SECTION  B  
 
DEVELOPMENT:  APPROACHES  TO    
DISCOURSE  ANALYSIS  

  44  
B1  THREE  WAYS  OF  LOOKING  AT  DISCOURSE  
 
Over  the  years  people  have  approached  the  study  of  discourse  in  many  different  
ways,  and  throughout  this  book  we  will  explore  some  of  these  ways  of  analyzing  
discourse  and  practice  applying  them  to  texts  and  conversations  from  our  own  
lives.  For  now  it  is  enough  to  say  that  people  who  analyze  discourse  have  
basically  gone  about  it  from  three  different  perspectives  based  on  three  different  
definitions  of  what  discourse  is.    
 
Some  have  taken  a  formal  approach  to  discourse,  defining  it  simply  as  ‘language  
above  the  level  of  the  clause  or  sentence.’  Those  working  from  this  definition  
often  try  to  understand  the  kinds  of  rules  and  conventions  that  govern  the  ways  
we  join  clauses  and  sentences  together  to  make  texts.    
 
Others  take  a  more  functional  approach,  defining  discourse  as  ‘language  in  use’.  
This  definition  leads  to  questions  about  how  people  use  language  to  do  things  
like  make  requests,  issue  warnings,  and  apologize  in  different  kinds  of  situations  
and  how  we  interpret  what  other  people  are  trying  to  do  when  they  speak  or  
write.    
 
Finally,  there  are  those  who  take  what  we  might  call  a  social  approach,  defining  
discourse  as  a  kind  of  social  practice.  What  is  meant  by  this  is  that  the  way  we  
use  language  is  tied  up  with  the  way  we  construct  different  social  identities  and  
relationships  and  participate  in  different  kinds  of  groups  and  institutions.  It  is  
tied  up  with  issues  of  what  we  believe  to  be  right  and  wrong,  who  has  power  
over  whom,  and  what  we  have  to  do  and  say  to  ‘fit  in’  to  our  societies  in  different  
ways.    
 
Although  these  three  different  approaches  to  discourse  are  often  treated  as  
separate,  and  are  certainly  associated  with  different  historical  traditions  and  
different  individual  discourse  analysts,  the  position  we  will  be  taking  in  this  book  
is  that  good  discourse  analysis  requires  that  we  take  into  account  all  three  of  
these  perspectives.  Instead  of  three  separate  definitions  of  discourse,  they  are  
better  seen  as  three  interrelated  aspects  of  discourse.  The  way  people  use  
language  cannot  really  be  separated  from  the  way  it  is  put  together,  and  the  way  
people  use  language  to  show  who  they  are  and  what  they  believe  cannot  be  
separated  from  the  things  people  are  using  language  to  do  in  particular  
situations.    
 

Language  above  the  clause  

 
The  use  of  the  term  ‘discourse’  to  mean  language  above  the  level  of  the  sentence  
or  the  clause  probably  originated  with  the  linguist  Zellig  Harris,  who,  back  in  the  
1950s,  wanted  to  take  the  study  of  linguistics  to  a  new  level.  Before  this,  linguists  
had  come  a  long  way  in  understanding  how  sounds  are  put  together  to  form  
words  and  how  words  are  put  together  to  form  sentences.  What  Harris  wanted  

  45  
to  do  was  to  understand  how  sentences  are  put  together  to  form  texts.    
 
The  idea  that  texts  could  be  analyzed  in  terms  their  formal  structure  was  actually  
very  popular  in  the  early  and  mid  20th  century,  even  before  Harris  invented  the  
term  ‘discourse  analysis’,  especially  in  the  field  of  literature.  One  group  of  
literary  critics  called  the  Russian  Formalists,  for  example,  tried  to  apply  the  same  
kinds  of  methods  people  used  to  analyze  the  grammar  of  sentences  to  analyzing  
stories  and  novels.  Perhaps  the  most  famous  was  Vladimir  Propp,  who  tried  to  
come  up  with  a  ‘grammar  of  stories’  by  studying  Russian  folk  tales.    
 
The  method  that  Harris  proposed  for  the  analysis  of  discourse,  which  he  called  
‘distributional  analysis’,  was  not  much  different  from  how  people  go  about  doing  
grammatical  analysis.  The  idea  was  to  identify  particular  linguistic  features  and  
determine  how  they  occurred  in  texts  relative  to  other  features,  that  is,  which  
features  occurred  next  to  other  features  or  ‘in  the  same  environment’  with  them.  
However,  as  you  will  see  from  the  excerpt  from  Harris’s  seminal  paper  reprinted  
in  Section  D1,  his  ambitions  went  beyond  simply  understanding  how  linguistic  
features  are  distributed  throughout  texts.  He  was  also  interested  in  
understanding  how  these  features  correlate  with  non-­‐linguistic  behavior  beyond  
texts,  that  is,  how  the  form  that  texts  take  is  related  to  the  social  situations  in  
which  they  occur.    It  was  really  left  to  discourse  analysts  who  came  after  him,  
however,  to  figure  out  exactly  how  the  relationship  between  texts  and  the  social  
contexts  in  which  they  are  used  could  be  fruitfully  studied.    
 
When  focusing  on  the  formal  aspect  of  discourse,  we  are  mostly  interested  in  
how  the  different  elements  of  texts  or  conversations  are  put  together  to  form  
unified  wholes.  In  this  respect,  we  usually  look  for  two  kinds  of  things.  We  look  
for  linguistic  features  (words  and  grammar),  which  help  to  link  different  parts  of  
the  text  or  conversation  together,  and  we  look  at  the  overall  pattern  of  the  text  or  
conversation.  As  we  said  in  Section  A2,  we  can  refer  to  these  two  things  as  1)  
cohesion  (how  pieces  of  the  text  are  ‘stuck  together’)  and  2)  coherence  (the  
overall  pattern  or  sequence  of  elements  in  a  text  or  conversation  that  conforms  
to  our  expectations  about  how  different  kinds  of  texts  or  interactions  ought  to  be  
structured).  We  will  deal  with  these  two  concepts  in  more  detail  in  the  Section  
B2.    

 
Language  in  use  
 
The  second  aspect  of  discourse  that  discourse  analysts  focus  on  is  how  people  
actually  use  language  to  get  things  done  in  specific  contexts.  In  fact,  as  was  
pointed  out  in  Section  A1,  it  is  often  very  difficult  to  understand  what  a  piece  of  
language  means  without  referring  to  the  social  context  in  which  it  is  being  used  
and  what  the  person  who  is  using  it  is  trying  to  do.    
 
This  view  of  discourse  grew  out  of  the  work  of  a  number  of  important  scholars  
including  Michael  Halliday,  whose  approach  to  the  study  of  grammar  differed  
markedly  from  earlier  approaches  by  focusing  less  on  the  forms  language  takes  
and  more  on  the  social  functions  accomplished  by  language,  and  the  work  of  the  

  46  
British  philosophers  John  L.  Austin  and  Paul  Grice  who  laid  the  foundation  for  
what  we  call  pragmatics  (the  study  of  how  people  do  things  with  language).  
Another  important  figure  who  promoted  this  view  of  discourse  is  the  applied  
linguist  H.G.  Widdowson,  who  approached  the  whole  problem  of  language  use  
from  the  perspective  of  language  learning,  noting  that  learning  a  foreign  
language  requires  more  than  just  learning  how  to  make  grammatical  sentences;  
it  also  involves  being  able  to  use  the  language  to  accomplish  things  in  the  world.    
 
There  are  a  number  of  ways  to  study  language  in  use.  One  way  is  to  consider  
discourse  itself  as  a  kind  of  action,  and  to  explore  how,  when  we  say  things  or  
write  things,  we  are  actually  doing  things  like  apologizing,  promising,  
threatening  or  making  requests  (as  we  noted  in  Section  A5).  Another  way  to  
consider  language  in  use  is  to  explore  the  role  of  discourse  in  certain  kinds  of  
activities  and  to  examine  how  different  kinds  of  discourse  make  certain  kinds  of  
actions  or  activities  either  easier  or  more  difficult  to  perform  (an  idea  we  
elaborated  on  in  Section  A8).    Finally,  we  might  consider  how  people  use  
discourse  strategically  to  try  to  communicate  their  interpretation  of  a  situation  
or  to  manage  their  relationships  with  the  people  with  whom  they  are  
communicating  (as  we  discussed  in  Section  A6).    
 

 Language  and  ‘Social  practice’  

 
The  third  aspect  of  discourse  has  to  do  with  the  role  of  language  in  ‘social  
practice’.  Language  is  seen  not  just  as  a  system  for  making  meaning,  but  as  part  
of  larger  systems  through  which  people  construct  social  identities  and  social  
realities.  Different  people  use  language  in  different  ways.  An  English  teacher  
talks  differently  than  a  hip-­‐hop  artist.  And  these  different  ways  of  talking  help  to  
show  who  we  are  and  also  reflect  our  different  ideas  about  the  world,  different  
beliefs,  and  different  values.    
 
This  view  of  discourse  probably  owes  the  most  to  the  French  philosopher  Michel  
Foucault,  who  argued  that  discourse  is  the  main  tool  through  which  we  construct  
‘knowledge’  and  exert  power  over  other  people.  Different  kinds  of  discourse  (or  
‘discourses’)  are  associated  with  different  kinds  of  people  and  different  ‘systems  
of  knowledge’.  Foucault  spoke,  for  example,  of  ‘clinical  discourse,  economic  
discourse,  the  discourse  of  natural  history,  (and)  psychiatric  discourse’  (1972:  
121).  The  American  discourse  analyst  James  Gee  uses  a  capital  ‘D’  to  distinguish  
this  view  of  discourse  from  the  others  we  have  talked  about.  For  him,  
‘Discourses’  are  ‘ways  of  being  in  the  world,  or  forms  of  life  which  integrate  
words,  acts,  values,  beliefs,  attitudes,  and  social  identities’  (1996:  127).  
 
This  aspect  of  discourse  leads  us  to  explore  how  people  use  language  to  advance  
certain  versions  of  reality  and  certain  relationships  of  power,  and  also  how  our  
beliefs,  values  and  social  institutions  are  constructed  through  and  supported  by  
discourse.  A  central  principle  of  this  view  of  discourse  is  that  discourse  is  always  
‘ideological’,  meaning  that  discourse  always  has  ‘an  agenda’,  that  it  always  ends  
up  serving  the  interests  of  certain  people  over  those  of  others.    
 

  47  
As  was  stated  above,  it  is  difficult  to  look  at  discourse  in  any  meaningful  way  
from  only  one  of  these  perspectives.  Simply  looking  at  how  texts  are  put  
together,  for  example,  while  it  may  be  interesting,  has  limited  practical  value.  At  
the  same  time,  you  cannot  really  make  broad  statements  about  ‘power’  or  
‘ideology’  in  a  text  without  first  understanding  some  basic  things  about  how  the  
text  is  put  together  and  how  people  are  actually  using  it  in  specific  social  
contexts  to  perform  specific  actions.    
 
  Look  deeper  into  this  topic  online  

  48  
B2  COHESION  AND  COHERENCE  
 

Earlier  we  said  that  one  of  the  most  basic  tasks  for  a  discourse  analyst  is  to  
future  out  what  makes  a  text  a  text  and  what  makes  a  conversation  and  
conversation,  in  other  words,  to  figure  out  what  gives  text  and  conversations  
texture.  Texture,  we  said,  comes  from  cohesion  and  coherence.  Cohesion  primarily  
has  to  do  with  linguistic  features  in  the  text,  and  coherence  has  to  do  with  the  
kind  of  ‘framework’  with  which  the  reader  approaches  the  text  and  what  he  or  
she  wants  to  use  the  text  to  do.    

This  is  perhaps  a  bit  misleading,  possibly  making  you  think  that,  when  it  comes  
to  cohesion,  the  reader  doesn’t  have  to  do  any  work,  and  in  the  case  of  coherence  
the  expectations  in  the  mind  of  the  reader  are  more  important  than  what  is  
actually  in  the  text.  This  is  not  the  case.  In  fact,  what  creates  cohesion  is  not  just  
the  linguistic  features  within  the  text  alone,  but  the  fact  that  these  features  lead  
readers  to  perform  certain  mental  operations  –  to  locate  and  take  note  of  earlier  
or  later  parts  of  the  text  as  they  are  going  through  it.    

For  example,  if  I  were  to  say,  ‘Lady  Gaga  doesn’t  appeal  to  me,  but  my  sister  loves  
her’,  in  order  to  understand  the  meaning  of  ‘her’  in  the  second  clause,  you  have  
to  do  some  mental  work.  Not  only  do  you  need  to  refer  back  to  the  first  clause,  
you  also  have  to  be  smart  enough  to  know  that  ‘her’  refers  to  Lady  Gaga  and  not  
my  sister.  Thus,  cohesion  is  the  quality  in  a  text  that  forces  you  to  look  either  
backward  or  forward  in  the  text  in  order  to  make  sense  of  the  things  you  read,  
and  through  your  acts  of  looking  backward  and  forward  the  text  takes  on  a  
quality  of  connectedness.  

Similarly,  to  say  that  coherence  is  a  matter  of  the  ‘frameworks’  or  sets  of  
expectations  that  we  bring  to  texts,  does  not  mean  that  what  is  actually  in  the  
text  is  any  less  important.  Concrete  features  must  exist  in  the  text  which  are  
often  arranged  in  a  certain  order  and  conform  to  or  ‘trigger’  those  expectations.  
For  example,  for  me  to  interpret  a  text  as  a  shopping  list,  it  must  have  a  certain  
structure  (a  list),  certain  kinds  of  words  (generally  nouns),  and  those  words  
must  represent  things  that  I  am  able  to  purchase  (as  opposed  to  abstract  things  
like  ‘world  peace’  or  unaffordable  items  like  the  Golden  Gate  Bridge).      

Cohesion  
 
Halliday  and  Hasan,  whose  work  is  excerpted  in  Section  D2,  describe  two  broad  
kinds  of  linguistic  devices  that  are  used  to  force  readers  to  engage  in  this  process  
of  backward  and  forward  looking  which  gives  them  a  sense  of  connectedness  in  
texts.  One  type  depends  on  grammar  (which  they  call  grammatical  cohesion)  and  
the  other  type  depends  more  on  the  meanings  of  words  (which  they  call  lexical  
cohesion).  

  49  
  
Devices  used  to  create  grammatical  cohesion  include:  

·  Conjunction  (using  ‘connecting  words’)  

·  Reference  (using  a  pronoun  to  refer  to  another  word)  
·  Substitution  (substituting  one  word  or  phrase  for  another  word  or  
phrase)  
·  Ellipses  (leaving  something  out)  

Lexical  cohesion  involves  the  repetition  of  words  or  of  words  from  the  same  
semantic  field  (e.g.  milk,  tomatoes,  rocket).  

Conjunction  refers  to  the  use  of  various  ‘connecting  words’  (such  as  conjunctions  
like  and  and  but  and  conjunctive  adverbs  like  furthermore  and  however)  to  join  
together  clauses  and  sentences.  Conjunction  causes  the  reader  to  look  back  to  
the  first  clause  in  a  pair  of  joined  clauses  to  make  sense  of  the  second  clause.  The  
important  thing  about  these  ‘connecting  words’  is  that  they  do  not  just  establish  
a  relationship  between  the  two  clauses,  but  that  they  tell  us  what  kind  of  
relationship  it  is.    

‘Connecting  words’,  then,  can  be  grouped  into  different  kinds  depending  on  the  
relationship  they  establish  between  the  clauses  or  sentences  that  they  join  
together.  Some  are  called  additive,  because  they  add  information  to  the  previous  
clause  or  sentence.  Examples  are  ‘and’,  ‘moreover’,  ‘furthermore’,  ‘in  addition’,  ‘as  
well’.  Others  are  called  contrastive  because  they  set  up  some  kind  of  contrast  
with  the  previous  sentence  or  clause.  Examples  are  ‘but’,  ‘however’.  Still  others  
are  called  causative  because  they  set  up  some  kind  of  cause  and  effect  
relationship  between  the  two  sentences  or  clauses.  Examples  of  these  are  
‘because’,  ‘consequently’,  ‘therefore’.  Finally,  some  are  called  sequential  because  
they  indicate  the  order  facts  or  events  come  in.  Examples  are  ‘firstly’,  
‘subsequently’,  ‘then’  and  ‘finally’.  In  the  two  examples  below,  the  first  uses  a  
contrastive  connective  and  the  second  uses  a  causative  connective.  

He  liked  the  exchange  students.  She,  however,  would  have  nothing  to  do  
with  them.  

He  liked  the  exchange  students.  She,  therefore,  would  have  nothing  to  do  
with  them.  

All  connecting  words  cause  the  reader  to  look  back  to  a  previous  clause  (or  
sentence)  in  order  to  understand  the  subsequent  clause  (or  sentence),  and  the  
kind  of  connecting  word  used  guides  the  reader  in  understanding  the  
relationship  between  two  clauses  (or  sentences).  In  the  first  example  given  
above,  the  word  however  causes  the  reader  to  look  back  at  the  first  sentence  to  
find  out  what  the  difference  is  between  her  and  him.  In  the  second  example,  the  
word  therefore  causes  the  reader  to  look  back  at  the  first  sentence  to  find  out  
why  she  won’t  have  anything  to  do  with  the  exchange  students.    

Another  very  common  way  we  make  our  texts  ‘stick  together’  is  by  using  words  
that  refer  to  words  we  used  elsewhere  in  the  text.  This  kind  of  cohesive  device  is  

  50  
known  as  reference.  The  two  examples  above,  besides  using  connecting  words,  
also  use  this  device.  The  word  them  in  the  second  sentence  refers  back  to  the  
exchange  students  in  the  first  sentence,  and  so,  to  make  sense  of  it,  the  reader  is  
forced  to  look  back.  He  and  she  are  also  pronouns  and  presumably  refer  to  
specific  people  who  are  probably  named  at  an  earlier  point  in  the  longer  text  
from  which  these  sentences  were  taken.  The  word  or  group  of  words  that  a  
pronoun  refers  to  is  called  its  antecedent.  What  reference  does,  then,  is  help  the  
reader  to  keep  track  of  the  various  participants  in  the  text  as  he  or  she  reads  
(Eggins,  1994:  95).    

There  are  basically  three  kinds  of  reference:  1)  anaphoric  reference  –  using    
words  that  point  back  to  a  word  used  before:  

 After  Lady  Gaga  appeared  at  the  MTV  Music  Video  Awards  in  a  dress  made  
completely  of  meat,  she  was  criticized  by  animal  rights  groups.      

2)  cataphoric  reference:  Using  words  that  point  forward  to  a  word  that  has  not  
been  used  yet:  

When  she  was  challenged  by  reporters,  Lady  Gaga  insisted  that  the  dress  
was  not  intended  to  offend  anyone.        

3)  Using  words  that  point  to  something  outside  the  text  (exophoric  reference):  

If  you  want  to  know  more  about  this  controversy,  you  can  read  the  
comments  people  have  left  on  animal  rights  blogs.  

The  definite  article  (the)  can  also  be  a  form  of  anaphoric  reference  in  that  it  
usually  refers  the  reader  back  to  an  earlier  mention  of  a  particular  noun.    

Lady  Gaga  appeared  in  a  dress  made  completely  of  meat.  The  dress  was  
designed  by  Franc  Fernandez.  

Substitution  is  similar  to  reference  except  rather  than  using  pronouns,  other  
words  are  used  to  refer  to  an  antecedent,  which  has  either  appeared  earlier  or  
will  appear  later.  In  the  sentence  below,  for  example,  the  word  one  is  used  to  
substitute  for  dress.    

Besides  wearing  a  meat  dress,  Lady  Gaga  has  also  worn  a  hair  one,  which  
was  designed  by  Chris  March.    

Substitution  can  also  be  used  to  refer  to  the  verb  or  the  entire  predicate  of  a  
clause,  as  in  the  example  below.    

  If  Lady  Gaga  was  intending  to  shock  people,  she  succeeded  in  doing  so.    

Ellipsis  is  the  omission  of  a  noun,  verb,  or  phrase  on  the  assumption  that  it  is  
understood  from  the  linguistic  context.  In  order  to  fill  in  the  gap(s),  readers  need  
to  look  back  to  previous  clauses  or  sentences,  as  in  the  example  below.    

  51  
 There  is  much  to  support  the  view  that  it  is  clothes  that  wear  us,  and  not  
we,  them.  (Virginia  Woolf)  

All  of  the  devices  mentioned  above  are  examples  of  grammatical  cohesion,  the  
kind  of  cohesion  that  is  created  because  of  the  grammatical  relationship  between  
words.  Lexical  cohesion  occurs  as  a  result  of  the  semantic  relationship  between  
words.  The  simplest  kind  of  lexical  cohesion  is  when  words  are  repeated.  But  a  
more  common  kind  is  the  repetition  of  words  related  to  the  same  subject.  We  call  
these  ‘chains’  of  similar  kinds  of  words  that  run  through  texts  lexical  chains.  In  
the  following  text,  for  example,  besides  the  use  of  reference  (who,  it,  she),  the  
clauses  are  held  together  by  the  repetition  of  the  verb  ‘to  wear’  and  of  other  
words  having  to  do  with  clothing  and  fashion  (bikini,  Vogue  –  a  famous  fashion  
magazine,  dress,  and  outfits).    

Lady  Gaga,  who  came  under  fire  recently  for  wearing  a  meat  bikini  on  the  
cover  of  Vogue  Hommes  Japan,  wore  a  raw  meat  dress  at  last  night's  
VMAs.  It  was  one  of  many  outfits  she  wore  throughout  the  night.  
(Oldenberg,  2010)    

Taken  together,  these  words  form  a  lexical  chain,  which  helps  to  bind  the  text  
together.    Lexical  chains  not  only  make  a  text  more  cohesive  but  also  highlight  
the  topic  or  topics  (such  as  ‘fashion’,  ‘entertainment’,  ‘technology’)  that  the  text  is  
about  –  and  so  can  provide  context  for  determining  the  meaning  of  ambiguous  
words  (such  as  ‘rocket’  in  the  example  of  the  shopping  list  given  in  Section  A2).  
In  fact,  searching  for  lexical  chains  is  one  the  main  techniques  used  in  computer  
automated  text  categorization  and  summarization.    

Some  texts  may  make  use  of  a  lot  of  these  devices,  whereas  others  may  use  very  
few  of  them.  Halliday  and  Hasan  (1976:  297)  refer  to  texture  in  text  as  being  
either  ‘tight’  –  meaning  that  there  are  many  cohesive  devices  –  or  ‘loose’,  –
meaning  that  there  are  fewer.  What  often  determines  the  extent  to  which  these  
devices  are  used  is  how  much  they  are  needed  for  readers  to  make  the  kinds  of  
connections  they  need  to  make  to  understand  the  text.  Communication  generally  
operates  according  to  the  principle  of  ‘least  effort’.  There  is  no  need,  for  example,  
for  me  to  insert  the  word  ‘and’  after  every  item  in  my  shopping  list  for  me  to  
know  that  I  need  to  buy  tomatoes  in  addition  to  buying  milk.  One  of  the  
challenges  for  people  who  are  producing  texts,  therefore,  is  figuring  out  what  
kinds  of  connections  readers  can  make  for  themselves  by  invoking  what  they  
already  know  about  the  world  and  about  this  particular  kind  of  text  (coherence)  
and  what  connections  need  top  be  spelled  out  explicitly  in  the  text  (cohesion).      
 
  Look  deeper  into  this  topic  online  
 

Coherence  
 
As  the  shopping  list  we  discussed  in  Section  A2  illustrates,  what  makes  a  text  a  
text  is  often  as  much  a  matter  of  the  interpretative  framework  that  the  reader  
brings  to  the  text  as  it  is  of  anything  internal  to  the  text.  The  relationship  

  52  
between  the  words  ‘tomatoes’  and  ‘rocket’  becomes  meaningful  to  a  reader  
based  on  his  or  her  understanding  of  what  a  shopping  list  is  and  what  it  is  used  
for.  This  aspect  of  texture  is  known  as  coherence,  and  it  has  to  do  with  our  
expectations  about  the  way  elements  in  a  text  ought  to  be  organized  and  the  
kinds  of  social  actions  (like  shopping)  that  are  associated  with  a  given  text.  
 
The  text  below    (Figure  B2.1)  is  a  good  example  of  how  we  sometimes  need  to  
apply  our  experience  with  past  texts  and  with  certain  conventions  that  have  
grown  up  in  our  society  in  order  to  understand  new  texts  we  encounter.  
 

 
  Figure  B2.1  Advertisement  from  Body  Coach.Net  
 
 
For  most  people,  as  soon  as  they  see  the  words  ‘before’  and  ‘after’,  a  certain  body  
of  knowledge  is  ‘triggered’  based  on  texts  they  have  seen  in  the  past  which  
contain  these  words  such  as  advertisements  for  beauty  products.  In  such  texts,  
‘before’  is  usually  portrayed  as  ‘bad’  and  ‘after’  is  usually  portrayed  as  ‘good’,  and  
the  product  being  advertised  is  portrayed  as  the  ‘agent’  that  causes  the  
transformation  from  ‘before’  to  ‘after’.  This  structure  is  a  variation  on  what  
Michael  Hoey  (1983)  has  called  the  ‘Problem-­‐Solution’  pattern,  which  underlies  
many  texts  from  business  proposals  to  newspaper  editorials.    
 
The  challenge  this  ad  presents  for  the  reader  is  that  there  is  no  explicit  
information  about  what  is  meant  by  ‘before’  and  ‘after’  other  than  a  curved  line  
drawn  down  the  center  of  the  page.  In  order  to  interpret  this  line,  we  must  make  
reference  to  the  smaller  words  in  the  lower  right  corner  which  give  the  name  of  
the  advertiser:  Body  Coach.Net,  and  the  slogan:  For  a  perfect  body.  This  
information  creates  for  readers  an  interpretive  framework  based  on  their  
knowledge  of  the  kind  of  business  such  a  company  might  be  engaged  in  and  
cultural  notions  of  what  a  ‘perfect  body’  might  look  like.  Once  this  framework  is  
triggered,  most  readers  have  no  trouble  interpreting  the  space  formed  on  the  
‘before’  side  of  the  ad  as  portraying  the  stomach  of  an  overweight  person,  and  
the  space  formed  on  the  ‘after’  side  as  the  ‘hourglass’  shape  associated  (at  least  

  53  
in  the  culture  in  which  this  ad  appeared)  with  female  beauty,  and  of  the  company  
–  Body  Coach.Net  and  the  product  that  it  sells–  as  the  agents  of  this  
transformation.    
 
There  are  a  number  of  different  kinds  of  interpretative  frameworks  that  we  use  
to  make  sense  of  texts.  One  kind,  which  we  will  discuss  further  in  the  next  
section,  we  might  call  a  generic  framework.  This  kind  of  framework  is  based  on  
the  expectations  we  have  about  different  kinds  of  texts,  the  kinds  of  information  
we  expect  to  encounter  in  texts  of  different  kinds  and  the  order  in  which  we  
expect  that  information  to  be  presented,  along  with  other  kinds  of  lexical  or  
grammatical  features  we  expect  to  encounter.  In  the  example  above,  for  instance,  
it  is  partially  our  knowledge  of  the  structure  of  ‘before  and  after  ads’  that  helps  
us  to  make  sense  of  this  particular  ad.    
 
Part  of  what  forms  such  generic  frameworks  is  that  different  parts  of  a  text  are  
not  just  grammatically  and  lexically  related,  but  that  they  are  conceptually  and  
procedurally  related  –  in  other  words,  that  they  appear  in  a  certain  logical  or  
predictable  sequence.  Texts  following  the  ‘Problem  –  Solution’  pattern,  for  
example,  begin  by  presenting  a  problem  and  then  go  on  to  present  one  or  more  
solutions  to  the  problem.  This  important  principle  in  discourse  analysis  has  its  
origins  largely  in  cognitive  science  and  early  research  in  artificial  intelligence  by  
people  like  Schank  and  Abelson  (1977),  who  pointed  out  that  many  human  
activities  are  governed  by  conventional,  sequentially  ordered,  multi-­‐step  
procedures  (which  they  called  ‘scripts’),  and  Rumelhart,  (1975),  who  pointed  out  
that,  in  a  similar  way,  texts  like  narratives  also  exhibit  conventional  structures  
based  on  predictable  sequences  of  actions  and  information  (which  he  called  
‘schema’).  An  excerpt  from  Rumelhart’s  classic  article  ‘Notes  on  a  Schema  for  
Stories’  is  reprinted  in  Section  D2.    
 
But  not  all  of  the  knowledge  we  use  to  make  sense  of  texts  comes  from  our  
knowledge  about  the  conventions  associated  with  different  kinds  of  texts.  Some  
of  this  knowledge  is  part  of  larger  conceptual  frameworks  that  we  build  up  based  
on  our  understanding  of  how  the  world  works.  We  will  use  the  term  cultural  
models  to  describe  these  frameworks.  James  Paul  Gee  (2010)  calls  cultural  
models  ‘videotapes  in  the  mind’  based  on  experiences  we  have  had  and  depicting  
what  we  take  to  be  prototypical  (or  ‘normal’)  people,  objects  and  events.  To  
illustrate  the  concept  he  points  out  that  we  would  never  refer  to  the  Pope  as  a  
‘bachelor’,  even  though  the  Pope,  as  an  unmarried  adult  male,  fulfills  the  
conditions  for  the  dictionary  definition  of  the  word,  because  he  does  not  fit  into  
our  cultural  model  of  what  a  bachelor  is.    
 
Cultural  models  regarding  both  the  kind  of  work  ‘coaches’  do  and  about  what  
constitutes  a  ‘perfect  body’  are  central  to  our  ability  to  interpret  the  ad  above,  
and  especially  for  our  understanding  of  the  meaning  of  the  two  shapes  formed  by  
the  line  drawn  down  the  center  of  the  page.    
 
The  important  thing  to  remember  about  cultural  models  (and,  for  that  matter,  
generic  frameworks)  is  that  they  are  cultural.  In  other  words,  they  reflect  the  
beliefs  and  values  of  a  particular  group  of  people  in  a  particular  place  at  a  

  54  
particular  point  in  history.  The  ad  reprinted  above  would  be  totally  
incomprehensible  for  people  in  many  societies  outside  of  our  own  because  they  
would  not  share  either  the  knowledge  of  ‘before  and  after  ads’  or  the  beliefs  
about  physical  attractiveness  that  we  have.  It  is  even  more  important  to  
remember  that  such  texts  do  not  just  reflect  such  expectations,  values  and  
beliefs,  but  also  reinforce  them.  Every  time  we  encounter  a  text  like  the  one  
above,  these  generic  frameworks  and  cultural  models  and  the  habitual  ways  of  
looking  at  the  world  associated  with  them  are  strengthened.    
 
  Look  deeper  into  this  topic  online  

  55  
B3  ALL  THE  RIGHT  MOVES  
 
Texts  that  are  structured  according  to  particular  generic  frameworks  are  called  
genres.  But,  as  we  noted  in  Section  A3,  genres  are  more  than  just  texts;  they  are  
means  by  which  people  get  things  done,  and  the  way  they  are  structured  depends  
crucially  on  what  the  particular  people  using  a  genre  want  or  need  to  do.  In  other  
words,  what  determines  the  way  a  particular  genre  is  put  together  is  its  
communicative  purpose,  and  so  this  must  be  our  central  focus  in  analyzing  
genres.    
 
Usually,  the  overall  communicative  purpose  of  a  genre  can  be  broken  down  into  
a  number  of  steps  that  users  need  to  follow  in  order  to  achieve  the  desired  
purpose  —  rather  like  the  steps  in  a  recipe  —  and  typically  the  most  important  
constraints  and  conventions  regarding  how  a  genre  is  structured  involve  1)  
which  steps  must  be  included,  and  2)  the  order  in  which  they  should  appear.  In  
the  field  of  genre  analysis  these  steps  are  known  as  moves.    
 
John  Swales,  the  father  of  genre  analysis,  illustrated  the  idea  of  moves  in  his  
analysis  of  introductions  to  academic  articles.  Instead  of  asking  the  traditional  
question:  ‘how  is  this  text  structured?’,  Swales  asked  ‘What  do  writers  of  such  
texts  need  to  do  in  order  to  achieve  their  desired  purpose?  (which,  in  the  case  of  
an  introduction  to  an  academic  article,  is  mainly  getting  people  to  believe  that  
the  article  is  worth  reading).  In  answering  this  question,  Swales  identified  four  
moves  characteristic  of  such  texts.  An  introduction  to  an  academic  article,  he  said,  
typically:      
 
1.  Establishes  the  field  in  which  the  writer  of  the  study  is  working;  
2.  Summarizes  the  related  research  or  interpretations  on  one  aspect  of  
the  field;  
3.  Creates  a  research  space  or  interpretive  space  (a  ‘niche’)  for  the  
present  study  by  indicating  a  gap  in  current  knowledge  or  by  raising  
questions;  and  
4.  Introduces  the  study  by  indicating  what  the  investigation  being  
reported  will  accomplish  for  the  field.  (adapted  from  Swales  1990)    
 
Of  course,  not  all  introductions  to  academic  articles  contain  all  four  of  these  
moves  in  exactly  the  order  presented  by  Swales.  Some  article  introductions  may  
contain  only  some  of  these  moves,  and  some  might  contain  different  moves.  
Furthermore,  the  ways  these  moves  are  realized  might  be  very  different  for  
articles  about  engineering  and  articles  about  English  literature.  The  point  that  
Swales  was  trying  to  make,  however,  was  not  that  these  moves  are  universal  or  
in  some  way  obligatory,  but  that  these  are  the  prototypical  moves  one  would  
expect  to  occur  in  this  genre,  and  understanding  these  default  expectations  is  the  
first  step  to  understanding  how  ‘expert  users’  might  creatively  flout  these  
conventions.    
 

  56  
At  the  same  time,  it  is  important  to  remember  that  not  all  genres  are  equally  
‘conventionalized’;  while  some  genres  have  very  strict  rules  about  which  moves  
should  be  included  and  what  order  they  should  be  in,  other  genres  exhibit  much  
more  variety  (see  for  example  the  weblog  entries  discussed  Section  C3).    
 
One  genre  which  has  a  particularly  consistent  set  of  communicative  moves  is  the  
genre  of  the  ‘personal  advertisement’  (sometimes  called  the  ‘dating  
advertisement’)  which  sometimes  appears  in  the  classified  sections  of  
newspapers  and,  increasingly,  on  online  social  media  and  dating  sites.  The  
following  is  an  example  given  by  Justine  Coupland  in  her  1996  study  of  dating  
advertisement  in  British  newspapers:    
 
Sensual,  imaginative  brunette,  25,  artistic,  intelligent,  with  a  sense  of  
humour.  Enjoys  home  life,  cooking,  sports,  country  life.  No  ties,  own  
home.  Seeking  a  tall,  strong,  intelligent  fun  companion  with  inner  depth  
for  passionate,  loving  romance,  25-­‐35.  Photo  guarantees  reply.  Must  feel  
able  to  love  Ben  my  dog  too.  London/anywhere.  (Coupland,  1996:  187)  
 
Advertisements  like  this  tend  to  consist  of  five  moves:    
 
1) The  advertiser  describes  himself  or  herself  (Sensual,  imaginative  
brunette…);    
2) The  advertiser  describes  the  kind  of  person  he  or  she  is  looking  for  
(Seeking  tall,  strong,  intelligent…);  
3) The  advertiser  describes  the  kind  of  relationship  or  activities  he  or  
she  wishes  to  engage  in  with  the  target  (for  passionate,  loving  romance);  
4) The  advertiser  gives  additional  information,  makes  a  humorous  remark  
or  issues  a  challenge  (Photo  guarantees  reply.  Must  feel  able  to  love  Ben  my  
dog  too);  and  
5) The  advertiser  indicates  how  he  or  she  can  be  contacted  (by,  for  
example,  giving  a  telephone  number,  an  email  address,  or  a  post  office  box  
–  this  move  is  not  present  in  the  excerpt  Coupland  gives,  but  was  
presumably  present  in  some  form  in  the  original  ad).    

Of  course,  as  we  will  see  below,  dating  ads  in  other  contexts  might  have  slightly  
different  move  structures,  but  all  of  these  moves  will  likely  be  present  in  one  
form  or  another.  The  reason  for  this  is  that  these  moves  (especially  1,  2,  3,  and  5)  
are  essential  if  the  overall  communicative  purpose  of  finding  a  partner  is  to  be  
achieved.    
 
Such  ads  also  tend  to  have  certain  regularities  in  style  and  the  kinds  of  language  
that  is  used  to  realize  these  five  moves.  If  they  appear  in  newspapers,  for  
example,  they  are  often  written  in  a  kind  of  telegraphic  style,  which  omits  non-­‐
essential  function  words  (since  advertisers  usually  have  a  word  limit  or  are  
charged  by  the  word).  In  most  cases,  self-­‐descriptions  and  other-­‐descriptions  
contain  information  about  things  like  age,  appearance,  and  personality  expressed  
in  lists  of  positive  adjectives  (like  young,  fit,  fun-­loving),  and  the  goal  is  almost  
always  a  romantic  or  sexual  relationship  or  activities  (like  opera,  candlelight  

  57  
dinners,  quiet  evenings  at  home)  which  are  normally  associated  with  or  act  as  
euphemisms  for  sex  or  romance.    
 
In  a  sense,  such  advertisements  not  only  serve  the  communicative  purpose  of  
individual  members  of  a  discourse  community  to  find  suitable  partners,  but  they  
also  serve  to  define  and  reinforce  the  values  of  the  discourse  community  as  a  
whole  regarding  what  kinds  of  partners  and  activities  are  considered  desirable.  
Therefore,  being  able  to  compose  such  ads  successfully  is  not  just  about  
portraying  oneself  as  desirable,  but  also  about  portraying  oneself  as  a  competent  
member  of  a  particular  community  of  users.    
 
Of  course,  many  different  kinds  of  discourse  communities  use  this  genre  for  
different  purposes,  and  so  one  might  identify  ‘sub-­‐genres’  of  the  personal  
advertisement  for  communities  of  heterosexual  singles,  gay  men,  seniors,  and  
any  number  of  other  groups,  each  with  different  conventions  and  constraints  on  
what  kind  of  information  should  be  included  and  how  it  should  be  structured.  
One  such  ‘sub-­‐genre’  is  the  matrimonial  advertisement  found  in  communities  of  
South  Asians,  an  example  of  which  is  given  below:    
 
A well-settled uncle invites matrimonial correspondence from slim, fair,
educated South Indian girl, for his nephew, 25 years, smart, M.B.A., green
card holder, 5’6". Full particulars with returnable photo appreciated. (Nanda,
2000: 196-204)
 
The  most  obvious  difference  in  this  ad  from  the  first  example  given  is  that  the  
advertiser  is  not  the  person  who  will  be  engaging  in  the  sought  after  
relationship,  but  rather  a  family  member  acting  as  an  intermediary.  Another  
important  difference  has  to  do  with  the  kinds  of  information  included  in  the  
descriptions.  Ads  of  this  sub-­‐genre  often  include  information  such  as  
immigration  status,  educational  attainment,  income,  caste,  and  religion,  
information  that  is  not  a  common  feature  of  dating  ads  in  other  communities.    
Another  rather  unique  sub-­‐genre  of  personal  ads  are  ads  placed  by  lesbians  in  
search  of  reproductive  partners,  such  as  those  examined  by  Susan  Hogben  and  
Justine  Coupland  in  their  2000  study.  Here  is  an  example  of  such  an  ad:    
 
Loving,  stable  lesbian  couple  require  donor.  Involvement  encouraged  but  
not  essential.  HIV  test  required.  London.  BoxPS34Q.  (Hogben  and  
Coupland  2000:  464)
 
What  is  interesting  about  this  ad  and  many  of  those  like  it  is  that  there  is  no  
elaborate  description  of  the  kind  of  person  sought  or  what  he  or  she  is  sought  for  
beyond  the  use  of  the  term  ‘donor’,  a  term  which,  in  this  community,  presumably  
communicates  all  of  the  necessary  information.  Another  interesting  aspect  of  this  
sub-­‐genre  is  that  the  ‘commenting  move’,  a  move  which  in  typical  heterosexual  
dating  ads  is  usually  of  the  least  consequence,  in  these  ads  often  includes  vital  
information  about  legal  and  health  issues  that  are  central  to  the  practice  of  
surrogate  parenthood.    
 

  58  
The  most  important  point  we  can  take  from  these  two  examples  is  that  generic  
variation  is  not  just  a  matter  of  the  different  values  or  styles  of  different  
discourse  communities,  but  is  also  very  often  a  function  of  differences  in  the  
overall  communicative  purpose  of  the  sub-­‐genre  (finding  a  sexual  partner,  a  wife,  
a  reproductive  partner).    

 
Bending  and  Blending  
 
Despite  the  stylistic  variety  in  personal  advertisements  among  different  
discourse  communities,  this  genre  nevertheless  remains  very  conventionalized,  
with  fairly  strict  constraints  on  what  is  considered  a  relevant  contribution.    
Advertisers  must  describe  themselves,  describe  the  kind  of  person  they  are  
seeking,  and  describe  the  kind  of  relationship  they  want  to  have.  Ironically,  
however,  the  strongly  conventionalized  nature  of  this  genre,  the  fact  that  nearly  
all  examples  of  it  have  more  or  less  the  same  structure,  has  the  potential  to  work  
against  the  overall  communicative  purpose,  which  is  attracting  the  attention  of  
interested  (and  interesting)  readers.  Consequently,  it  is  not  uncommon  for  
‘expert  users’  of  this  genre  to  try  to  make  their  ads  stand  out  by  ‘playing  with’  the  
conventions  of  the  genre.    
 
One  way  of  ‘playing  with’  generic  conventions,  which  Bhatia  (1997)  calls,  genre  
bending,  involves  flouting  the  conventions  of  a  genre  in  subtle  ways  which,  while  
not  altering  the  move  structure  substantially,  makes  a  particular  realization  of  a  
genre  seem  creative  or  unique.  One  way  writers  of  personal  advertisements  
sometimes  bend  this  genre  is  by  flouting  the  expectations  for  self-­‐
aggrandizement  associated  with  it.  The  following  example  comes  from  a  study  by  
Jones  on  gay  personal  ads  in  Hong  Kong:  
 
CHINESE,  20,  STILL  YOUNG,  but  not  good-­‐looking,  not  attractive,  not  sexy,  
not  hairy,  not  fit,  not  tall,  not  experienced,  not  mature,  not  very  intelligent  
but  Thoughtful  and  Sincere,  looking  for  friendship  and  love.  (Jones  2000:  
46)  
 
Another  way  of  ‘playing  with’  generic  conventions  is  to  mix  the  conventions  of  
one  genre  with  another,  a  process  which  Bhatia  (1997)  refers  to  as  genre  
blending.  In  the  following  example  from  Coupland’s  study,  for  instance,  the  
advertiser  blends  the  conventions  of  the  dating  ad  genre  with  the  conventions  of  
another  genre,  namely  ads  for  automobiles.    
 
CLASSIC  LADY  limousine,  mint  condition,  excellent  runner  for  years  seeks  
gentleman  enthusiast  45+  for  TLC  and  excursions  in  the  Exeter  area  BOX  
555L.  (Coupland  1996:  192)  
 
Ironically,  what  both  of  these  writers  are  doing  by  flouting  the  conventions  of  the  
genre  is  subtly  distancing  themselves  from  the  discourse  community  of  users  
while  at  the  same  time  identifying  with  it.  This  seemingly  odd  strategy  is  less  
surprising  when  one  considers  that  most  people  who  post  such  ads  feel  some  
ambivalence  about  identifying  themselves  as  members  of  the  community  of  

  59  
people  who  have  resorted  to  such  means  to  find  a  partner.  By  ‘playing  with’  the  
genre  they  succeed  in  resisting  the  commodifying  nature  of  the  genre  (Coupland  
1996)  and  humanizing  themselves,  one  through  modesty,  and  the  other  through  
humor.  It  is  a  way  of  saying,  ‘even  though  I  am  posting  a  personal  ad,  I  am  not  the  
usual  kind  of  person  who  posts  such  ads.’  
 
While  membership  in  other  discourse  communities  does  not  usually  involve  the  
same  kind  of  ambivalence,  ‘tactical’  aspects  of  using  genres  like  bending  and  
blending  are  common  in  nearly  all  communities,  and,  indeed,  are  often  markers  
of  users’  expertise.  Of  course,  in  order  for  blending  to  be  effective  it  must  result  
in  some  sort  of  enhancement  that  contributes  to  the  overall  communicative  
purpose  being  achieved  more  effectively  or  more  efficiently.  Similarly,  when  
bending  a  genre,  one  must  be  careful  not  to  bend  it  to  the  point  of  breaking.  
Whether  a  particular  use  of  a  genre  is  considered  a  creative  innovation  or  an  
embarrassing  failure  is  ultimately  a  matter  of  whether  or  not  the  original  
communicative  purpose  of  the  genre  is  achieved.    
 
 
Modes,  Media  and  Context  
 
A  number  of  other  important  factors  determine  how  genres  are  used  and  how  
they  change.  One,  which  we  deal  with  in  more  detail  in  a  later  Sections  A9  and  
B9,  has  to  do  with  the  different  modes  (e.g.  writing,  graphics,  video)  that  are  
available  for  constructing  the  genre.  Another,  which  we  will  discuss  in  Section  
B8,  has  to  do  with  the  media  through  which  genres  are  produced  and  distributed.    
 
Both  of  these  factors  are  important  in  relation  to  the  genre  we  have  been  
discussing,  personal  advertisements,  given  the  fact  that  recently  this  genre  has,  
to  a  large  extent,  migrated  online.  Nowadays  it  is  more  likely  that  one  would  
encounter  such  an  advertisement  on  the  Internet  than  in  a  newspaper.  As  a  
result  of  this  migration,  the  genre  itself  has  changed  dramatically.    
First,  it  has  changed  in  terms  of  the  different  modes  that  are  available  to  users  to  
realize  the  moves  discussed  above.  Because  it  is  so  easy  to  upload  digital  
photographs  and  even  video,  self-­‐descriptions  in  online  personal  advertisements  
are  not  dependent  on  text  alone.    
 
Second,  websites  that  host  such  advertisements  often  require  users  to  fill  out  
web  forms,  which  specify  exactly  which  information  should  be  included  and  
render  that  information  in  a  predetermined  format.  Such  standardization  leads  
to  more  uniformity,  but  also  makes  it  easier  for  users  to  electronically  search  
through  thousands  of  ads  using  keywords.    
 
Third,  Internet-­‐based  dating  advertisements  include  all  kinds  of  ways  for  the  
advertiser  and  target  to  interact,  including  sending  online  messages,  engaging  in  
real-­‐time  video  chat,  or  exchanging  forms  of  communication  unique  to  this  
medium  like  virtual  ‘kisses’,  ‘pokes’  and  ‘hugs’,  ‘winks’  and  ‘hearts’  (see  Jones  
2009a).    
 

  60  
Finally,  with  the  development  of  mobile  technologies,  users  of  such  genres  can  
access  them  anywhere  through  their  mobile  phones  and  use  GPS  tools  to  search  
for  suitable  partners  within  a  certain  radius  of  their  present  location.    
The  point  is  that  genres  inevitably  change,  either  because  the  communicative  
goals  of  users  change  or  because  technologies  for  the  production  or  distribution  
of  texts  introduce  new,  more  efficient  ways  of  fulfilling  old  communicative  goals.  
Every  time  a  genre  changes,  however,  new  sets  of  conventions  and  constraints  
are  introduced,  and  users  need  to  invent  new  ways  to  operate  strategically  
within  these  constraints  and  to  bend  or  blend  the  genre  in  creative  ways.    
 
  Look  deeper  into  this  topic  online  

  61  
B4  CONSTRUCTING  REALITY  
 

Participants  and  Processes    

 
The  main  ways  authors  of  texts  promote  ideologies  is  by  constructing  for  us  
versions  of  reality  in  which  certain  kinds  of  participants  are  excluded,  and  those  
that  are  included  are  linked  to  each  other  in  certain  relationships,  often  based  on  
the  actions  they  are  portrayed  as  engaging  in  (which  we  have  called  processes).    
 
Different  kinds  of  processes  link  participants  in  different  ways.  As  we  have  seen,  
processes  involving  some  kind  of  physical  action  often  link  participants  in  ways  
in  which  one  participant  is  portrayed  as  doing  something  to  or  for  the  other.  
Processes  involving  saying  or  writing,  on  the  other  hand,  often  link  participants  
in  ways  in  which  one  participant  takes  the  position  of  the  speaker  or  writer  and  
the  other  takes  the  position  of  the  listener  or  reader.  Processes  involving  
thinking  and  feeling  usually  link  participants  to  ideas  or  emotions.    
 
Participants  can  also  be  linked  in  various  ways  that  show  their  relationship  with  
each  other:  they  might  be  portrayed  as  equal  or  equivalent  with  linking  verbs  
like  ‘to  be’  or  ‘to  seem’  (as  in  ‘this  ice-­‐cream  is  my  dinner’);  one  participant  might  
be  portrayed  as  possessing  another  with  words  like  ‘to  have’  or  ‘to  contain’  (as  in  
‘this  ice-­‐cream  contains  nuts’);  and  participants  might  be  linked  to  each  other  in  
other  kinds  of  relationship  like  cause  and  effect  with  words  like  ‘to  cause,’    ‘to  
lead  to,’  or  ‘to  result  in’  (as  in  ‘ice-­‐cream  leads  to  obesity’).    
 
Finally,  processes  themselves  can  sometimes  be  transformed  into  participants  
and  linked  to  other  participants  or  other  processes  (as  in  ‘eating  ice-­‐cream  
caused  my  divorce’).  In  this  last  example,  one  process,  the  action  of  eating  ice  
cream,  is  linked  in  a  cause  and  effect  relationship  with  another  process,  the  
speaker’s  getting  divorced  from  his  or  her  spouse.  Turning  a  process  into  a  
participant  is  known  as  nominalization  and  is  often  a  characteristic  of  technical  
or  academic  texts.      
 
One  example  of  how  participants  and  processes  can  be  combined  in  texts  to  
create  certain  versions  of  reality  can  be  found  in  the  warning  labels  that  most  
governments  require  appear  on  cigarette  packets.  These  requirements  did  not  
always  exist,  and  when  the  U.S.  government  first  instituted  them  in  1966,  there  
was  fierce  negotiation  between  legislators  and  tobacco  companies,  who  at  that  
time  still  had  considerable  influence  with  the  government  on  how  the  warnings  
should  be  worded.  The  first  warning  labels  that  appeared  on  cigarettes  in  the  U.S.  
read:    
 
Caution:  Cigarette  Smoking  May  be  Hazardous  to  Your  Health  
 
Here,  the  process  type  used  is  relational  (‘to  be’),  linking  the  nominalized  process  
of  smoking  cigarettes  to  a  possible  attribute:  ‘hazardousness’.  This  link  is  

  62  
weakened  by  the  use  of  the  modal  verb  ‘may’,  which  reduces  the  certainty  of  the  
statement.  One  can  almost  hear  the  voice  of  the  government  competing  with  the  
voice  of  the  tobacco  companies  in  this  statement,  the  one  working  to  claim  that  
cigarette  smoking  is  risky  and  the  other  working  to  undermine  this  claim.    
In  1970,  the  U.S.  Congress  passed  new  legislation,  which  revised  the  warning  to  
read:    
 
Warning:  The  Surgeon  General  Has  Determined  that  Cigarette  Smoking  is  
Dangerous  to  Your  Health  
 
The  first  difference  we  can  notice  about  this  statement  is  that  it  is  now  
characterized  as  a  ‘warning’  rather  than  just  a  ‘caution’.  The  second  thing  we  can  
notice  is  that  the  participants  and  processes  have  changed.  Now  the  main  
participant  is  no  longer  the  nominalized  process  of  ‘cigarette  smoking’  but  a  
person,  the  Surgeon  General,  the  chief  medical  officer  of  the  United  States,  
engaged  in  a  mental  process,  that  of  ‘determining’.  The  statement  in  the  previous  
warning  about  cigarettes  being  hazardous  has  itself  become  a  participant,  the  
thing  that  the  surgeon  general  has  determined,  and  has  been  slightly  modified,  
the  relational  link  becoming  more  certain  (‘is’  rather  than  ‘may  be’),  and  the  
attribute  changed  to  ‘dangerous’.  On  the  one  hand,  it  is  easy  to  see  how  this  
warning  is  in  some  ways  stronger  than  the  previous  one:  ‘dangerous’  seems  
more  serious  than  ‘hazardous’  and  the  voice  of  the  Surgeon  General  seems  to  add  
authority  to  the  statement.  On  the  other  hand,  the  statement  about  the  risk  of  
cigarette  smoking  is  no  longer  the  main  clause  of  the  sentence,  but  has  been  
‘demoted’  to  the  status  of  a  participant.  In  other  words,  while  the  previous  
warning  was  about  cigarette  smoking  and  its  ‘hazardousness,’  this  sentence  is  
about  the  Surgeon  General  and  what  he  (at  the  time,  a  man)  had  determined.    
In  1985,  the  warning  label  was  again  changed  to  read:    
 
SURGEON  GENERAL'S  WARNING:  Smoking  Causes  Lung  Cancer,  Heart  
Disease,  Emphysema,  And  May  Complicate  Pregnancy.  
 
Here,  the  main  participant  is  once  again  the  nominalized  process  ‘smoking’,  but  a  
new  process  has  been  introduced,  the  process  of  ‘causing’.  This  process  is  also  a  
relational  one,  but  it  portrays  a  different  kind  of  relationship.  Rather  than  simply  
talking  about  an  attribute  of  smoking,  it  places  smoking  in  a  cause  and  effect  
relationship  with  a  number  of  serious  diseases  (‘lung  cancer,’  ‘heart  disease’  and  
‘emphysema’).  Smoking’s  relationship  with  ‘pregnancy’,  however,  is  more  
uncertain.  First,  the  modal  verb  ‘may’  weakens  the  relationship  created  by  the  
process.  Second,  the  process  itself,  also  one  of  causation  (‘complicate’  means  ‘to  
cause  to  be  complicated’),  is  much  more  vague.  It  is  uncertain  exactly  how  
pregnancy  might  be  made  ‘complicated’  and  what  the  implications  of  that  might  
be.    
 
It  is  interesting  to  note  that  the  cigarette  warnings  mandated  by  the  government  
in  the  United  States,  the  country  where  most  to  the  world’s  biggest  tobacco  
companies  are  based,  tend  to  portray  cigarettes  as  either  having  certain  
attributes  (being  ‘harmful’)  or  being  in  certain  other  kinds  of  relationships  with  
other  participants  rather  than  doing  things  to  people.  Although  ‘cause’  may  seem  

  63  
to  be  about  doing  something,  it  is  actually  more  about  the  relationship  between  
two  things,  one  thing  leading  to  another  thing.  Warnings  used  in  other  countries,  
on  the  other  hand,  often  use  ‘action  processes.’    One  warning  used  in  Australia,  
for  example  is:    
 
Smoking  harms  unborn  babies    
 
Here  ‘smoking’  is  portrayed  as  doing  something  (harming)  to  someone  (unborn  
babies).  Similarly,  since  2003,  cigarettes  in  the  European  Union  have  carried  
warnings  like:      
 
Smoking  seriously  harms  you  and  others  around  you  
 
Smoking  while  pregnant  harms  your  child  
 
and  the  direct  and  unambiguous  statement:    
 
Smoking  Kills  
 
In  the  examples  above  it  is  clear  how  the  use  of  different  kinds  of  participants  
and  processes  constructs  very  different  versions  of  the  risk  of  cigarette  smoking.  
At  the  same  time,  it  is  important  to  caution  that  searching  for  ideology  in  texts  is  
usually  not  simple  or  straightforward.  One  cannot,  for  example,  say  that  certain  
process  types  or  other  grammatical  features  like  nominalization  always  result  in  
certain  kinds  of  effects.  Rather,  grammar  is  a  resource  that  authors  draw  upon  to  
represent  reality  in  particular  ways.    

 
Constructing  Relationships  
 
Constructing  reality  is  not  just  a  matter  of  representing  what  is  going  on.  It  is  
also  a  matter  of  the  author  of  a  text  constructing  a  certain  kind  of  relationship  
with  the  reader  or  listener  and  communicating  something  about  the  relevance  of  
what  is  going  on  to  him  or  her.  As  stated  before,  one  way  this  is  done  is  to  use  the  
language’s  system  of  modality.  The  use  of  the  modal  verb  ‘may’  in  the  statement  
‘Cigarette  smoking  may  be  hazardous  to  your  health,’  for  example,  creates  in  the  
reader  some  doubt  about  the  certainty  of  the  statement.    
 
Another  way  authors  might  construct  a  relationship  with  readers  is  through  the  
use  of  pronouns  like  ‘you’  and  we’.  By  using  the  possessive  pronoun  ‘your’  in  the  
above  statement,  for  example,  the  authors  of  the  statement  make  the  potential  
‘hazardousness’  of  cigarette  smoking  relevant  to  readers.  Similarly,  the  
statement,  ‘Smoking  seriously  harms  you  and  others  around  you’  makes  the  
harm  of  cigarettes  directly  relevant  by  making  the  reader  a  participant  in  the  
statement  and,  particularly,  the  participant  to  which  the  act  of  ‘harming’  is  being  
done.  This  statement  also  constructs  readers  as  socially  responsible  by  implying  
that  they  would  not  only  wish  to  avoid  harm  to  themselves,  but  also  harm  to  
those  around  them.  An  even  more  striking  example  of  this  technique  can  be  seen  
in  the  Australian  warning  label  below:  

  64  
  
Protect  children:  don't  make  them  breathe  your  smoke  
 
In  this  example  the  message  is  also  personalized  by  making  the  reader  a  
participant.  In  contrast  to  the  warning  above,  however,  which  positions  the  
reader  as  a  victim  of  cigarette  smoking,  this  warning  positions  the  reader  as  the  
potential  agent  of  harm,  making  children  breathe  smoke  which  is  explicitly  
portrayed  as  ‘belonging  to’  him  or  her.    
 
Finally,  texts  create  relationships  between  authors  and  readers  through  the  use  
of  what  we  have  been  calling  ‘social  languages.’  Consider  the  two  examples  
below.    
 
1)  Smoking  when  pregnant  harms  your  baby  (European  Union)    
 
2)  SURGEON  GENERAL’S  WARNING:  Smoking  By  Pregnant  Women  May  
Result  in  Fetal  Injury,  Premature  Birth,  And  Low  Birth  Weight  (United  
States)    
 
Both  of  these  examples  are  about  the  same  thing:  smoking  by  pregnant  women.  
This  first  text,  however,  constructs  a  reader  who  is  herself  a  pregnant  woman,  
whereas  the  second  constructs  a  reader  who,  while  he  or  she  may  be  interested  
in  ‘pregnant  women’,  may  not  be  one.  Furthermore,  the  first  example  uses  
common,  everyday  language  and  few  nominalizations,  constructing  the  author  as  
a  person  not  so  different  from  the  reader,  someone  akin  to  a  friend  or  a  relative.    
 
The  second  example,  on  the  other  hand,  uses  very  dense  scientific  language  and  
nominalizations  like  ‘fetal  injury’  in  which  the  process  of  ‘harming’  from  the  first  
example  is  transformed  into  a  noun,  and  the  participant  ‘your  baby’  is  
transformed  into  an  adjective  modifying  that  noun  (‘fetal’).  This  sort  of  language  
constructs  the  author  as  some  kind  of  expert,  perhaps  a  doctor  or  a  research  
scientist,  and  creates  a  considerable  distance  between  him  or  her  and  the  reader.    
 
As  can  be  seen  from  these  examples,  social  languages  and  other  interpersonal  
aspects  of  texts  work  to  portray  the  authors  of  the  texts  as  certain  kinds  of  
people  and  also  construct  readers  of  the  texts  as  certain  kinds  of  people.  Another  
way  to  say  this  is  that  texts  make  available  certain  ‘reading  positions’  (Hodge  and  
Kress,  1988)  that  situate  readers  in  relation  to  the  authors  of  the  text,  the  topic  
that  the  text  deals  with,  and  other  people  or  institutions  relevant  to  the  topic.  
The  extent  to  which  readers  are  able  and  willing  to  occupy  these  ‘reading  
positions’  helps  to  determine  the  kind  of  ideological  effect  the  texts  will  have.    
 
While  the  kind  of  textual  analysis  illustrated  here  can  tell  us  something  about  the  
versions  of  reality  that  texts  construct  and  about  the  kinds  of  reading  positions  
they  make  available,  it  is  impossible  to  say  for  certain  just  by  analyzing  texts  
what  their  actual  effect  will  be  on  readers.  Some  pregnant  women,  for  example,  
might  respond  more  readily  to  the  plain  familiar  language  and  the  personal  
approach  in  example  1  above.  Others  might  be  more  persuaded  by  the  
authoritative  voice  of  example  2.    

  65  
  
To  really  understand  how  people  actually  interpret  texts,  or,  for  that  matter,  how  
ideologies  end  up  finding  their  way  into  texts  in  the  first  place,  it  is  necessary  to  
go  beyond  texts  themselves  and  analyze  both  discourse  practices,  the  practices  
authors  engage  in  when  creating  texts  and  the  practices  readers  engage  in  when  
interpreting  them,  and  social  practices,  the  activities,  norms,  and  social  
relationships  that  make  up  readers’  social  worlds.    
 
The  more  we  know  about  the  negotiations  that  went  on  between  big  tobacco  
companies  and  politicians  in  the  United  States  in  the  late  1960’s,  for  example,  the  
better  we  can  understand  why  early  cigarette  warnings  were  worded  the  way  
they  were;  and  the  more  we  understand  readers’  experiences  of  and  knowledge  
about  smoking  and  the  status  of  smoking  in  their  circle  of  acquaintances,  the  
better  we  will  be  able  to  understand  the  effects  warnings  on  cigarette  packages  
might  have  had  on  their  behavior,  For  this  reason,  people  who  are  interested  in  
studying  ideology  in  discourse,  known  as  critical  discourse  analysts,    are  
increasingly  supplementing  textual  analysis  with  more  ethnographic  research  
techniques  like  interviews,  observations,  and  historical  research.  

  66  
B5  THE  TEXTURE  OF  TALK  
 
In  our  analysis  of  how  people  make  sense  of  written  texts,  we  introduced  the  
concept  of  texture.  Texture,  we  said,  basically  comes  from  two  things:  the  ways  
different  parts  of  a  text  are  related  to  one  another,  and  the  various  expectations  
that  people  have  about  texts.  Making  sense  of  conversations  also  involves  these  
two  aspects  of  communication:  the  structure  and  patterning  of  the  
communication  and  the  broader  expectations  about  meaning  and  human  
behavior  that  participants  bring  to  it.  Generally  speaking,  conversation  analysis  
focuses  more  on  the  first  aspect,  and  pragmatics  focuses  more  on  the  second.    
The  basis  of  pragmatics  is  the  idea  that  people  enter  into  conversations  with  the  
assumption  that  the  people  they  are  conversing  with  will  behave  in  a  logical  way.  
The  philosopher  Herbert  Paul  Grice  called  this  assumption  the  cooperative  
principle.  When  people  engage  in  conversation,  he  said,  they  do  so  with  the  idea  
that  people  will:      
 
Make  (their)  conversational  contribution  such  as  is  required,  at  the  stage  at  
which  it  is  occurs,  by  the  accepted  purpose  or  direction  of  the  talk  exchange  
in  which  you  are  engaged.  (Grice  1975:  45)  
 
What  he  meant  by  this  was  that  when  people  talk  with  each  other  they  generally  
cooperate  in  making  their  utterances  understandable  by  conforming  to  what  
they  believe  to  be  the  other  person’s  expectations  about  how  people  usually  
behave  in  conversation.  Most  people,  he  said,  have  four  main  expectations  about  
conversational  behavior:    
 
1.  That  what  people  say  will  be  true  (the  maxim  of  quality)    
2.  That  what  people  say  will  be  relevant  to  the  topic  under  discussion  (the  
maxim  of  relevance)  
3.  That  people  will  try  to  make  what  they  mean  clear  and  unambiguous  
(the  maxim  of  manner)    
4.  That  people  will  say  as  much  as  they  need  to  say  to  express  their  
meaning  and  not  say  more  that  they  need  to  say  (the  maxim  of  quantity)    
 
Grice  called  these  four  expectations  maxims.  Maxims  are  not  rules  that  must  be  
followed;  rather,  they  are  general  statements  of  principle  about  how  things  
should  be  done.  In  actual  conversations,  however,  people  often  violate  or  ‘flout’  
these  maxims:  they  say  things  that  are  not  true;  they  make  seemingly  irrelevant  
statements;  they  are  not  always  clear  about  what  they  mean;  and  they  
sometimes  say  more  than  they  need  to  or  not  enough  to  fully  express  their  
meaning.  The  point  that  Grice  was  making  was  not  that  people  always  follow  or  
even  that  they  ‘should’  follow  these  maxims,  but  that  when  they  do  not  follow  
them,  they  usually  do  so  for  a  reason:  the  very  fact  that  they  have  flouted  a  
maxim  itself  creates  meaning,  a  special  type  of  meaning  known  as  implicature,  
which  involves  implying  or  suggesting  something  without  having  to  directly  
express  it.  When  people  try  to  make  sense  of  what  others  have  said,  they  do  so  
against  the  background  of  these  default  expectations.  When  speakers  do  not  

  67  
behave  as  expected,  listeners  logically  conclude  that  they  are  trying  to  imply  
something  indirectly  and  try  to  work  out  what  it  is.    
 
If  your  friend  asks  you  if  you  think  her  new  boyfriend  is  good  looking,  but  you  do  
not  think  he  is,  you  might  say  something  like,  ‘He  has  a  lovely  personality,’  
violating  the  maxim  of  relevance  (her  question  was  about  his  appearance,  not  his  
personality),  or  you  might  say  something  rather  vague  which  communicates  that  
you  do  not  think  he  is  very  good  looking  but  which  avoids  saying  this  explicitly,  
violating  the  maxim  of  manner.  
 
The  obvious  question  is,  why  do  people  do  this?  Why  don’t  they  simply  
communicate  what  they  mean  directly?  One  reason  is  that  implicature  allows  us  
to  manage  the  interpersonal  aspect  of  communication.  We  might,  for  example,  
use  implicature  to  be  more  polite  or  avoid  hurting  someone’s  feelings.  We  might  
also  use  implicature  to  avoid  making  ourselves  too  accountable  for  what  we  have  
said  -­‐-­‐    in  other  words,  to  say  something  without  ‘really  saying’  it.      
 
Of  course,  the  fact  that  someone  says  something  that  is  not  true  or  is  not  entirely  
clear  does  not  necessarily  mean  they  are  creating  implicature.  Sometimes  people  
simply  lie.  You  might,  for  example,  tell  your  friend  that  you  think  her  boyfriend  is  
very  handsome.  In  this  case,  you  have  not  created  any  indirect  meaning.  Your  
meaning  is  very  direct.  It  is  just  not  true.  Another  example  can  be  seen  in  the  
often-­‐quoted  exchange  below:    
 
A      Does  your  dog  bite?  
B      No.  
A      [Bends  down  to  stroke  it  and  gets  bitten]  Ow!  You  said  your  dog    
doesn’t  bite.  
B      That  isn’t  my  dog.  
 
Here  A  has  violated  the  maxim  of  quantity  by  saying  too  little,  but,  in  doing  so,  he  
has  not  created  implicature.  He  has  simply  said  too  little.  And  so  for  the  flouting  
of  a  maxim  to  be  meaningful,  it  must  be  done  within  the  overall  framework  of  the  
cooperative  principle.  The  person  flouting  a  maxim  must  expect  that  the  other  
person  will  realize  that  they  are  flouting  the  maxim  and  that  the  meaning  created  
by  this  is  not  too  difficult  to  figure  out.    
 
 

How  we  do  things  with  words  

 
Another  important  aspect  of  pragmatics  concerns  how  people  accomplish  
various  social  actions  when  they  talk,  such  as  requesting,  promising  and  
threatening.  The  philosopher  John  Austin  pointed  out  that  certain  utterances,  
when  they  are  spoken,  have  the  effect  of  actually  performing  some  action  in  the  
physical  world.  When  the  officiant  at  a  wedding  ceremony,  for  example,  says,  ‘I  
now  pronounce  you  husband  and  wife,’  it  is  by  this  pronouncement  that  the  
couple  becomes  married,  or  when  a  judge  says,  ‘I  sentence  you  to  five  years  in  

  68  
prison,’  it  is  by  this  utterance  that  the  person  to  whom  this  is  uttered  is  
sentenced.  Austin  called  these  kinds  of  utterances  performatives.    
 
While  Austin’s  insight  might  seem  rather  obvious  now,  it  was  quite  revolutionary  
at  the  time  he  was  writing,  when  most  philosophers  of  language  were  mainly  
focused  on  analyzing  sentences  in  terms  of  whether  or  not  they  were  ‘true’.  
Austin  pointed  out  that,  for  many  utterances,  their  ‘truth  value’  is  not  as  
important  as  whether  or  not  they  are  able  to  perform  the  action  they  are  
intended  to  perform.    
 
The  more  Austin  thought  about  this  idea  of  performatives,  the  more  he  realized  
that  many  utterances  —  not  just  those  containing  phrases  like  ‘I  pronounce…’  
and  ‘I  declare…’  and  ‘I  command...’  —  have  a  performative  function.  If  somebody  
says  to  you,  ‘Cigarette  smoking  is  dangerous  to  your  health,’  for  example,  he  or  
she  is  usually  not  just  making  a  statement.  He  or  she  is  also  doing  something,  that  
is,  warning  you  not  to  smoke.    
 
Austin  called  these  utterances  that  perform  actions  speech  acts.  The  important  
thing  about  these  kinds  of  utterances,  he  said,  is  not  so  much  their  ‘meaning’  as  
their  ‘force’,  their  ability  to  perform  actions.  All  speech  acts  have  three  kinds  of  
force:  locutionary  force,  the  force  of  what  the  words  actually  mean,  illocutionary  
force,  the  force  of  the  action  the  words  are  intended  to  perform,  and  
perlocutionary  force,  the  force  of  the  actual  effect  of  the  words  on  listeners.    
 
One  of  the  problems  with  analyzing  speech  acts  is  that,  for  many  of  the  same  
reasons  speakers  express  meanings  indirectly  by  flouting  conversational  
maxims,  they  also  express  speech  acts  indirectly.  In  other  words,  the  locutionary  
force  of  their  speech  act  (the  meaning  of  the  words)  might  be  very  different  from  
the  illocutionary  force  (what  they  are  actually  doing  with  their  words).  We  have  
already  discussed  a  number  of  examples  of  this,  such  as  the  question  ‘Do  you  
have  a  pen?’  uttered  to  perform  the  act  of  requesting.    
 
And  so  the  problem  is,  how  do  we  figure  out  what  people  are  trying  to  do  with  
their  words?  For  Austin,  the  main  way  we  do  this  is  by  logically  analyzing  the  
conditions  under  which  a  particular  utterance  is  produced.  He  called  the  ability  
of  an  utterance  to  perform  a  particular  action  the  ‘felicity’  (or  ‘happiness’)  of  the  
utterance,  and  in  order  for  speech  acts  to  be  ‘happy’,  certain  kinds  of  conditions  
must  be  met,  which  Austin  called  felicity  conditions.    
 
Some  of  these  conditions  relate  to  what  is  said.  For  some  speech  acts  to  be  
felicitous,  for  example,  they  must  be  uttered  in  a  certain  conventional  way.  The  
officiant  at  a  wedding  must  say  something  very  close  to  ‘I  now  pronounce  you  
husband  and  wife’  in  order  for  this  to  be  a  pronouncement  of  marriage.  Some  of  
the  conditions  have  to  do  with  who  utters  the  speech  act  —  the  kind  of  authority  
or  identity  they  have.  Only  someone  specially  empowered  to  do  so,  for  instance,  
is  able  to  perform  marriages.  If  a  random  person  walked  up  to  you  and  your  
companion  on  the  street  and  said,  ‘I  now  pronounce  you  husband  and  wife,’  this  
would  not  be  considered  a  felicitous  pronouncement  of  marriage.  Some  of  these  
conditions  concern  the  person  or  people  to  whom  the  utterance  is  addressed.  

  69  
They  must  generally  be  able  to  decipher  the  speech  act  and  comply  with  it.  
People  under  a  certain  age,  for  example,  cannot  get  married,  and  so  the  
pronouncement  of  marriage  given  above  would  not  succeed  as  a  speech  act.    
Similarly,  if  the  two  people  to  whom  this  pronouncement  is  uttered  are  not  
willing  to  get  married,  the  pronouncement  would  also  lack  felicity.  Finally,  some  
of  these  conditions  may  have  to  do  with  the  time  or  place  the  utterance  is  issued.  
Captains  of  ships,  for  example,  are  only  empowered  to  make  pronouncements  of  
marriage  aboard  their  ships.    
 
And  so,  according  to  Austin  and  his  followers,  the  main  way  we  figure  out  what  
people  are  trying  to  do  when  they  speak  to  us  is  by  trying  to  match  the  
conditions  in  which  an  utterance  is  made  to  the  conditions  necessary  for  
particular  kinds  of  speech  acts.  So,  when  somebody  comes  up  to  me  in  a  bar  and  
says,  ‘Hey  mate,  I  suggest  you  leave  my  girlfriend  alone,’  I  use  my  logic  to  try  to  
figure  out  what  he  is  doing  and  what  he  is  trying  to  get  me  to  do.  At  first  I  might  
think  that  he  is  making  a  suggestion  to  me.  But,  when  I  consider  the  conditions  of  
the  situation,  I  realize  that  this  utterance  does  not  fulfill  the  necessary  conditions  
of  a  suggestion,  one  of  which  is  that  whether  or  not  I  follow  the  suggestion  is  
optional.  I  can  tell  quite  clearly  from  the  expression  on  this  fellow’s  face  that  
what  he  is  ‘suggesting’  is  not  optional.  I  also  realize  that  there  will  probably  be  
unpleasant  consequences  for  me  should  I  fail  to  comply.  Given  this  condition,  I  
can  only  conclude  that  what  he  is  doing  with  his  words  is  not  making  a  
suggestion  but  issuing  a  threat.  
 
The  important  thing  about  this  example  is  that  I  must  use  both  of  the  tools  
introduced  above.  I  must  make  use  of  the  cooperative  principle  to  realize  that  he  
is  flouting  the  maxim  of  quality  (he  is  not  making  a  suggestion)  and  that  there  
must  be  some  reason  for  this,  and  I  must  be  able  to  analyze  the  conditions  in  
which  this  utterance  is  made  to  figure  out  what  the  speaker  is  actually  trying  to  
do.        
 
 
  Look  deeper  into  this  topic  online  
 

Sense  and  sequencing:  Conversation  analysis  

 
Whereas  pragmatics  begins  with  the  assumption  that  conversations  are  logical,  
conversation  analysis  begins  with  the  assumption  that  they  are  orderly.  What  
orderly  means  is  that  they  follow  a  certain  predictable  pattern,  with  some  kinds  
of  utterances  necessarily  coming  before  or  following  other  kinds  of  utterances.    
Conversation  analysts  also  see  utterances  as  actions.  Where  they  differ  is  in  their  
ideas  about  how  we  interpret  these  actions  —  what  gives  ‘force’  to  our  words.    
Whereas  followers  of  Austin  consider  the  speaker’s  intentions  and  the  conditions  
under  which  the  words  are  uttered  to  be  the  most  important  things,  
conversation  analysts  consider  the  utterance  that  occurred  prior  to  the  utterance  
in  question,  and  the  one  that  occurs  afterwards  to  be  more  important.  In  other  
words,  they  believe  we  interpret  utterances  chiefly  based  on  how  they  ‘fit’  
sequentially  with  other  utterances  in  a  conversation.    

  70  
  
The  core  of  conversation  analysis,  then,  is  the  exploration  of  the  sequential  
structure  of  conversation.  According  to  Schegloff    and  Sacks  (1973),  social  
interaction  is  often  arranged  in  pairs  of  utterances  —  what  one  person  says  
basically  determines  what  the  next  person  can  say.  They  call  these  sequences  of  
‘paired  actions’  adjacency  pairs.  Examples  of  common  adjacency  pairs  are  
'question/answer',  'invitation/acceptance’,  and  'greeting/greeting'.    
 
The  most  important  thing  about  the  two  utterances  that  make  up  an  adjacency  
pair  is  that  they  have  a  relationship  of  conditional  relevance.  In  other  words,  one  
utterance  is  dependent  on  (conditioned  by)  the  other  utterance.  The  first  
utterance  determines  what  the  second  utterance  can  be  (a  question,  for  example,  
should  be  followed  by  an  answer,  and  a  greeting  should  be  followed  by  a  
greeting).  In  the  same  way,  the  second  utterance  also  determines  what  the  first  
utterance  has  been  understood  to  be.  If  I  have  given  you  an  answer,  this  provides  
evidence  that  I  have  taken  your  preceding  utterance  to  be  a  question.  This  is  a  
very  big  difference  between  conversation  analysis  and  the  speech  act  theory  of  
Austin.  For  speech  act  theory,  the  conditions  for  whether  or  not  an  utterance  is  a  
particular  speech  act  include  things  like  the  intentions  and  identities  of  the  
speakers  and  the  context  of  the  situation.  For  conversation  analysts,  the  
conditions  that  determine  how  an  utterance  should  be  interpreted  must  exist  
within  the  conversation  itself.  
 
At  the  same  time,  conversation  analysis  also  focuses  on  how  speakers  make  use  
of  the  default  expectations  people  bring  to  conversations  in  order  to  make  
meaning.  The  main  difference  is  that  important  expectations  are  not  so  much  
about  the  content  of  utterances  (whether  or  not,  for  example,  they  are  ‘true’  or  
‘clear’),  but  rather  about  the  structure  of  conversation,  and  particularly  the  ways  
that  utterances  should  ‘fit’  with  previous  utterances.  The  idea  behind  adjacency  
pairs  is  that  when  one  person  says  something,  he  or  she  creates  a  'slot'  for  the  
next  person  to  ‘fill  in’  in  a  particular  way.  If  they  fill  it  in  in  the  expected  way,  this  
is  called  a  'preferred  response'.  If  they  do  not  fill  in  this  slot  in  the  expected  way,  
their  interlocutor  ‘hears’  the  preferred  response  as  being  ‘officially  absent’.  As  
Schegloff  (1968:  1083)  put  it:    
 
Given  the  first,  the  second  is  expectable.  Upon  its  occurrence,  it  can  be  
seen  to  be  the  second  item  to  the  first.  Upon  its  non-­‐occurrence,  it  can  be  
seen  to  be  officially  absent.    
 
Take  for  example  the  following  exchange  between  a  woman  and  her  boyfriend:  
   
  A:  I  love  you.    
  B:  Thank  you.    
 
The  reason  this  exchange  seems  odd  to  us,  and  undoubtedly  seems  odd  to  A,  is  
that  the  preferred  response  to  an  expression  of  love  is  a  reciprocal  expression  of  
love.  When  this  response  is  not  given,  it  creates  implicature.  Thus,  the  most  
important  thing  about  B’s  response  is  not  the  meaning  that  he  expresses  
(gratitude),  but  the  meaning  that  is  absent  from  the  utterance.    

  71  
  
All  first  utterances  in  adjacency  pairs  are  said  to  have  a  ‘preferred’  second  
utterance.  For  example,  the  preferred  response  to  an  invitation  is  an  acceptance.  
The  preferred  response  to  a  greeting  is  a  greeting.  What  makes  a  preferred  
response  preferred  is  not  that  the  person  who  offered  the  first  utterance  would  
‘prefer’  this  response  (the  preferred  response  for  an  accusation,  for  example,  is  a  
denial),  but  rather  that  this  is  the  response  which  usually  requires  the  least  
additional  conversational  work.  So  the  preferred  response  is  the  most  efficient  
response.  When  we  issue  dispreferred  responses,  we  often  have  to  add  
something  to  them  in  order  to  avoid  producing  unintended  implicature.    For  
example,  if  you  ask  me  to  come  to  your  party  and  I  accept  your  invitation,  all  I  
have  to  do  is  say  ‘Sure!’  But  if  I  want  to  refuse  the  invitation,  I  cannot  just  say  
‘No!’  If  I  do,  I  create  the  implicature  that  I  do  not  much  like  you  or  care  about  
your  feelings.  If  I  want  to  avoid  communicating  this,  I  have  to  supplement  it  with  
other  things  like  an  apology  (‘I’m  really  sorry…’)  and  an  excuse  or  account  of  why  
I  cannot  come  to  your  party  (‘I  have  to  do  my  discourse  analysis  homework’).  
 
You  can  divide  almost  any  conversation  into  a  series  of  adjacency  pairs.  
Sometimes,  though,  adjacency  pairs  can  be  quite  complicated,  with  pairs  of  
utterances  overlapping  or  being  embedded  in  other  pairs  of  utterances.  
Nevertheless,  for  conversation  analysts,  it  is  this  underlying  ‘pair  wise  
organization’  of  utterances  that  helps  us  to  make  sense  of  our  conversations  and  
use  them  to  accomplish  actions  in  an  orderly  way.    
 
  Look  deeper  into  this  topic  online  
 
There  is,  of  course,  a  lot  more  to  both  pragmatics  and  conversation  analysis  than  
has  been  covered  in  this  brief  summary.  Pragmatics,  for  example,  has  much  more  
to  say  about  the  various  cognitive  models  that  people  bring  to  interaction,  and  
conversation  analysis  has  much  to  say  about  how  people  manage  things  like  
turn-­‐taking,  topic  negotiation,  openings  and  closings,  and  repair  in  
conversations.  What  we  have  focused  on  here  is  primarily  how  each  of  these  
approaches  addresses  the  problem  of  ambiguity  in  spoken  discourse  —  the  
problem  that  people  do  not  always  say  what  they  mean  or  mean  what  they  say.    

  72  
B6  NEGOTIATING  RELATIONSHIPS  AND  ACTIVITIES  
 

Power  and  politeness  

 
Whenever  we  interact  with  someone  we  always  communicate  something  about  
our  relationship  with  them.  We  do  this  by  using  various  discursive  strategies,  
which  we  have  divided  into  two  categories:  involvement  strategies  and  
independence  strategies.  Involvement  strategies  are  strategies  people  use  to  
communicate  friendliness  or  solidarity,  and  independence  strategies  are  
strategies  people  use  to  communicate  respect  or  deference.    
 
In  many  cases,  both  parties  in  an  interaction  share  a  fairly  clear  idea  about  how  
close  they  are  and  whether  one  has  more  power  than  the  other,  but  in  other  
cases,  participants  in  interaction  need  to  negotiate  their  relationship.  Such  
negotiations  are  common,  for  example,  as  people  move  from  more  distant  to  
closer  relationships,  or  when  one  person  wishes  to  challenge  another  person’s  
assertion  of  power  or  dominance.    
 
Regardless  of  whether  or  not  a  relationship  is  seen  as  ‘negotiable’,  we  always  
approach  interactions  with  certain  sets  of  expectations  about  how  independence  
and  involvement  strategies  will  be  used  to  communicate  information  about  
power  and  intimacy.  We  call  these  expectations  face  systems.  Although  
expectations  about  when  independence  and  involvement  strategies  are  
appropriate  and  what  they  mean  vary  across  cultures  and  groups,  most  people  
enter  interaction  with  three  basic  ideas:  1)  in  interactions  where  the  parties  are  
socially  distant  but  relatively  equal,  both  parties  are  likely  to  use  independence  
strategies  (deference  face  system);  2)  in  interactions  where  people  are  close  and  
relatively  equal,  they  are  likely  to  use  involvement  strategies  (solidarity  face  
system);  and  3)  in  interactions  in  which  one  person  has  more  power  than  the  
other  (regardless  of  their  social  distance)  the  more  powerful  one  is  more  likely  
to  use  involvement  strategies  and  the  less  powerful  one  is  more  likely  to  use  
independence  strategies.    
 
Like  the  conversational  maxims  we  discussed  in  the  last  section,  these  ‘systems’  
should  not  be  treated  as  ‘rules’,  but  rather  as  broad  sets  of  expectations  people  
draw  on  to  decide  how  to  act  towards  other  people  and  how  to  interpret  others’  
behavior  towards  them.  Since  power  and  distance  are  relative  rather  than  
absolute,  and  because  interaction  often  involves  the  negotiation  of  power  and  
distance,  people  usually  employ  both  independence  and  involvement  strategies,  
mixing  them  tactically  depending  on  the  situation  and  what  they  are  trying  to  
accomplish  in  the  interaction.    
 
An  example  of  the  way  participants  often  strategically  mix  independence  and  
involvement  strategies  can  be  seen  in  the  following  conversation  between  a  
senior  engineer  (Martin)  and  his  subordinate  (Ollie)  reported  in  Ladegaard  
(2011):    

  73  
  
 
Martin:  Happy  birthday  or  (0.2)  whatever  it  is  (laughing)  
Ollie:  thank  you  (0.2)  it’s  actually  a  while  ago  
Martin:  okay  eh:  Ollie//  
Ollie:  //there’s  Danish  pastry  over  there  if  you’re  interested  (0.2)  
Martin:  thanks  ah:  (0.6)  (talks  about  tape  recorder)  
Martin:  okay  well  to  cut  a  long  story  short  Sam  called  (0.2)  and  I’m  not  
sure  how  busy  you  are  or  what  you’re  doing  right  now  (0.4)  
Ollie:  ah:  we’re  just  about  to  launch  the  [name]  project  and  ah:  
Martin:  okay  
Ollie:  so  this  is  where  we  are  [xxx]  quite  busy  (0.5)  but  Sam  called  you  
said  
Martin:  yes  (0.2)  
Ollie:  and  he?  (0.3)  
Martin:  he  needs  some  help  here  and  now  (0.2)  he  needs  someone  to  
calculate  the  price  of  rubber  bands  (0.3)  for  the  [name]  project  in  India  
Ollie:  okay  
Martin:  they  expect  the  customer  to  sign  today  (1.3)  
Ollie:  okay        
(Ladegaard  2011:  14-­‐15)  
 
In  this  example,  Martin,  the  more  powerful  participant,  begins  using  involvement  
strategies,  wishing  Ollie  happy  birthday  (although  it  is  not  his  birthday)  and  
laughing.  Ollie,  on  the  other  hand,  though  friendly,  uses  more  independence  
strategies,  accepting  the  inappropriate,  birthday  wish  and  then  using  words  like  
‘actually’  and  ‘a  while’  to  soften  his  revelation  that  it  is  not  his  birthday,  and  then  
offering  Martin  some  pastry  in  a  way  which  is  designed  not  to  impose  on  him  
(‘…if  you’re  interested’).  Were  Martin  and  Ollie  equals  and  friends,  the  
inappropriate  birthday  wishes  might  have  been  answered  in  a  more  direct  way  
like,  ‘What  are  you  talking  about?  My  birthday  was  ages  ago!’,  and  the  offer  of  
pastry  might  have  been  more  insistent  (Have  some  Danish!).  In  other  words,  the  
mixture  of  involvement  and  independence  strategies  in  the  beginning  of  the  
conversation  are  what  one  might  expect  within  a  hierarchical  face  system.    
 
What  happens  next  in  the  conversation,  however,  is  rather  interesting.  Martin,  
the  more  powerful  person,  changes  to  independence  strategies,  asking  Ollie  how  
busy  he  is  and  making  it  clear  that  he  does  not  wish  to  impose  on  him.  In  fact,  he  
acts  so  reluctant  to  make  the  request  that  Ollie  practically  has  to  drag  it  out  of  
him  (‘but  Sam  called  you  said…  and  he?’).  This,  in  fact,  is  the  opposite  of  what  one  
might  expect  in  a  hierarchical  relationship.  Of  course,  this  shift  in  politeness  
strategies,  with  the  more  powerful  participant  using  independence  strategies  
and  the  less  powerful  one  showing  more  involvement  does  not  really  reflect  a  
shift  in  power.  Rather,  it  is  a  clever  strategy  Martin  has  used  to  make  it  more  
difficult  for  Ollie  to  refuse  the  request  by  putting  him  in  the  position  of  soliciting  
it.    
 

  74  
The  point  of  this  analysis  is  that,  even  though  our  expectations  about  face  
systems  form  the  background  to  how  we  communicate  about  relationships,  
people  often  strategically  confound  these  expectations  to  their  own  advantage.    
 
One  further  factor,  that  determines  which  strategy  a  person  will  use  to  
communicate  his  or  her  relationship  with  another  person  is  the  topic  of  the  
conversation  he  or  she  is  engaged  in.  In  cases  in  which  the  topic  of  the  
conversation  is  serious  or  potentially  embarrassing  for  either  party,  or  in  which  
the  weight  of  imposition  is  seen  to  be  great,  independence  strategies  will  be  
more  common,  whereas  in  situations  where  the  topic  is  less  serious,  the  outcome  
more  predictable  and  the  weight  of  imposition  seen  to  be  relatively  small,  
involvement  strategies  are  more  common.    
 
As  can  be  seen  in  the  example  above,  rather  than  as  simple  reflections  of  a  priori  
relationships  of  power  and  distance  or  the  ‘weightiness’  of  a  particular  topic,  face  
strategies  can  be  regarded  as  resources  that  people  use  to  negotiate  social  
distance,  enact  power  relationships,  and  sometimes  manipulate  others  into  
doing  things  which  they  may  not  normally  be  inclined  to  do.  A  person  might  use  
involvement  strategies  with  another  not  because  they  are  close,  but  because  he  
or  she  wants  to  create  or  strengthen  the  impression  that  there  is  a  power  
difference.  Similarly,  a  person  might  use  independence  strategies  not  to  create  a  
sense  of  distance  from  the  person  they  are  interacting  with,  but  rather  to  endow  
the  topic  under  discussion  with  a  certain  ‘weightiness’.  In  other  words,  face  
strategies  are  not  just  reflections  of  the  expectations  about  relationships  that  
people  bring  to  interactions  but  resources  they  make  use  of  to  manage  and  
sometimes  change  those  relationships  on  a  moment-­‐by-­‐moment  basis.  
 
  Look  deeper  into  this  topic  online  
 
 

Framing  and  contextualization  cues  

 
As  we  have  seen  above,  conversational  strategies  like  involvement  and  
independence  are  not  just  ways  that  we  communicate  and  manage  our  
relationships  with  other  people,  but  also  ways  that  we  communicate  something  
about  what  we  are  doing  (the  degree,  for  example,  to  which  we  think  we  are  
imposing  on  other  people).  We  also  have  other  ways  of  signaling  to  people  what  
we  think  we  are  doing  in  an  interaction,  whether  we,  for  example,  are  arguing,  
joking,  commiserating  or  making  small  talk.  Whenever  we  speak,  in  fact,  we  
communicate  not  just  the  message  contained  in  (or  implied  by)  our  words,  but  
also  information  about  what  we  think  we  are  doing  and,  therefore,  how  our  
words  should  be  interpreted.  We  call  the  signals  we  use  to  communicate  this  
information  contextualization  cues.  
 
In  Section  A6  we  said  that  there  are  basically  two  kinds  of  frames:  broader  
primary  frameworks  which  consist  of  the  relatively  stable  sets  of  expectations  we  
bring  to  particular  situations  (like  lectures  or  medical  consultations),  and  
smaller,  more  dynamic  interactive  frames,  which  consist  of  our  negotiated  ideas  

  75  
about  what  we  are  doing  moment  by  moment  in  a  conversation,  ideas  which  
often  change  rapidly  in  the  course  of  an  interaction.  Although  contextualization  
cues  are  often  important  in  signaling  primary  frameworks,  they  are  particularly  
important  in  the  role  they  play  in  helping  us  to  manage  and  negotiate  interactive  
frames.    
 
Sometimes  contextualization  cues  are  verbal,  that  is,  we  signal  what  we  are  
doing  through  our  choice  of  topic,  vocabulary,  grammar,  or  even  the  language  
that  we  use.  For  example,  in  her  analysis  of  the  talk  of  teachers  in  bilingual  
classrooms,  Angel  Lin  (1996),  has  pointed  out  that  when  English  teachers  in  
Hong  Kong  are  focusing  on  teaching,  they  tend  to  use  English,  but  when  they  are  
engaged  in  reprimanding  their  students,  they  tend  to  switch  to  Cantonese.  
Sometimes  these  verbal  cues  involve  adopting  a  particular  social  language  (see  
Sections  A4,  B4)  or  certain  genres  (see  Sections  A3,  B3)  associated  with  
particular  kinds  of  activities.  A  doctor,  for  example,  might  begin  a  consultation  
with  a  period  of  small  talk  in  which  the  language  might  be  extremely  informal  
and  the  topic  might  range  from  the  weather  to  a  local  sports  team  before  he  or  
she  ‘shifts  gears’  and  starts  ‘talking  like  a  doctor’.    
 
One  of  the  most  obvious  ways  we  signal  shifts  in  frames  verbally  is  through  what  
are  known  as  discourse  markers.  These  are  words  or  phrases  that  often  rather  
explicitly  mark  the  end  of  one  activity  and  the  beginning  of  another.  A  lecturer,  
for  example,  might  move  from  the  pre-­‐lecture  chatting  and  milling  around  frame  
to  the  formal  lecture  frame  with  words  like  ‘Okay,  let’s  get  started…’  Similarly,  
the  doctor  might  move  from  small  talk  to  the  more  formal  medical  examination  
by  saying  something  like  ‘So,  how  are  you  feeling?’  Discourse  markers  typically  
consist  of  words  like  okay,  so,  well,  and  anyway,  as  well  as  more  formal  
connectors  like  first,  next,  and  however.  It  is  important  to  remember  that  
discourse  markers  do  not  always  signal  a  shift  in  frame  –  sometimes  they  signal  
other  things  like  the  relationship  between  one  idea  and  another  (see  Section  B2).    
 
These  verbal  strategies  are  not  the  only  ways,  or  even  the  most  common  ways,  
people  signal  what  they  are  doing  when  they  talk.  Contextualization  cues  also  
include  non-­verbal  signals  delivered  through  things  like  gestures,  facial  
expressions,  gaze,  our  use  of  space,  and  paralinguistic  signals  delivered  through  
alterations  in  the  pitch,  speed,  rhythm  or  intonation  of  our  voices.  For  this  
reason,  people  who  study  frames  and  contextualization  cues  often  pay  a  lot  of  
attention  to  marking  things  like  stress,  intonation  and  pausing  and  even  facial  
expressions,  gestures  and  other  movements  when  they  produce  transcripts  of  
the  conversations  they  are  studying.    
 
These  non-­‐verbal  and  paralinguistic  contextualization  cues  are  sometimes  much  
more  subtle  than  verbal  strategies  and  so  more  easily  misunderstood.  The  way  
they  are  used  and  interpreted  might  also  vary  considerably  from  group  to  group  
or  even  person  to  person.  In  one  of  his  most  famous  studies,  Gumperz  (1982a:  
173-­‐174)  found  a  mismatch  between  the  ways  South  Asian  servers  in  a  staff  
canteen  in  a  British  airport  used  intonation  as  a  contextualization  cue  and  the  
ways  their  British  customers  interpreted  them.  The  South  Asian  servers,  for  
example,  used  falling  intonation  when  asking  customers  if  they  wanted  gravy  on  

  76  
their  meat  (consistent  with  the  conventions  of  their  variety  of  English),  but  the  
British  customers,  expecting  the  rising  intonation  they  associated  with  a  polite  
offer,  interpreted  the  servers’  behavior  as  rude.  What  this  example  tells  us  is  that  
contextualization  cues  do  not  in  themselves  contain  information  about  what  we  
think  we  are  doing  –  rather,  they  activate  culturally  conditioned  assumptions  
about  context,  interactional  goals  and  interpersonal  relationships  that  might  be  
different  for  different  people.    
 
As  we  said  above,  interactive  frames  are  not  static,  but  can  change  rapidly  in  the  
course  of  an  interaction.  They  are  also,  as  their  name  implies,  interactive  –  that  is,  
they  are  always  a  matter  of  negotiation  between  participants  in  the  conversation,  
and  the  way  they  are  used  and  interpreted  often  has  a  great  deal  to  do  with  
things  that  happened  previously  in  the  conversation  and  with  the  history  of  the  
relationship  between  those  involved.  In  other  words,  just  as  face  strategies  of  
involvement  and  independence,  while  primarily  providing  information  about  
relationships,  also  give  clues  as  to  what  we  think  we  are  doing  and  our  attitude  
towards  it,  framing  strategies,  while  primarily  signaling  what  we  think  we  are  
doing,  also  play  an  important  role  in  managing  relationships.    
 
In  an  article  called  ‘Talking  the  Dog’,  in  which  she  examines  how  people  use  pets  
to  frame  and  reframe  their  utterances  in  interaction,  Deborah  Tannen  (2004)  
gives  the  following  example  of  a  conversation  between  a  woman,  Clara,  and  her  
husband,  Neil,  in  the  presence  of  their  dog,  Rickie.    
 
Clara:  You  leave  the  door  open  for  any  reason?  
((short  pause,  sound  of  door  shutting))  
—>  <babytalk>  Rickie,  
—>  he’s  helpin  burglars  come  in,  
—>  and  you  have  to  defend  us  Rick.>      
Tannen  2004:413)  
 
In  this  example,  Clara  shifts  frames  from  talking  to  her  husband  to  talking  to  the  
dog  by  altering  her  voice  quality  (adopting  the  high  pitched  and  playful  tone  of  
‘baby  talk’).  In  a  sense,  though,  she  is  still  talking  to  her  husband,  communicating  
to  him  ‘through’  the  dog  the  potential  seriousness  of  leaving  the  door  open.  By  
addressing  her  remarks  to  the  dog,  however,  and  by  adopting  a  different  tone  of  
voice,  she  shifts  the  frame  from  scolding  to  playing,  allowing  her  to  get  the  
message  across  without  threatening  her  husband’s  face.    
 
Sometimes  participants  in  an  interaction  will  experience  disagreement  regarding  
‘what’s  going  on’.  The  way  one  person  frames  the  conversation,  for  example,  may  
be  at  odds  with  the  other  person’s  wishes,  expectations  or  interpretation  of  the  
situation.  In  some  cases,  they  may  simply  accept  the  framing  that  has  been  
imposed  by  the  other  person,  or  they  may  contest  or  resist  it  by  either  
attempting  to  reframe  the  conversation  using  their  own  contextualization  cues  
or  by  breaking  the  frame  altogether  and  engaging  in  a  ‘meta-­‐conversation’  about  
‘what’s  going  on’.  
 

  77  
The  film  When  Harry  Met  Sally  (1988  Castle  Rock  Pictures)  contains  a  number  of  
good  examples  of  characters  competitively  negotiating  frames  in  interaction.  In  
the  following  example,  Harry,  who  is  going  out  with  Sally’s  best  friend,  tells  Sally  
that  he  thinks  she  is  attractive,  and  what  ensures  is  a  negotiation  about  what  
such  a  statement  means  based  on  what  he  was  ‘doing’  when  he  said  it.    
 
HARRY:  You’re  a  very  attractive  person.    
SALLY:  Oh,  thank  you.    
HARRY:  Amanda  never  said  you  were  so  attractive.    
SALLY:  Maybe  she  doesn’t  think  I’m  attractive.    
HARRY:  It’s  not  a  matter  of  opinion.  Empirically  you  are  attractive.    
SALLY:  Harry,  Amanda  is  my  friend.    
HARRY:  So?    
SALLY:  So  you’re  going  with  her.    
HARRY:  So?    
SALLY:  So  you’re  coming  on  to  me.    
HARRY:  No  I  wasn’t.    
HARRY  (continuing):  What?  Can’t  a  man  say  a  woman  is  attractive  
without  it  being  a  come-­‐on?    
HARRY  (continuing):  All  right.  Let’s  just  say  for  the  sake  of  argument  it  
was  a  come-­‐on.  Okay.  What  do  you  want  me  to  do?  I  take  it  back.  All  right,  
I  take  it  back.    
SALLY:  You  can’t  take  it  back.    
HARRY:  Why  not?    
SALLY:  It’s  already  out  there.    
An  awkward  pause  
HARRY:  Ohm  jeez.  What  are  we  supposed  to  do  now?  Call  the  cops?  It’s  
already  out  there.    
SALLY:  Just  let  it  lie,  okay?    
HARRY:  Right,  right.  Let  it  lie.  That’s  my  policy.  Let  it  lie…  So,  you  want  to  
spend  the  night  in  the  motel?    
HARRY  (continuing):  See  what  I  did?  I  didn’t  let  it  lie.    
SALLY:  Harry  –-­‐-­‐  
HARRY:  I  said  I  would  and  then  I  didn’t  -­‐-­‐  -­‐-­‐  
SALLY:  Harry  -­‐-­‐-­‐-­‐  
HARRY:  I  went  the  other  way  -­‐-­‐-­‐-­‐  
SALLY:  Harry  -­‐-­‐-­‐-­‐  
HARRY:  Yes?    
SALLY:  We  are  just  going  to  be  friends,  okay?    
HARRY:  Yeah.  Great.  Friends.  Best  thing.    
 
In  this  example,  Harry  tries  to  frame  his  initial  compliment  as  an  ‘objective  
observation’  using  formal  language  like  ‘empirically’.  Sally,  however,  labels  what  
he  is  doing  as  a  ‘come  on’,  a  label  which  he  first  resists  with  the  question,  ‘Can’t  a  
man  say  a  woman  is  attractive  without  it  being  a  come-­‐on?’,  framing  the  
accusation  as  unreasonable  and  possibly  sexist.  He  then  half  accepts  her  framing  
and  offers  to  ‘take  it  back’.  This  acceptance  is  only  partial  because  he  frames  it  as  
‘hypothetical’  (‘Let’s  just  say  for  the  sake  of  argument  it  was  a  come-­‐on…’).  Sally,  
however,  does  not  accept  his  retraction,  framing  a  ‘come  on’  as  an  irreparable  

  78  
breech  in  decorum,  which  Harry  responds  to  by  again  shifting  frames  from  
conciliation  to  mocking  (‘What  are  we  supposed  to  do  now?  Call  the  cops?’).  
What  happens  after  this,  however,  is  particularity  interesting.  After  agreeing  to  
‘let  it  lie’,  that  is,  abandon  this  particular  negotiation  about  framing,  Harry  then  
issues  what  is  unambiguously  a  ‘come-­‐on’,  and  then  deflects  her  objections  by  
again  engaging  in  meta-­‐conversation  about  his  own  framing  (‘See  what  I  did?  I  
didn’t  let  it  lie…  I  said  I  would  and  then  I  didn’t…  I  went  the  other  way…’).    
 
Part  of  the  humor  in  this  scene  lies  in  the  fact  that  it  foregrounds  the  process  of  
framing  itself,  a  process  which  is  usually  left  tacit  in  conversations.  It  also  shows  
how  complex  and  contentious  negotiations  of  framing  can  be,  with  parties  not  
only  shifting  frames,  breaking  frames,  and  attempting  to  reframe  the  utterances  
of  themselves  and  others,  but  also  superimposing  frames  on  top  of  other  frames  
in  order  to  create  strategic  ambiguity  (as  when  Harry  imposes  a  ‘hypotheictal’  
frame  onto  his  admission  of  guilt).    

  79  
B7  THE  SPEAKING  MODEL  
 

Speech  acts,  speech  events  and  speech  situations  

 
The  main  unit  of  analysis  for  the  ethnography  of  speaking  is  the  speech  event,  
which  can  be  defined  as  a  communicative  activity  that  has  a  clear  beginning  and  
a  clear  ending  and  in  which  people’s  shared  understandings  of  the  relevance  of  
various  contextual  features  remain  fairly  constant  throughout  the  event.  
Examples  of  speech  events  are  such  things  as  religious  ceremonies,  lessons,  
debates,  and  conversations.  Speech  events  occur  within  broader  speech  
situations,  and  are  made  up  of  smaller  speech  acts  of  the  type  we  have  already  
discussed  (including  such  things  as  greeting,  questioning,  promising  and  
insulting,  see  Section  B5).  For  example,  a  university  lecture  can  be  considered  a  
speech  event  which  occurs  within  the  speech  situation  of  a  school  day  and  is  made  
up  of  smaller  speech  acts  like  asking  and  answering  questions,  giving  
explanations  and  illustrations  of  certain  concepts,  and  even  joking  or  
threatening.  Similarly,  the  speech  event  of  a  conversation  may  occur  within  the  
larger  speech  situation  of  a  party  and  may  include  smaller  speech  acts  like  joking.  
Notice  that  the  same  speech  act,  joking  for  example,  can  take  place  in  many  
different  kinds  of  speech  events,  and  that  different  speech  events,  conversations  
for  example,  can  occur  in  many  different  kinds  of  speech  situations.    
 
What  distinguishes  a  speech  event  from  a  speech  situation  is  not  just  its  size  and  
the  fact  that  speech  events  tend  to  have  clearer  boundaries.  The  main  distinction  
is  coherence  (see  Section  B2):  participants  tend  to  approach  speech  events  with  
consistent  sets  of  expectations  that  remain  the  same  throughout  the  speech  
event,  whereas  participants’  expectations  about  the  relevant  features  of  context  
may  undergo  dramatic  changes  throughout  a  speech  situation:  students  eating  
lunch  at  the  university  canteen  during  a  school  day,  for  example,  are  likely  to  pay  
attention  to  different  sorts  of  things  than  they  do  in  a  lecture  during  the  same  
day.  The  way  to  distinguish  between  a  speech  situation  and  a  speech  event,  then,  
is  to  ask  if  the  same  rules  of  SPEAKING  apply  throughout  the  phenomenon.  If  so  
it  can  be  regarded  as  a  speech  event.    
 

SPEAKING  
 
One  potentially  confusing  aspect  of  the  ethnography  of  speaking  is  that  it  does  
not,  as  its  name  implies,  focus  so  much  on  rules  and  expectations  about  speaking  
so  much  as  rules  and  expectations  about  the  circumstances  in  which  certain  
kinds  of  speaking  takes  place  (or,  does  not  take  place).  In  fact,  one  of  the  most  
famous  studies  using  this  approach,  Keith  Basso’s  examination  of  silence  among  
the  Western  Apache  in  the  United  States,  explored  the  conditions  under  which,  
for  members  of  this  speech  community,  not  speaking  is  considered  the  most  
appropriate  behavior.    
 

  80  
Ron  and  Suzanne  Scollon  have  used  the  term  ‘the  Grammar  of  Context’  to  refer  to  
a  model  very  much  like  Hymes’s  speaking  model  (Scollon,  Scollon  and  Jones  
2011).  Their  reasons  for  comparing  the  rules  and  expectations  associated  with  
context  to  the  kinds  of  rules  and  expectations  associated  with  the  grammar  of  a  
language  are  twofold:  first,  to  highlight  that  the  same  difference  between  
competence  and  performance  which  we  see  in  grammar  also  occurs  in  rules  and  
expectations  associated  with  context:  not  everyone  performs  in  particular  
speech  events  exactly  in  accordance  with  how  people  in  their  speech  community  
(including  themselves)  think  they  should;  and  second,  to  introduce  the  notion  of  
markedness  into  the  analysis  of  context.  
 
The  idea  of  'unmarked’  (the  usual  or  normal  way  of  saying  or  doing  something)  
vs.  'marked'  (an  unusual  or  deviant  way  of  saying  or  doing  something)  was  
introduced  into  structural  linguistics  by  the  Prague  School  of  linguists,  which  
included  such  figures  as  Roman  Jakobson  (see  Jakobson  1990:  134-­‐40).  Although  
the  concept  is  quite  complex,  the  general  idea  is  that  when  people  deviate  from  
the  default  or  expected  way  of  using  language,  the  result  is  often  the  expression  
of  some  special,  more  precise  or  additional  meaning.  This  is  an  idea  we  have  
already  encountered  in  our  discussion  of  pragmatics  and  the  cooperative  
principle.  When  it  is  applied  to  ‘context’,  it  reminds  us  that  communicative  
competence  does  not  refer  to  a  set  of  ‘rules’  that  must  be  followed,  but  rather  to  a  
set  of  expectations  that  experienced  speakers  can  sometimes  manipulate  in  
order  to  strategically  manage  the  meanings  of  speech  acts,  the  relationships  
among  participants,  or  the  outcomes  of  the  speech  event.    
 
The  components  of  the  SPEAKING  model  devised  by  Hymes,  therefore,  are  not  
meant  to  provide  an  objective  list  of  those  elements  of  context  which  need  to  be  
taken  into  account  by  the  analyst,  but  rather  a  set  of  guidelines  an  analyst  can  
use  in  attempting  to  find  out  what  aspects  of  context  are  important  and  relevant  
from  the  point  of  view  of  participants.  In  other  words,  in  any  given  speech  event,  
different  elements  will  be  afforded  different  weight  by  participants,  and  some  
might  be  regarded  as  totally  unimportant.    
 
The  first  component  in  the  model  is  setting,  which  refers  to  the  time  and  place  of  
the  speech  event  as  well  as  any  other  physical  circumstances.  Along  with  the  
physical  aspects  of  setting,  Hymes  included  what  he  called  the  ‘psychological  
setting’  or  the  ‘cultural  definition’  of  a  scene.  The  unmarked  setting  for  a  
particular  speech  event,  for  example,  might  be  in  a  church.  A  church  has  
particular  physical  characteristics,  but  it  is  also  likely  to  have  certain  associations  
for  people  in  a  particular  culture  so  that  when  they  enter  a  church  they  are  
predisposed  to  speak  or  behave  in  certain  ways.  Thus,  the  component  of  setting  
can  have  an  effect  on  other  components  like  key  and  instrumentalities  (see  
below).    
 
The  second  component  in  the  SPEAKING  model  is  participants.  Most  of  the  
approaches  to  spoken  discourse  we  have  looked  at  so  far,  including  conversation  
analysis  and  pragmatics,  begin  with  the  assumption  of  an  essentially  didactic  
model  of  communication  in  which  the  participants  are  the  speaker  and  the  
hearer.  Ethnographic  work,  however,  indicates  that  many  if  not  most  speech  

  81  
events  involve  many  kinds  of  participants,  not  just  speakers  and  hearers,  but  also  
participants  like  audiences  and  bystanders.  Furthermore,  groups  differ  in  their  
ideas  of  which  participants  in  speech  events  are  considered  legitimate  or  
relevant  (for  example,  maids,  pets,  supernatural  beings).  Besides  identifying  the  
relevant  participants,  the  different  kinds  of  identities,  roles  and  rights  different  
participants  have  are  also  important.  These  aspects,  of  course,  will  depend  on  
things  like  the  genre  of  the  speech  event,  and  may  change  over  the  course  of  the  
speech  event  in  accordance  with  a  particular  act  sequence  (see  below).    
 
The  third  component  of  the  model  is  ends,  which  refers  to  the  purpose,  goals  and  
outcomes  of  the  event,  which,  of  course,  may  be  different  for  different  
participants  (the  goals  of  a  teacher,  for  example,  are  not  always  the  same  as  the  
goals  of  his  or  her  students),  and  the  fourth  component  is  act  sequence,  the  form  
the  event  takes  as  it  unfolds,  including  the  order  of  different  speech  acts  and  
other  behaviors.  Both  of  these  components  are  intimately  connected  not  just  
with  expectations  about  participant  roles,  but  also  with  the  genre  of  the  speech  
event.  
 
The  fifth  component  in  the  model  is  key,  by  which  is  meant  the  overall  ‘tone’  or  
mood  of  the  speech  event.  Key  is  important  because  it  provides  an  attitudinal  
context  for  speech  acts,  sometimes  dramatically  altering  their  meaning  (as  with  
sarcasm).  At  the  same  time,  key  is  often  signaled  in  very  subtle  ways  that  are  
sometimes  outside  the  purview  of  most  linguistic  analysis.  We  have  already  
explored  some  of  these  signals  in  our  discussion  of  contextualization  cues  in  
Section  B6.    
 
The  sixth  component  is  instrumentalities,  meaning  the  ‘message  form’  –  the  
means  or  media  through  which  meaning  is  made.  Speech,  for  example,  might  be  
spoken,  sung,  chanted  or  shouted,  and  it  may  be  amplified  through  microphones,  
broadcast  through  electronic  media,  or  written  down  and  somehow  passed  back  
and  forth  between  participants.  Typically,  speech  events  include  complex  
combinations  of  instrumentalities  that  interact  with  one  another  and  with  the  
other  components  in  the  model.  In  the  next  strand  on  mediated  discourse  
analysis  we  will  explore  in  more  detail  the  effect  different  instrumentalities  can  
have  on  speech  acts  and  speech  events.    
 
The  seventh  component  is  norms,  which  can  be  divided  into  norms  of  interaction  
and  norms  of  interpretation.  These  are  the  common  sets  of  understandings  that  
participants  bring  to  events  about  what  is  appropriate  behavior  and  how  
different  actions  and  utterances  ought  to  be  understood.  The  important  thing  
about  norms  is  that  they  may  be  different  for  different  participants  (a  waiter  vs.  
a  customer,  for  example)  and  that  the  ‘setting  of  norms’  is  often  a  matter  of  
power  and  ideology  (see  Section  A4).    
 
Finally,  the  eighth  component  is  genre,  or  the  ‘type’  of  speech  event.  We  have  
already  dealt  at  length  with  the  concept  of  genre  (see  Section  B3),  and,  although  
Hymes’s  understanding  of  genre  is  slightly  different  from  that  of  genre  analysts  
like  Swales  and  Bhatia,  much  of  what  was  said  before  about  community  
expectations,  form,  and  communicative  purpose  applies  here.  The  most  

  82  
important  aspect  of  this  component  is  the  notion  that  certain  speech  events  are  
recognizable  by  members  of  a  speech  community  as  being  of  a  certain  type,  and  
as  soon  as  they  are  ‘labeled’  as  such,  many  of  the  other  components  of  the  model  
like  ends,  act  sequence,  participant  roles  and  key  are  taken  as  givens.    
 
It  should  be  clear  from  this  brief  rundown  of  the  components  of  the  SPEAKING  
model  that  none  of  them  can  really  be  considered  alone:  each  component  
interacts  with  other  components  in  multiple  ways.  The  most  important  job  of  an  
analyst  using  this  model,  then,  is  not  just  to  determine  the  kinds  of  knowledge  
about  the  different  components  members  of  speech  communities  need  to  
successfully  participate  in  a  given  speech  event,  but  also  to  determine  how  the  
different  components  are  linked  together  in  particular  ways  for  different  speech  
events.  For  it  is  in  these  linkages,  the  ways,  for  example,  different  kinds  of  
participants  are  associated  with  different  genres,  or  different  settings  are  seen  as  
suitable  for  different  purposes,  or  different  forms  of  discourse  or  media  are  
associated  with  different  keys,  that  the  analyst  can  begin  to  get  an  understanding  
of  deeper  cultural  assumptions  about  people,  places,  values,  power  and  
communication  itself  that  exist  in  a  particular  speech  community.    

  83  
B8  MEDIATION  
 
Cultural  tools  
 
The  starting  point  for  mediated  discourse  analysis  is  the  concept  of  mediation.  
The  traditional  definition  of  mediation  is  the  passing  of  a  message  through  some  
medium,  which  is  placed,  between  two  or  more  people  who  are  communicating.  
When  we  think  of  media,  we  usually  think  of  things  like  newspapers,  television  
and  computers.  Lots  of  people  have  pointed  out  that  when  messages  pass  
through  media,  they  change  fundamentally.  Different  kinds  of  media  favor  
different  kinds  of  meanings.  The  kinds  of  meaning  people  can  make  in  a  
newspaper  article,  for  example,  are  different  from  those  they  can  make  in  a  
television  broadcast.  This  fact  led  the  media  scholar  Marshall  McLuhan  
(1964/2001)  to  make  the  famous  pronouncement:  ‘the  medium  is  the  message.’  
 
Mediated  discourse  analysis  is  also  interested  in  how  different  media  like  
televisions  and  computers  affect  the  way  people  use  discourse,  but  it  takes  a  
rather  broader  view  of  media  and  mediation.  This  view  comes  from  the  work  of  
the  Russian  psychologist  Lev  Vygotsky.  Vygotsky  (1981)  had  the  idea  that  all  
actions  that  people  take  in  the  world  are  somehow  mediated  through  what  he  
called  cultural  tools.  Cultural  tools  can  include  technological  tools  like  televisions,  
computers,  and  megaphones,  but  also  include  more  abstract  tools  like  languages,  
counting  systems,  diagrams  and  mental  schema.  Anything  an  individual  uses  to  
take  action  in  the  world  can  be  considered  a  cultural  tool.    
 
The  important  thing  about  cultural  tools  is  that  they  make  it  easier  to  perform  
some  kinds  of  actions  and  communicate  some  kinds  of  meanings,  and  more  
difficult  to  take  other  kinds  of  actions  and  communicate  others  kinds  of  
meanings.  In  other  words,  all  tools  come  with  certain  affordances  and  constraints.  
Writing  a  letter  or  an  email,  for  example,  allows  us  to  do  things  that  we  cannot  do  
when  we  are  producing  spoken  discourse  in  the  context  of  a  conversation,  things  
such  as  going  back  and  deleting  or  revising  things  we  have  written  before.  But  it  
is  more  difficult  to  do  other  things  like  gauge  the  reaction  of  other  people  to  what  
we  are  writing  as  we  are  writing  it  (as  we  can  do  with  spoken  language  in  face-­‐
to-­‐face  conversations).  A  microphone  makes  it  easier  to  talk  to  a  large  group  of  
people,  but  more  difficult  to  say  something  private  to  a  person  standing  next  to  
you  (as  some  politicians  have  rather  painfully  learned).  Most  instant  messaging  
programs  make  it  easy  to  have  a  real  time  conversation,  but  more  difficult  to  
interrupt  one’s  conversational  partner  in  the  middle  of  an  utterance  the  way  we  
can  do  in  face-­‐to-­‐face  conversations.    
 
What  this  idea  of  affordances  and  constraints  means  for  discourse  analysis  is  that  
the  kinds  of  discourse  and  other  tools  we  have  available  to  us  affect  the  kinds  of  
actions  that  we  can  take.  In  many  situations,  for  example,  such  as  ordering  lunch  
in  a  restaurant  in  a  remote  area  of  China,  access  to  the  ‘tool’  of  the  Chinese  
language  will  allow  us  to  do  different  things  than  we  couéld  do  with  the  tool  of  

  84  
English.  Different  modes  and  media  also  allow  us  to  do  different  things:  we  can  
perform  different  actions  with  pictures  and  gestures  than  we  can  with  words  
(see  Section  A9),  and  we  can  do  different  things  with  mobile  telephones  than  we  
can  with  landlines.  Even  genres  and  social  languages  have  affordances  and  
constraints.  A  résumé,  for  example,  might  be  effective  for  getting  a  job,  but  less  
effective  for  getting  a  date  with  a  man  or  a  woman  whom  we  fancy.  Even  in  the  
context  of  getting  a  job,  the  genre  may  make  it  easy  to  communicate  things  about  
formal  education,  credentials  and  work  experience,  but  more  difficult  to  
communicate  things  about  more  informal  learning  and  non-­‐work  experience.    
 
And  so,  when  we  perform  mediated  discourse  analysis,  we  first  identify  the  
actions  that  are  important  to  a  particular  social  actor  in  a  particular  situation  and  
then  attempt  to  determine  how  the  cultural  tools  (such  as  languages  and  other  
modes,  media,  genres  and  social  languages)  contribute  to  making  these  actions  
possible  and  making  other  kinds  of  actions  impossible  or  more  difficult.  Of  
course,  we  also  have  to  recognize  that  many  of  the  cultural  tools  we  use  to  
perform  actions  are  not  discursive.  If  you  want  to  put  together  a  piece  of  
furniture  you  have  bought  at  IKEA,  while  some  discourse  such  as  the  instructions  
for  assembly  might  be  very  important,  if  you  lack  access  to  technological  tools  
like  a  hammer  and  a  screwdriver,  no  amount  of  discourse  can  make  it  possible  
for  you  to  perform  the  actions  you  need  to  perform.    
 
This  simple  idea  that  having  access  to  different  kinds  of  tools  makes  it  easier  or  
more  difficult  to  perform  social  actions  has  important  implications.  Earlier,  for  
example,  we  discussed  how  people  sometimes  try  to  use  discourse  to  advance  
certain  ideologies  or  versions  of  reality  in  order  to  try  to  affect  what  people  
think.  Mediated  discourse  analysis  highlights  the  fact  that  discourse  does  not  just  
have  a  role  in  affecting  what  we  think,  but  also,  in  a  very  practical  way,  in  
affecting  what  we  can  do.  If  we  do  not  have  the  proper  tools  available  to  us,  there  
are  certain  things  that  we  simply  cannot  do.  And  so  people  who  have  access  to  
particular  tools  (such  as  languages,  genres,  electronic  media)  can  often  exert  
certain  power  over  people  who  do  not  in  very  concrete  ways.  If  we  also  consider  
that  our  social  identities  are  created  through  the  actions  that  we  can  take,  we  
come  to  the  conclusion  that  the  tools  we  have  available  to  us  and  how  we  use  
them  help  to  determine  not  just  what  we  can  do,  but  who  we  can  be.    
 
At  the  same  time,  human  beings  are  extremely  creative  in  their  use  of  tools.  If  I  
do  not  have  a  screwdriver  to  put  together  my  IKEA  table,  I  might  try  using  a  
butter  knife.  If  the  genre  of  the  résumé  does  not  allow  me  to  showcase  my  
talents,  I  might  try  to  bend  that  genre  or  blend  it  with  another  genre.  In  fact,  one  
important  focus  of  mediated  discourse  analysis  is  in  exploring  the  tension  that  
exists  between  the  affordances  and  constraints  built  in  to  different  cultural  tools  
and  the  ways  people  creatively  appropriate  and  adapt  those  tools  into  different  
situations  to  achieve  different  goals.    
 
One  example  of  the  way  technological  tools  can  affect  the  kinds  of  things  we  can  
do  when  we  communicate,  and  the  creative  ways  people  adapt  to  these  
affordances  and  constraints,  can  be  seen  in  the  way  people  use  online  personal  
ads  and  dating  sites.  In  Section  B3  we  considered  print  based  personal  ads  as  a  

  85  
kind  of  genre  and  discussed  how,  by  mastering  the  structure  of  the  genre,  people  
claim  membership  in  different  discourse  communities.  Nowadays,  however,  
most  people  rely  on  electronic  personal  ads  (or  ‘profiles’)  rather  than  print  based  
ones.  These  new  online  ‘profiles  introduce  a  new  set  of  affordances  and  
constraints  for  users.  In  older,  print-­‐based  ads,  for  example,  users  had  much  
more  control  over  self-­‐presentation,  which  was  performed  entirely  through  text.  
While  such  ads  often  invited  the  later  exchange  of  pictures,  users  had  a  way  of  
vetting  potential  prospects  before  sending  a  picture,  thereby  maintaining  some  
degree  of  privacy.  On  online  dating  sites,  where  the  inclusion  of  a  picture  is  
expected  up  front,  users  relinquish  their  control  over  physical  self-­‐description  
and  risk  disclosing  their  identities  to  unintended  parties  (such  as  employers,  co-­‐
workers,  or  family  members).  In  other  words,  the  inclusion  of  pictures  in  profiles  
introduces  both  affordances  (it  allows  for  a  fuller  an  more  accurate  presentation  
of  self)  and  constraints  (it  makes  it  more  difficult  for  the  owners  of  profiles  to  
control  information  about  themselves).  In  my  own  study  of  online  gay  dating  
sites  (Jones  2009),  I  discovered  how  users  creatively  deal  with  this  challenge  by  
posting  pictures  designed  to  strategically  obscure  their  identities  or  by  posting  
pictures  of  celebrities,  cartoon  characters  or  even  inanimate  objects  which  reveal  
something  about  their  personalities  or  interests  without  giving  away  their  
identities.  There  are  two  important  points  to  take  from  this  example.  The  first  is  
that  every  new  affordance  in  a  cultural  tool  also  brings  along  with  it  some  new  
constraint,  and  second,  experienced  users  of  tools,  like  experienced  users  of  
genres  (see  Section  B3),  often  find  ways  to  creatively  adapt  to  the  affordances  
and  constraints  of  the  tools  they  are  using.    
 

Context  revisited:  Sites  of  engagement  

 
Mediated  discourse  analysts  call  the  situations  into  which  tools  are  appropriated  
sites  of  engagement.  Sites  of  engagement  are  moments  when  different  kinds  of  
social  actors,  different  kinds  of  cultural  tools  and  different  kinds  of  social  
relationships  come  together  to  make  certain  actions  possible.    
 
In  the  previous  section  we  explored  the  importance  of  ‘context’  in  the  production  
of  meaning.  The  problem  with  the  idea  of  ‘context’  from  the  point  of  view  of  
mediated  discourse  analysis  is  that  it  takes  ‘texts’  as  its  reference  point.  The  idea  
of  sites  of  engagement  takes  social  actions  as  its  reference  point.  Instead  of  
making  an  artificial  distinction  between  discourse  and  everything  else,  it  
considers  all  cultural  tools  (texts,  furniture,  objects,  machines)  that  are  available  
to  social  actors  at  a  particular  time  in  a  particular  place  and  explores  how  they  
contribute  to  making  possible  certain  kinds  of  actions.    
 
Ron  and  Suzanne  Wong  Scollon  (2004)  say  that  all  social  actions  occur  at  the  
nexus  or  ‘coming  together’  of  three  crucial  elements:  1)  the  discursive  resources  
and  other  cultural  tools  that  people  have  available  for  action  (which  they  call  
‘discourses  in  place’),  2)  the  social  relationships  among  the  people  involved  
(which  they  call  the  ‘interaction  order’),  and  the  knowledge,  abilities  and  
experiences  of  the  individual  social  actor  (which  they  call  the  ‘historical  body’).    
 

  86  
To  illustrate  how  these  three  elements  come  together  to  form  the  site  of  
engagement  of  a  social  action  we  can  take  the  example  of  crossing  a  busy  city  
street  (see  figure  B8.1).  There  is  normally  a  lot  of  discourse  available  to  people  in  
this  situation.  There  are  things  like  street  signs,  traffic  signals  and  zebra  stripes  
painted  on  the  pavement  to  assist  pedestrians  in  crossing  the  street;  there  is  also  
a  lot  of  discourse  such  as  shop  signs  and  advertisements  that  might  actually  
interfere  with  the  action  of  successfully  crossing  the  street.  And  so  one  of  the  
most  important  things  for  people  engaged  in  performing  this  action  is  
determining  which  discourse  to  attend  to  and  which  discourse  to  ignore.    
 
The  second  element  is  the  interaction  order,  the  relationships  people  have  with  
the  people  with  whom  they  are  crossing  the  street.  If  we  are  crossing  the  street  
alone,  for  example,  we  might  take  extra  care  in  checking  for  on-­‐coming  traffic,  
whereas  if  we  are  part  of  a  large  crowd  of  people,  we  might  pay  more  attention  
to  the  actions  of  other  pedestrians  to  decide  when  to  cross  simply  by  following  
them.  If  we  are  with  someone  else,  we  might  find  we  need  to  distribute  our  
attention  between  the  action  of  crossing  the  street  and  some  other  action  such  as  
carrying  on  a  conversation  or  making  sure  our  companion  (if  they  are,  for  
instance,  a  small  child)  gets  across  the  street  safely.  
 
Finally,  the  action  of  crossing  the  street  depends  on  people’s  knowledge  and  
experience  of  crossing  city  streets,  the  habits  and  mental  models  they  have  built  
up  around  this  social  practice,  which  the  Scollons  refer  to  as  the  ‘historical  body’.  
Most  of  the  time  we  do  things  like  crossing  the  street  in  a  rather  automatic  way.  
When  we  find  ourselves  in  unfamiliar  situations,  however,  our  habitual  ways  of  
doing  things  sometimes  do  not  work  so  well.  Most  of  us  have  found  ourselves  
having  some  difficulty  crossing  streets  in  cities  where  conventions  about  which  
discourses  in  place  pedestrians  ought  to  attend  to  or  what  kind  of  behavior  is  
expected  from  drivers  are  different  from  those  in  the  city  in  which  we  live.    

 
Figure  B8.1  Crossing  the  street    
 

  87  
And  so  the  main  differences  between  the  ideas  of  ‘site  of  engagement’  and  
‘context’  are,  first,  that  while  ‘contexts’  take  ‘texts’  as  their  points  of  reference,  
sites  of  engagement  take  actions  as  their  points  of  reference,  and  second,  that  
while  contexts  are  usually  considered  to  be  external  to  the  social  actor,  sites  of  
engagement  are  a  matter  of  the  interaction  among  the  texts  and  other  cultural  
tools  available  in  a  social  situation,  the  people  that  are  present,  and  the  habits,  
expectations  and  goals  of  individuals.    
 
 
    Find  additional  examples  online  

  88  
  

B9  MODES,  MEANING  AND  ACTION  

 
Multimodal  discourse  analysis,  the  analysis  of  how  multiple  modes  of  
communication  interact  when  we  communicate,  tends  to  be  divided  into  two  
broad  approaches,  one  which  focuses  on  ‘texts’  (like  magazines  and  web  pages)  
and  the  other  which  focuses  on  ‘real  time’  interactions.  One  important  concept  
that  is  common  to  both  of  these  approaches  is  the  idea  that  different  modes  have  
different  affordances  and  constraints.  We  introduced  the  idea  of  affordances  and  
constraints  in  the  last  section  when  we  said  that  different  ‘cultural  tools’  make  
some  actions  easier  and  other  actions  more  difficult.    
 
When  we  say  that  different  modes  have  different  affordances  and  constraints,  we  
mean  both  that  they  have  different  sets  of  ‘meaning  potential’  and  that  they  allow  
us  to  take  different  kinds  of  actions.  For  example,  in  written  text  and  spoken  
language  we  must  present  information  in  a  sequential  way  governed  by  the  logic  
of  time.  Thus,  an  author  or  speaker  can  manipulate  the  order  and  speed  at  which  
information  is  given  out,  perhaps  withholding  certain  facts  until  later  in  the  text  
or  conversation  for  strategic  purposes.  Images,  on  the  other  hand,  are  governed  
by  the  logic  of  space.  The  producer  of  the  image  presents  all  of  the  elements  in  
the  image  all  at  once  and  has  limited  control  over  the  order  in  which  viewers  
look  at  those  elements.  Similarly,  images  allow  for  the  communication  of  very  
fine  gradations  of  meaning  when  it  comes  to  things  like  shape  and  color  –  the  
exact  shade  of  pink  in  someone’s  cheeks,  for  example  –  whereas  language  forces  
us  to  represent  things  in  terms  of  types  –  the  word  ‘pink’  for  example,  cannot  
represent  an  exact  color,  but  only  a  range  of  colors  within  a  particular  class.    
The  fact  that  different  modes  make  different  kinds  of  meanings  more  possible  
and  others  less  possible  is  one  of  the  reasons  why  people  strategically  mix  
different  modes  when  they  are  communicating,  so  that  the  constraints  of  one  
mode  are  balanced  out  by  the  affordances  of  others.  While  there  are  some  things  
that  ‘just  cannot  be  expressed  in  words,’  it  might  be  possible  to  express  them  
with  a  carefully  timed  facial  expression  or  a  carefully  placed  image.    
 

Communicative  functions  of  modes  

 
In  Section  A4  we  introduced  Halliday’s  idea  that  language  has  three  basic  
functions:  it  is  used  to  represent  our  experience  of  the  world;  it  is  used  to  
communicate  something  about  the  relationship  between  us  and  the  people  with  
whom  we  are  communicating;  and  it  is  used  to  organize  ideas,  representations  
and  other  kinds  of  information  in  ways  that  people  can  make  sense  of.    We  called  
these  three  functions  the  ideational  function,  the  interpersonal  function  and  the  
textual  function.  Although  these  three  functions  were  originally  conceived  of  as  a  
model  for  understanding  language,  Kress  and  van  Leeuwen  insist  that  they  
provide  a  useful  starting  point  for  studying  all  modes.  In  their  book  Reading  
images:  The  grammar  of  visual  design,  for  example,  they  explore  how  images  also  

  89  
fulfill  these  three  functions,  but  do  so  in  a  rather  different  way  than  language  
(see  Section  D9).      
 
Ideational  function  
 
As  noted  in  Section  A4,  the  ideational  function  of  language  is  accomplished  
through  the  linking  together  of  participants  (typically  nouns)  with  processes  
(typically  verbs),  creating  what  Gee  (2011)  calls  ‘whos  doing  whats’.  In  images,  
on  the  other  hand,  participants  are  generally  portrayed  as  figures,  and  the  
processes  that  join  them  together  are  portrayed  visually.    
 
Images  can  be  narrative,  representing  figures  engaged  in  actions  or  events,  
classificatory,  representing  figures  in  ways  in  which  they  are  related  to  one  
another  in  terms  of  similarities  and  differences  or  as  representatives  of  ‘types’,  
or  analytical,  representing  figures  in  ways  in  which  parts  are  related  to  wholes.    
 
In  narrative  images,  action  processes  are  usually  represented  by  what  Kress  and  
van  Leeuwen  call  vectors,  compositional  elements  that  indicate  the  directionality  
of  an  action.  In  figure  B9.1,  for  example,  the  arm  of  the  boxer  on  the  left  
extending  rightward  towards  the  head  of  the  other  boxer  portrays  the  process  of  
‘hitting’.  There  are  also  other  processes  portrayed.  For  example,  the  upward  
gazes  of  the  figures  in  the  background  create  vectors  connecting  the  spectators  
with  the  fighters.    
 

 
Figure  B9.1  Warriors  (photo  credit  Claudio  Gennari)  
 

  90  
Like  this  image,  many  images  actually  represent  multiple  processes  
simultaneously.  Figure  B9.2,  for  example,  also  involves  action  processes,  the  face  
on  the  left  (representing  a  library  user)  joined  to  the  different  kinds  of  resources  
he  or  she  ‘consumes,  uses,  evaluates,  creates,  combines,  and  shares’.  This  image,  
however,  is  more  abstract,  and  so  the  vectors  are  represented  as  labeled  arrows  
rather  than  visual  representations  of  these  actions.  At  the  same  time,  the  image  
also  contains  classificatory  relationships  –  the  objects  portrayed  under  the  
headings  ‘Information  Literacy’,  ‘Media  Literacy’  and  ‘Digital  Literacy’  
representing  distinct  classes  of  things  -­‐-­‐  and  analytical  relationships  -­‐-­‐  the  
smaller  faces  in  the  lower  right  corner,  for  example,  portrayed  as  parts  of  a  
larger  social  network.    
 

 
 
Figure  B9.2  Using  information,  media  and  digital  literacy  (credit  Karin  Dalziel)  
 
Interpersonal  function  
 
Another  important  function  of  any  mode  is  to  create  and  maintain  some  kind  of  
relationship  between  the  producer  of  the  message  and  its  recipient.  As  we  said  in  
Section  A4,  in  language  these  relationships  are  usually  created  through  the  
language’s  system  of  modality,  as  well  as  through  the  use  of  different  ‘social  
languages’  or  ‘registers’.    
 

  91  
In  images,  viewers  are  placed  into  relationships  with  the  figures  in  the  image,  
and,  by  extension,  the  producers  of  the  image,  through  devices  like  perspective  
and  gaze.  The  image  of  the  child  in  figure  B9.3  illustrates  both  of  these  devices.  
The  camera  angle  positions  the  viewer  above  the  child  rather  than  on  the  same  
level,  creating  the  perspective  of  an  adult,  and  the  child’s  direct  gaze  into  the  
camera  creates  a  sense  of  intimacy  with  the  viewer,  though  the  expression  on  the  
child’s  face  does  denote  some  degree  of  uncertainty.  Another  important  device  
for  expressing  the  relationship  between  the  viewer  and  the  figures  in  an  image  is  
how  close  or  far  away  they  appear.  Long  shots  tend  to  create  a  more  impersonal  
relationship,  whereas  close-­‐ups  tend  to  create  a  feeling  of  psychological  
closeness  along  with  physical  closeness.    
 

 
 
Figure  B9.3  Child  (photo  credit  Denis  Mihailov)  
 
 ‘Modality’  in  images  is  partially  realized  by  how  ‘realistic’  the  image  seems  to  the  
viewer.  Photographs,  for  example,  generally  attest  more  strongly  to  the  ‘truth’  of  
a  representation  than  drawings  or  paintings.  However,  this  is  not  always  the  
case.  Scientific  diagrams  and  sketches,  for  example,  are  often  regarded  as  having  
even  more  ‘authority’  than  photographs,  and  black  and  white  images  like  those  
often  found  in  newspapers  are  often  regarded  as  more  ‘realistic’  than  highly  
saturated  color  images  in  magazine  advertisements.    
 
Textual  function  
 
As  we  said  above,  while  texts  are  organized  in  a  linear  fashion  based  on  
sequentiality,  images  are  organized  spatially.  Figures  in  an  image,  for  example,  
can  be  placed  in  the  center  or  periphery  of  the  image,  on  the  top  or  the  bottom,  
the  left  or  the  right,  and  in  the  foreground  or  in  the  background.  Although  

  92  
producers  of  images  have  much  less  control  than  producers  of  written  texts  over  
how  viewers  ‘read’  the  image,  they  can  create  pathways  for  the  viewer’s  gaze  by,  
for  example,  placing  different  figures  in  different  places  within  the  frame  and  
making  some  more  prominent  and  others  less  prominent.    
 
One  obvious  way  to  do  this  is  by  creating  a  distinction  between  foreground  and  
background,  the  figures  which  seem  closer  to  the  viewer  generally  commanding  
more  prominence.  Another  way  is  to  place  one  or  more  figures  in  the  center  of  
the  image  and  others  on  the  margins.  Many  images  make  use  of  the  
center/margin  distinction  to  present  one  figure  or  piece  of  information  as  the  
center  or  ‘nucleus’  of  the  image  and  the  marginal  figures  as  somehow  dependent  
upon  or  subservient  to  the  central  figure  (Kress  and  van  Leeuwen,  2006).    
 
Two  other  important  distinctions  in  the  composition  of  images,  according  to  
Kress  and  van  Leeuwen  (2006)  are  the  distinction  between  the  left  side  and  the  
right  side  of  the  image,  and  the  distinction  between  the  upper  part  and  the  lower  
part.  Taking  as  their  starting  point,  Halliday’s  idea  that  in  language,  ‘given’  
information  (information  that  the  reader  or  hearer  is  already  familiar  with)  
tends  to  appear  at  the  beginning  of  clauses,  and  new  information  tends  to  appear  
closer  to  the  end  of  clauses,  they  posit  that,  similarly,  the  left  side  of  an  image  is  
more  likely  to  contain  ‘given’  information  and  the  right  side  to  contain  ‘new’  
information.  This  is  based  on  the  assumption  that  people  tend  to  ‘read’  images  in  
the  same  way  they  read  texts,  starting  at  the  left  and  moving  towards  the  right.  
This,  of  course,  may  be  different  for  people  from  speech  communities  that  are  
accustomed  to  reading  text  from  right  to  left  or  from  top  to  bottom.    
 
The  distinction  between  the  upper  part  of  an  image  and  the  lower  part  is  related  
to  the  strong  metaphorical  connotations  of  ‘up’  and  ‘down’  in  many  cultures  
(Lakoff  and  Johnson,  1980).  According  to  Kress  and  van  Leeuwen,  the  top  part  of  
the  image  is  often  used  for  more  ‘ideal’,  generalized  or  abstract  information,  and  
the  bottom  for  ‘real’,  specific  and  concrete  information.  They  give  as  an  example  
advertisements  in  which  the  upper  section  usually  shows  ‘the  “promise  of  the  
product”,  the  status  of  glamour  it  can  bestow  on  its  users’  and  the  lower  section  
tends  to  provide  factual  information  such  as  where  the  product  can  be  obtained  
(2006:  186).    
 
Both  of  these  principles  can  be  seen  in  figure  B9.4.  In  order  to  make  sense  of  the  
‘narrative’  of  HIV  transmission  that  the  text  tells,  one  must  begin  at  the  far  left  of  
the  advertisement  and  move  to  the  far  right.  The  figures  of  the  man  and  the  
woman  on  the  left  of  the  image  constitute  ‘given’  information,  while  the  virus  on  
the  right  of  the  image  constitutes  the  ‘new’  information.  There  is  also  a  clear  
demarcation  between  the  upper  half  of  the  text  and  the  lower  half.  While  the  
upper  half  does  not  portray  the  positive  ‘promise’  of  a  particular  product  as  
many  advertisements  do,  it  does  represent  a  kind  of  idealized  hypothetical  
situation  which  the  viewer  is  invited  to  imagine.  Rather  than  a  ‘promise’,  
however,  it  is  something  more  akin  to  a  ‘threat’.  And  the  lower  half  of  the  image,  
rather  than  giving  information  about  where  the  product  portrayed  in  the  upper  
half  can  be  obtained,  it  gives  information  on  how  this  hypothetical  situation  can  

  93  
be  avoided,  along  with  specific  information  about  such  things  as  the  name  of  the  
organization  that  produced  the  ad  and  the  condom  company  that  sponsored  it.    
 

 
 
Figure  B9.  4  AIDS  prevention  advertisement  (Abrasco,  Brazil)    
 
This  text  also  illustrates  how  images  and  words  often  work  together.  The  words  
‘Joy  Stick’,  ‘Play  Station’  and  ‘Game  Over’  tell  the  viewer  how  the  images  are  to  be  
interpreted,  and  the  slogan  in  the  lower  half  of  the  text  (‘You  only  have  one  life:  
use  a  condom’),  explains  the  image  of  the  condom  above  it.  Finally,  this  text  
shows  how  multimodality  can  be  effective  in  getting  viewers  to  make  
connections  between  different  ‘Discourses’.  While  the  images  belong  to  the  
‘Discourse  of  biomedicine’,  the  words  invite  the  viewer  to  interpret  these  images  
within  the  framework  of  the  ‘Discourse  of  video  games’.  
 
    Find  additional  examples  online  
 

Multimodality  in  Interaction  

 
Modes  in  face-­‐to-­‐face  interaction  such  as  gaze  and  gesture  also  fulfill  these  three  
functions.  The  mode  of  gaze,  for  example,  has  an  obvious  interpersonal  function,  
creating  a  relationship  between  the  gazer  and  whomever  or  whatever  is  the  
object  of  the  gaze.  It  also  carries  ideational  meaning,  conveying  that  the  gazer  is  
looking  at,  watching  or  paying  attention  to  something.  Finally,  gaze  is  often  an  
important  textual  resource,  helping  people  to  manage  things  like  turn-­‐taking  in  
conversations.  
 
While  the  ‘inter-­‐modal’  relationships  (the  ways  multiple  modes  work  together)  
in  static  texts  like  the  advertisement  analyzed  above  can  be  complicated,  they  
can  be  even  more  complicated  in  dynamic  interactions.  One  of  the  problems  with  

  94  
analyzing  real  time,  face-­‐to-­‐face  interactions  is  that  participants  have  so  many  
modes  available  to  them  to  make  meaning.  There  are  what  Norris  (2005)  calls  
‘embodied’  modes  such  as  gaze,  gesture,  posture,  head  movement,  proxemics  
(the  distance  one  maintains  from  his  or  her  interlocutor),  spoken  language,  and  
prosody  (features  of  stress  and  intonation  in  a  person’s  voice).  And  there  are  also  
‘disembodied’  modes  like  written  texts,  images,  signs,  clothing,  the  layout  of  
furniture  and  the  architectural  arrangement  of  rooms  and  other  spaces  in  which  
the  interaction  takes  place.  All  of  these  different  modes  organize  meaning  
differently.  Some,  like  spoken  language  and  gaze  tend  to  operate  sequentially,  
while  others  like  gesture  and  prosody  tend  to  operate  globally,  often  helping  to  
create  the  context  in  which  other  modes  like  spoken  language  are  to  be  
interpreted  (see  Section  B6).  Not  all  of  these  modes  are  of  equal  importance  to  
participants  at  any  given  moment  in  the  interaction.  In  fact,  different  modes  are  
likely  to  take  on  different  degrees  of  importance  at  different  times.  How  then  is  
the  analyst  to  determine  which  modes  to  focus  on  in  a  multimodal  analysis?  
 
Another  problem  with  analyzing  multimodality  in  face-­‐to-­‐face  interactions  is  
that  the  spatial  boundaries  of  interactions  are  not  always  as  clear  as  the  spatial  
boundaries  of  texts.  While  the  frame  of  an  image  clearly  marks  what  should  be  
considered  as  belonging  to  the  image  and  what  should  be  considered  external  to  
it,  a  conversation  in  a  coffee  shop  is  not  so  clearly  bounded.  In  analyzing  such  an  
interaction,  how  much  of  the  surrounding  modes  should  be  taken  into  account?  
Should  the  analyst  consider,  for  example,  the  signs  and  posters  on  the  walls,  the  
conversations  occurring  at  other  tables,  the  ambient  music  playing  over  the  p.a.  
system  and  the  sounds  of  milk  being  steamed?  What  about  the  smell  and  taste  of  
the  coffee?    
 
Norris  (2005)  solves  these  two  problems  by  adopting  the  practice  of  mediated  
discourse  analysis  (see  Sections  A8  and  B8)  and  taking  action  as  her  unit  of  
analysis.  Thus,  in  determining  which  modes  to  focus  on,  the  analyst  begins  by  
asking  what  actions  participants  are  engaged  in  and  then  attempts  to  determine  
which  modes  are  being  used  to  accomplish  these  actions.    
 
As  we  said  in  Section  A8,  actions  are  always  made  up  of  smaller  actions  and  
themselves  contribute  to  making  up  larger  actions.  Norris  divides  actions  into  
three  types:  lower-­level  actions,  the  smallest  pragmatic  meaning  units  of  
communicative  modes  (including  things  like  gestures,  postural  shifts,  gaze  shifts,  
and  tone  units),  higher-­level  actions  (such  as  ‘having  a  cup  of  coffee’),  and  frozen  
actions  (previously  performed  actions  that  are  instantiated  in  material  modes—a  
half  eaten  plate  of  food,  for  example,  or  an  unmade  bed).    
 
One  of  the  goals  of  multimodal  interaction  analysis,  then,  is  to  understand  how  
participants  in  interaction  work  cooperatively  to  weave  together  lower-­‐level  
actions  like  gestures,  glances,  and  head  and  body  movements  into  higher-­‐level  
actions,  and,  in  doing,  so  help  to  create  and  reinforce  social  practices,  social  
relationships  and  social  identities  (see  Section  C9).    
 
  Look  deeper  into  this  topic  online  

  95  
B10  PROCEDURES  FOR  CORPUS-­‐ASSISTED  DISCOURSE  
ANALYSIS  
 
Conducting  a  corpus-­‐assisted  discourse  analysis  requires  a  number  of  steps,  
which  include  building  a  corpus,  cleaning  and  tagging  the  corpus,  analyzing  the  
corpus  with  computer  tools  using  a  number  of  procedures,  and,  finally,  
interpreting  the  data.  These  last  two  steps  tend  to  be  cyclical  and  recursive.  That  
is,  usually  the  results  of  several  procedures  need  to  be  combined  when  we  are  
interpreting  the  data,  and  often  our  interpretations  lead  us  to  re-­‐performing  
these  procedures  or  performing  other  procedures.    
 
The  first  step  in  building  a  corpus  is  deciding  what  kinds  of  texts  you  want  to  
include  in  it  and  making  sure  that  you  can  include  a  representative  sample  of  
those  kinds  of  texts.  For  very  specialized  corpora,  such  as  the  works  of  a  
particular  author,  this  is  easy  since  there  are  a  limited  number  of  texts  and  you  
can  simply  include  them  all.  This  is  more  difficult  the  less  specific  the  corpus  is.  
For  example,  if  you  want  to  build  a  corpus  of  business  letters,  you  need  to  decide  
what  kind  of  letters  (sales  letters,  complaint  letters,  etc.)  you  want  to  include,  
what  kinds  of  companies  these  letters  will  come  from,  and  what  countries  
and/or  ‘cultures’  will  be  included.  You  might  choose  texts  based  on  some  
predetermined  criteria  like  topic  or  the  inclusion  of  some  keyword.  Baker  and  
McEnery,  in  their  study  reprinted  in  Section  D10,  for  example,  chose  the  texts  for  
their  corpus  on  the  basis  of  whether  or  not  they  contained  the  words  refugee  or  
refugees  or  the  phrases  asylum  seeker  or  asylum  seekers.    
 
Another  important  decision  is  how  many  texts  you  are  going  to  include  in  your  
corpus.  Generally  with  corpus-­‐assisted  analysis,  the  bigger  the  corpus  the  easier  
it  will  be  for  you  to  make  generalizations  from  your  results.  However,  it  is  also  
possible  to  have  very  small  corpora.    
 
You  will  probably  also  need  a  reference  corpus.  A  reference  corpus  is  another  
corpus  that  you  will  compare  your  primary  corpus  with.  It  is  usually  made  up  of  
a  broader  spectrum  of  texts  or  conversations  than  the  corpus  you  are  analyzing.    
You  might,  for  example,  use  one  of  the  large  corpora  like  the  British  National  
Corpus,  or  you  might  choose  another  specialized  corpus  with  a  broader  sample  
of  texts.    
 
Nowadays  it  is  actually  quite  easy  to  build  a  corpus  since  so  many  texts  are  
already  in  electronic  format  on  the  Internet.  But  is  it  important  that  you  go  
though  these  texts  carefully  and  take  out  any  HTML  code  or  formatting  that  
might  have  been  attached  to  them,  which  might  interfere  with  your  analysis.  You  
also  might  want  to  attach  new  code  to  certain  parts  of  the  text  or  to  certain  
words  to  aid  your  analysis.  This  later  process  to  called  ‘tagging’.  Analysts,  for  
example,  sometimes  insert  code  to  indicate  different  parts  of  a  text  (like  
introduction,  body  and  conclusion),  and  others  tag  individual  words  based  on  
their  grammatical  function  so  they  can  detect  grammatical  patterns  in  their  

  96  
analysis  along  with  lexical  patterns.  It  is  important  that  each  text  in  your  corpus  
is  saved  in  separate  text  file.    
 
The  analysis  of  the  corpus  is  carried  out  with  a  computer  program,  and  there  are  
a  number  of  such  programs  available  for  free  on  the  internet.  The  most  widely  
used  commercial  program  is  called  WordSmith  Tools  (http://www.lexically.net/  
wordsmith/index.html),  but  there  is  also  a  very  good  free  program  available  
called  AntConc,  developed  by  Laurence  Anthony,  which  works  on  both  Windows  
and  Macintosh  operating  systems  (http://www.antlab.sci.waseda.ac.jp/  
antconc_index.html).  In  the  explanations  and  examples  below  I  will  describe  how  
to  perform  the  relevant  procedures  using  AntConc.    
 
After  your  corpus  has  been  ‘cleaned’  and  ‘tagged’,  you  need  to  import  it  in  the  
form  of  text  files  into  your  analysis  program.  In  AntConc  this  is  done  by  using  the  
commands  File  >  Open  File(s)  (or  Cntrl  F).  You  may  choose  as  many  files  as  you  
wish.  If  you  would  like  to  open  a  directory  of  files,  choose  Open  Dir  (or  Cntrl  D).    
 
While  there  are  a  whole  host  of  different  operations  that  can  be  performed  on  
corpora  using  this  software,  the  six  most  basic  procedures  useful  for  the  
discourse  analyst  are  as  follows:    
 
1)  Generating  word  frequency  lists  
2)  Calculating  type  token  ratio  
3)  Analyzing  concordances  
4)  Analyzing  collocation  
5)  Analyzing  keywords  
6)  Creating  dispersion  plots  
 
Most  of  these  procedures  can  be  performed  on  their  own,  but  it  is  usually  a  good  
idea  to  perform  them  together  with  the  other  procedures  since  the  results  from  
one  procedure  can  often  inform  your  interpretation  of  the  results  from  the  
others.  
 
Word  frequency  and  type  token  ratio  
 
One  of  the  most  basic  pieces  of  information  you  can  get  about  your  corpus  from  a  
computer  aided  analysis  is  information  about  the  frequency  with  which  different  
words  occur.  In  AntConc  a  word  frequency  list  for  a  corpus  can  be  generated  by  
clicking  on  the  Word  List  tab  and  then  clicking  the  Start  button.  Unless  you  have  
a  good  reason  to  treat  words  in  different  cases  (e.g.  refugee  vs.  Refugee)  as  
separate  words,  it  is  a  good  idea  to  tick  ‘Treat  all  data  as  lower  case’  in  the  
Display  Options.  Words  in  frequency  lists  can  be  sorted  by  rank,  frequency  or  
word,  so  an  analyst  can  easily  determine  not  just  the  most  or  least  frequently  
occurring  words,  but  also  check  the  frequency  of  specific  words.    
 
After  a  word  list  is  generated,  the  information  necessary  to  calculate  type  token  
ratio  appears  at  the  top  of  the  AntConc  window.  Type  token  ratio  is  basically  a  
measure  of  how  many  different  kinds  of  words  occur  in  the  text  in  relation  to  the  
total  number  of  words,  and  so  can  give  some  indication  of  the  lexical  complexity  

  97  
of  texts  in  a  corpus.  It  is  calculated  by  dividing  the  number  of  types  by  the  
number  of  tokens.  A  low  type  token  ration  generally  indicates  a  relatively  
narrow  range  of  subjects,  a  lack  of  lexical  variety  or  frequent  repetition.  A  high  
type  token  ratio  indicates  a  wider  range  of  subjects,  greater  lexical  variation,  
and/or  less  frequent  repetition.  In  the  British  National  Corpus,  the  type  token  
ratio  for  the  corpus  of  written  texts  is  45.53,  whereas  the  type  token  ratio  for  the  
corpus  of  spoken  texts  is  32.96.  This  confirms  a  number  of  things  we  already  
know  about  the  differences  between  speech  and  writing,  in  particular,  that  
writing  tends  to  involve  a  much  more  varied  and  complex  vocabulary,  and  that  
speech  tends  to  involve  frequent  repetition.    
 
Usually  the  most  frequent  words  in  any  text  are  function  words  (articles,  
prepositions,  pronouns  and  other  grammatical  words)  such  as  ‘the’  and  ‘a’.  While  
looking  at  function  words  can  be  useful  in  helping  you  to  understand  
grammatical  patterns,  style  and  register  in  the  corpus,  content  words  like  nouns,  
verbs  and  adjectives  are  usually  more  relevant  to  finding  evidence  of  
‘Discourses’.    
 
Concordances  
 
Concordances  show  words  in  the  context  of  the  sentences  or  utterances  in  which  
they  were  used.  Usually  we  use  frequency  lists  to  give  us  an  idea  of  what  some  of  
the  important  words  in  a  corpus  might  be,  and  then  we  do  a  concordance  of  
those  words  in  order  to  find  out  more  information  about  them.  Concordances  
can  be  sorted  alphabetically  based  on  the  words  either  to  the  right  or  left  of  the  
word  that  you  searched  for,  and  playing  around  with  this  sorting  system  is  often  
a  good  way  to  spot  patterns  in  word  usage.  For  example,  in  Baker  and  McEnery’s  
study  reprinted  in  Section  D10,  the  alphabetical  sorting  of  words  directly  to  the  
left  and  right  of  the  target  word  (refugee)  helped  reveal  that  refugees  were  
commonly  described  in  newspaper  articles  in  terms  of  quantification  (using  
numerals  or  terms  like  tens  of  thousands,  more  and  more).  
 
In  AntConc  concordances  are  created  by  typing  a  word  or  phrase  into  the  Search  
Term  box,  generating  a  list  of  instances  in  which  this  word  appears  in  the  corpus  
listed  in  their  immediate  contexts.  The  search  word  appears  in  the  concordance  
in  the  center  of  the  page  highlighted  in  blue,  with  what  occurs  before  and  after  
appearing  to  the  left  and  the  right  of  the  word.  The  Kwic  Sort  dialogue  can  be  
used  to  sort  the  concordance  alphabetically  based  on  the  word  one,  two,  three,  
etc.  places  to  the  left  or  the  right  of  the  search  term.    
 
Collocation  analysis  
 
Collocation  has  to  do  with  the  fact  that  certain  words  tend  to  appear  together.  
Often  words  take  on  a  negative  or  positive  meaning  based  on  the  kinds  of  words  
they  are  often  grouped  with.  As  Firth  (1957)  put  it,  ‘You  shall  know  a  lot  about  a  
word  from  the  company  it  keeps.’  For  example,  the  verb  ‘commit’  is  nearly  
always  associated  with  negative  words  like  ‘crime’.  We  don’t  ‘commit’  good  
deeds,  we  ‘perform’  them.  Thus  we  find  phrases  like  ‘commit  random  acts  of  
kindness’  humorous.    

  98  
  
Analyzing  the  kinds  of  words  that  appear  together  with  other  words  is  an  
especially  useful  way  to  understand  the  ‘Discourses’  that  are  expressed  in  a  
corpus  because  they  can  reveal  patterns  of  association  between  different  kinds  
of  words  or  concepts.  In  their  study  of  the  portrayal  of  refugees  in  the  British  
press,  for  example,  Baker  and  McEnery  note  not  just  that  the  word  stream  is  used  
frequently  in  their  corpus  to  describe  the  movement  of  refugees,  but  that  in  the  
British  National  Corpus  this  word  frequently  collocates  with  the  words  tears,  
blood,  sweat,  water  and  rain,  giving  it  a  generally  negative  connotation.  Baker  
(2006)  refers  to  the  situation  where  patterns  can  be  found  between  words  and  
various  sets  of  related  words  in  ways  that  suggest  a  ‘Discourse’  as  discourse  
prosody.  Others  (see  for  example  Sinclair  1991)  refer  to  this  as  semantic  prosody.    
 
In  order  to  perform  a  collocation  analysis  with  AntConc,  click  the  Collocate  tab  
and  enter  your  chosen  search  term.  You  will  also  need  to  determine  the  span  to  
the  left  or  right  of  the  search  term  within  which  you  want  to  check  for  collocates.  
This  can  be  set  from  any  number  of  words  to  the  left  of  the  search  term  to  any  
number  of  words  to  the  right  of  the  search  term  using  the  Window  Span  
dialogue.  The  result  will  be  a  list  of  collocates,  their  rank,  overall  frequency,  and  
the  frequency  with  which  they  occur  to  the  left  of  the  search  term  and  to  the  
right  of  the  search  term.    
 
Keyword  analysis  
 
Word  frequency  lists  can  only  tell  you  how  frequently  certain  words  occur  in  the  
corpus.  Some  words,  however,  like  articles,  occur  frequently  in  nearly  every  text  
or  conversation.  The  frequency  with  which  a  word  occurs  in  a  corpus  is  not  in  
itself  necessarily  meaningful.  What  is  more  important  is  whether  or  not  a  word  
occurs  more  or  less  frequently  than  ‘normal’.  This  is  what  keyword  analysis  is  
designed  to  determine.      
 
The  difference  between  keywords  and  frequent  words  is  that  keywords  are  
words  that  appear  with  a  greater  frequency  in  the  corpus  that  you  are  studying  
than  they  do  in  a  ‘reference  corpus’.  Reference  corpora  usually  consist  of  a  
broader  sampling  of  texts  or  conversations.  Many  people,  for  example,  use  large  
publically  available  corpora  like  the  British  National  Corpus.    
 
In  order  to  generate  a  list  of  keywords  for  your  corpus  with  AntConc,  it  is  first  
necessary  to  load  your  reference  corpus.  This  is  done  using  the  Keyword  List  
preferences  (Tool  Preferences  >  Keyword  List).  The  reference  corpus  can  be  
loaded  either  as  a  list  of  files  or  as  a  directory.  Once  it  is  loaded,  the  keyword  list  
is  generated  by  choosing  the  Keyword  List  tab  and  clicking  Start.  The  result  will  
be  a  list  of  keywords,  their  rank,  frequency  and  a  number  measuring  their  
keyness.  The  keyness  value  indicates  the  degree  to  which  the  word  occurs  more  
frequently  than  expected  in  your  primary  corpus  (taking  the  reference  corpus  as  
representing  a  ‘normal’  pattern  of  frequency).    Some  programs  also  allow  you  to  
calculate  negative  keyness,  that  is,  to  determine  which  words  occur  less  
frequently  than  expected.    
 

  99  
Dispersion  Plots  
 
Dispersion  plots,  referred  to  in  AntConc  as  concordance  plots,  can  give  you  
information  about  where  words  occur  in  texts.  This  can  be  particularly  useful  if  
an  analyst  is  interested  in  the  structure  of  texts  or  conversations.  A  genre  
analyst,  for  example,  might  be  interested  in  the  kinds  of  words  or  phrases  that  
occur  in  a  section  of  a  text  associated  with  a  particular  move,  or  a  conversation  
analyst  might  want  to  explore  the  kinds  of  words  that  occur  in  different  parts  of  a  
conversation  such  as  the  opening  or  the  closing.    
 
In  AntConc,  concordance  plots  are  generated  by  clicking  the  Concordance  Plot  
tab,  typing  in  a  search  term  and  clicking  Start.  The  result  is  a  series  of  bars,  each  
representing  a  text  in  the  corpus  with  lines  representing  where  the  search  term  
has  appeared.  Figure  B10.1  shows  the  dispersion  plots  generated  by  searching  
for  the  word  ‘love’  in  a  corpus  of  Lady  Gaga  songs.    
 

 
 
Fig.  B10.1  Concordance  plots  for  Lady  Gaga  songs  
 
 
      Look  deeper  into  this  topic  online  

  100