  between  second  language
and  additional  languages.  Cenoz  (2000)
states  that  the  main  differences  between
SLA  and  TLA  are:  (a)  the  order  in  which
languages are learned, (b) the sociolinguistic
factors,  and  (c)  the  psycholinguistic
processes involved.
 
Acquisition orders
In  SLA,  when  two  languages  are  involved,
we  only  have  two  possible  acquisition
orders: either the L2 is acquired after L1, or
the  two  languages  are  learned
simultaneously.  In  the  case  of  TLA  there  is
greater  diversity  and  there  are  at  least  four
possible  acquisition  orders:  (a)the  three
languages  can  be  acquired  consecutively
(L1→L2→L3),  or  (b)the  two  languages
(L2/L3)  can  be  acquired  simultaneously
after  the  L1  has  been  acquired  (L1→Lx/
Ly),  or(c)the  two  languages  (L1/L2)  can  be
acquired  simultaneously  before  the  L3  is
acquired  (Lx/Ly→L3),  or  (d)  the  three
languages  can  be  acquired  simultaneously
(Lx/Ly/ Lz) (Cenoz, 2000).
 
Sociolinguistic factors
The other difference between SLA and TLA
refers  to  a  set  of  contextual  and  linguistic
factors  (Cenoz,  2000).  They  are  subdivided
into:
 
Context of acquisition
One  variable  which  increases  this
complexity  is  'the  context  of  acquisition',
because more than two language acquisition
situations  are  involved.  L2  acquisition  can
take  place  formally,  naturally  or  by  a
combination  of  both.  But  in  TLA  this
situation is more complex than SLA (Cenoz,
2000).
 
Linguistic typology
Another  variable  is  'linguistic  typology'  or
'the  type  of  languages'  (Cenoz,  2000)
involved  in  TLA  which  can  present  an
important  variation  since  "languages
typologically  closer  to  the  target  language
may facilitate its acquisition or favour code-mixing procedures" (Jordà, 2005, p. 19).  
 
Sociocultural status  
The other variable is 'the sociocultural status
of  the  languages'  or  'their  ethnolinguistic
vitality' (Cenoz, 2000) which present greater
diversity.  This  situation  is  seen  in  diglossic
societies "where the L2 is used in the media,
for  educational  purposes  and  the  like,  while
members of these societies resort to their L1
and  L3  in  their  everyday  conversations  (at
work,  with  their  families  and  the  like)"
(Jordà, 2005, p. 20).  
 
Psycholinguistic processes
The  third  factor  influencing  TLA  refers  to
the  'psycholinguistic  processes'  involved
(Cenoz,  2000).  Indeed,  the  acquisition  of
more  than  two  languages  complicates  the
cognitive  and  linguistic  processes  involved.
In  sum,  TLA  presents  more  diversity  and
complexity  than  SLA  resulting  in  situations
which are unique in language acquisition.
 
Bilingualism
Becoming  bilingual  “is  a  way  of  life.  Your
whole  person  is  affected  as  you  struggle  to
reach  beyond  the  confines  of  your  first
language  and  into  a  new  language,  a  new
culture, a new way of thinking, feeling, and
acting" (Brown, 1994, p. 1).  
 
The  narrowest  definition  was  perhaps
suggested  by  Bloomfield  who  defined  a
bilingual  person  as  an  individual  who  has
"native-like  control  of  two  or  more
languages"  (Bloomfield,  as  cited  in  Butler,
2013,  p.  111).  In  other  words,  according  to
Bloomfield, having an extensive vocabulary
as  well  as  perfect  skills  in  reading,  writing,
listening  and  speaking  is  a  prerequisite  for
being bilingual.
 
On  the  other  hand,  Haugen  defined
bilinguals  "as  individuals  who  are  fluent  in
one language but who can produce complete
meaningful utterances in the other language"
(cited in Butler, 2013, p. 111). According to
Haugen,  native-like  proficiency  is  not  a
prerequisite condition for being bilingual.
 
Currently,  many  researchers  employ  a
broader view of bilinguals (e.g. Macnamara,
1967;  Hakuta,  1986;  Valdés  &  Figueroa,
1994;  Mohanty  &  Perregaux,  all  cited  in
Butler,  2013).  Similarly,  in  this  study,  the
broader  notion  of  ‘bilinguals’  was  adopted
as  individuals  or  groups  of  people  "who
obtain  communicative  skills,  with  various
degrees of proficiency, in oral and/or written
forms,  in  order  to  interact  with  speakers  of
one  or  more  languages  in  a  given  society"
(Bhatia & Ritchie, 2004, p. 115).  
 
Studies conducted in Iran
In  Iran,  a  multicultural  and  multilingual
society  (Khadivi  &  Kalantari,  2010;
Kalantari,  2012),  the  studies  on  the
relationship  between  bilingualism  and  the
acquisition  of  L3  have  produced
contradictory  results.  On  the  one  hand,
Keshavarz  and  Astaneh  (2004),
Modirkhamene  (2008),  Farhadian  et  al.
(2010),  Kassaian  and  Esmae’li  (2011)  and
Saeidi and Mazoochi (2013), to name only a
few,  concluded  that  the  third  language
learners outperformed their second language
counterparts.  
 
For example, Keshavarz and Astaneh (2004)
investigated  the  relationship  between
bilinguality of second language learners and
their  vocabulary  achievement  in  the  target
language.  They  compared  three  groups  of
female  students,  Turkish–Persian  bilinguals
studying  only  one  language  (Persian)
academically  in  Tabriz,  Armenian–Persian
bilinguals  studying  both  languages
academically  in  Tehran,  and  Persian
monolinguals.  The  authors  used  the
Controlled  Productive  Ability  Test  (CPAT)
to  measure  the  participants'  vocabulary
knowledge.  The  results  of  the  data  analyses
showed that:
 
Native  speakers  of  Turkish  and
Armenian who speak Persian as their
second  language  performed  better  in
the  English  vocabulary  test  than  the
Persian  monolingual  learners  of
English. This can be attributed to the
positive  effect  of  the  subjects'
bilinguality  on  their  third  language
vocabulary  achievement.  (Keshavarz
& Astaneh, 2004, p. 295)
 
By  applying  Nation's  Vocabulary  Levels
Test  to  measure  the  breadth  of  vocabulary
knowledge and the Burt Word Reading Test
to  measure  the  participants'  word  reading
skill,  Kassaian  and  Esmae’li  (2011)
investigated    the    relationship    between  
bilingualism     and    the    breadth    of  
vocabulary  knowledge   and  word  reading  
skills. Two groups of female students at two
different  pre-university  centers  were

compared;  Armenian-Persian  bilinguals,
learning  Armenian  as  their  L1  and  Persian
as  their  L2,  and  Persian  monolinguals.  The
results  of  the    data    analyses    revealed    that  
"bilinguality    is    highly    correlated    with  
breadth    of    vocabulary    knowledge    and
reading  skill.  In  other  words,  bilingual
participants  have  larger  size  of  vocabulary
knowledge  and  they  enjoy  better  word
reading skill" (p. 966).
 
On  the  other  hand,  Talebinezhad  and
Mehrabi  (2007),  Gooniband  Shooshtari
(2009),  Amerian  and  Maghsoudi  (2009),
Saffarian,  Gorjian  and  NejadFazel  (2013),
and  Khany  and  Bazyar  (2014),  among
others, found that bilingual and monolingual
learners  did  not  perform  significantly
differently from each other.  
 
Employing  an  experimental,  within-group
design,  Talebinezhad  and  Mehrabi  (2007)
studied  three  languages,  Persian  as  the
mother tongue of the participants, English as
the  second  language  they  learned  at  high
school,  and  German  as  their  third  language
to "explore the manner in which word forms
are  connected  to  the  other  words  in  the
multilingual minds…check the claim  made
by  many  first  language  researchers  that
multilingual  people's  first  languages  play  a
privileged  role  in  the  acquisition  of
subsequent  languages"  (p.1).  A  translation
deletion  task,  consisting  of  four  parts,
carried out in a separate session. The authors
concluded that "the multilingual participants
were  processing  the  second  and  third
language  at  the  same  speed…and  first
language  does  not  seem  to  have  a
determining  role  in  the  development  of  a
third language" (pp.1-10-11).  
 
Gooniband  Shooshtari  (2009)  conducted  a
comparative  study  in  light  of  the
syntactically-based  generative  models  of
SLA,  namely,  Full  Access  Full  Transfer
(FAFT)  and  the  Failed  Functional  Feature
Hypothesis  (FFFH)  in  order  to  investigate
the acquisition of two syntactic properties of
head and operator movements in English by
L2  and  L3  learners  within  Universal
Grammar  (UG)  framework.  The  study  was
undertaken  among  Arabic-Persian  bilingual
and Persian monolingual learners of English
in Khuzestan. Findings indicated that:
 
the  bilingual  and  monolingual
learners did not perform significantly
different  from  each  other  with
respect  to  the  resetting  of  the  two
parameters  of  head  and  operator
movements….the  findings  of  the
study  with  respect  to  language
transfer  in  L3A  give  rise  to  the
conclusion  that  the  source  of  cross-linguistic  influence  in  L3A  is
probably  more  of  the  learners'  L2
than their L1, evidence in support of
the  prediction  of  FTFA  hypothesis
which  argues  for  the  availability  of
the  all  sources  available  to  language
learner. (pp. 136-138)
 
Moreover,  Bahrainy  (2007)  investigated
both  lexical  and  syntactic  knowledge  and
concluded  that  monolinguals  outperformed
bilinguals.  A  grammatically  judgment  test
along  with  a  correction  task  were  used  to
examine  two  structures:  preposition-stranding  and  pied-piping.  The  results
revealed  that  monolinguals  outperformed
bilinguals  in  both  vocabulary  and  syntax.
The author believed that:
 
Perhaps  the  most  important  reason
for  such  unexpected  finding  is  that
Turkish-Persian  subjects  had
learned  their  L1  only  orally  in  a
natural setting. They did not receive
schooling  in  Turkish  and  their
academic language was Persian, the
native  language  of  the  majority
 
linguistic group. So it can be argued
that  Persian  is  the  more  dominant
language  among  the  bilinguals.
(p.17)     
 
Theoretical framework
Hypotheses concerning L3 syntax
Bardel  (cited  in  Falk  &  Bardel,  2010),
Leung  (cited  in  Falk  &  Bardel,  2010),
Sjögren (cited in Falk & Bardel, 2010),  and
Vinnitskaya  et  al.  (cited  in  Falk  &  Bardel,
2010)  investigated  L3  syntax  and  transfer
from  both  L1  and  L2  and  the  results
revealed  that  both  L1  and  L2  play  a
significant role in this respect. These studies
have  been  further  discussed  and  the  results
have been meticulously re-analysed (Falk &
Bardel,  2010)  resulting  in  three
syntactically-based  generative  models  of
TLA:  the  Cumulative  Enhancement  Model
(CEM; Flynn et al., 2004; Flynn, 2009), the
L2 Status Factor Hypothesis (LSFH; Bardel
&  Falk,  2007;  Falk  &  Bardel,  2010,  2011;
Falk,  submitted)  and  the  Typological
Primacy Model (TPM; Rothman, 2010).
 
The  CEM  was  suggested  by  Flynn  et  al.
(2004).  This  model  made  use  of  the
Vinnitskaya  et  al.’s  (2003)  study.  The
authors  investigated  the  acquisition  of  three
types  of  English  restrictive  relative  clauses
by  comparing  three  groups  of  English
learners,  (a)  Kazakh  native  speakers  whose
L2  is  Russian,  (b)  Spanish  native  speakers
and (c) Japanese native speakers. Kazakh (a
Turkish  language),  like  Japanese,  is  a  head-final, left-branching language while English,
Spanish  and  Russian  are  head-initial,  right-branching  languages.  They  wanted  to
investigate  whether  "the  role  of  a  first
language  is  privileged  over  the  role  of  a
second  language  in  the  development  of  a
third  language  or  it  is  possible  that  all
languages  known  can  play  a  role  in
subsequent  language  acquisition"
(Vinnitskaya  et  al.,  2003,  p.  2).  The  results
showed  that  the  Kazakh  speakers  behaved
like  the  Spanish  speakers  and  contrasted
strongly  with  the  Japanese  speakers.
Therefore,  the  authors  concluded  that  "no
one  language  maintains  a  privileged  role
with respect to next or subsequent language
learning"  (Vinnitskaya  et  al.,  2003,  p.  2).
According  to  this  hypothesis,  language
acquisition  can  be  said  to  be  cumulative  as
the learner can fall back on not only one, but
all,  previously  acquired  languages  in  L3
acquisition (Vinnitskaya et al., 2003).
 
The LSFH is based on the properties shared
by languages learned in the classroom. This
hypothesis  originates  from  Williams  and
Hammarberg's  (1998)  study  on  third
language acquisition of content and function
words, but it is proposed by Bardel and Falk
(2007,  2011).  In  two  studies,  Bardel  and
Falk  (2007,  2011)  compared  learners  with
different  L1s  and  L2s.  In  one  study,  the
participants  were  in  the  initial  state  of  L3
acquisition,  in  another  one  they  were  at  an
intermediate  level.  Their  data  support  the
hypothesis  that  "the  L2  factor  is  stronger
than  the  typology  factor  in  L3  acquisition."
In  other  words,  "in  L3  acquisition,  the  L2
acts  like  a  filter,  making  the  L1
inaccessible" (Bardel & Falk, 2007, p. 480).
In  other  words,  The  LSFH  generally
suggests  that,  in  the  acquisition  of  an  L3,  a
general  tendency  is  to  activate  a  previously
learned  (second)  language  rather  than  to
activate the L1 (Bardel & Falk, 2007, 2011).
 
According  to  the  TPM  by  Rothman  (2010),
(psycho)typology  determines  whether  the
L1  or  the  L2  will  be  transferred  in  TLA.
This  model  is  a  modification  to  the  CEM
based  on  the  suggestions  of  (psycho)-typological factors by Rothman and Cabrelli
(2009).  Rothman  (2010)  investigated  the
acquisition  of  the  syntactic  and  semantic
properties  of  the  Romance  DPs  in  two
groups, Italian native L2 learners of English
 
learning  Spanish  as  an  L3  and  English
native  L2  learners  of  L2  Spanish  learning
Portuguese  as  an  L3  in  order  to  determine  
whether    or    not    'L2    status    factor'  or
'linguistic  typology'  between  the languages
is    the    most  explanatory    account    of  
transfer  in  TLA. The data revealed that:
 
Neither  the  order  of  acquisition  nor
the  L1/L2  status  effects  determine
the  source  of  transfer.  The  L3
Spanish  learners  transfer  from  their
L1  Italian  while  the  L3  Brazilian
Portuguese  learners  rely  on  their  L2
Spanish,  not  their  L1  English.  Both
transfer  sources  are  (psycho)-typologically  similar  to  the  L3,
unlike  English,  a  non-Romance
language (Hermas, 2010, p.4).
 
The  present  study  builds  on  the  notion  of
Flynn  et  al.'s  (2004)  CEM,  Bardel  and
Falk’s (2007) LSFH and Rothman's (2010)
TPM for syntactic learning. It compares and
contrasts  the  status  of  relative  clauses  in
Persian  native  speakers  of  L2  English  and
Azeri  native  speakers  of  L2  Persian  and  L3
English.  
 
Research questions  
The  following  research  questions  were
formulated:
 
1)  Does  bilinguality  affect  the  non-native  comprehension  of  English  L3
relative  clauses  by  Azeri-Persian
bilingual learners?
2)  Does  bilinguality  affect  the  non-native  production  of  English  L3
relative  clauses  by  Azeri-Persian
bilingual learners?
 
Context of the study
Persian  is  an  Indo-European  language-a
southwestern  Iranian  language  from  the
Indo-Iranian  branch.  Persian  is  Iran’s
official  language,  the  language  of  education
and instruction.  
 
Azerbaijani  or  Azeri  is  a  member  of  the
Oghuz  branch  of  the  Turkic  languages.  In
Iran,  Azeri  uses  the  Perso-Arabic  script,
although  the  spelling  and  orthography  are
not yet standardized.  
 
In  Iran,  English  is  regarded  as  an  academic
subject  in  the  formal  context  of  classrooms.
In  some  parts  of  Iran,  where  learners  are
members  of  linguistic  backgrounds  like
Arabs,  Turks,  and  Kurds,  English  is
regarded  as  an  L3,  which  is  acquired  after
the acquisition of L1 and L2.
 
Method  
This  study  adopted  an  ex  post  facto  design
to  see  if  English  L3  learners’  distinct
language  background  causes  them  to
develop  interlanguage  patterns  which  are
different  or  similar  to  those  of  monolingual
learners of English.
 
Participants
A  total  of  200  female  high  school  students
studying at the second grade were randomly
selected  from  two  educational  districts  of
Tabriz  and  Shiraz.  The  bilingual  group
(Azeri–Persian) were studying English as an
L3  academically  in  Tabriz  (an  Azeri-speaking  city  in  Iran),  and  the  monolingual
participants (Persian) were studying English
as  an  L2  in  Shiraz.  The  number  of  students
in the bilingual group was 100, between the
ages  of  15  and  18  years  (M=15.68,
SD=.62),  and  in  the  monolingual  group  the
number was 100, between the ages of 15 and
17 years (M=15.71, SD=.61).  
 
All  the  participants  had  Persian  as  the
language  of  instruction,  and  they  also
studied  English  as  a  school  subject.
Although  Persian  is  the  language  of
 
instruction  in  Iran,  Azeri  is  the  language  at
the community level in a city like Tabriz.
 
The  participants  in  both  groups  were
homogeneous  in  terms  of  the  educational
context:  both  groups  attended  public  high
schools;  they  were  taught  using  the  same
material; i.e. the textbooks and methodology
for  teaching  English  as  a  foreign  language
were  the  same  (sanctioned  by  the  Ministry
of  Education).  Both  groups  had  the  same
number  of  hours  of  instruction,  which  was
one  session  (one  hour  and  thirty  minutes)
every  week.  They  were  homogeneous  in
terms  of  their  English  proficiency  level
(with  respect  to  the  result  of  the  OPT),  sex,
and age too.  
 
Target structure
The  target  structure  selected  for  the  study
was  the  English  relative  clause  (RC).  In
Iranian  high  schools,  the  formation  of  RCs
appears  as  a  grammar  item  in  grade  two.
The  third  unit  in  the  students'  English  book
highlights this structure.  
 
RC  is  "a  clause  which  modifies  a  noun  or
noun  phrase"  (Richards  &  Schmidt,  2010,
p.494)  and  is  typically  introduced  by  a
relative pronoun/adverb such as that, which,
who,  when,  or  where.  Celce-Murcia  and
Larsen-Freeman  define  a  RC  as  "a  type  of
complex  post-nominal  adjectival  modifier
that  is  used  in  both  written  and  spoken
English"  (cited  in  Abdolmanafi,  2012,
p.196).  The  literature  on  the  acquisition  of
RCs  has  concentrated  on  four  particular
types: (a) subject-subject (SS) relatives (The
boy  who  speaks  English  is  my  cousin),  (b)
subject-object  (SO)  relatives  (The  woman
whom  you  met  is  my  mother),  (c)  object-subject  (OS)  relatives  (I  know  the  boy  who
speaks  English),  (d)  object-object  (OO)
relatives  (I  read  the  book  that  my  teacher
mentioned).
 
RC,  as  one  of  the  subordinate  clauses,  has
attracted  the  attention  of  SLA  researchers
and  educators  "due  to  its  complex
structures"  (Gass  &  Selinker,  as  cited  in
Abdolmanafi & Rahmani, 2012). Moreover,
RC  which  is  considered  as  "a  universal
linguistic  phenomenon  in  languages  of  the
world, have unique syntactic properties, and
are  frequent  in  everyday  use  of  language"
(Izumi,  as  cited  in  Marefat  &  Rahmany,
2009,  p.  1)  has  been  a  very  important  issue
in linguistic studies. The complexity of these
structures “is related to their intrinsic nature
of  subordination  which  is  a  basic,  universal
linguistic  process"  (Sheldon,  as  cited  in
Abdolmanafi & Rahmani, 2012).
 
Therefore,  the  first  reason  for  choosing  this
structure  was  the  fact  that  RCs  present  a
major  obstacle  for  Iranian  EFL  learners
(Bahrami  &  Ketabi,  2013).  Pedagogically,
due  to  their  structural  complexity  (Gass  &
Selinker,  as  cited  in  Abdolmanafi  &
Rahmani, 2012), it seems that  English RCs
present  a  number  of    problems  for  Iranian
EFL  learners  (Ghaemi  &  Bagherzadeh,
2012;  Abdolmanafi  &  Rahmani,  2012;
Marefat & Rahmany, 2009; Abdolmanafi &
Rezaee,  2012;  Bahrami  &  Ketabi,  2013).
Relativization is often considered “to be the
last hurdle for students to overcome since it
involves  complex  grammatical  rules
(Yabuki-Soh,  as  cited  in  Abdolmanafi,
2012, p.197).   
 
Instruments
Background information questionnaire
A  questionnaire  was  used  to  elicit
information  about  the  participants'
background  and  about  the  language
repertoire  of  the  participants.  To  provide  a
better  picture  of  the  context  in  which
participants  were  learning  the  languages
they  knew,  they  were  asked  to  provide
information  about  the  educational  level  of
their  parents,  their  families'  native
 
languages, and how many years of education
they had received in the L1,  L2 and  L3,  the
language(s) they use at school as well as the
city they came from.
 
To  avoid  participants'  English  proficiency
impinging  upon  their  ability  to  fill  in  the
questionnaire,  the  questionnaire  was  written
in Persian. Although Persian is the language
of  instruction  in  Iran,  Azeri  is  the  language
used  at  the  community  level  in  a  city  like
Tabriz.  The  bilingual  children  learn  and
speak  their  L1  (Azeri)  at  home;  like  other
Iranian  students  they  start  learning  Persian
literacy  skills  (reading  and  writing)  at  the
age  of  seven.  As  a  result,  they  become
bilingual  by  speaking  their  mother  tongue,
Azeri,  from  birth  and  by  learning  to  speak
and  write  in  Persian  at  school  (Bahrainy,
2007).
 
The  questionnaire  was  piloted  on  a  pilot
group,  an  intact  class  consisting  of  25
students,  was  selected  from  one  of  the
assigned  schools;  participants  had  similar
characteristics  to  those  of  students  in  the
main  study.  The  reliability  of  the  test  was
found  to  be  .87,  based  on  the  Cronbach
Alpha  coefficient.  Furthermore,  two
university  lecturers  with  a  PhD  degree  in
teaching English as a foreign language were
invited  to  appraise  whether  content  validity
was  present.  They  were  asked  to  offer  any
comments  regarding  the  relevance  of  items
to  the  purpose  of  the  questionnaire,  the
wording,  and  interpretation  problems  and
the instructions.
 
Standard general English proficiency test
A  standard  general  English  proficiency  test
was  used  to  ensure  the  homogeneity  of  the
participants.  As  grammar  and  vocabulary
are  heavily  focused  on  in  the  Iranian  EFL
curriculum,  we  decided  to  use  the  grammar
and  vocabulary  sections  of  the  OPT.  The
test  consisted  of  50  multiple  choice  items
with  an  estimated  time  of  45  minutes  for
completion, as determined by the OPT.  
 
The  reliability  of  the  OPT  was  calculated
using the Cronbach Alpha coefficient, which
was  found  to  be  .616.  Besides,  for  the
purpose of measuring the concurrent validity
of  this  test,  it  was  correlated  with  an
achievement  test  developed  by  the  Ministry
of  Education  for  second-grade  centers.  The
correlation  coefficient  calculated  between
the  achievement  test  and  the  OPT  appeared
to be .91.   
 
Grammaticality judgment task (GJT)
The  comprehension  task  was  a  grammatical
judgment  task  which  was  used  to  tap  into
the  participants’  actual  mental
representation of English relative clauses.  
 
The  test  comprised  24  English  relative
clause sentences, with an even split of three
grammatical  and  three  ungrammatical
sentences  in  each  category,  plus  three
distractors  (adapted  from  Azar,  2000).  The
distractor  sentences  were  added  so  that  the
students  could  not  predict  that  only  their
relative  clause  knowledge  was  being
assessed.  To  control  for  the  ordering  effect,
three  versions  were  provided  with  different
orders  of  the  test  items.  The  distribution  of
each  type  of  relative  clause  was  random.
Only  one  error  was  included  in  each
ungrammatical  sentence  so  that  the
participants would not be distracted by other
errors.  Furthermore,  vocabulary  was
controlled  for  and  should  not  have  caused
any problems for the participants. Therefore,
the  participants  were  allowed  to  ask  the
meaning  of  the  words  they  did  not  know.
The  time  allotted  for  the  test  was  15
minutes.  
 
In  order  to  meet  the  internal  consistency
reliability,  the  Cronbach  Alpha  coefficient
was  calculated  which  was  found  to  be  .59.

Furthermore,  regarding  the  content  validity
of  the  GJT,  the  test  was  evaluated  by  two
university  lecturers  with  a  PhD  degree  in
TEFL.  They  were  asked  to  comment  on  the
relevance  of  items  to  the  purpose  of  the
GJT.  This  resulted  in  a  few  adjustments  to
the questions in the GJT.
 
Moreover,  the  GJT  was  coded  into  ‘hits’,
‘misses’  and  ‘skips’.  A  ‘hit’  was  either  a
correct  acceptance  of  a  grammatical
sentence or a correct rejection. A ‘miss’ was
either  an  incorrect  rejection  of  a
grammatical  sentence  or  an  incorrect
acceptance of an ungrammatical sentence. A
‘skip’  (a  missing  value)  was  when  no
answer  was  given  to  an  item  (Falk,
submitted).
 
Cloze task (CT)
The  production  task  was  a  close  task  which
was  used  to  measure  the  English  relative
pronoun proficiency of high school students.
The  students  were  asked  to  fill  in  the
omitted words with one word and they were
allowed  to  ask  the  meaning  of  any  words
they did not know.
 
The  cloze  task  was  also  piloted  on  the  pilot
group.  The  participants'  comments  and  the
process  of  responding  to  the  test  led  the
researchers to increase the task time from 15
minutes  to  20  minutes.  Furthermore,  the
Cronbach  Alpha  was  used  in  order  to
estimate  the  reliability  of  the  cloze  task,
according to which the reliability of the test
was .62.   
 
It  should  be  noted  that  the  cloze  task  was
scored  according  to  the  appropriate  scoring
method;  that  is,  answers  which  were
grammatically  appropriate  were  accepted  as
correct and were given a score of one. Those
answers which were incorrect were assumed
incorrect  and  were  given  a  zero  score  and
those  answers  which  were  left  blank  were
coded as a skip.  
 
Furthermore, the researchers decided to set a
hypothetical  level  for  acquisition  in  both
tasks,  in  line  with  other  acquisition  studies.
In  this  study,  the  level  of  acquisition  at  an
accuracy rate of 75% was chosen, following
Neeleman and Weerman (as cited in Falk &
Bardel, 2011) who assume that:
 
Deviations    from    the    perfect    score  
are  due  to  performance  factors  and  
other  variables    that    are    not    under  
our    control.    A  subject  might  accept
an  ungrammatical  sentence  because
he  or  she  can  assign  a  pragmatically
plausible  interpretation  to  it  or
because he or she is simply confused.
Of  course,  such  factors  are  irrelevant
from  our  perspective  and  hence  we
should  somehow  correct  for  their
influence  when  considering  the  test
results.  In  order  to  do  so  we  assume
that  a  subject  has  knowledge  of  a
particular  construction  if  he  or  she
reaches a score of 75%.  
 
Data collection procedures
The  data  collection  phase  comprised  the
administration  of  four  instruments.  During
the  first  phase  of  the  study,  after  carrying
out  the  sampling  procedure  and  choosing
subjects randomly, the researchers used oral
description  to  explain  the  study  to  the
students,  giving  brief  instructions  for  all
phases  of  the  study.  The  questionnaire  was
then distributed.  
 
Next, the OPT was taken by the participants.
The  test  was  administered  according  to  the
test  instructions  (45  minutes),  and  the
participants  were  found  to  be  at  lower-intermediate level of proficiency.
 
In the next step, the participants were asked
to  judge  the  (un)grammaticality  of  English
sentences.  They  were  asked  to  respond  as
quickly  as  possible,  because  of  the  15-minute time limit.  
 
The last phase was the administration of the
cloze  task.  The  participants  were  asked  to
spend 20 minutes to complete this task. The
tasks  were  all  carried  out  during  regular
lesson-time  and  administrated  by  their
teachers so as to avoid self-selection bias.  
 
Data analysis
The  results  obtained  were  analyzed  using
the SPSS software. First the main test items
were  coded  and  given  value.  The  values  of
similar  variables  were  computed  in
percentage  in  order  to  have  more  organized
data.  That  is,  first  the  exact  distribution  in
the  learners’  responses,  their  percentages
along  with  the  mean  ratings  of  accurate
responses  for  relative  clauses  and  relative
pronouns,  across  the  two  groups,  were
tabulated.
 
Furthermore,  the  independent-samples  t-tests  (two-tailed)  were  calculated  and
between-group  comparisons  were
conducted.  
 
Results  
Figure  1  exhibits  the  percentages  the
learners  obtained  on  GJT.  As  Figure  1
displays,  the  bilingual  group  had  65.67%
acceptance  and  1.5%  rejection  out  of  1200
responses  to  grammatical  sentences  while
the  monolingual  learners  had  54.83%
acceptance and 2.67% rejection out of 1200
responses  to  grammatical  sentences.  There
is  also  a  category  including  skipped  items.
Bilinguals had 32.83% skipped items out of
1200  responses  while  the  monolingual
learners  had  42.5%  skipped  items  out  of
1200 responses.  
 
It  should  be  noted  that  the  total  responses
were  2400  sentences.  The  descriptive
statistics  of  grammatical  stimuli  for  relative
clauses by both (bilingual and monolingual)
groups are presented in Table 1.
 
Comparison  of  the  means  provided  by  the
participants  in  both  groups  shows  the
superiority of the performance scores of the
bilingual  participants  since  the  bilingual
group (M = 7.88) gained a higher mean than
the monolingual group (M = 6.58).To probe
the significant differences between the mean
scores  of  the  monolingual  and  bilingual
groups,  an  independent-samples  t-test  (two-tailed)  on  the  grammatical  hits  was  applied.
The  magnitude  of  the  difference  (mean
difference=  1.3,  95%  CI:  .71  to  1.88)  was
moderate (eta squared=.088). The results are
presented in Table 2.

Furthermore,  the  bilingual  learners  had
1.25%  acceptance  and  34.58%  rejection  out
of  1200  responses  to  ungrammatical
sentences  while  the  monolingual  learners
had 4.67% acceptance and 32.08% rejection
out  of  1200  responses  to  ungrammatical
sentences.  Also,  bilinguals  had  64.17%
skipped  items  out  of  1200  responses  while
the  monolingual  learners  had  63.25%
skipped  items  out  of  1200  responses  (see
Figure 2).  
 
Table  3  shows  the  results  of  the  descriptive
statistics  of  ungrammatical  stimuli  for
relative  clauses  by  the  bilingual  and
monolingual groups.
 
As  shown  in  Table  3,  the  bilingual  group
gained  the  lowest  mean  score  (M  =  .15)
whereas  the  monolingual  group  gained  the
highest  mean  score  (M  =  .56).  An
independent-samples  t-test  (two-tailed)  on
the ungrammatical hits was run to probe the
significant  differences  in  the  scores  of  the
two  groups.  The  magnitude  of  the
differences in the means (mean difference =
-.41,  95%  CI:  -.68  to  -.14)  was  small  (eta
squared = .042). The results are presented in
Table 4.  
 
In  total,  we  found  803  hits  (both
grammatical  and  ungrammatical)  out  of
2400 grammatical sentences in the bilingual
group  and  714  hits  (both  grammatical  and
ungrammatical)  out  of  2400  grammatical
sentences in the monolingual group. In other
words,  as  shown  in  Figure  3,  the  overall
accuracy  rates  for  bilinguals  are  33.46  %
and  for  monolinguals  are  29.75%.  Besides,
as  Figure  4  shows  the  bilingual  group
revealed  a  higher  mean  rating  in  total
grammatical  relative  clauses  (hits)  than  the
monolingual group: 8 vs. 7.15. To probe the
significant  differences  between  the  mean
scores  of  the  two  groups,  an  independent-samples  t-test  (two-tailed)  was  utilized.  The
magnitude  of  the  difference  (mean
difference=  .85,  95%  CI:  .18  to  1.52)  was
moderate (eta squared=.030). The results are
presented in Table 5.
 

These  results  suggest  that  bilingual  learners
performed  significantly  differently  to
monolingual  learners.  Yet,  according  to  the
75%  criterion  (Neeleman  &  Weerman,  as
cited  in  Falk  &  Bardel,  2011),  the
participants in both groups did not reach the
high proficiency level, i.e. structures such as
relative clauses cannot be said to have been
fully  acquired.  They  are  learning  the  target
language, but the majority cannot be said to
be at a high proficiency level of acquisition.
It is therefore hardly surprising that they do
not behave like native speakers.
 
Findings from the cloze test analysis
As  Figure  5  displays,  the  bilingual  group
had  32.90%  correct  responses  and  40.80%
incorrect  responses  (out  of  2000  responses)
while  the  monolingual  group  had  25.65%
correct  responses  and  40.50%  incorrect
responses  (out  of  2000  responses).  There  is
also  a  category  including  skipped  items;
bilinguals had 26.30% skipped items (out of
2000  responses)  while  the  monolingual
learners  had  33.85%  skipped  items  (out  of
2000 responses).  
 
Figure  6  presents  the  mean  percentages
obtained  by  both  groups  for  the  relative
pronouns  on  the  CT.  Comparison  of  the
means  of  the  answers  provided  by  the
participants  in  both  groups  shows  the
superiority of the performance scores of the
bilingual  group.  The  bilingual  group  (M  =
6.58)  gained  a  higher  mean  than  the
monolingual  group  (M  =  5.13).  An
independent-samples  t-test  (two-tailed)  was
run  to  probe  the  significant  differences
between  the  mean  score  of  the  two  groups.
The  magnitude  of  the  differences  in  the
means  (mean  difference  =  .85,  95%  CI:  .77
to  2.13)  was  moderate  (eta  squared  =
.08). The results are presented in Table 6.

Discussion
Two  points  are  worthy  of  being  mentioned
from  the  findings  of  this  study.  First,  the
results reveal that the bilingual group gained
a  higher  mean  than  the  monolingual  group
in the CT: 6.58 vs. 5.13. Also, the results of
the  independent-samples  t-tests  revealed  a
significant  difference  between  bilingual  and
monolingual  groups  not  only  in  the
comprehension task (GJT) (t (198) = 2.50, p
= .013) but also in the production task (CT)
(t  (198)  =  4.22,  p  =  .000).  These  results
support  the  claim  that  the  bilingual  group
generally  had  a  better  performance  and
significantly outperformed the L2  group in
both  the  GJT  and  CT.  It  can  be  assumed
from  this  study  that  bilinguals  gain  an
advantage  in  knowing  two  languages  when
learning a third one.
 
According  to  the  TPM  theory  (Rothman
2010), both L1 and L2 fhave the potential to
play  a  stronger  role  in  TLA  based  on  their
(psycho)-typological  proximity  to  the  L3.
Consequently, typology is argued to have an
impact on the transfer source, such that "the
more  typologically  proximate  the  L2  or  the
L1  is  to  the  L3,  the  more  likely  it  is  to  be
transferred" (Bardel & Falk, 2007, p. 474).
 
Typologically,  English  and  Persian  belong
to  the  same  Indo-European  family  while
Azeri  is  classified  as  Altaic-Turkic-Southeastern/Oghuz.  Azeri  does  not  match
English  and  Persian  in  its  head  direction.
With regard to relative clause constructions,
both  Persian  and  English  are  post-nominal.
Azeri  is  basically  pre-nominal,  although  it
has a borrowed form from Persian, which is
post-nominal  (the  persified  head-initial
construction).  The  native  relative  clause
construction  is  the  most  typical  type  of
relative clause, whereas the borrowed one is
not.  Therefore,  based  on  the  typological
proximity  of  Persian  to  English,  it  seems
that  Persian  has  a  stronger  role  in  learning
the L3 (English).
 
The  results  conform  to  the  studies  by
Schachter (as cited in Ellis, 2008) and Flynn
et  al.  (2004).  In  a  quantitative  study,
Schachter  (as  cited  in  Ellis,  2008)
investigated  the  relative  clause  structures  in
L2  and  focused  on  four  groups  of  students
with  different  L1  backgrounds  -  Arabs,
Persians,  Japanese  and  Chinese.  Results
showed  that  Arab  and  Persian  learners  used
relative clauses two or three times more than
the  Japanese  and  Chinese  students.
Schachter  suggested  that  the  reason
responsible  for  the  relatively  greater  use  of
the  relative  clauses  was  right-branching
relative  clause  structures  in  Arabic  and
Persian.  
 
In  another  study  investigating  the
acquisition  of  the  English  Complementizer
Phrase  (CP),  more  specifically  restrictive
relative  clauses  in  L3,  Flynn  et  al.  (2004)
suggested  that  "prior  CP  development  can
influence the development of CP structure in
subsequent  language  acquisition"  (Flynn,
2009, p. 80).
 
Similarly,  in  this  study  the  bilingual  group
judged  grammatical  relative  clauses  more
than  the  monolingual  group  in  GJT  (803
correct  responses  by  the  bilingual  group  vs.
714  correct  responses  by  the  monolingual
group). In other words, the overall accuracy
rates  for  bilinguals  were  33.46  %  and  for
monolinguals  were  29.75%.    Furthermore,
the  bilingual  group  used  correct  relative
pronouns  more  than  the  monolingual  group
in  CT  (658  correct  responses  by  the
bilingual group vs. 513 correct responses by
the monolingual  group).  In other words, the
overall  accuracy  rates  for  bilinguals  were
32.90  %  and  for  monolinguals  were  25.65
%.  Therefore,  it  seems  that  right-branching
relative clause structures in Persian, which is
 
the  same  in  English,  were  responsible  for
the  relatively  greater  use  of  the  relative
clauses  and  relative  pronouns  by  the
bilingual  group  because  the  structure  of  the
relative  clause  structure  is  dependent  on  a
language's  head-directionality  (Flynn  et  al,
2004).  To  put  it  differently,  in  line  with
Flynn  et  al.’s  (2004)  suggestion,  it  seems
that  the  bilingual  group  can  benefit  from
their  L2  Persian  due  to  having  the  post-nominal  relative  clause  structure  which  is
typologically  similar  to  the  L3  English
relative  clause  structure  since  prior  post-nominal  relative  clause  development  can
influence  the  development  of  post-nominal
relative  clause  structure  in  learning  L3
English.  
 
The  second  point  is  that  the  results  do  not
clearly rule out evidence of transfer from the
bilinguals’  L1.  The  distribution  of
acceptance  and  rejection  responses  can  be
ascribed  to  transfer  from  Azeri  native  pre-nominal  relative  clause  structure.
Considering  the  misses  in  the  GJT  by  the
bilingual  group,  we  found  that  1.5%  of  the
grammatical  sentences  were  judged  in  an
incorrect  way  and  the  ungrammatical
sentences  received  34.58%  misses.  These
numbers  might  be  the  result  of  participants
transferring  L1  Azeri  relative  clause
structure.  According  to  Ellis  (1994),  errors
have  been  considered  as  one  of  the
manifestations  of  language  transfer.  Where
L1  and  L2  features  are  identical,  learning
can  take  place  easily  through  positive
transfer of the L1 features, but where L1 and
L2  features  are  different,  learning  difficulty
can arise, and errors resulting from negative
transfer are more likely (Ellis, 2002).
 
According  to  the  results  of  GJT  and  CT
analysis,  it  can  be  concluded  that  the
bilingual  group  generally  performed  better
(i.e.  the  overall  accuracy  of  Azeri-Persian
bilingual  learners  was  higher  than  the
Persian  monolingual  participants)  and
significantly outperformed the L2  group  in
both  the  GJT  and  CT.  The  results  of  this
study  are  in  line  with  the  findings  of  other
studies,  which  suggest  that  "becoming
bilingual,  either  as  a  result  of  home  or
school  experiences,  can  positively  influence
aspects of cognitive functioning" (Cummins,
1976, p. 11).  
 
Yet,  according  to  the  75%  criterion
(Neeleman  &  Weerman,  as  cited  in  Falk  &
Bardel,  2011),  the  participants  had  not
reached  high  proficiency  levels;  i.e.,
structures such as relative clauses cannot be
said  to  have  been  fully  acquired.  It  is
therefore  hardly  surprising  that  they  do  not
behave like native speakers. Nevertheless, it
can  be  concluded  that  having  a  second
language  has  an  effect  on  the  acquisition  of
English  L3  relative  clauses  by  bilingual
learners. The findings are supported by other
studies  which  have  demonstrated  that
bilingualism results in more efficient foreign
language learning (Ringbom, 1987; Thomas,
1988;  Klein,  1995;  Sanz,  2000;  Hoffman,
2001;  Keshavarz  &  Astaneh,  2004;
Modirkhamene, 2008; Jaensch, 2009; Dibaj,
2011;  Kassaian  &  Esmae’li, 2011; Seifi  &
Abdolmanafi,  2013;  Saeidi  &  Mazoochi,
2013; and Zare & Mobarakeh, 2013).
 
Zobl (as cited in Falk  & Bardel, 2010) used
a  grammaticality  judgment  test  to  measure
several  structures  by  monolingual  and
multilingual  learners  of  English.  Zobl’s
study indicated that multilinguals were at an
advantage when learning English.  
 
Klein  (1995)  conducted  a  study  with
monolinguals  and  multilinguals  learning
English  and  tested  specific  verbs  and  their
prepositional  complements  and  preposition
strandings.  Multilinguals  presented
significantly  higher  scores  in  both
constructions.

Thomas  (1988)  concluded  that  the  Spanish-English  bilinguals  who  were  only  orally
proficient  in  the  two  languages
outperformed  their  monolingual  English-speaking  counterparts  in  learning  French
vocabulary.  
 
Keshavarz  and  Astaneh  (2004)  found  that
the  Azeri-Persian  speakers  outperformed
their  Persian  peers  on  a  CPAT  at  the  2000
and  3000  word  levels.  Furthermore,
Modirkhamene (2008) explored the possible
differences  between  the  performance  of
Persian  monolingual  and  Turkish-Persian
bilingual  learners  on  metalinguistic  tasks  of
ungrammatical  structures  and  translation,
and  found  that  bilinguals  outperformed
monolingual learners.  
 
Moreover,  in  Dibaj’s  (2011)  study,  the
Azeri-Persian  speakers  outperformed  their
Persian  counterparts  on  two  incidental  and
four  intentional  vocabulary  learning
exercises.  In  Saeidi  and  Mazoochi’s  (2013)
study,  the  results  also  indicated  that
bilinguals  were  superior  in  terms  of  their
linguistic  intelligence.  They  claim  that  the
participants'  bilingualism  can  enhance  their
cognitive  development.  Seifi  and
Abdolmanafi  (2013)  also  indicated  that
bilinguals  had  an  advantage  over
monolinguals in terms of using strategies.  
 
Conclusion
The  main  purpose  of  this  study  was  to
ascertain  whether  bilinguals  would  perform
better  than  monolinguals.  The  theoretical
framework  suggests  that  language
background  is  an  important  factor  in  TLA
(Thomas,  1988;  Klein,  1995;  Hoffman,
2001;  Sanz,  2000;  Keshavarz  &  Astaneh,
2004;  Modirkhamene,  2008;  Jaensch,  2009;
Saeidi  &  Mazoochi,  2013;  Seifi  &
Abdolmanafi,  2013;  Zare  &  Mobarakeh,
2013).The  data  were  obtained  through  a
questionnaire  and  two  syntactic  structure
tests  (GJT  and  CT).  The  respondents  were
100  female  Azeri-Persian  bilingual  high
school  students  and  100  female  Persian
monolingual high school students.  
 
The  first  research  question  concerned  the
effect  of  bilinguality  on  the  non-native
comprehension  of  English  L3  relative
clauses  by  Azeri-Persian  bilingual  learners.
The bilingual group judged the acceptability
of  the  grammatical  and  ungrammatical
sentences on the relative clauses to a higher
degree  than  the  monolingual  group.
Furthermore, the results of the independent-samples  t-test  on  the  total  hits  revealed  a
significant  difference  between  the  bilingual
and  monolingual  groups  in  the  GJT.  The
second  research  question  concerned  the
effect  of  bilinguality  on  the  non-native
production of English L3 relative clauses by
Azeri-Persian  bilingual  learners.  The  results
showed  the  superiority  of  the  performance
scores of the bilingual group. Moreover, the
results  of  the  independent-samples  t-test  on
the  correct  responses  revealed  a  significant
difference  between  the  bilingual  and
monolingual groups in the CT.  
 
The overall results of the study revealed that
bilingualism  presents  a  significant
advantage  in  TLA.  This  difference  can  be
explained in this way: "The more languages
one  has  acquired,  the  more  beneficial  it
would  be  for  the  acquisition  of  additional
non-native  languages"  (Leung,  2005,
p.1351).
 
The  L3  Azeri-Persian  learners  are  supposed
to  benefit  from  their  unique  language
experience  in  two  ways:  the  privilege  of
having knowledge of two separate  grammar
systems  (Azeri  and  Persian)  and  the
availability  of  the  relative  clause  similarity
between  the  target  language,  English,  and
their  L2,  Persian.    In  other  words,  they
already  have  access  to  knowledge  from
 
more  than  one  language  system,  which
results  in  ‘multi  competence’  defined  by
Cook  as  ‘knowledge  of  two  or  more
languages in the same mind’ (1992). Cook’s
notion  of  ‘multi  competence’  refers  to
multilingual  linguistic  competence
characterized  by  greater  creativity  and
cognitive  flexibility  and  more  diversified
mental abilities (Cook, 1992).
 
All  in  all,  the  predictions  of  particular
syntactically-based  TLA  theories  were
tested in this study. The results appear to be
compatible  with  TPM  theory  (Rothman,
2010),  which  states  that  both  L1  and  L2
have  the  potential  to  play  a  role  in  TLA,  as
determined by their typological proximity to
the L3. In other words, in TLA, the language
which  is  typologically  similar  to  the  L3
seems  to  be  a  determining  factor  in  the
shape  and  speed  of  TLA  (Rothman,  2010).
With  regard  to  typological  proximity  it  is
possible  that  Azeri-Persian  bilinguals  may
benefit  from  L2  Persian  since  prior  post-nominal  relative  clause  development  can
influence  the  development  of  post-nominal
relative  clause  structure  when  learning  L3
English. Moreover, in line with TPM theory,
the  results  indicate  non-facilitative  transfer
based  on  the  distribution  of  the  acceptance
and rejection responses.
 
Implications of the study   
A  clear  implication  of  this  study  is  that
students'  sensitivity  to  the  differences  and
similarities  between  the  languages  they
know should be increased. In fact, “the more
aware  learners  are  of  the  similarities  and
differences between their mother tongue and
the target language, the easier they will find
it to adopt effective learning and production
strategies" (Swan, 1997, p.178).  
 
In  multilingual  educational  settings,
similarities  and  differences  between
languages can be concentrated on in order to
increase  metalinguistic  awareness  in  both
teachers  and  students.  It  is  thought  that
learners’  awareness  of  similarities  and
differences between their mother tongue and
additional  languages  will  pave  the  way  for
effective  learning.  Therefore,  a  method  of
teaching  foreign  languages  that
demonstrates  cross-linguistic  similarities
among  languages  seems  to  be  an  effective
way in preparing language learners for more
successful learning (Modirkhamene, 2008).   
 
Further research
The findings in this study open up a range of
new  research  questions  that  should  be
answered  in  the  future.  As  scholars  (see
Falk,  2010)  argue,  research  outcomes  are
sensitive to the data collection method. This
study involved GJTs. One of the advantages
of using GJTs in research on the acquisition
of  syntax  is  that  they  are  "handy  tools  with
which  we  can  construct  any  structures  that
we are interested in and force the informant
to respond to the sentence" (Falk, & Bardel,
2011,  p.  76).    It  is  a  matter  for  future
research to determine whether having a very
high  accuracy  rate  depends  on  the
proficiency level of the participants or not.
 
This  study  only  examined  two  genetically
and typologically similar languages (English
and  Persian)  in  relation  to  another  language
(Azeri).  Different    combinations  could  be
adopted to test whether language distance  is
really  an  important  factor  underlying  cross-linguistic  influence  among  languages,  such
as    three    totally    distant  languages,    three  
closely  related  languages,  or  the  L1  and  
L3    being    more    closely  related    than    the  
L2.
 
Research manageability made it necessary to
delimit  the  study  in  terms  of  the  age  and
gender  of  the  participants.  Thus,  the  results
obtained  from  this  study  cannot  be
generalized  to  other  age  ranges  and  male
 
learners.  Therefore,  more  studies  may  be
conducted  with  different  age  groups  and
with male participants.  
 
Limitations of the study
The current study has shed some light on the
effect  of  bilingualism  on  learning  an
additional language, especially in the area of
syntactic  learning,  but  it  has  certain
limitations and more studies in this area may
be worthwhile.
 
The  most  obvious  limitation  present  in  this
study  originates  in  the  test  administered  to
evaluate  the  participants’  proficiency.  The
test  only  included  a  structure  section  and  a
vocabulary section.  
 
The  second  limitation  is  that  an  error
analysis  was  not  carried  out.  Error  analysis
is  useful  because  it  reveals  problematic
areas.  
 
The third limitation involves the proficiency
level  of  the  participants  in  their  L1  Azeri.
Proficiency in the source language is argued
to  be  an  important  factor  (e.g.,  De  Angelis,
2007). However, the proficiency level of the
bilingual  learners  in  their  L1  language
(Azeri) was not possible to measure.