Genre analysis of literature research article abstracts: A cross-linguistic, cross-cultural study

Introduction
Genre  analysis  has  been  in  the  limelight  for
more  than  two  decades.  The  increasing
interest  in  this  discipline  is  motivated  by  a
need  to  supply  models  of  academic  and
scientific  texts  for  the  students,  so  that  they
can  produce  those  texts  appropriately.
Likewise,  scholars  and  scientists  need  to
communicate  their  ideas  and  findings  using
publications,  and  it  requires  them  to  have  a
full  grasp  of  the  discourse  community’s
conventions  (Martin,  2003).  English  has
long  been  established  as  the  language  of
scientific  communication  (Flowerdew  &
Dudley-Evans,  2002);  it  is  a  norm  for
journals  published  in  any  other  language  to
require  the  authors  to  provide  English
abstracts  for  their  articles  (Lores,  2004).
Unfortunately,  lack  of  knowledge  of  text
structures  and  audience  expectations  has
caused  non-native  writers  to  be  relatively
unsuccessful in the international community
(Connor, as cited in Martin, 2003).
 
Since  Swales’s  (1981)  work  RA
Introduction  section,  there  have  been
numerous studies on different sections of the
RA,  such  as  Brett  (1994)  and  Williams
(1999) on the Results; Hopkins and Dudley-Evans  (1988),  Holmes  (1997)  and  Fallahi
and Erzi (2003) on the Discussion section.
 
The  abstract  is  one  of  the  most  important
sections  of  the  RA;  it  can  determine  the
acceptance  or  rejection  of  an  article  for
conferences,  and  its  selection  by  readers.
Nevertheless,  as  Swales  (1990)  also
maintains,  the  abstract,  as  a  genre,  has
received  insufficient  attention  from  the
researchers.
 
A  number  of  genre  analysis  studies  have
been  conducted  on  RA  abstracts  in  English
(Anderson  &  Maclean,  1997;  Cross  &
Oppenheim,  2006;  Duncan,  2008;  Lores,
2004;  Salager-Meyer,  1991;  Stotesbury,
2003)  and  other  languages  such  as  German
(Busch-Lauer,  1995)  and  Spanish  (Martin,
2003);  yet  it  seems  that  abstracts  in  the
Persian language have been left unexplored.
Moreover,  one  of  the  fields  highly
disregarded  by  genre  scholars  is  literature.
In  fact,  only  two  studies  (Afshari,  2005;
Stotesbury,  2003)  relating  to  literature  were
known to the researchers.  
 
Lores  (2004)  hypothesized  that,  despite  the
general  belief  that  all  abstracts  follow  the
IMRD  structure,  there  are  two  types  of
abstracts  which  she  named  "informative"
and  "indicative",  and  suggested  that  while
the  former  conformed  to  IMRD,  the  latter
corresponded  with  CARS  model.  However,
these  models  have  never  been  applied  to
literature  abstracts  to  see  whether  they  can
thoroughly  represent  their  generic  structure.
The  only  report  is  the  study  conducted  by
Stotesbury  (2003)  which  indicated  a
“different rhetorical structure and style” (p.
330) in literature abstracts, but the degree of
difference was not the focus of the study.
 
In  addition,  Martin  (2003)  links  the
deviations  he  found  in  Spanish  abstracts
from  the  international  standards  to  “the
relationship  between  the  writer  and  the
discourse  community  s/he  addresses,  which
is  different  both  in  terms  of  numbers  and
expectations” (p. 42). Similarly, Tahririan
and  Jalilifar  (2004)  speculate  that
“sociocultural  factors  condition  the  way
academic  writers  write  abstracts” (p. 140).
In  the  case  of  Persian  writers  writing  in
English,  there  are  studies  (Afshari,  2005;
Tahririan  &  Jalilifar,  2004)  revealing  some
deviations  from  the  international  standards,
but  since  no  study  has  investigated  the
norms  and  conventions  of  Persian  writing,
these  deviations  can  never  be  confidently
linked to cultural differences.
 
The  present  study  aimed  at  filling  the
abovementioned  gaps  in  the  literature  by
comparing  the  generic  standards  of  Persian
and  international  communities,  and  by
discovering  to  which  community  the
Persian-speaking  writers  writing  in  English
belong.  Besides,  the  predictive  value  of
CARS  and  IMRD  models  was  examined  to
evaluate  their  appropriateness  for  literature
abstracts.
 
Background  
Genre  was  first  introduced  in  the  area  of
ESP  in  the  1980s.  Various  influences  on
Genre Analysis have been listed by scholars,
namely  the  examination  of  children’s
writings  in  Australia,  composition  studies
and new rhetoric in North America, and also
Miller’s (as cited in Paltridge, 2007, p. 931)
notion of “genre as social action”.
 
Definition of genre
Defining  genre  is  a  “fuzzy” task (Swales,
1990).  Traditionally,  the  word  indicated
various  kinds  of  literary  and  artistic  works;
however,  its  use  was  extended  by  linguists
to  cover  “classes  of  language  use  and
communication in all areas of life” (Allison,
1999, p. 144).  
 
For  Swales  (1990,  p.  58),  a  genre
“comprises a class of communicative events,
the  members  of  which  share  some  set  of
communicative  purposes.”  For  him,
particular  genres  share  similarities  in  their
structure,  style,  content,  intended  audience,
and rhetorical movement.
 
Elements in genre
Genres,  as  perceived  in  linguistic
approaches,  are  characterized  in  terms  of
communicative  functions  they  serve,  and
can be analyzed  into “generic  structures”
(Flowerdew  &  Dudley-Evans,  2002)  or
obligatory  and  optional  elements  which
comprise  these  functions.  Swales  (1990)
classified these elements as follows:
 
1. Moves
Moves represent the writer’s social purpose
and  include  steps.  Move  is  defined  by
Nwogu (1997) as “a text  segment  made  up
of  a  bundle  of  linguistic  features  .  .  .  which
give  the  segment  a  uniform  orientation  and
signal the content of discourse in it” (p.
122).
 
2. Steps
Steps  are  optional  textual  elements,  which
may or may not exist in any specific text.
 
Pedagogic potentials of genre analysis
Genre  Analysis  is  known  for  its  various
pedagogic  implications.  Kay  and  Dudley-Evans (1998) asserted that genre is a “very
powerful pedagogic tool” because it defines
the  kinds  of  discourse  the  students  need  to
be  able  to  produce,  and  also—considering
its  social  context  and  purpose—it  can
explain “why a discourse is the way it is” (p.
310).
 
Poole (2002), too, considered genre analysis
to  be  the  “best-realized  link  between
discourse  analysis  and  contemporary  L2
pedagogy” (p.  76)  because  it  aids  writing
instructors via  yielding analyses of different
academic texts, and also helps them provide
appropriate  discourse  awareness  for  their
students.  
 
Cross-cultural  and  cross-linguistic  studies
involving Persian
The  number  of  cross-cultural  genre  studies
of  different  sections  of  RA  written  by
Persian  native  speakers  is  relatively  small
(Atai  &  Falah,  2005;  Atai  &  Sadr,  2006;
Bandary,  1999;  Keshavarz,  Ataei,  &
Barzegar,  2007),  and  the  researchers  have
encountered  only  one  instance  of  such
studies  focusing  on  abstracts:  Tahririan  and
 
Jalilifar  (2004)  conducted  a  study  on
Applied  Linguistics  abstracts  written  by
native speakers of Persian, English and other
languages,  and  speculated  that  the
differences might be linked to first language
interference and sociocultural factors.
 
Thus  far,  the  researchers  have  come  across
only  one  cross-linguistic  genre  study
involving Persian RAs: Falahi Moghimi and
Mobasher  (2007)  studied  the  Introduction
section  of  60  English  and  60  Persian
Mechanical  Engineering  RAs,  and  found  a
significant  difference  between  the
frequencies of steps of these two categories.
No  study  on  Persian  abstracts  has  been
conducted.
 
This study
To  fill  the  abovementioned  gaps  in  the
literature, assess the predictive power of the
existing  models,  and  investigate  the  cross-linguistic,  cross-cultural  similarities  and
differences  of  RA  abstracts  in  the  field  of
literature,  two  questions  were  addressed  in
the present study:
 
1)  Is  there  any  association  between  the
frequency  of  moves  used  in  the
corpus,  i.e.  abstract  sections  of
Persian  research  articles  written  by
Persian  native  speakers  (PPs),
English  articles  written  by  Persian
native  speakers  (EPs),  and  English
articles  written  by  English  native
speakers  (EEs)  in  the  field  of
literature, and the models for abstract
(IMRD and CARS)?
 
2)  Is  the  frequency  of  the  moves
incorporated into the abstracts of EPs
the same as those of the PPs or those
of the EEs?
 
Corpus
A total of 90 abstracts were employed in this
study. The corpus in Persian consisted of 30
literature  abstracts written by  Persian native
speakers  and  published  in  Iranian  academic
journals.  Ten  journals  were  randomly
selected, and three articles were chosen from
each.  
 
The  English  corpus  was  composed  of  two
groups.  The  first  group  included  30
literature abstracts, written by Persian native
speakers,  and  published  in  Iranian  journals.
Four  journals  were  chosen;  nine  articles
were  picked  out  from  the  first  journal;
another nine from the second; eight from the
third, and four from the fourth.
 
The  second  group  consisted  of  30  abstracts
in  the  discipline  of  literature  written  by
English  native  speakers  and  published  in
international ISI journals. Ten journals were
decided  upon,  and  three  articles  were
extracted from each.
 
All  the  journals  were  available  online,  and
the  articles  were  chosen  from  the  most
recent  issues  of  the  journals.  The  articles
were  all  checked  in  terms  of  the  nationality
of their authors and those written by Persian
and  English  native  speakers  were  selected.
Table  1  below  presents  a  summary  of  the
corpus characteristics.

 Procedure
The  analysis  of  the  data  was  carried  out  in
two  main  stages.  The  IMRD  model  for
informative abstracts and Swales’s  (2004)
CARS model for Introduction section which
applies  to  indicative  abstracts  as  mentioned
by  Lores  (2004)  were  used  as  the  basis  of
analysis.  
 
In the first stage, the abstracts were scanned
for  the  presence  of  Introduction,  Method,
Results,  and  Discussion  moves,  following
Lores’s definition of each (Table 2). 

In  the  second  stage,  the  abstracts  were
checked  against  Swales’s  (2004)  CARS
model  (Table  3)  to  test  its  predictive  value
for literature abstracts.

It is to be noted that each of the I, M, R and
D  sections  corresponds  to  the  following
models,  respectively:  Swales’s  (2004)
CARS model for Introduction, Lim’s (2006)
for Method, Brett’s (1994) for Results, and
Hopkins  and  Dudley-Evans’s (1988) model
for  Discussion.  These  models  were  later
utilized  in  a  step  analysis  to  find  out  if  any
steps  existed  in  the  literature  abstracts  not
predicted by the models.
 
Sample analysis
An  instance  of  the  move  analysis  of  an
English abstract written by an English native
speaker will be presented below.  
 
In  the  first  stage,  the  abstract  was  inspected
to  find  out  whether  it  manifested  the  four
moves  of  I,  M,  R,  and  D  and  was,  thus,  an
“informative" abstract. It was found that all
four moves existed in it:
 
1 (Introduction)  
This  essay  reassesses  James  Fenimore
Cooper's  literary  relationship  to  Walter
Scott…
 
2 (Method)
 … by examining the depiction of Scots in
The  Last  of  the  Mohicans  (1826)  and  The
Prairie (1827).
 
(Results)
 Read  as  companion  texts,  these  novels
represent the imperial migrations of Scots as
a  cause  of  Native  Americans'  unfortunate,
but  for  Cooper  seemingly  inevitable,
eradication.  They  also  trace  the
development  of  an  American  identity  that
incorporates  feudal  chivalry  and  savage
fortitude and that is formed through cultural
appropriation rather than racial mixing. The
Last  of  the  Mohicans'  Scottish  protagonist,
Duncan  Heyward,  learns  to  survive  in  the
northeastern  wilderness  by  adopting  the
Mohicans'  savage  self-control  as  a
complement  to  his  own  feudal  chivalry;  in
turn,  The  Prairie's  Paul  Hover  equips
himself  for  the  challenges  of  westward
expansion by adopting both the remnants of
this  chivalry  and  the  exilic  adaptability  and
colonial  striving  that  Cooper  accords  to
Scots
 
(Discussion)  
I  suggest  that  the  cultural  appropriation
through  which  Heyward  and  Hover  achieve
an  American  identity  that  incorporates
Scottish  chivalry  and  savage  self-command
offers  a  model  for  the  literary  relationship
between  Cooper's  and  Scott's  historical
romances.  The  Leather  stocking  Tales
borrow  selectively  from  the  Waverely
Novels, rejecting their valorization of feudal
chivalry  while  incorporating  their
representation  of  cultural  appropriation  as
a  mechanism  of  teleological  social
development.
 
The  same  abstract  was  then  matched  with
CARS  to  see  if  it  was  similar  to  an
“indicative” abstract. It showed one move:
I3 (Presenting Present Research).  
 
Results  
The first step to answer the questions was to
analyze  the  abstracts  and  note  all
occurrences  of  moves  based  on  IMRD  and
CARS models.  
 
After the first stage of analysis and in order
to  achieve  a  better  understanding  of  the
nature  of  literature  abstracts,  and  also  to
assess the appropriateness of the two models,
the researchers matched  the corpus with the
models  once  more  to  identify  the  steps  as
well.  During  this  stage,  a  number  of  new
steps  were  discovered  which  will  be
discussed in detail.
 
Question  No.  1  The  first  question  was  an
attempt  to  analyze  the  abstracts  using  two
models: IMRD and CARS.  
 
Analysis  based  on  IMRD  moves
To  answer  the  first  part  of  question  number
1,  i.e.  the  existence  of  any  association
between the frequency of moves used by the
three  groups  based  on  IMRD,  the  abstracts
were  subjected  to  a  move  analysis  to  check

 
the  existence  of  the  moves  I,  M,  R  and  D.
The  results  are  summarized  in  Table  4
below. (The percentages have been  rounded
up and Critical χ2 for df of 2 is 5.99.)

As the table suggests, the three groups’ use
of  the  I  and  R  sections  is  close  to  the
prediction  made  by  the  model,  i.e.  the  three
groups  follow  the  IMRD  model  in  their
incorporation  of  Introduction  and  Results.
Nevertheless,  all  groups  make  use  of  the  M
and  D  sections  significantly  less  than  the
model predicts.
 
Concerning  the  sequence  of  moves,  only
13%  of  the  abstracts  manifested  the  I-M-R-D  sequence;  the  most  frequent  sequences
were I (12 PPs) and I-R (7 EPs and 5 EEs).
 
Analysis  based  on  CARS  moves
To  answer  the  second  part  of  question
number  1,  i.e.  the  existence  of  any
association between the frequency of moves
used in the three groups based on CARS, the
abstracts  were  scanned  for  the  existence  of
the  CARS  moves.  The  results  have  been
summarized in Table 5 below.

As  the  table  shows,  the  three  groups
incorporate  I1  and  I3  into  their  abstracts  as
frequently as the model predicts, but the use
of I2 has been far less than the prediction of
the model.  
 
In respect to sequences, 33% of the abstracts
showed I1-I2-I3 sequence (2 PPs, 2 EPs, and
6 EEs); the most frequent move sequence
was I1-I3 (25 PPs, 18 EPs, and 11 EEs).
 
Step analysis
In  order  to  further  elucidate  the  nature  of
literature  abstracts  and  also  the  predictive
power  of  the  models,  the  corpus  was
subjected  to  a  deeper  analysis  which
identified  the  constituent  steps  as  well.
Several points were revealed after this stage
of analysis.
 
1.  Four  new  steps,  not  predicted  by  the
models,  were  discovered  in  the  corpus.  The
researchers  named  them  “Solution”,
“Counter-claiming”,  “Significance”  and
“Implications”, and they existed in two (EP
and  EE),  three  (EP  and  EE),  two  (PP  and
EP)  and  five  (PP  and  EP)  abstracts
respectively. (It is worth noting that the step
“Counter-claiming”  had  previously  been
considered  in  Swales’s  1990  version  of
CARS,  yet  removed  from  the  2004  version
concerned  in  this  study.)  Examples  of  these
new steps are included in Appendix A.

2.  There  was  a  shortcoming  in  Swales’s
(2004)  CARS  model  regarding  the  first  two
steps, i.e. I1s1 “Claiming Centrality” and
I1s2  “Making  Topic  Generalizations”.
Swales  (1990)  defines  the  function  of  the
former  as  follows:  “centrality  claims  are
appeals  to  the  discourse  community
whereby  members  are  asked  to  accept  that
the research about to be reported is part of a
lively,  significant  or  well-established
research area” (p. 144). Regarding the latter,
he writes: “Step 2. . . represents a more
neutral  kind  of  general  statement  than  Step
1” (p. 146).
 
The difference between these two steps was
sometimes  clear,  as  in  the  following
example chosen from the corpus:
 
(I1s1)  Scholars  have  long  been  fascinated
with  the  performance  of  Richard  II  on  the
eve of the Essex “rising”—an episode where
the  interface  between  drama  and  politics  is
particularly broad and responsive. . . .  
 
(I1s2) The incident is intriguing because we
only know about it from the chance survival
of  three  newsletters  reporting  that,  in  early
August  1628,  a  performance  of
Shakespeare’s  play  Henry  VIII  was
“bespoken of purpose” at the Globe by the
duke of Buckingham.
 
Nonetheless,  in  eight  cases  this  dichotomy
did not seem to be efficient in describing the
steps.  For  instance,  in  the  next  case  taken
from  the  corpus,  the  two  opening  steps
cannot be named I1s1 and I1s2.
 
Step1:  In  the  text-based  disciplines,
psychoanalysis  and  Marxism  have  had  a
major influence on how we read,  
 
Step2:  and  this  has  been  expressed  most
consistently  in  the  practice  of  symptomatic
reading,  a  mode  of  interpretation  that
assumes that a text’s truest meaning lies in
what  it  does  not  say,  describes  textual
surfaces  as  superfluous,  and  seeks  to
unmask  hidden  meanings.  For  symptomatic
readers,  texts  possess  meanings  that  are
veiled, latent, all but absent if it were not for
their irrepressible and recurring symptoms.
 
Here,  the  difference  is  not  related  to  the
persuasive  or  neutral  tone  of  the  writer,  but
is  more  of  a  “general”  versus  “specific”
background  information,  which  is  not
captured  by  Swales’s  (2004)  I1s1-I1s2
dichotomy.  It  is  worth  mentioning  that  this
difference  had  been  noted  in  Dudley-Evans’s 1989 Introduction model (as cited in
Bandary, 1999, p. 55) via these three moves:
“Introducing  the  field”,  “Introducing  the
general  topic  (within  the  field)”  and
“Introducing the particular topic (within the
general topic)”.
 
3.  There  were  six  cases  (all  within  the  PP
group)  where  one  step  was  confined  within
the  boundaries  of  another  step.  In  other
words, one step subsumed another step. This
phenomenon  was  exclusive  to  the
Introduction  section  where  the  I3s1  step
contained  either  an  I2s2  or  an  I3s4  step
when  matched  with  CARS  model,  or  their
corresponding  Method  steps  when  matched
with  IMRD.  Nevertheless,  these  cases  were
all  counted  as  I3s1  steps  while  doing  the
statistical analyses.
 
Question No. 2
The  second  question  explores  the
similarities between the groups. The purpose
was  to  see  whether  the  patterns  utilized  in
abstracts  written  in  English  by  Persian
writers  were  similar  to  those  in  abstracts
written  in  Persian  by  Persian  writers,  or  to
the patterns employed in abstracts written in
English  by  English  writers.  Question
number 1 had delved into the comparison of
the  abstracts  with  the  model;  this  time,

however,  the  groups  were  compared  with
one  another.  Thus,  the  frequencies  were
calculated  from  the  data,  not  based  on  the
models.
 
Analysis based on IMRD
The  frequency  of  occurrence  of  IMRD
moves in the three groups were counted and
compared,  and  the  next  table  summarizes
the results.

With  regard  to  Table  6  above,  it  can  be
concluded that there is no difference among
the groups in the incorporation of the IMRD
moves  into  their  abstracts.  All  groups
employed  I  almost  all  the  time  but  D  to  a
low  degree,  but  there  are  differences  in  the
extent  to  which  each  group  employs  R  and
D and M.  
 
Analysis based on CARS
Table  7  below  presents  the  results  for  the
frequency  of  occurrence  of  CARS  moves.
As can be seen, the frequency of occurrence
of the I1 and I3 moves is equally high across
the  three  groups.  Respecting  move  I2,  the
EP  group  is  different  from  PPs  and  EEs  in
that  it  uses  I2  significantly  more  than  the
other  groups.  Therefore,  one  can  conclude
that  the  three  groups  are  similar  to  one
another  except  for  I2  move,  where  EPs  are
significantly  different  from  the  other  two
groups.

Discussion
This study had three major aims:
 
1)  evaluating Lores’s (2004) claim that
different  types  of  abstracts
correspond with different models;
2)  probing into the generic structures of
literature RA abstracts; and
3)  checking  if  Persian  scholars  are
under  the  influence  of  Persian
culture  and/or  community
expectations  regarding  RA  genres
when they write in English.
 
In  so  doing,  a  number  of  questions  were
raised.  To  answer  these  questions  a  corpus
of  90  literature  RA  abstracts  written  in
English  and  Persian  by  English  and  Persian
native  speakers  were  subjected  to  move
analysis.
 
The first question to deal with in the present
study  was  whether  there  is  any  association
between  the  frequency  of  moves  in  the
abstracts and the IMRD and CARS models.  
The  results  for  the  match  with  the  IMRD
model  illustrated  that  the  abstracts  follow
the  model  only  in  their  Introduction  and
 
Results  sections.  Only  13%  of  the  abstracts
had the I-M-R-D pattern.
 
The results for the match against the CARS
model  revealed  that  the  corpus  follows  the
model  in  the  use  of  I1  (Establishing
Research  Territory)  and  I3  (Presenting
Present  Research)  moves,  but  does  not
include I2 (Establishing a Niche). Also, 33%
of  the  abstracts  had  the  I1-I2-I3  move
sequence.  
 
Thus,  it  can  be  concluded  from  the  results
that  none  of  the  models  is  able  to  describe
the  corpus  reliably.  However,  CARS  seems
to be a better model for this field than IMRD
since  (a)  two  out  of  its  three  moves  are
incorporated  into  the  abstracts,  and  (b)  it
predicts  the  move  sequence  of  abstracts
more accurately than IMRD does.
 
In  regard  to  the  first  aim,  one  may  thus  be
able to claim that in line with Lores’s (2004)
idea,  some  abstracts  (here,  literature
abstracts)  tend  to  follow  CARS  model  and
not  the  IMRD  model  and  are,  in  her  words,
of the “indicative” type, not “informative.”
Still,  even  CARS  is  not  a  strong  predictor
for these abstracts.  
 
As  for  the  second  aim,  i.e.,  comprehending
the nature of literature RA abstracts, one can
argue  that  the  writers  tend  to  focus  on
Introduction  and  Results,  yet  neglect
Method and, to a greater extent, Discussion.
Also,  they  do  not  generally  mention  the
niche in the previous works which led them
to carry out the study.
 
The second question of this study dealt with
the  similarities  between  the  groups.  The
focus here is on the English abstracts written
by Persian native speakers (EPs), and to find
out  which  group  they  are  more  similar  to,
the EE or the PP group.
 
Concerning  the  IMRD  model,  the  three
groups  were  similar  to  one  another.
Respecting CARS model, on the other hand,
a  discrepancy  was  observed:  the  EP  group
used  I2  move  significantly  more  than  both
EEs  and  PPs,  and  was  –  surprisingly  –
different from the other two. In other words,
literature  abstracts written by  Persian native
speakers are not entirely under the influence
of  either  Persian  or  English  communities,
but define a standard of their own.
 
The third aim of the study is thus gained: the
argument  advanced  by  Martin  (2003)  and
Tahririan  and  Jalilifar  (2004)  about  the
influence of sociocultural factors on the way
academicians write abstracts is corroborated;
however,  this  influence  is  not  necessarily
exhibited in terms of similarity to  any other
community  rather  than  their  own.  In  other
words,  one  should  be  cautious  in
interpreting  the  abovementioned  hypothesis
as  predictor  of  similarities  between
communities  sharing  the  same  native
language.
 
Theoretical implications
The  present  study  aimed  at,  among  others,
evaluating  Lores’s  (2004)  claim  that
different  types  of  abstracts  conformed  to
different  models.  She  made  a  distinction
between  “informative”  and  “indicative”
abstracts,  and  held  that  the  former
 
corresponded  with  IMRD  while  the  latter
with  CARS  model.  The  findings  of  this
study  corroborate  this  claim:  literature
abstracts  are  mostly  of  the  indicative  type
and generally follow CARS model. Still, the
abstracts  exhibited  marked  deviations  from
CARS  and  it  cannot,  therefore,  be
considered  as  a  reliable  predictor  of  these
abstracts.  
 
Another aim of this study  was to clarify the
nature  of  literature  RA  abstracts.  Based  on
the  results,  it  is  evident  that  literature
scholars fail to mention the reason why they
are  conducting  the  study,  and  often  avoid
presenting  and  discussing  the  results.
Furthermore, a number of deficiencies in the
models  were  described  by  this  study,  and
four new steps were identified.
 
This study also evaluated the hypothesis put
forward  by,  among  others,  Martin  (2003)
and  Tahririan  and  Jalilifar  (2004)  that  the
way  academicians  write  is  under  the
influence  of  sociocultural  factors,  by
demonstrating how the abstracts by Persian-speaking  literature  scholars  writing  in
English  differ  from  those  by  English-speaking  writers.  Nonetheless,  one  cannot
confidently  link  this  deviation  to  native
language  differences,  since  the  same
discrepancy  was  shown  between  the  two
groups of Persian-speaking scholars too. The
influencing factor is not the native language
but  rather  the  norms  of  the  community  for
which the scholars write.
 
Pedagogical implications
Besides  theoretical  implications,  this  study
has  a  number  of  implications  for  teaching
English as a foreign language.
 
It  is  generally  believed  (Martin,  2003)  that,
in  order  to  be  accepted  within  the  scientific
communities, scholars must be familiar with
international  generic  conventions  of  their
field. Thus, it may be necessary for syllabus
designers to develop ESP courses on generic
structures  in  university  programs  to  make
sure  Persian  native  speakers  are  familiar
with generic norms of writing.
 
The results of this study may also serve as a
guide  for  literature  scholars  with  other
native  languages  who  wish  to  write  in
English, by delineating the generic structure
of  literature  RA  abstracts  published  in
international journals.
 
This research study, besides answering some
questions,  raises  some  others  which  can  be
dealt with in other studies:
 
1)  The  researcher  focused  on  RA
abstracts  written  on  the  subject  of
literature.  A  similar  procedure  may
be  replicated  with  abstracts  written
by  English  and  Persian  native
speakers  on  other  similarly
disregarded  subjects,  in  order  to
discover  the  generic  conventions  of
those disciplines.
 
2)  The  present  study  focused  on  RA
abstracts.  It  is  possible  that  thesis
abstracts  behave  in  different  ways;
 
 
48          Genre analysis of literature research article abstracts
therefore,  it  is  necessary  to  inspect
them as well.
 
3)  A  similar  research  study  can  be
carried  out  using  models  other  than
those  utilized  in  this  study,  in  order
to  find  a  better  model  for  describing
literature RA abstracts.
 
4)  This  study  was  confined  to  90  RA
abstracts.  A  similar  study  with  a
larger  corpus  may  lead  to  more
reliable results.
5)  It  is  also  rewarding  to  examine  the
lexico-grammatical  features  of
Persian  and  English  RA  abstracts  to
more  clearly  perceive  the  difference
between  Persian  and  international
community norms.