The development of a Persian reading span test for the measure of L1 Persian EFL learners’ working memory capacity

What is working memory?
Working memory (WM) can be defined as a
cognitive workspace with a limited capacity
pool  of  attentional  resources  for  the
temporary  storage  of  information  while
performing higher order cognitive tasks such
as  reasoning,  learning  and  comprehension
(Baddeley  &  Hitch,  1974;  Baddeley  &
Logie,  1999).  Baddeley  and  his  colleagues
view  WM  as  that  which  simultaneously
maintains and processes the input it receives
through  different  channels  of
communications  (e.g.,  touch,  long-term
memory,  sight,  and  hearing)  (Baddeley,
1986, 1996, 2003, 2007; Baddeley & Hitch,
1974;  Baddeley  &  Logie,  1999;  Gathercole
&  Baddeley,  1993).  A  three-component
model  of  WM  was  proposed  by  Baddeley
and  Hitch  (1974).  This  model  consists  of  a
central  executive  and  two  “slave”
components,  the  phonological  loop  and  the
visuo-spatial  sketchpad.  This  model  was  in
use until 2000, when Baddeley added a new
component  to  it,  the  episodic  buffer,  to
account for the studies on densely amnesiac
patients  with  long-term  memory  deficits.
This  model,  as  shown  in  Figure  1,  specifies
a  functional  role  of  memory  as  well  as  an
economical  and  coherent  account  of
information on each memory component.

Baddeley’s (2000) model of WM, revised to
incorporate  links  with  long-term  memory
(LTM)  by  way  of  both  the  subsystems  and
the newly proposed episodic buffer.
 
The  most  important  component  in  this
model is the central executive or supervisory
attentional  system,  which  is  a  limited
capacity  pool  of  general  resources.
According to N. Ellis, (2001), “It regulates
information  flow  within  WM,  activates  or
inhibits  the  whole  sequences  of  activities,
and resolves potential  conflicts between on-going  schema-controlled activities” (p., 33).
The  reading  or  listening  span  tests  are
usually  used  to  measure  central  executive
and give an index for WM.  
 
The  phonological  loop  is  in  charge  of  the
temporary  storage  and  processing  of  verbal
information. It plays a role as a phonological
store  by  holding  phonological
representations of auditory information for a
brief  period  of  time,  and  as  an  articulatory
rehearsal  system  by  enabling  the  reader  to
use  inner  speech  to  refresh  the  decaying
representations  in  the  phonological  store
(Baddeley,  2000,  2007;  N.  Ellis,  2001).
Phonological  loop  is  often  measured  by
presenting  spoken  lists  of  words  (word
span), digits (digit span) or non-words (non-word span), and asking participants to recall
the  lists  of  words  and/or  digits  in  the  order
in  which  they  are  presented.  The  maximum
number  of  items  that  the  individual  can
correctly  recall  is  considered  to  be  their
phonological memory score.
 
The  visuo-spatial  sketchpad  is  an  interface
between  visual  and  spatial  information
received  either  through  the  senses  or  from
long-term  memory  (Baddeley  &  Hitch,
1974,  p.,  854).  It  is  also  involved  in
generating  visual  images,  temporarily
maintaining  them,  and  manipulating
information  with  visual  or  spatial
dimensions. Furthermore, it can be activated
by  spoken  words  by  using  long-term
knowledge to convert the auditory presented
words  into  visuo-spatial  code  (Baddeley,
2007;  N.  Ellis,  2001).  To  measure  visual
memory,  Della  Sala,  Gray,  Baddeley,
Allamano  & Wilson’s (1999) pattern span
test  is  usually  used  by  researchers.  In  this
test,  the  individual  is  presented  with  2  x  2
matrixes,  with  two  of  the  cells  filled.  Then
after  3  seconds,  the  individual  is  asked  to
indicate  which  cells  were  filled  in  the
stimulus  matrix,  using  an  empty  2  x  2
matrix.  The  size  of  the  matrix  is  increased
by  two  cells  every  three  trials,  with  half  of
the  cells  of  each  matrix  being  randomly
filled.  The  individual’s  pattern  span  is
determined  by  the  maximum  number  of  the
cells  that  the  participant  is  able  to  recall
correctly.
 
The  Corsi  Block  task  is  typically  used  to
measure  spatial  memory  (Milner,  1971).  In
this  test,  the  subject  is  presented  with  an
array  of  nine  cubes  arranged  at  random
locations  on  a  board  placed  between  the
tester and the participant. The test starts with
the tester initially tapping two of the blocks
one  after  the  other  and  then  asking  the
subject  to  imitate  the  sequence.  The
sequence  of  taps  gradually  increases  to  a
point at which performance breaks down.
 
The  episodic  buffer  (Baddeley,  2000)  is  a
limited  capacity  temporary  storage  system.
According to Baddeley (2007), “It combines
information  from  the  loop,  the  sketchpad,
long-term  memory,  or  indeed  from
perceptual input into a coherent episode” (p.,
148). Moreover, it plays a role in interfacing
between  WM  and  long-term  memory
through  the  central  executive,  interacting
phonological  loop  and  sketchpad.  It  is  also
proposed  that  retrieval  from  the  episodic
buffer  is  through  conscious  awareness.
However,  no  method  of  measurement  has
 
 
been  proposed  yet  to  assess  the  episodic
buffer (Baddeley, 2007).
 
Rationale of the study
Since an important role for working memory
has  been  found  in  the  first  language
acquisition  (e.g.,  Daneman,  1991;  Daneman
&  Green,  1986;  Waters  &  Caplan,  1996),
research  on  the  role  of  working  memory  is
emerging  as  an  area  of  concern  for  second
language  acquisition  (e.g.,  Atkins  &
Baddeley, 1998; Miyake & Freidman, 1998;
Robinson, 2002, 2005). Working memory is
typically  measured  by  a  reading  span  test
(RST) or listening span test in L1 or L2.  
 
The Reading span tests were first introduced
by Daneman & Carpenter (1980). They were
used  to  measure  and  give  an  index  for
working  memory  capacity  (WMC).  In  a
reading  span  test  (RST),  participants  are
asked to read sets of sentences, report on the
semantic  acceptability  of  each  sentence
(processing  assessment),  and  then  recall  the
final word of  each sentence when prompted
(storage  assessment).  Since  the  introduction
of  the  RST  by  Daneman  and  Carpenter
(1980),  many  researchers  have  used  either
Daneman and Carpenter’s original RST or
the modified versions of that in their studies
(e.g.,  Alptekin  &  Erçetin,  2009;  Chun  &
Payne,  2004;  Daneman  &  Carpenter,  1980;
Harrington  &  Sawyer,  1992;  Lesser,  2007;
Osaka  &  Osaka,  1992;  Swanson,  1993;
Walter,  2004).These  studies  measured  WM
either  through  an  L1  RST  (Chun  &  Payne,
2004; Lesser, 2007), an L2 RST (Alptekin &
Erçetin,  2009),  or  both  L1  and  L2  RSTs
(Harrington  &Sawyer,  1992;  Walter,  2004).
As  prior  research  indicated  that  WM  is
language  independent  (e.g.,  Miyake  &
Freidman,  1998;  Osaka  &  Osaka,  1992;
Osaka,  Osaka  &  Groner,  1993),  measuring
WM  in  L1  was  then  became  popular  in
cognitive  psychology  and  studies  in  second
language  learning.  This  could  also  help  to
avoid  conflating  WM  and  L2  proficiency.
However,  while  there  may  be  considerable
number  of  L1  RSTs  for  some  languages;
there  are  few  L1  RSTs  in  some  others.  In
Persian,  there  may  be  few  reliable  versions
of  RST,  and  if  any,  none  of  them  has  been
published  or  accessible  for  the  use  in  other
L2 studies. This issue points to the need for
the  development  of  a  RST  in  this  language
for  the  research  with  L1  Persian  EFL
learners.  The  present  study  focused  on  the
process of development and validation of an
L1  Persian  RST  for  the  use  in  second
language learning studies. More specifically,
this study describes the stages at which RST
items  were  developed,  piloted,  revised,  and
finally  employed  in  the  research  with  L1
Persian participants.
 
Methodology
Subjects
74  L1  Persian  EFL  learners  at  three
proficiency  levels  participated  in  three  pilot
studies.  Then  the  newly  developed  test  was
administered to 140 L1 Persian EFL learners
in  an  experimental  study.  They  included
both  males  and  females,  16-35  years  old,
studying  English  as  a  foreign  language  in  a
private language school in Iran.
 
Material
A  corpus  of  Persian  sentences  was
constructed  by  an  expert  in  the  Persian
language.  The  sentences  contained  general
information, and lacked of any technical and
scientific  content.  64  sentences  were
selected  from  this  corpus  to  form  the  RST.
This  test  included  10  practice  session
sentences and 54 test sentences, all of which
were  in  an  active  and  affirmative  form
within  a  range  of  13-16  words.  Half  of  the
sentences  were  constructed  as  ‘nonsense’
sentences.  This  was  done  by  rearranging  a
few words in such a way that sentences were
semantically  anomalous  (Chun  &  Payne,
2004;  Harrington  &  Sawyer,  Lesser,  2007,
                          
Turner  &  Engle,  1989;  Waters  &  Caplan,
1996).  This  was  to  make  sure  that  the
participants  processed  sentences  for
meaning  without  focusing  only  on  the
retention  of  recall  items.  This  test  was
administered individually using a computer-based  format.  Because  Persian  sentences
follow  SOV  syntax  (the  sentences  initiate
with  a  subject  followed  by  an  object  and  a
verb  respectively),  each  sentence  ends  in  a
verb,  similar  to  the  reading  span  tests  in
Japanese  (Osaka  &  Osaka,  1992)  and
German  (Osaka  et  al.,  1993;  Roehr  &
Ganem-Gutierrez,  2008).  Each  verb
appeared  only  once  in  the  test.  Therefore,
the final words in this test were 64 different
verbs.  The  verbs  in  each  set  were  not
semantically  related.  The  sentences  in  the
test  were  arranged  in  three  sets  of  3,  4,  5,
and  6  sentences.  Half  of  the  sentences  in
each set were nonsense.
 
Test procedure
After  the  initial  form  of  the  RST  was
developed,  three  pilot  studies  were
administered  to  three  groups  of  L1  Persian
EFL  learners.  This  was  to  identify  the
potential  problems  with  the  test.  Then  the
newly  developed  test  was  used  in  an
experimental  study  for  the  measurement  of
working memory capacity.  
 
The  test  was  in  a  PowerPoint  format  and
was taken individually. It assessed two WM
components,  processing  and  storage  (e.g.,
Chun  &  Payne,  2004;  Daneman  &
Carpenter,  1980;  Harrington  &  Sawyer,
1992;  Lesser,  2007;  Waters  &  Caplan,
1996).  The  participants  had  to  read  each
sentence, judge whether or not it made sense
and  say  their  judgment  aloud  while  their
answer was recorded. This was the measure
of  WM  processing.  They  also  had  to
remember the last word of each sentence up
to  the  end  of  the  set  until  a  visual  prompt
(three  hash  keys)  along  with  a  two-second
auditory  prompt  appeared  on  the  computer
screen. The pilot study results suggested that
these  two  simultaneous  prompts  could  well
put  a  clear  boundary  between  the  sets  and
help  the  participants  not  to  miss  the  recall
time.  At  this  time,  the  participants  had  to
recall the sentence-final words and say them
out  loud  while  their  answers  were  recorded
by  the  researcher.  This  was  the  measure  of
the  WM  storage  component.  To  control  the
recency  effect,  the  participants  were
required  to  recall  the  words  in  the  order  in
which  they  appeared  (Baddeley  &  Hitch,
1993; Waters & Caplan, 1996).
 
A test instruction  guide  followed by  an oral
explanation  which  included  an  example  set
of  three  sentences  was  given  to  the
participants prior to the test. Then they were
given  a  practice  session  consisting  of  10
sentences  in  two  sets  of  three  and  a  set  of
four  sentences.  Then  the  test  began  with  a
set of 3 sentences, and as the test progressed,
the  number  of  sentences  presented  on  each
trial  increased  successively  from  three  to
six, with three trials being presented at each
series  length.  The  prompt  slide  transitions
increased accordingly from 12 to 18 seconds
based on the length of each set.  
 
Pilot studies
To  identify  the  potential  problems  with  the
RST,  three  pilot  studies  were  administered
to  three  different  groups  of  L1  Persian  EFL
learners.  In  the  first  pilot,  a  group  of  12  L2
participants completed the RST, followed by
a  retrospective  report.  In  their  retrospective
report,  they  all  reported  that  the  transition
time, 6 seconds, for each slide was too short
to  read  through  the  sentence.  They  also
wrote  that  a  few  sentences  were  too  vague
for  them  to  determine  whether  they  made
sense or not. The results of an item analysis
indicated  that  there  were  some  poor  test
items  in  the  test.  They  were  identified  as
being  too  difficult.  These  results  indicated
 
 
that  the  participants  had  performed  poorly
on  both  the  processing  and  recall
components.  The  sentences  which  the
students  had  identified  as  too  vague  were
located  among  the  ones  which  had  been
identified  as  too  difficult  by  the  item
analysis.  In  consultation  with  the  Persian
language expert, these sentences were either
revised  or  replaced  with  new  sentences.
Then  the  transition  time  for  each  slide  was
increased to 8 seconds as well.
 
In  the  second  pilot  study,  similar  to  the
procedure in the first pilot study, a group of
18  L1  Persian  EFL  learners  completed  the
revised  RST  followed  by  a  retrospective
report.  In  their  retrospective  report,  they
wrote  that  they  had  had  sufficient  time  to
read through the sentence on  each slide  and
even  rehearse  the  sentence  final  words
(target).  They  also  reported  a  case  where
two  sentence  final  words  were  semantically
related,  and  they  had  been  able  to  make  an
association  between  them  for  better  recall.
The  results  of  this  study  supported  the
participants’ claims. Their performance on
the RST was better than the prior group’s.
Most  of  them  were  also  able  to  obtain  the
scores  for  the  two  semantically  related
targets.  Since  the  participants’  rehearsing
could  have  inflated  the  recall  scores,  the
transition  time  for  each  slide  was  decreased
to  7  seconds.  Furthermore,  one  of  the
sentences  including  a  semantically  related
word  was  replaced  with  a  new  sentence
including  a  different  target  word.  The  new
sentence was developed and proposed by the
same Persian language expert.  
 
In  the  third  pilot  study,  the  revised  reading
span test was administered to a  group of 44
participants.  They  reported  that  the
transition  time  for  each  slide  was  just
enough  to  read  the  sentence  through  and
decide whether it made sense or not. No one
reported  any  opportunity  for  rehearsing  the
targets.  Moreover,  they  believed  that  all
sentences  throughout  the  test  had  been
neither  too  easy  nor  too  difficult  for  them.
The  results  of  the  item  analysis  also
indicated  that  each  item  made  a  good
contribution  to  the  test.  The  internal
reliability  for  this  test,  as  indicated  by
Cronbach’s Alpha, was .834 & .737 for the
RST recall and processing respectively.  
 
Application  of  the  newly  developed
reading span test in L2 research
The  final  test  was  used  in  an  experimental
study  conducted  by  the  researcher.  This
study  investigated  the  relationship  between
WM  and  L2  reading  ability  on  140  L1
Persian  EFL  learners  at  three  proficiency
levels.  The  sentences  in  the  test  were
arranged  in  three  sets  of  3,  4,  5,  and  6
sentences.  Half  of  the  sentences  in  each  set
were  nonsense.  Each  sentence  appeared  on
screen  for  7  seconds,  when  the  computer
transitioned to the next slide. After each set,
a  slide  with  3  hash  keys  and  a  two-second
auditory  prompt  appeared.  This  was  to
signal  to  the  participants  to  recall  the  final
word of each sentence in the set.
 
To score the test, one mark was allocated to
the participants’ correct judgment and one
mark to their correct recall of the test session
items, with the total of 54 each. Thus, since
there  were  54  sentences  across  all  the  trial
sets,  the  range  of  the  participants’
processing  and  recall  scores  was  between  0
and  54  for  each  participant.  No  marks  were
given to the practice session items. This was
consistent with the scoring method in recent
studies  (e.g.,  Alptekin  &  Erçetin,  2009).
Then a composite WM score was used as an
indicator of the participants’ WMC (e.g.,
Lesser,  2007;  Waters  &  Caplan,  1996).  The
composite  WM  was  obtained  by  adding  the
processing  and  recall  z-scores.  This  is  a
more  reliable  scoring  method  of  WMC
compared  to  the  traditional  span  scores  that
                          
quantify  the  highest  set  size  completed  or
the  number  of  words  in  correct  sets
(Freidman  &  Miyake,  2005).  An  item
analysis was conducted on this measure. The
internal  reliability  for  this  measure,  as
indicated by Cronbach’s Alpha, was .844
and  .790  for  the  RST  processing  and  recall
respectively.  This  suggests  that  the  newly
developed RST is reliable enough and could
be  used  for  the  measurement  of  WM  in
future studies.   
 
Conclusion
This  study  described  developing  an  L1
Persian  reading  span  test  for  the
measurement of L1 Persian EFL learners’
working  memory  capacity.  The  Persian
reading span test was developed, piloted and
successfully  used  in  a  study  with  140
participants. As the internal reliability of this
measure was quite high, the test can be used
to  measure  working  memory  capacity  in
future second language learning studies. The
same  procedure  could  also  be  used  to
develop  a  reading  span  test  for  speakers  of
other languages.