Reassembling Formal Features in Articles by L1 Persian Learners of L2 English

I. Introduction
 
In  her  seminal  article,  Lardiere  (2005)
challenged  the  prevailing  notion  of  parameter
resetting,  arguing  that  the  formal  task  facing  a
learner  is  much  more  complex  than  the  simple
parametric  selecting  of  a  new  feature.  She
attributed  the  existing  cross-linguistic  variations
to  permutations  in  the  configuration  of  morpho-syntactic  features  which  make  up  grammatical
categories.    Difficulties  in  learning  a  second
language,  in  the  framework  of  parameter  setting
and  resetting,  are  explained  through  the
availability of features. If a feature is selected in
the L1, it would be easy for the learner to acquire
that feature in the L2. On the other hand, features
that  are  morphologically  selected  in  the  L2  but
not  in  the  L1  would  be  no  longer  available  and,
therefore,  unacquirable,  resulting  in  a  fossilized
grammar.  In  the  later  proposed  framework  of
feature reassembly (Lardiere, 2005, 2007, 2009),
the difficulties are attributed to the massive task
of determining how to reconfigure features from
the  way  they  are  mapped  in  the  L1  into  new
configurations  which  may  be  realized  on  quite
different lexical items in the L2.
However,  as  White  (2009)  suggests,  there  are
two  parts  to  this  proposal.  If  a  feature  exists  in
the first language, sometimes the learners should
associate that feature with different lexical items
in  the  second  language.  An  example  would  be
the  acquisition  of  articles  in  a  second  language
like  English  or  Greek  by  learners  whose  L1  is
article-less such as Mandarin Chinese which has
the  feature  definiteness  but  this  feature  is
realized  on  other  determiners  such  as
demonstratives. So, Chinese learners’ task would
be  to  discover  that  the  feature  is  realized  on
different  lexical  items  in  English:  the  and  a.  In
the  second  place,  if  feature  combinations  in  L1
and  L2  differ,  the  learner  will  have  to  acquire  a
new language-specific configuration, assembling 
 
features into different bundles in the L2 from the
L1.  An  example  would  be  the  way  the  four
features  of  plural,  definite,  human,  and  animate
are assembled in English and Korean resulting in
different  realizations  of  plural  in  each  language
(Choi,  2009).  There  are  other  cases  which
involve different combinations of features across
two  languages  and  the  learning  of  such
grammatical  points  will  depend  on  successful
reassembly  on  the  part  of  learners  (see  for
example  Renaud,  2009; Yuan  &  Zhao,  2009).  It
is this second form of reassembly which is more
complicated  and  more  likely  to  pose  serious
problems to learners (Lardiere, 2005).
One  area  of  grammar  that  has  been  studied  in
research  on  L2  acquisition  is  the  article  system.
It  is  of  particular  interest  in  testing  hypotheses
about  L2  acquisition  because  of  its  complexity
and  the  number  of  features  that  play  a  role  in
determining the correct article to be used. In case
of  English,  those  features  include  definiteness,
specificity,  countability,  genericity  and  number.
So, much information is encoded in the two little
words the and a. Wakabayashi (2009) noted this
intricacy  of  marking  definiteness  in  English  and
believed  that  it  has  been  largely  ignored  in  the
existing  SLA  research  which  has  focused  on
only  one  or  two  features  that  determine  article
use.  Examples  of  such  studies  are  Ionin,
Zubizarreta  and  Maldonado  (2008),  Ko,  Ionin
and  Wexler  (2009)  and  Snape,  Leung  and  Ting
(2006).  This  gave  this  study  the  impetus  to
present  a  comprehensive  account  of  how  the
system  works  in  English  with  regard  to  the
above mentioned features and to find out if those
features  also  operate  in  Persian.  Also,  the
acquisition  of  the  article  system  in  English  by
Persian  speakers  seems  to  be  particularly
suitable  for  testing  the  theoretical  predictions  of
the  feature  reassembly  hypothesis  (FRH
henceforth)  because  many  features  seem  to  be
bundled  into  the  articles  (lexical  items)  in  both
languages.  The  discrepancy  in  feature
combination  may  explain  Persian  speakers’
reported failure to achieve native-like fluency in
the  use  of  English  articles  (Afzali,  2008;
Rahmani,  2010;  Momenzade  &  Youhanaee,
2013).  The  aim,  then,  was  to  seek  explanation
for the observed pattern of article use by Persian
learners.  If  it  can  be  shown  that  the
configurations  of  the  system  differ  between  the
two  languages  but  the  features  are  present  at
both  scenes,  then  the  reassembly  hypothesis
might  explain  the  process  of  acquisition  as  it
really is. 
II. Properties of the article system in English
and Persian
In  the  following,  the  article  system  in  English
and  Persian  is  elaborated  on  in  order  to  clarify
how  the  features  of  definiteness,  specificity,
genericity,  number  and  countability  actually
affect article use. Each point is presented with an
example.  Similarities  and  differences  between
the  two  languages  are  highlighted  and
predictions  are  made  as  to  the  performance  of
Persian learners of English on each point.  
a. Definiteness and specificity
The  English  article  system  marks  definiteness,
not  specificity  (Ionin,  2006).  Therefore,  definite
determiner  phrases  (henceforth  DPs)  in  English
are  marked  by  the  and  indefinite  DPs  by  a.  As
for  specificity,  DPs  in  English  may  come  with
the or a in both [±specific] situations (see Ionin,
Ko & Wexler, 2007). Persian, on the other hand,
does  not  have  a  definite  article  in  its  formal
written  form-  though  there  is  an  enclitic  used  to
mark  definiteness  in  the  colloquial  language
(Ghomeshi,  2003)-  but  marks  indefinite  NPs  in
three ways (using an  enclitic  -i noun finally, the
word  yek  before  the  noun,  or  both).  Indefinite
NPs  in  Persian  can  bear  specificity  as  well  (see
Karimi, 1999). Table 1 presents examples for the
two  types  of  NPs  in  the  two  corresponding
languages:

It  is,  then,  reasonable  to  predict  that  Persian
learners  should  have  no  difficulty  with  the
indefinite  article  but,  since  an  overt  article  is
lacking  in  Persian,  they  may  have  problems  in
learning how to use the. However, any difficulty
should  be  of  short  duration  on  the  grounds  that
the  concepts  exist  and  are  implemented  in
Persian  and  that  there  is  an  informal  definite
enclitic in that language. 
Besides  definiteness  and  specificity,  there  are
still  a  number  of  other  features  that  affect  how
articles  are  used  in  English.  They  include
genericity,  number,  and  the  count/mass
distinction.  The  same  features  also  exist  in
Persian  but  the  two  languages  do  not  behave
similarly (as presented in the following sections).
How  those  features  influence  article  use  in
English  plus  the  difference  there  is  with  Persian
is explained in the following.
b. Genericity
In  English,  the  concept  of  genericity  can  be
expressed using the, a/an, or Ø. The three forms
are  not  always  interchangeable,  a  point  that  is
not  the  concern  of  the  present  paper  (for  more
information  see  Quirk,  Greenbaum,  Leech  &
Svartvik,  1972).  In  Persian,  however,  only  bare
singular  or  bare  plural  nouns  can  be  interpreted
as  generic  (Moin,  1990).  To  identify  the
acquisition  problem  for  Persian  learners,  one
should bear in mind that English uses three ways
to  denote  genericity  but  Persian  uses  two.
Examples are:

It  has  to  be  mentioned  that  bare  plurals  in
Persian  depend  on  context  for  their
interpretation.  They  may  also  be  interpreted  as
definite.  So,  example  (13)  above  can  also  be
interpreted as ‘The flowers need light.’ when the
speaker is referring to some flowers in his office
for example.   
 
 
Although  Persian  has  an  overt  indefinite  article,
it  cannot  be  used  in  a  generic  NP  which  is  a
point of divergence from English. It is also of the
same  difficulty  to  Persian  learners  to  figure  out
that  the  definite  article  can  be  used  in  a  generic
NP.  It  seems,  then,  reasonable  for  Persian-speakers  to  be  able  to  best  understand  and
produce  the  first  type  of  generic  sentences  in
English  (bare  plurals)  since  the  same  structure
exists in Persian as well. For the other two types,
difficulties  are  expected  at  low  levels  of
proficiency  simply  because  the  other  two  types
involve  article  use  that  makes  them  different
from  their  L1  structure.  At  higher  levels,
however,  learners  are  expected  to  comprehend
and  produce  the  other  two  forms  as  they  master
the articles. 
c. Number
As  to  number,  both  languages  select  the  feature
but  the  realization  is  different  again.  In  English,
singular  nouns  can  be  accompanied  by  definite,
indefinite,  or  zero  articles  (in  case  of  mass
nouns).  Plural  nouns  also  may  occur  with
definite  or  zero  articles  (in  case  they  are
generic).  In  Persian,  however,  the  indefinite
article  may  precede  plural  nouns  as  well  as
singular  nouns.  If  nouns  (singular,  plural,  or
mass)  are  definite  or  generic,  then  they  are
accompanied  by  a  zero  article.  Another  point  to
mention  is  that  in  informal  spoken  Persian,  the
definite enclitic is only used with singular nouns
(Ghomeshi,  2003).  So,  whereas  in  English  the
definite  article  is  used  with  both  singular  and
plural  nouns,  in  Persian  it  is  used  only  with
singular nouns. Similarly, whereas in English the
indefinite  article  can  only  be  used  with  singular
nouns,  it  is  used  with  both  singular  and  plural
nouns  in  Persian.  Examples  are  displayed  in
Table 3

To  make  things  more  complicated,  bare  plural
NPs  in  English  are  either  indefinite  or  generic
whereas the same construction can be generic or
definite. As a result, the learning task for Persian
learners  of  English  would  be  to  use  the  definite
article  for  singular  and  plural  nouns  and  to  use
the indefinite article just for singular nouns. It is,
then, expected that Persian speakers should have
no  serious  or  long-lasting  problems  since  both
features  (definiteness  and  number)  exist  in  their
L1. 
d. Countability
The last feature to be discussed is the count/mass
distinction on nouns. In English, mass nouns can
be  definite  and,  thus,  preceded  by  the.  A  zero
article  can  also  come  before  a  mass  noun  in  a
generic  or  indefinite  sense  though  in  the
indefinite reading words like some are preferred.

In Persian, on the other hand, mass nouns cannot
take  any  article  other  than  zero  in  which  case
they  can be interpreted  as definite, indefinite, or
generic.  Although  on  occasions  people  use  the
indefinite  article  with  a  mass  noun  in  the
informal spoken language (Moin, 1990), it is not
acceptable  in  the  formal  written  variety  we  are
discussing.  There  is,  however,  a  very  important
difference  between  English  and  Persian  in  that
mass  nouns  can  be  pluralized  in  Persian,  a
construction  which  is  impossible  in  English.
(Table 4)

Again, it seems that differences between the two
languages  are  not  due  to  absence  of  some
features but to how those features are realized on
lexical  morphemes  cross-linguistically.  If  we
take  the  view  that  absence  of  features  is  the
source of difficulty, Persian learners should have
no  serious  problems.  If  they  face  difficulties,
then,  the  responsibility  falls  on  the  different
assembly of features in the two languages.
 III. Review of literature
As  a  difficult  linguistic  element  to  be  acquired
by learners, studies abound on the acquisition of
the English article system and on the difficulties
learners  face.  These  pieces  of  research  can  be
generally  divided  into  two  groups  based  on
learners’  L1  background:  if  they  speak  a
[+article]  or  a  [–article]  first  language.  Two
notions  dominate  such  studies:  transfer  and
fluctuation. That is, either learners have access to
UG settings and, therefore, fluctuate between the
two  settings  of  using  articles  for  specificity  or
for definiteness, or they don’t have such access
and transfer their L1 settings into their pattern of
L2  article  use.  Studies  on  learners  with  article-based L1s have shown that the second alternative
works  and  the  transfer  option  is  dominant
(Snape,  Leung  &  Ting,  2006;  Ionin,  Zubizarreta
&  Maldonado,  2008;  Jaensch  &  Sarko,  2009;
Garcia  Mayo,  2009).  Learners  from  article-less
L1s, however, have demonstrated that, for them,
fluctuation  is  the  prevailing  factor  and  that  they
use  English  articles  sometimes  to  mark
specificity and other times definiteness (Guella ,
Déprez  &  Sleeman,  2008;  Zdorenko  &  Paradis,
2008).The  few  studies  conducted  on  the  use  of
articles for generic reference have given support
for the strong  role  L1 plays in  L2 acquisition of
articles as well (Perez-Leroux, Munn, Schmitt &
DeIrish,  2004;  Ionin  &  Montrul,  2009;  Snape,
Garcia Mayo & Gürel, 2009).
Most  studies  that  have  investigated  learning  of
the  articles  by  Persian  learners  have  approached
the problem from  a pedagogical perspective  and
so,  do  not  contribute  to  this  research  (for
instance,  Afzali,  2008;  Dabaghi  &  Tavakoli,
2009;  Farrokhi  &  Sattarpour,  2012;  Geranpaye,
1995;  Mobini,  2006;  Pashazade  &  Marefat,
2010; Sarani, 2009).Two studies, however, have
tried  to  provide  a  theoretical  account  for  article
acquisition  using  the  UG  framework.  Rezai  and
Jabbari  (2010)  were  able  to  show  that  the
learners  in  their  study  did  acquire  the  English
definite  article  though  some  fluctuation  was 
 
observed  in  their  teasing  apart  the  definiteness
and  specificity  features  in  singular  contexts.
They  believed  that  their  study  provided  another
piece  of  support  for  the  interpretability
hypothesis  (Tsimpli  &  Dimitrakopoulou,  2007)
which  claims  that  interpretable  features  are
accessible to the L2 learner while uninterpretable
features  are  difficult  to  identify  and  analyze  in
the L2 input due to persistent L1 effects on adult
L2  grammars.  The  definiteness  feature  is
interpretable,  they  argue,  at  both  conceptual
level  (LF)  and  phonetic  level  (PF)  in  English.
The same feature is, however, interpretable only
at  the  LF  in  Persian.  Nevertheless,  the  fact  that
definiteness  is  an  interpretable  feature  in  both
languages  poses  no  serious  problems  to  learners
and  the  study  supported  this  claim.  Elsewhere,
Rezai (2012) tested 50 learners of English on the
use  of  articles  for  mass  nouns  and,  again,  the
results  substantiated  the  interpretability
hypothesis.  He  also  concluded  that  article
suppliance created more learnability problems in
the plural and indefinite mass contexts compared
to the count singular ones.
Considering  research  on  feature  reassembly
hypothesis,  on  the  other  hand,  numerous  studies
exist all of which point to the explanatory power
of the FRH for the acquisition failure of learners.
In  fact,  Montrul  and  Yoon  (2009,  p.  291)  called
selecting  and  reassembling  of  features  ‘the
logical problem of L2 acquisition’. Dominguez,
Archi  and  Myles,  (2011)  studied  aspect  in
Spanish where English learners had to overcome
a  morphological  contrast  between  the  two
languages  and  reported  that  cases  where  a
meaning-requiring  reassembly  was  necessary,
students  failed  even  at  the  advanced  level.  In
their  study  of  English  learners’  acquisition  of
Chinese  resultative  constructions  where  there
was  an  asymmetry  between  reconfiguration  of
syntactic  and  thematic  features,  Yuan  and  Zhao
(2008)  concluded  that  the  learners  were  able  to
reassemble  syntactic  structures  but  not  the
thematic  ones.  The  writers,  then,  suggested  that
reconfiguration  of  syntactic  and  semantic
features  did  not  develop  in  a  uniform  fashion  in
L2  acquisition.  Choi  (2009)  also  added  to  the
literature by studying the interpretation of wh-in-situ  expressions  in  L2  Korean  by  adult  native
speakers  of  English,  a  structure  that  posed
difficulty  to  English  learners  of  Korean.  She
adequately  showed  that  the  difficulty  for  such
learners  did  not  reside  in  parametric  selection
because  both  L1  and  L2  selected  the  relevant
features.  Instead,  the  learners  had  difficulty  in
reconfiguring  such  features  into  a  different  L2
configuration.  Other  areas  of  grammar  which
have  been  considered  ideal  for  testing  the
predictions  of  the  FRH  include  the  existential
quantifier  (Gil  &  Marsden,  2013),  wh-questions
(Muroya,  2013a),  accusative  clitics
(Shimanskaya  &  Slabakova,  2014),  VP  adverbs
(Muroya,  2013b),  past  participle  agreement
(Renaud,  2009),  inflectional  morphology
(Muroya, 2012), and spatial predicates (Stringer,
2012)  all  of  which  have  lent  support  to  the
hypothesis.
Concerning  studies  on  English  articles,  Lardiere
(2005)  presented  data  from  a  Chinese  learner  of
English (Patty) who had apparently been able to
use  the  article  system  to  mark  definiteness  and
number.  These  two  features  are  lacking  in
Chinese  and  the  researcher  argued  that  ‘the
acquisition of definiteness and plural marking is
not  a  matter  of  mere  parameter  resetting  from  a
minus  value  in  Chinese  to  a  plus  value  in
English.  Rather,  it  involved ‘a  more  painstaking
process  of  reassembling  the  relevant  features
from the way they were conditioned and realized
in  the  L1  to  that  of  the  L2’ (p.  183).  In  another
study, Kaku (2006) provided evidence by testing
a  group  of  L1  Japanese  learners  of  English.
Japanese does not have an article system but the
concept  of  in/definiteness  is  expressed  through
other  means  (demonstratives,  pronouns,  etc).  In
other  words,  the  two  features  required  for  the
English  articles  (definiteness  and  referentiality)
are  expressed  by  case  markers  and
demonstratives.  Using  an  elicitation  task  and  a
translation task, she showed that, overall, a good
proportion of articles had been supplied correctly
especially  by  the  advanced  learners.  The
researcher,  then,  suggested  that  it  was  plausible
that  the  features  of  the  English  articles  could  be  
 
acquired  (based  on  the  performance  of  the
advanced  students)  and  that  the  variability  in
article  choice  among  the  intermediate  learners
could  be  because  of  unsuccessful  reassembly  of
features  in  the  L2.  However,  this  study  was
based  on  data  from  only  five  participants  (three
advanced  and  two  intermediate)  so  it  is  difficult
to make generalizations based on it.
The  article  system,  therefore,  is  an  ideal
grammatical  category  for  comprehensive  testing
of the FRH since it involves lexical encoding of
more  than  one  feature.  We  believe  that
investigating  the  acquisition  of  articles  by
Persian speakers is a fruitful case because of the
variety  of  features  involved,  the  complexities
that  exist  in  the  Persian  article  system,  and  the
cross-linguistic  comparability  of  the  two
systems. 
IV. The focus of the present study
Previous  studies  on  Persian  L2  learners  of
English have documented the article system as a
major  barrier  to  native-like  proficiency  even  at
the  advanced  level  (Mobini,  2006;  Momenzade
&  Youhanaee,  2013)  and  have  sought  to
alleviate  the  problem  by  providing  pedagogical
suggestions  (Dabaghi  &  Tavakoli,  2009;
Pashazade  &  Marefat,  2010).  Nevertheless,
Persian is an article-based language which marks
definiteness  and  which  bundles  some  other
features  into  its  article  system  besides
definiteness.  The  main  aim  of  this  study  was,
then,  finding  out  whether  Persian  learners  of
English  are  able  to  reconfigure  the  assembly  of
features  from  their  L1  to  the  L2  and  if  so,  at
what  level  of  grammatical  proficiency.  Besides,
we were interested in finding out if Persian gives
clues  as  to  what  article  should  be  used  with
nouns in English. To that purpose, the following
questions were formed: 
  Are Persian EFL learners eventually able
to  reassemble  definiteness,  specificity,
genericity,  number,  and  countability  on
English articles?
  Is  there  any  order  of  acquisition  with
regard  to  the  features  related  to  article
use?
  Does  the  function  of  nouns  as
subject/object help correct use of  articles
in English?
V. Method
Participants
Forty  three  Persian-speaking  learners  of  English
at  Sheikhbahaee  University  took  part  in  the
present  study.  Initially,  they  took  the  Oxford
Placement Test (2001, version 1) based on which
their  level  of  grammatical  knowledge  was
determined  and  three  groups  of  grammatical
proficiency  were  identified.  In  the  highly
advanced  group,  learners  who  scored  55-60  on
the OPT were placed. They had an age range of
25-40.  This  group  included  11  people.  The
intermediate  group  consisted  of  those
participants  who  could  score  37-47  on  the  OPT.
They  were  14  in  number  and  19-27  years  old.
The  elementary  group  included  18  participants
who  scored  18-27  on  the  placement  test.  They
had  the  same  age  range  as  the  intermediate
group.  It  should  be  mentioned  that  wider  gaps
were  considered  between  the  groups  in  order  to
make  sure  that  they  belonged  to  different  levels
of  grammatical  knowledge.  Fifteen  native
speakers  were  also  included  in  the  study  as  the
control group. They were 9 males and 6 females
and  aged  8-34,  from  school  children  to  adults.
Actually, they were the only available natives to
the  researchers.  They  all  were  originally
Americans having lived in their home county all
their  lives  and  volunteered  to  take  our  English
test. Table 5 summarizes the specifications of the
participants.

Instruments
Data  elicitation  in  the  present  study  was  based
on  an  OPT  test,  a  grammaticality  judgment  test
and  a  translation  test,  the  description  of  each  is
presented in the following.
a. The Oxford Quick Placement Test (OPT)
The  quick  placement  test  used  in  this  study  is  a
flexible  test  of  English  developed  by  Oxford
University Press and Cambridge ESOL (2001) to
give  teachers  a  reliable  and  time-saving  method
of finding a student’s level of English. The paper
and pencil version of the test used here includes
60  multiple-choice  items  to  be  answered  within
30  minutes.  According  to  the  publishers,
elementary  students  are  the  ones  who  score  18-29 on the test. Intermediate learners score 30-47
and  those  who  score  within  the  range  of  48-54
are  considered  as  advanced.  Any  learner  who
scores above 55 (55-60) is a very advanced one.
The  test  also  has  a  writing  task  with  a  different
scoring  method  which  was  not  used  in  the
present study. 
b. The grammaticality judgment (GJ) test
To  tap  the  participants’  understanding  of
grammaticality/ungrammaticality  of  the  forms
associated  with  the  English  article  system,  a
grammaticality  judgment  (GJ)  test  was
developed  and  used  for  the  purposes  of  this
study.  The  test  included  100  items.  There  were
eleven categories on the test which measured the
five  features  that  have  an  effect  on  article  use.
For  each  category,  both  grammatical  and
ungrammatical  items  were  included.  The  reason
was  that  acquisition  implies  knowledge  of  both
grammatical and ungrammatical forms. So, only
when  learners  know  the  correct  and  incorrect
form  of  a  grammatical  item,  they  can  be  said  to
have  acquired  it.  To  distract  the  participants’
attention  from  the  point  that  was  tested,  some
filler items were also included. The categories on
the test and the number of items in each category
are presented in Table 6 below.

Two points have to be mentioned: first, an effort
was  put  into  considering  all  types  of  nouns  in
English with which an article should be used, so
we  came  up  with  eleven  categories.  Second,  the
number  of  ungrammatical  items  across
categories was not equal because the cases were
different. For example, considering plural nouns,
whether  definite  or  indefinite,  the  use  of  a  was
completely  out  of  the  question.  So,  the  number
of  ungrammatical  items  in  such  groups  was  4
(zero  article  +N  for  definite  plurals  and  the  +N
for  indefinite  ones)  whereas  for  singular  nouns,
more ungrammatical items could be included. To
further  clarify  the  point,  consider  the  sixth
No  Categories  Grammatical   Ungrammatical   Total 
          
category  (indefinite,  singular,  non-specific
nouns).  Ungrammatical  items  included  four
sentences  containing  the+N  and  four  including
zero  article+N.  Finally,  the  ninth  category
included  no  ungrammatical  sentences  simply
because  the  corresponding  ungrammatical  items
(zero  article  +N)  were  included  in  the  8th
category. So, there was no need to make the test
lengthier than it was. 
Each  item  on  the  test  included  two  sentences.
Following  the  stem,  the  test  takers  had  three
choices.  They  were  supposed  to  judge  the
grammaticality  or  ungrammaticality  of  the
second  sentence  in  each  pair  based  on  the  first
one. They were asked to choose √ if the second
sentence  was,  in  their  opinion,  grammatically
correct  and  *  if  incorrect.  They  were  also  given
the  option  of  choosing  ?  if they didn’t know or
were  not  sure  of  the  answer.  Examples  of  each
category  on  the  test  are  provided  below.  They
are all selected from the grammatical items:
1.  I  saw  an  accident  on  my  way  back  home.  A
man was injured in the accident. 
2. I’m going to the boss today to ask him for a
raise.  The  man  who  comes  with  me  will  not
regret it.    
3.  Mr.  Peterson  owns  two  houses.  He  tries  to
keep the houses in good condition.
4.  I  returned  from  the  shopping  center  to  find
that  my  car  had  been  punctured.  I  must  find
the wrongdoers whoever they are.       
5.  Alice  is  so  angry.  A  pig  has  stepped  on  her
front garden.
6.  My  pen  ran  out  of  ink.  Will  you  give  me  a
pen, please?
7.  Whenever  I  can’t  find  something,  my  dog
does it for me. Dogs are very intelligent.
8. When you are at a zoo, you should stay away
from  lions’  cages.  The  lion  is  a  dangerous
animal.
9.  Large  animals  usually  give  birth  to  their
children.  But  an  ostrich,  which  is  the  biggest
bird of all, lays eggs.
10.  What  would  you  like  best  for  desert?  Oh,  I
prefer ice cream to the rest.
11. We’ll never go to that restaurant again.  The
food was awful.
c. The translation test
The  test  was  made  up  of  72  items  and  included
categories  some  of  which  were  similar  to  the
ones  in  the  GJ  test.  There  were  still  some
categories  in  the  translation  test  that  were  not
found  in  the  other  test.  The  reason  was  that  this
test  was  constructed  based  on  the  concept  of
definiteness  and  how  it  is  assembled  with  other
features  in  Persian.  Therefore,  there  were  cases
which  existed  in  Persian  but  not  in  English  (see
Tables  1-4).  The  purpose  of  developing  the
translation test was to detect possible L1 transfer
effects  on  how  Persian-speaking  EFL  learners
used the article system in English. The following
table summarizes the details of the test:

The definite, indefinite, and generic items on the
test  were  evenly  distributed  for  number,
specificity,  countability,  and  function
(subject/object). Considering the nouns in object
position,  the  test  included  both  direct  objects
(with  the  object  marker  ra)  and  indirect  objects
(without  the  object  marker  ra)  so  that  the
influence  of  the  object  marker  in  correct  article
use  could  be  detected.  A  sample  item  from  the
test is:
وقتی از اتاق بیرون میرفتم بچه ها گریه کردند. 

The underlined word is plural, definite, specific
and in the subject position.  
 
Data Collection, scoring, and analysis
Prior  to  data  collection, a  pilot  testing  was  done
with  15  language  learners.  Few  inconsistencies
that  existed  in  the  tests  were  modified  and  an
approximate time limit was set for each one. The
language  of  instruction  was  decided  to  be
Persian.  Also,  two  experts  in  the  field  of
language  acquisition  scrutinized  the  tests  and
validated  the  content  of  each  one.  After  data
collection,  the  internal  consistency  was
measured  for  each  test,  using  a  Cronbach  alpha
coefficient. For the GJ test, the alpha coefficient
was  .96.  It  was  calculated  to  be  .87  for  the
translation  test.  Therefore,  both  tests  could  be
considered  reliable  with  the  specified  sample  of
the study. 
The  L2  learners  voluntarily  participated  in  the
study. Initially ninety-five learners took the OPT
based on which forty-three were selected and put
into  three  quite  distinct  levels.  The  participants
took  the  GJ  test  in  the  following  week.  On  the
third week, they were asked to do the translation
test. Because the tests were rather long and they
had to be done at times other than class hours, a
one  week  interval  was  observed  between  the
tests.  So,  the  participants  met  once  a  week  for
three  weeks.  The  second  test  was  the  GJ  test.
Each  correct  answer  was  worth  one  score  and
each incorrect one was given a zero score. Also,
for  each  incorrect  item  on  the  test,  only  if  the
participant  had  circled  *  and  supplied  the  right
answer,  s/he  was  scored  1.  To  put  it  differently,
just  circling  *was  not  enough  to  indicate  the
participant  necessarily  knew  which  part  of  the
sentence made it ungrammatical. For the correct
answers  as  well,  the  participants  had  to  mark  √
to  merit  1. They  weren’t granted any scores if
they  had  left  the  sentence  intact  or  if  they  had
chosen ?.On the translation test, the concern was
with the noun phrase in question. It didn’t matter
if  the  whole  sentence  was  translated  correctly.
Each correct answer was granted 1and the wrong
answers were given a zero. 
The  1/0  coded  data  was  submitted  to  the
Statistical  Packages  in  Social  Sciences  (SPSS,
16)  software  for  the  purpose  of  analysis.  For
each  category  on  the  tests,  the  mean  percentage
for each individual participant and later for each
proficiency  group  was  calculated.  Since  there
were  four  groups  of  participants  and  one
independent  variable  on  each  category,  one-way
between  groups  ANOVA  was  performed  as  the
proper  statistical  procedure  to  see  if  inter-group
differences existed with regard to the features in
question.  Wherever  necessary,  the  paired
samples  t-test  was  conducted  to  detect
significant  intra-group  differences  in
performance.  Also,  in  cases  where  comparisons
needed  to  be  made  between  two  groups,  the
independent samples t-test was used.  
 
VI. Results
The  first  research  question  asked  if  the  Persian
learners  could  eventually  reassemble
definiteness,  specificity,  genericity,  number,  and
countability  on  English  articles.  The  results  are
presented  separately  for  each  feature  so  that
group  performance  can  be  compared.  The  first
feature  to  examine  is  definiteness.  Table  8
presents results on the GJ test. 

For  the  sake  of  comparison,  a  one-way  between
groups  ANOVA  was  used.  It  showed  that  the
groups’  performance  on  this  test  was
significantly different from each other (Table 9).

At the p< .05 level, the four groups’ performance
on  the  GJ  test  was  significantly  different  from
each  other  both  on  the  definite  items  (F  =
160.49,  p  =  0.00,  r  =  0.8)  and  on  the  indefinite
items  (F  =  75.54,  p  =  0.00,  r  =  0.8).  A  Sheffe
post-hoc test was also conducted which revealed
that  the  learner  groups  were  significantly
different  from  each  other  and  from  the  native
group.  On  the  indefinite  items  of  the  test,
however,  the  advanced  group  could  catch  up
with  the  natives  as  the  difference  between  the
two  groups  was  not  significant  (p=  0.08).
Elementary  and  intermediate  learners  were
significantly  different  from  both  advanced
learners and native speakers. 
The  second  feature  involved  in  article  selection
was  specificity.  Table  10  summarizes
performance on [± specific] items. 

Table 10
Specificity (%) on the GJ Test

Performance  in  the  [definite  non-specific]
context  was  not  as  good  as  that  in  [definite
specific] for all proficiency groups which shows
that  our  participants  had  more  doubts  about
using  the  for  non-specific  definite  nouns  and
were  clearly  non-native-like  even  at  the
advanced  level.  Considering  the  indefinite
context, using a was unaffected by specificity for
the  elementary  group  as  no  difference  could  be
detected.  Evidence  for  these  conclusions  came
from  a  one-way  between  groups  ANOVA  the
results of which are tabulated below.

At  p<  .05,  there  was  a  statistically  significant
difference  among  the  groups  in  their
performance  in  the  four  contexts  (p=  0.00).  The
effect size (r) was large for all contexts. Post-hoc
comparisons  using  the  Scheffe  test  indicated
that,  in  all  four  contexts,  the  elementary  and
intermediate  groups  performed  significantly
different  from  the  native  speakers  (p  =  0.00  for
the  elementary  group  and  p  =  0.00  for  the
intermediate  group).  The  advanced  participants
performed  similar  to  the  natives  in  [definite,
specific]  and  [indefinite,  non-specific]  
(p  =  0.28  and  p  =  0.75  respectively).  In  the  two
suspected contexts of [definite, non-specific] and
[indefinite, specific], in which learners are prone
to  fluctuation,  even  the  advanced  learners
performed  significantly  different  from  the
natives (p = 0.00 and p = 0.00 respectively). 
The  next  feature  to  be  checked  is  genericity.
Performance  on  singular  and  plural  generic  NPs
is displayed in Table 12.

In  order  to  see  if  the  observed  differences  in
performance on singular and plural generics was
meaningful, a one-way between groups ANOVA
was  calculated.  The  results  (Table13)  revealed
significant  differences  between  the  groups.  The
effect size, using eta squared, was large for both
singular and plural generic nouns.

Further post-hoc calculation using a Scheffe test
revealed  that,  considering  singular  generic  NPs,
none  of  the  learner  groups  conformed  to  the
native  speaker  norm  as  there  was  a  significant
difference  detected  (p  =  0.00  for  all  learner
groups).  For  the  plural  generic  items,  however,
the  advanced  group  performed  similarly  to  the
native speakers (p = 0.00 for the elementary and
intermediate  groups  but  p  =  0.38  for  the
advanced group). 
The  next  feature  to  be  examined  is  number.
Group performance is presented in Table 14.

Again,  differences  in  performance  among  the
four  groups  were  tested  by  a  one-way  between
groups  ANOVA  which  pointed  to  significant
gaps (Table 15). 

In other words, at  p< .05, the participant  groups
were  significantly  different  from  each  other  in
using  the  correct  article  for  both  singular  (F  =
93.35,  p  =  0.00,  r  =  0.8)  and  plural  nouns  (F  =
79.40, p = 0.00, r = 0.8). In both cases there was
a large effect size: using eta squared, 80% of the
total  variance  could  be  accounted  for  by  the
groups’ knowledge state. The results of a post-hoc  Scheffe  test  also  revealed  that  the  learner
groups  differed  from  each  other  and  from  the 
 
native  controls  in  performance  on  both  singular
and plural nouns (p = 0.00 on both singular and
plural nouns for all groups). 
 
The  last  feature  to  put  under  scrutiny  is
countability. Only the participants’  performance
on  definite  items  was  analyzed  mainly  because
the  tests  did  not  include  indefinite  mass  nouns.
Table  16  summarizes  group  performance  on  the
two types of nouns:

At p< .05, the groups were significantly different
from  each  other  in  using  the  correct  article  for
both  count  (F  =  123.93,  p  =  0.00,  r  =  0.8)  and
non-count  nouns  (F  =  67.59,  p  =  0.00,  r  =  0.7)

based on a one-way between groups ANOVA.

Furthermore,  Persian  learners  were  significantly
different  from  the  native  control  group  in
performance  on  both  count  and  mass  NPs.  This
was  revealed  using  a  Scheffe  post-hoc  test  (p  =
0.00).
Based  on  the  results  presented  so  far,  it  is  clear
that  the  participants  in  this  study  could  not  be
claimed to have acquired the features in question
at  none  of  the  three  proficiency  levels,  so  the
answer to the first research question is negative.
To  arrive  at  an  answer  to  the  second  research
question  concerning  the  order  of  acquisition  of
the  features,  performance  of  the  participants  in
the  advanced  group  was  further  analyzed  to
compare  their  degree  of  mastery  over  the  five
features.  Mastery  was  defined  as  correct  article
use  over  90%  for  each  feature  in  question  in
which  case  no  significant  difference  could  be
detected  between  natives  and  non-natives.  The
findings on the GJ test are presented in Table 18.
The + mark shows that learner performance was
accurate over 90%

The  observation  we  made  was  that  definiteness
was  the  only  feature  in  which  five  of  the
advanced  participants  were  accurate.  Three
participants  could  satisfy  the  accuracy  criterion
in  article  use  for  the  features  of  specificity  and
number.  Three  participants  were  accurate  in
article use for generic expressions. And only two
of  the  participants  could  achieve  the  accuracy
criteria  for  the  feature  of  countability.  So,  the
order  of  definiteness  >  specificity/  number  >
genericity  >  countability  was  found  as  the
answer to the second research question. It has to
be mentioned that the learner participants in this
study,  as  will  be  discussed  in  the  following
section, could not be considered to have acquired 
 
those  features  so  the  order  that  is  found  is,  at
best, tentative.
The  last  research  question  was  formed  to  see  if
the  grammatical  function  of  the  noun  as  subject
or object in Persian would be of  any help to the
learners  in  appropriate  article  use  in  English.
Results  on  the  translation  test  can  help  arrive  at
an answer to that question. (Figure 1)

As the figure depicts, the participants were more
successful  in  indefinite  article  use  either  for
subject  or  object  NPs.  This  is  not  surprising
because  it  was  previously  concluded  that  the
Persian  learners  had  fewer  problems  using  the
indefinite article and were actually native-like at
the  advanced  level  (Table  8).However,
performance  of  the  learners  showed  that  they
were more successful in using the definite article
for  NPs  as  objects.  Although  the  results  of  a
series of paired samples t-tests showed that there
was  no  significant  difference  in  definite  article
use  for  subject/object  NPs  for  the  elementary
group  (t(17)=  1.25,  p
other  two  groups  had  a  better  performance  on
object  nouns  (t(13)  =  2.58,  p
for  the  intermediate  group  and  t(10)  =  3.12,
p
answer  to  the  third  research  question  is,  then
partially  positive.  NP  function  helped  the
participants  in  using  the  correct  definite  article
for  subject  NPs  at  higher  levels  of  grammatical
knowledge.  
 
VII. Discussion
The  first  question  in  this  study  concerned  the
acquisition  of  English  article  system.
Comparisons between our learner groups and the
native  group  revealed  that  Persian  learners  were
significantly  different  form  the  natives  in  their
article  use,  a  difference  which  was  observed  for
all  the  features  that  affect  article  use.  In  other
words,  the  study  revealed  that  Persian  learners,
regardless  of  their  L2  state  of  grammatical
knowledge,  were  non-native-like  in  their  article
use  for  ±definite,  ±specific,  ±singular,  ±generic,
and ±count nouns. These results were not in line
with  the  ones  reported  by  Rezai  and  Jabbari
(2010). They reported that their Persian subjects
were  able  to  acquire  the  definite  article  and,
hence,  could  provide  support  for  the
interpretability  hypothesis.  According  to  them,
definiteness  is  interpretable  in  both  English  (at
both LF and PF) and Persian (just at LF) and so,
it  should  not  pose  serious  problems  to  the
learners.  More  than  that,  following  Ghomeshi
(2003), we assume that there is a definite enclitic
in  the  spoken  form  of  Persian.  So,  definiteness
would  be  considered  as  interpretable  at  both  LF
and  PF  (quite  like  English)  and  it  should  not  be
an obstacle to learning anyway. Results from the
present study, however,  do not seem compatible
with Rezai and Jabbari (2010), henceforth R & J.
We  suggest  that  this  is  due  to  the  scope  and
methodology  of  their  study  and  the  present  one.
First,  R  &  J  considered  only  two  features  of
definiteness  and  specificity  whereas  we  looked
at  the  combination  of  three  more  grammatical
features  and  more  importantly  their  assembly  in
English  and  Persian  articles.  Second,  the  main
source  of  their  data  was  an  elicitation  task.  In
fact,  in  the  present  study  we  also  used  an
elicitation  task  the  results  of  which  showed  no
difference  between  native  and  non-native 
 
speakers, which is compatible with R & J results.
However,  as  we  found  significant  differences
between  the  performances  of  the  L2  learners  on
the other two tasks, we found that the elicitation
task  was  easy  and  its  results  did  not  perfectly
represent  the  status  of  knowledge  of  English
articles. This is not surprising given the fact that
the  context  of  English  L2  learning  in  Iran  is  an
instructed  setting,  where  English  is  taught
explicitly  in  most  schools  and  hence  all  Persian
learners  of  English  have  a  good  command  of
explicit  grammatical  knowledge.  Therefore,  we
concluded  that  this  pen  and  paper  task  tapped
only  the  learned  linguistic  knowledge  of  our
learners and decided to exclude its results. Third,
R  &  J  pooled  the  data  of  their  intermediate  and
advanced  learners,  which  actually  blurs  the
picture  additionally.  Finally,  the  lack  of  a  group
of  native  speakers  was  a  major  hindrance  to
make  detailed  comparisons  and  carry  out  in-depth  analyses.  In  fact,  considering  these
differences, we construe that R & J’s study does
not provide counter evidence to our study.   
The  results  of  the  present  study  then  indicated
divergence  between  native  speakers  and  non-native  L2  learners  due  to  morphological
variability  in  the  performance  of  L2  learners
which  seems  not  to  disappear  even  at  advanced
levels  hence  a  persistent  problem.  We  suggest
that  the  source  of  this  problem  could  be
morphological  feature  combination  and
conditioning  as  proposed  by  Lardiere  2005.  She
assumes  that  L2  learners  ‘initially  seek  the
morphological  equivalents  from  already
assembled  lexical  items  in  the  L1  to  analyze  L2
conditioning environment’ (2009,  p.  213).  Proof
of  this  assertion  was  found  in  our  study  where
the  participants  had  more  difficulty  in  using  the
definite  article  than  the  indefinite  one,  where
they  could  not  adequately  mark  [±  definite]  on
count/non-count  as  well  as  singular/plural  NPs,
and  where  they  could  not  use  the  correct  article
for  singular  generic  NPs.  We  would  like  to
argue,  then,  that  using  English  articles  to  mark
in/definiteness  by  Persian  learners  reflects  L1
transfer  effects.  This  effect  is  not  syntactic  in
nature,  though,  because  the  DP  structure  and
definiteness as a feature  exist in both languages.
There are definite, indefinite, and zero articles in
both  English  and  Persian.  And  article  use  is
affected by other features as specificity, number,
countability,  and  genericity  in  both  languages.
Despite all these cross-linguistic similarities, the
developmental  patterns  of  the  participants’
performance  did  not  bear  significant
resemblance  to  the  native  speaker  group.  They
obviously had problems in using English articles.
Why  would  such  learners,  regardless  of  their
language  proficiency,  have  persisting  problems
in article use? The answer relates to the fact that
the difference between the two languages is of a
morphological  nature  as  the  features  that  have  a
role  in  article  use  are  bundled  differently  and
into  different  lexical  items  in  English  and
Persian. The definite article (as a lexical item) in
Persian  is  not  overtly  realized  in  the  formal
language.  It  is  reasonable  for  Persian-speaking
learners  to  have  persistent  difficulty  with  the
English  definite  article.  Other  studies  have  also
pointed  to  the  difficulty  the  learners  face  when
one  or  both  of  English  articles  are  lacking  in
learners’  L1s  (Jaensch  &  Sarko,  2009;  Kaku,
2006).  As  for  the  indefinite  article,  it
encompasses  different  feature  combinations  in
Persian as well. The present study is also backed
up  by  Lardiere  (2005)  where  she  argues  that
knowledge  of  definiteness  feature  does  not
correlate  with  Definite  article  use  in  the
performance of a Chinese L2 learner of English.
Similarly,  Dominguez,  Archi  and  Myles  (2011)
claim  that  feature  reassembly  is  a  necessary
process  during  the  course  of  acquiring  a  second
language. The feature reassembly hypothesis can
best account for such problems as the one in the
present  study  because  it  does  not  consider  a
facilitative  role  for  the  L1  even  in  situations
where the point is seemingly identical across the
two  languages.  In  fact,  the  situation  for  L1
Persian learners in this study is exactly like this:
the  contexts  seem  to  be  identical  across  the  two
languages.  However,  the  participants’
performance  showed  that  the  identicality  is
superficial.  As  they  could  display  mastery  in
none  of  the  features  realized  in  article  use,
feature  reassembly  seems  to  provide  a 
 
reasonable  explanation  for  such  failure.  The
reason for lack of mastery in article use could be,
therefore,  the  combination  of  the  features  which
differed across the L1 and L2 and the inability of
the  learners  to  recombine  those  L1  features  so
that they would conform to the L2 configuration. 
The second research question, which asked about
the  order  of  acquisition  of  the  article  related
features, cannot be definitely answered based on
the  data  from  this  study  as  it  was  revealed  that
the  learners  at  all  three  levels  of  grammatical
knowledge  had  not  achieved  full  mastery  over
article  use  as  English  native  speakers.  Being  so,
no order of acquisition could be arrived at. It has
to  be  mentioned,  however,  that  our  participants
were  more  proficient  in  using  the  appropriate
article  to  mark  definiteness  and  specificity  on
English NPs and that they had the most difficulty
in  dealing  with  the  feature  of  countability.
Reasons for the observed pattern can be traced in
how  the  two  languages  realize  each  feature.
They  both  have  means  to  show  definiteness  and
specificity  on  nouns  (Table  1)  whereas  the
realization of definiteness on non-count nouns is
more  complicated  because  of  cross  linguistic
differences (Table 4). Also, it was elaborated on
in Table 2 that the two languages stand in sharp
contrast as to their generic forms. Again, generic
NPs in Persian are bare while this is not the case
in  English.  In  other  words,  difficulty  in  article
use occurred in situations where either the means
were different in the two languages or they were
absent in one language and present in the other. 
 
Moreover,  performance  of  these  participants
showed  their  sensitivity  to  NP  function  within
the  sentence  which  was  the  answer  to  the  third
research  question.  They  did  much  better  in
contexts where the NP in question was in object
position. Performance on subject NPs was not at
all  comparable  to  object  NPs,  showing  that  NP
function  affected  article  use  at  least  for  the
participants  in  this  study.  The  effect  of  noun
function  on  article  use  by  learners  from  other
language  backgrounds  is  yet  unexplored  since  it
is  not  documented  by  any  study  so  far.  In  case
that  no  such  effect  is  found  in  other  languages,
then,  reassembly  of  features  would  be  more
difficult  for  Persian  learners  because  they  have
to  additionally  be  sensitive  to  noun  function.
This  implies  that,  for  some  learners  with  a
specific  language  background,  reassembly  of
features would be intertwined with other aspects
of  grammar  (for  instance,  NP  function).  So,
reassembly  would  be  language  specific,  making
it more difficult for some learners but not others.   
 
VIII. Conclusion
The  central  goal  of  this  study  was  to  find  out  if
Persian-speaking  learners  of  English  were  able
to use the article system in a native-like fashion.
Analysis  of  the  data,  collected  through  two
different  measures,  showed  that  the  participants
had  not  mastered  appropriate  article  use  even  at
the very advanced level. Explanation was sought
from  the  feature  reassembly  hypothesis
according  to  which  the  source  of  such  failure
was  unsuccessful  reconfiguration  of  the  features
related  to  article  use  from  the  way  they  were
bundled  in  Persian  into  L2  English.  Our  data
were  in  line  with  Lardiere’s  (2008,  p.  114)
assertion  that  ‘persistent  L2  variable
morphological  production  (omission  and/or
faulty  use)  is  attributable  to  differences  in  the
‘conditioning environment’ for  the  assembly  of
features into lexical items between L1 and L2’.
The  findings  of  this  study  indicate  that  the
problem Persian-speaking learners faced was not
acquiring  new  features  but  rearranging  the
already  existing ones  into different lexical items
in their L2. Elsewhere, Lardiere (2009) suggests
that  ‘if  a  feature  contrast  is  detectable,  it  is
eventually  acquirable’  (p.  214).  Since  the
advanced learners in this study were shown to be
highly  proficient  in  other  areas  of  English
grammar  (scored  above  55  on  the  OPT),  the
question still remains as to how long they would
still  need  to  figure  out  the  contrasts  and  to  be
able  to  perform  like  a  native  speaker.  With  this
goal  in  mind,  attention  should  now  be  turned  to
conducting studies that focus on the pedagogical
aspect  and  look  for  more  fruitful  methods  of
instruction. The present study also paves the way
for  future  studies  that  focus  on  why  the  NP  
 
function  affects  article  use  or  on  other  possible
aspects of grammar that may play a role.