Exploring  the  meaning  of  gender:  

Evaluating  and  revising  the  Bem  Sex-­Role  
Inventory  (BSRI)  for  a  Swedish  research  
context  (BSRI-­SE)  
 
 

Roland  S.  Persson  

 

INSIKT  1999:1  
Vetenskapliga  rapporter  från  HLK  

 
 
 Abstract  
This  research  evaluates  the  Bem  Sex-­‐Role  Inventory  (BSRI)  for  use  in  a  
Swedish   setting,   and   in   the   process   of   so   doing   also   compares   the  
distribution   of   American   and   Swedish   gender   roles   as   elicited   by   the  
inventory.  A  sample  of  118  individuals  (48  males  and  70  females)  was  
used   in   order   to   arrive   at   norms   better   suited   to   a   Swedish   context  
than   those   provided   by   Bem   (1974).   Reliability   and   factor   analyses  
were   performed   and   a   revised   version   of   the   inventory,   with  
acceptable   psychometric   properties   for   the   Swedish   sample,   was  
arrived   at   and   termed   BSRI-­‐SE.   Comparisons   between   the   original  
American   and   the   Swedish   sample   show   that   differences   in   the   way  
that   femininity,   masculinity,   androgyny   and   undifferentiated   gender  
are  distributed  are  not  statistically  significant.  Raw  score  data  for  the  
BSRI-­‐SE,   as   well   as   the   scoring   sheet,   are   provided   for   continued  
standardisation  of  the  inventory.  Methods  for  classification  and  further  
research  are  discussed.  
 
 
 
 
 
 
____________________________________________________________  
Field  of  study:     Social  psychology  
Keywords:     Gender,  Androgyny,  Masculinity,  Femininity,  
Psychological  Measurement,  Cross-­‐cultural  
comparisons,  Identity,  Differentiation,  Personality.  

2
  
 
Index  
 
Introduction,                        4  
The  Bem  Sex-­Role  Inventory  (BSRI),              8  
The  BSRI  in  a  Swedish  context,             17  
  The  nature  of  the  Swedish  sample,           17  
  Reliability  of  the  inventory  scales,           20  
Procedure  for  classification  and  standardisation,       27  
  The  median  split  method,             28  
  The  M-­‐F  Difference  score,           29  
Are  Swedes  masculine,  feminine  or  androgynous?       32  
  Sex-­‐typing  in  the  sample  subgroups,         35  
Concluding  remarks,                 36  
References,                     40  
Appendices  
  I   Test  sheet  for  the  BSRI-­‐SE  
  II   Raw  scores  of  the  Swedish  sample  
 
 
 
 
 
 
 

3
  

Introduction  
Few   issues   in   recent   years   have   been   subject   to   so   much   debate,  
research,  exploitation  and  politics  as  have  gender  roles,  their  meaning  
and   social   impact.   While   researchers   agree   that   sex   is   biologically  
determined   they   do   not   agree   on   when,   why   and   how   psychological  
differentiation   between   gender   roles   occur.   To   argue   that   something   is  
typically  male  or  female  is  by  no  means  theoretically  straight-­‐forward  
(Sternberg,  1993),  and  there  is  a  great  deal  of  variety  across  cultures  of  
what  is  regarded  as  typically  male  or  female  behaviour.  Although  there  
are   some   universals   with   regard   to   gender,   societies   differ   in   the   value  
attached   to   maintenance   of   differences   between   the   sexes.   Generally,  
males   are   more   likely   to   be   physically   aggressive   and   express  
dominance  over  women  rather  than  vice  versa.  Women  are  more  likely  
in   a   global   perspective   to   conform,   defer,   comply   and   submit   to  
authority,   especially   if   the   authority   is   male   (Segall,   Dasen,   Berry   &  
Poortinga,   1990).   Men   and   women’s   self-­‐perceptions   are   less  
stereotypical   in   more   economically   and   socially   developed   countries,  
and   it   appears   that   as   sexual-­‐equality   ideals   spread,   for   example  
through   feminist   ideology,   behavioural   differences   between   the   sexes  
diminish   and   prevailing   stereotypes   are   increasingly   being   blurred  
(Williams  &  Best,  1989).    
Gender   is   a   multi-­‐facetted   phenomenon.   Most   would   think   of  
gender   in   terms   of   feminine   and   masculine,   thus   expressing   that  
gender   is   confined   to   a   dichotomy.   But   there   is   in   fact   no   agreement  
amongst   behavioural   scientists   on   how   many   genders   there   are!   Is   it  

4
feasible   to   have   as   many   genders   as   there   are   sexes,   namely   two?   Or  
are   there   also   “nuances”   of   both,   which   should   perhaps   be   regarded   as  
separate   genders?   Note   that   by   gender   is   meant   the   psychological  
array  of  behaviours,  that  to  a  large  extent  are  socially  determined,  and  
that   we   have   a   tendency   to   identify   as   typical   of   either   sex.   In   some  
cases   there   are   biological   causes   that   could   possibly   qualify   the  
conceptualisation   of   more   genders   than   the   traditional   two.   Sex  
chromosome  disorders  like  Turner’s  Syndrome  (individuals  born  with  
only   one   X   chromosome),   Klinefelter’s   Syndrome   (individuals   born  
with   three   X   chromosomes   or   two   X   chromosomes   and   one   Y  
chromosome)   or   the   XYY   Syndrome   are   all   genetic   disorders,   which  
have   distinctive   influences   on   an   individual’s   secondary   sexual  
characteristics  like  body  hair,  breasts,  body  build  and  so  on.  With  these  
disorders   follow   some   differentiation   in   cognitive   functioning   (cf.  
Willerman   &   Cohen,   1990   for   an   overview).   In   addition,   there   are   a  
number   of   different   sexual   behaviours   that   challenge   the   traditional  
division   of   gender   restricted   to   the   dichotomy   of   male   and   female.  
Transsexualism,   for   example,   describes   incongruity   between   gender  
identity   and   anatomical   sex,   the   aetiology   of   which   could   possibly   be  
due   to   endocrine   factors   rather   than   in   various   ways   social   learning  
(Hoenig   &   Kenna,   1974).   Similarly,   male   transvestism   could   be  
understood   as   the   co-­‐existence   of   two   selves:   male   and   female  
(Larsson,   1997).   Homosexuality   is   also   in   a   sense   a   challenge   to  
traditional   gender   roles   (Bailey,   1996;   Ross,   1985;   Williams,   1996),  
particularly   since   recent   research   is   increasingly   showing   that  
homosexuality   may   be   genetically   and/or   hormonally   determined  
(Ellis,   1996;   LeVay,   1993).   A   majority   of   researchers   tend   to   stress  

5
social   aspects   instrumental   to   developing   sexual   behaviour   (Howells,  
1987).   Money   and   Erhardt   (1972),   for   example,   reviewed   a   large  
number   of   studies   on   sex   hormone   anomalies   and   concluded   that   the  
most  important  factor  in  developing  sex  role  and  gender  identity  is  the  
assigned  sex  of  rearing.  However,  such  research  may  well  be  in  error  if  
understanding  homosexuality  as  merely  a  social  construction  of  no  or  
little   evolutionary   significance,   as   elegantly   argued   by   McKnight  
(1997).   Genetics   and   human   physiology   seem   to   play   important,   but  
hitherto   often   wilfully   ignored,   roles   in   the   development   of   a  
homosexual  identity.  Murphy  (1990)  provocatively  points  out  that  “the  
incentive  to  discover  the  origins  of  homosexuality  seems  to  belong  to  
those   who   find   homosexuality   a   pathological,   sinful,   immoral   or  
criminal  condition.  At  least  on  the  basis  of  these  views  there  is  reason  
to   try   and   understand   the   origins   of   homosexual   behaviour   if   only   to  
prevent   and   eliminate   it.   It   is   ordinarily   some   deficit   which   prompts  
medicine   and   the   rest   to   reach   for   a   causal   explanation   of   behaviour”  
(p.  134).  
Thus,   on   the   basis   of   culture,   and   invariably   at   some   level,   also  
biological   sex—which   at   times   by   no   means   is   always   straight  
forward—it  may  well  be  possible  to  construe  gender  roles  in  terms  of  a  
masculine   male,   a   feminine   male,   a   masculine   woman,   a   feminine  
woman  or  any  nuance  of  these.  
  Rudberg   and   Bjerrum-­‐Nielsen   (1994)   point   out,   that   “gender   is  
still  relevant  [in  modern  society].  There  is  nothing  in  either  the  family’s  
gender   socialisation   or   the   history   of   modernity   that   suggests   that  
gender   should   be   ‘suspended’   as   a   psychological   or   social   category;  
rather   it   seems   that   gender   will   survive   in   a   new   form”   (p.   49-­‐50).   The  

6
understanding   of   gender   is   a   field   of   knowledge   in   flux   not   only  
because   it   intrigues   social   scientists   due   to   its   complexity,   but   also  
because   the   results   of   research   on   possible   causes   and   effects   in   this  
context   have   strong   political   repercussions.   It   is   historically   true   in   the  
Western  world  that  women,  often  at  great  risk,  have  struggled  in  order  
to  obtain  the  same  rights  that  by  unquestioned  tradition  have  been  the  
privilege   of   men   only.   However,   during   the   process   by   which   such  
rights   and   privileges   are   increasingly   recognised   for   both   sexes,   the  
biased  function  of  rationalisation  should  also  be  recognised.    
While   it   is   politically   essential   to   argue   that   men   and   women   have  
equal   rights   in   a   democracy   it   is   simultaneously   paramount   for   science  
to   investigate   the   nature   of   differentiation.   But   where   the   political  
agenda   is   set   differences   between   gender   roles   would   appear   not   to   be  
a  favoured  object  of  discussion  or  debate—at  least  in  a  context  of  equal  
opportunities  on  the  job  market.  Rather  issues  are  focussed  which  may  
justify   a   certain   political   development.   In   such   a   situation   it   is   far   more  
opportune   to   focus   how   men   and   women   are   similar   rather   than  
different,   since   anything   different   from   what   is   considered   typically  
masculine   not   infrequently   is   interpreted   as   inferior.   Although   many  
stereotypes   about   gender   differentiation   are   false—like   for   example  
the   notion   prevalent   amongst   men   and   school   teachers,   that   women  
generally   are   believed   to   be   less   intelligent   than   men   (Broverman,  
Vogel  et  al.,  1972)—it  is  my  impression  that  real  differences  often  are  
underplayed   in   favour   of   making   women   more   like   men   (cf.   Lipman-­‐
Blumen,   1989).   Baumeister   (1988),   for   example,   argues   that  
differences   between   sexes   should   not   be   studied   at   all,   whereas  
McHugh,   Koeske   and   Frieze   (1986)   suggest   that   gender   differences  

7
should  only  be  reported  under  limited  circumstances  
  This   research   targets   the   social   differentiation   of   gender   believing  
it   to   be   essential.   First,   it   sets   out   to   evaluate   the   Bem   Sex-­‐Role  
Inventory   (Bem,   1974)   in   a   Swedish   context,   revising   it   to   make   it  
more   akin   to   Swedish   culture   rather   than   American   culture,   and   to  
provide   norms   by   which   the   revised   inventory   may   be   compared.  
Similar   efforts   have   been   done   in   several   Hispanic   settings   (e.g.  
Kaschak   &   Sharrat,   1983;   Kranau,   Green   &   Valencia-­‐Weber,   1982).  
Some   studies   have   used   the   BSRI   for   cross-­‐cultural   comparisons   in  
Israel   (Maloney,   Wilkof   &   Dambrot,   1981)   and   Australia   (Rowland,  
1977),   but   both   of   these   used   the   BSRI   original   norms   on   the   slightly  
dubious   assumption   that   the   understanding   of   gender   identity   and  
behaviour  is  roughly  the  same  as  in  the  United  States.  
 
  An  instrument  such  as  the  BSRI,  however,  has  value  for  research  
purposes,  since  gender  consideration  often  is  an  important  variable  in  
Social   Science   research.   Few   instruments   are   developed   or   translated  
and   standardised   for   a   Swedish   population.   For   example,   the  
publishing   branch   of   the   Swedish   Psychological   Society   in   their   1997  
catalogue   lists   111   different   tests   and   test   batteries,   which   should   be  
compared   to   The   American   Psychological   Association’s   estimate   that  
some  20.000  tests  are  developed  annually  (APA,  1993).    
Second,   and   a   result   of   the   process   of   evaluating   the   BSRI   in   a  
Swedish   context,   is   an   exploration   into   the   meaning   of   Swedish   gender  
roles.  How  does   the   Swedish   sample   compare   to   the   American   sample,  
which  served  as  the  basis  for  developing  the  American  norms  by  which  
BSRI   scores   are   compared?   Are   Swedes   more   or   less   masculine,  

8
feminine,  androgynous  than  Americans?  

The  Bem  Sex-­Role  Inventory  (BSRI)  

  The   BSRI   was   developed   by   Stanford-­‐psychologist   Sandra   Bem  
(1974;   1975;   1981)   in   an   effort   to   accommodate   the   fact   that   there   are  
more   nuances   to   being   masculine   or   feminine   than   the   theoretical  
constructs  at  the  time  would  allow  to  be  expressed  (e.g.  The  California  
Psychological   Inventory   by   Gough,   1957).   This   sex-­‐role   dichotomy,  
Bem   (1974)   argues,   “has   served   to   obscure   two   very   plausible  
hypotheses:   first,   that   many   individuals   might   be   ‘androgynous’;   that  
is,   they   may   be   both   masculine   and   feminine   ...   depending   on   the  
situational   appropriateness   of   these   various   behaviours;   and  
conversely,   that   strongly   sex-­‐typed   individuals   might   be   seriously  
limited  in  the  range  of  behaviours  available  to  them  as  they  move  from  
situation  to  situation”  (p.  155).  
 
  In  Bem’s  research  effort  masculinity  and  femininity  are  regarded  
as   two   independent   dimensions   rather   than   two   ends   of   a   single  
dimension.   This   conceptualisation   of   gender   allows   any   individual   to  
indicate   whether   he   or   she   is   high   on   both   dimensions   (and   thus  
androgynous),   low   on   both   dimensions   (and   therefore  
undifferentiated)   or   high   on   one   and   low   on   the   other   (either   feminine  
or   masculine).   Three   scales   were   originally   developed   in   order   to  
arrive   at   an   instrument,   which   would   allow   these   distinctions   to   be  
made   in   reference   to   stereotypical   gender   roles,   namely   Femininity,  
Masculinity  and  Social  Desirability.    
  An   item   pool   was   selected   by   Bem   and   her   students   on   the  

9
criterion   that   descriptions   of   masculinity   and   femininity   should  
describe   something   positive   and   stereotypically   male   or   female.   A  
personality   characteristic   qualified   as   typically   masculine   if  
independently   judged   by   both   females   and   males   to   be   significantly  
more   desirable   for   a   man   than   a   woman   (Table   1).   Similarly,   a  
characteristic   qualified   as   typically   feminine   if   judged   by   both   males  
and  females  to  be  significantly  more  desirable  for  a  woman  than  for  a  
man  (Table  2).      
  The   items   relating   to   social   desirability   were   culled   from  
characteristics,   which   seemed   to   be   neither   typically   feminine   nor  
masculine  but  rather  applicable  to  both  sexes  and  thus  neutral  in  terms  
of   sex-­‐appropriateness.   Of   these   half   were   positive   in   tone   and   half  
were   negative.   Items   were   judged   neutral   if   they   were   independently  
judged  by  both  males  and  females  to  be  no  more  desirable  for  one  sex  
than   for   the   other   (Table   3).   Note,   however,   that   this   scale   mainly  
served  an  initial  purpose  in  constructing  the  instrument  “to  insure  that  
the   inventory   would   not   simply   be   tapping   a   general   tendency   to  
endorse  socially  desirable  traits”  (Bem,  1974,  p.  156).  
 
 
 
 
 
 
 
 
 

10
Table   1.   Items   of   the   BSRI   Masculinity   Scale   (Bem,   1974;   1981).  
Translation   into   Swedish   included   as   well   as   their   variable   labels   by  
which  they  were  processed.  
 
English  original     Swedish  translation         Variable  
labels  
 
Self-­‐reliant       Full  av  självförtroende       M1SJLVFR  
Defends  own  beliefs   Försvarande  mina  övertygelser   M2FORSV  
Independent       Oberoende           M3OBERO  
Athletic         Sportslig             M4SPORT  
Assertive         Bestämd             M5BESTMD  
Strong  personality     En  stark  personlighet       M6STPER  
Forceful         Kraftfull             M7KRFTFL  
Analytical       Analytisk             M8ANALYT  
Has  leadership  abilities  En  ledartyp           M9LEDARE    
Willing  to  take  risks   Villig  att  ta  risker         M10RISK  
Makes  decisions  easily   Lätt  för  att  ta  beslut       M11BSLUT  
Self-­‐sufficient       Självtillräcklig         M12SJLVT  
Dominant       Dominant           M13DOMIN  
Masculine       Maskulin             M14MASK  
Willing  to  take  a  stand   Villig  att  fatta  beslut         M15BSLST  
och  stå  för  dem    
Aggressive       Aggressiv           M16AGGRS  
Act  as  a  leader     Att  agera  ledare         M17LEDAR  
Individualistic     Individualist           M18INDIV  
Competitive       Ha  en  tävlingsmentalitet     M19TAVL  
Ambitious       Ambitiös             M20AMBI  
 
 
 
 
 
 
 
 
 
 
 
 

11
  
 
 
Table   2.   Items   of   the   BSRI   Femininity   Scale   (Bem,   1974;   1981).  
Translation   into   Swedish   included   as   well   as   their   variable   labels   by  
which  they  were  processed.  
 
English  original     Swedish  translation     Variable  labels  
 
Yielding         Undfallande         F1UNDFAL  
Cheerful         Munter           F2MUNTR  
Shy           Blyg             F3BLYG  
Affectionate       Tillgiven           F4TILLGV  
Flatterable       Mottaglig  för  smicker     F5SMICKR  
Loyal         Lojal           F6LOJAL  
Feminine         Feminin           F7FEMINI  
Sympathetic       Sympatisk         F8SYMPAT  
Sensitive  to  the  needs     Känslig  för  andras  behov   F9KNSLBH  
of  others  
Understanding     Förstående         F10FORST  
Compassionate     Medlidsam         F11MDLID  
Eager  to  sooth  hurt   Angelägen  att  lindra     F12LINDR  
feelings           sårade  känslor  
Soft  spoken       Mild  och  stillsam       F13MILD    
Warm         Varm  person         F14VARM  
Tender         Ömsint           F15OMSIN  
Gullible         Lättlurad           F16LTLUR  
Childlike         Barnslig           F17BARNS  
Does  not  use  harsh     Använder  inte  svordomar   F18SVOR  
harsh  language  
Loves  children     Barnkär           F19BARKR  
Gentle         Varsam           F20VARSM  
 
 
 
 
 
 

12
Table   3.   Items   of   the   BSRI   Social   Desirability   Scale   (Bem,   1974;   1981).  
Translation   into   Swedish   included   as   well   as   their   variable   labels   by  
which  they  were  processed.  
 
English  original       Swedish  translation   Variable  labels  
 
Helpful           Hjälpsam           SD1HJALP  
Moody           Lynnig           SD2LYNN  
Conscientious         Samvetsgrann       SD3SMVTG  
Theatrical         Teatralisk         SD4TEATR  
Happy           Glad             SD5GLAD  
Unpredictable         Oförutsägbar         SD6OFRSB  
Reliable           Pålitlig           SD7PLIT  
Jealous           Svartsjuk           SD8SVART  
Truthful           Sanningsenlig         SD9SANN  
Secretive           Hemlighetsfull       SD10HEML  
Sincere           Uppriktig         SD11UPPR  
Conceited         Inbilsk           SD12INB  
Likeable           Trevlig           SD13TREV  
Solemn           Allvarlig           SD14ALLV  
Friendly           Vänlig           SD15VANL  
Inefficient         Ineffektiv         SD16INEF  
Adaptable         Anpassningsbar       SD17ANPA  
Unsystematic         Osystematisk         SD18OSYS  
Tactful           Finkänslig         SD19FINK  
Conventional         Konventionell         SD20KONV  
 
 
 
 
  During   construction   judges   used   a   seven-­‐point   Likert-­‐scale,  
ranging  from  one  (not  at  all  desirable)  to  seven  (extremely  desirable)  
when   rating   items.   Note   that   the   inventory   outlines   American  
stereotypes   of   gender   roles,   which   are   not   necessarily   transferable   to  
other  nations  and  cultures.  Bem  is  aware  of  the  fact  that  the  inventory  
outlines   feminine   and   masculine   traits   desirable   to   an   American  

13
population.   It   was   emphasised   during   the   construction   of   the  
instrument  that  participants  should  voice  what  they  believed  to  be  an  
opinion  general  to  American  society  rather  than  uniquely  their  own.  
  The  psychometric  properties  of  Bem’s  original  instrument  are  fair.  
Internal   consistency   was   evaluated   by   allowing   separate   samples   of  
men   and   women   provide   scores   for   both   the   feminine   and   masculine  
scales   respectively:   Both   scales   as   scored   by   women   show   that   the   F-­‐
scale   (α   =   .75)   as   well   as   the   M-­‐scale   (α   =   .78)   have   an   acceptable  
degree   of   internal   consistency.   Similarly,   both   scales   as   scored   by   the  
male   sample   show   an   acceptable   degree   of   consistency   as   well:   F-­‐scale  
(α  =  .87)  and  M-­‐scale  (α  =  .86).    
  Test-­‐retest   reliability   is   also   acceptable,   estimated   in   the   same  
way   with   men   and   women   providing   separate   scores.   Test-­‐retest  
reliability  ranges  from  r  =  .76  to  r  =    .94  (i.  e.  from  the  female  sample  
the   F-­‐scale   yields   r   =   .82   and   M-­‐scale   r   =   .94,   whereas   for   the   male  
sample  F-­‐scale  yields  r  =  .89  and  M-­‐scale  r  =  .76).      
  The   neutral   Social   Desirability   Scale   serves   the   purpose   of  
safeguarding   the   conceptualisation   of   masculinity   and   femininity   as  
separate  constructs,  but  serves  no  purpose  when  using  the  completed  
and   validated   inventory.   By   allowing   females   and   males   separately   to  
judge   both   the   F-­‐scale   and   the   M-­‐scale,   Bem   showed   that   sex-­‐
appropriate  behaviour—which  is  deemed  to  be  socially  desirable—as  
judged   by   either   sex,   was   significantly   higher   when   females   judged  
feminine  items  and  when  males  judged  masculine  items  in  comparison  
to   when   males   judged   feminine   items   and   women   judged   masculine  
item.   There   was,   however,   no   such   significant   difference   when   both  
sexes  judged  the  neutral  Social  Desirability  Scale.  This  prompted  Bem  

14
to   conclude   that   Femininity   and   Masculinity   are   indeed   separate  
constructs.   In   addition,   she   goes   on   to   show   that   both   the   participating  
men   and   women   are   in   significant   agreement   as   to   what   qualifies   as  
“sex-­‐appropriate”   behaviour   and   “sex-­‐inappropriate”   behaviour.   In  
other   words,   when   the   men   felt   that   a   certain   characteristic   is   sex-­‐
inappropriate   for   a   man   or   a   woman   the   female   participants   would  
tend   to   agree.   And   conversely,   when   women   felt   that   a   certain  
characteristic   is   sex-­‐inappropriate   for   a   man   or   a   woman   the  
participating  men  would  largely  concur.  
  Androgyny,   the   measure   of   which   is   the   ultimate   objective   of  
reconceptualising   masculinity   and   femininity,   reflects   the   relative  
levels   of   masculinity   and   femininity   that   any   individual   includes   in  
their   self-­‐description.   Androgyny   is   thus   expressed   as   an   index   that  
shows   the   extent   to   which   a   person   is   sex-­‐typed   on   the   basis   of   how  
they  score  on  the  M-­‐scale  and  F-­‐scale  together.  The  greater  the  value  of  
the   Androgyny   Index   (AI)   the   more   the   person   is   sex-­‐typed   or   sex-­‐
reversed.  The  smaller  the  value  of  AI  the  more  an  individual  could  be  
considered  androgynous.    
  Bem  has  proposed  three  ways  of  calculating  AI.  Originally  AI  was  
defined  as  Student’s  t  -­  ratio  for  the  difference  between  masculine  and  
feminine   self-­‐endorsement.   In   other   words   AI   is   the   difference  
between  an  individual’s  Masculinity  and  Femininity  scores  normalised  
with  respect  to  the  standard  deviations  of  the  scores  from  both  scales  
and   expressed   as   the   Student’s   t   -­   ratio.     In   this   line   of   reasoning  
individuals   are   classified   as   sex-­‐typed,   masculine   or   feminine   if   the  
androgyny  t    -­‐  ratio  reaches  statistical  significance  (|  t  |  ≥  2.2025,  df  =  
38,  p  <  .05).  An  individual  is  classified  as  androgynous  if  the  absolute  

15
value  of  the  t    -­‐  ratio  is  less  or  equal  to  one.  Bem  points  out  that  these  
cut-­‐off  points  are  somewhat  arbitrary  and  may  be  adjusted  if  need  be.      
  Bem   offers   a   simpler   way   of   assessing   the   AI.   If   multiplying   the  
difference   score   between   the   F-­‐scale   and   the   M-­‐scale   by   a   conversion  
factor  of  2.322  one  arrives  at  an  index  which  correlates  nigh  perfectly  
with   the   t   -­   ratio   calculation   (r   =   .98).   This   conversion   factor   was  
derived   from   the   original   American   normative   sample   (n   =   917),   but  
Bem   does   not   convey   by   what   means.   Note   that   this   “short-­‐cut”   was  
used   by,   for   example,   Hassler   (1991)   in   her   experimental   studies   of  
spatial   abilities,   musical   composition   ability   and   their   relationship   to  
the  Bem’s  construct  of  psychological  androgyny.  
  The   original   three-­‐way   classification   by   Bem   into   gender-­‐role   as  
masculine,   feminine   or   androgynous,   however,   was   criticised   by  
Spence,   Helmreich   and   Stapp   (1975).   They   argued   that   classification  
based   on   Student’s   t   -­   ratio   obscures   one   important   distinction   made  
possible   by   understanding   femininity   and   masculinity   as   separate  
constructs,  namely  the  fact  that  it  is  also  possible  to  score  high  on  both  
scales   or   low   on   both   scales   rather   than   preferring   one   to   the   other.  
Bem   (1977)   investigated   the   possibility   to   consider   a   four-­‐way  
classification,   and   found   that   the   critique   of   merely   a   three-­‐way  
classification  was  warranted.  At  a  later  stage  in  the  development  of  the  
inventory,   therefore,   as   the   BSRI   became   commercially   available,   the  
procedures   for   classification   slightly   changed.   Bem   (1981)   now  
recommends   that   classification   be   done   by   means   of   a   median   split  
classifying   individuals   into   four   categories   of   gender   role   orientation:  
Undifferentiated,  Androgynous,  Feminine  and  Masculine.  The  step-­‐by-­‐
step   procedure   by   which   to   apply   the   median   split   for   classification  

16
will  be  discussed  below  and  in  relation  to  the  Swedish  revision  of  the  
BSRI.  
  The  normative  samples  used  by  Bem  in  developing  the  instrument  
consists   of   approximately   900   individuals   made   up   of   the   following  
groups:   Caucasian   undergraduates   (n   =   32),   Afro-­‐American  
undergraduates   (n   =   63),   Hispanic   undergraduates   (n   =   35),  
Adolescents   aged   14   -­‐   17   (n   =   29),   Age   group   20   -­‐   30   (n   =   108),   Age  
group  31  -­‐  65  (n  =  55),  Older  adults  ages  >62  (n  =  60)  and  Psychiatric  
in-­‐patients  (n  =  55).  The  averaged  norms  for  these  groups  are  given  in  
Table  4  below.  
  The   distribution   of   categorisations   in   the   normative   sample  
according  to  Bem’s  (1974)  original  classification  scheme,  and  as  based  
on  Student’s  averaged  t  -­  ratios,  shows  that  amongst  males  7  %  are  sex-­‐
typed  as  feminine,  7  %  as  near-­‐feminine,  39  %  as  androgynous,  18  %  
as   near-­‐masculine   and   29   %   as   masculine.   Amongst   females   on   the  
other  hand,  37  %  are  sex-­‐typed  as  feminine,  14  %  as  near-­‐feminine,  32  
%  as  androgynous,  9  %  as  near-­‐masculine  and  8  %  as  masculine.    
 
 
 
 
 
 
 
 
 
 
 
 
 
 

17
Table   4.   Raw   score   means,   medians   and   standard   deviations   for   the   F-­
scale,   M-­scale   and   the   F-­minus-­M   difference   score   of   the   normative  
sample.   Note   that   the   category   sexes   combined   has   been   statistically  
weighted   in   order   to   correct   the   unequal   numbers   of   men   and   women  
making  up  the  sample  (Bem,  1981)  
 
 
              Females        Males    
Sexes  combined   (n  =  340)     (n  =  476)     t    
 
 
Femininity  
  Mean     4.82       5.05       4.59       11.95*  
  MD       4.90       5.10       4.60  
  SD          .59          .53          .55  
Masculinity  
  Mean     4.95       4.79       5.12        7.03*  
  MD       4.95       4.80       5.10  
  SD          .68          .66            .65  
F-­minus-­M  
  Mean     -­‐  .01          6.30     -­‐6.33     13.09*  
  MD            .97          6.83     -­‐6.50  
  SD       14.94     13.35     13.37  
 
*  p  <  .001  
 
Table  5.  Distribution  of  subjects  in  the  combined  normative  samples  into  
sex-­role   categories   according   to   the   four-­way   classification   using   the  
median  split  method    (Bem,  1974)  
 
 
                Females       Males  
Sex-­role  category           (%)         (%)  
 
 
Feminine               39         12  
Masculine             12         42  
Androgynous             30         20  
Undifferentiated           18         27  
 

18
   Distribution   of   categories   in   the   same   sample   but   as   based   on  
median  splits  are  accounted  for  in  Table  5.    
  It   is   possible   to   evaluate   how   the   two   classification   schemes  
compare   if   the   values   of   near-­‐feminine   and   near-­‐masculine   are  
combined  and  regarded—for  the  purpose  of  comparison—as  more  or  
less  the  same  as  being  undifferentiated.  In  the  male  sample  the  t  -­  ratio  
classification   corresponds   only   moderately   to   the   median   split  
classification   (r   =   0.45),   whereas   in   the   female   sample   there   is   little  
difference  between  two  schemes  (r  =  0.949).  
 
 
The  BSRI  in  a  Swedish  context  
 
The  nature  of  the  Swedish  sample  
The  Swedish  sample  (n  =  118)  is  relatively  small  in  comparison  to  the  
normative  sample  used  by  Bem.  The  objective  of  the  present  research,  
however,   is   mainly   to   investigate   how   the   BSRI   fares   in   a   Swedish  
context   rather   than   to   strictly   standardise   the   instrument   anew   in   a  
fully  fledged  large-­‐scale  effort.  On  the  other  hand,  Swedish  norms  will  
be  produced  and  the  present  data  may  indicate  the  psychometric  value  
of   the   instrument   as   such.   At   the   very   least   this   evaluation   may  
certainly   provide   a   base   of   data,   which   can   be   used   to   probe  
psychometric  properties  further.  For  this  purpose  the  raw  mean  scores  
are  appended  to  this  report  (Appendix  II).  A  breakdown  of  the  Swedish  
is   provided   in   Table   6.   Note   that   there   is   an   over-­‐representation   of  
females  in  the  sample  (41  %  males  and  59  %  females).  This  will  need  
to  be  considered  as  the  normative  scores  are  produced.  The  same  over-­‐

19
representation   problem   is   evident   in   almost   exactly   the   same  
proportions   in   Sandra   Bem’s   normative   sample.   However,   in   Bem’s  
sample  over-­‐representation  is  reversed  (42  %  are  female  and  58  %  are  
male).  
  The   original   American   BSRI   from   1974   was   translated   into  
Swedish   and   the   administered   answering   sheet   included   all   three  
scales   (see   Tables   1   -­‐   3   above   for   a   comparison   between   the   original  
American  items  and  their  translation  into  Swedish).  A  majority  of  the  
respondents   filled   in   the   translated   during   class   and   always   because  
course   content   qualified   or   included   a   discussion   of   gender-­‐roles   or  
socialisation.   The   inventory   was   always   completed   prior   to   any  
discussion   of   gender-­‐roles,   however,   and   respondents   were   only  
provided  with  the  information  needed  complete  the  task.    
  Scale  items  on  the  scoring  sheets  were  compiled  in  such  a  manner  
that  order  effects  were  avoided.  It  was  not  possible  for  participants  to  
perceive  that  in  fact  three  scales  rather  than  one  were  being  rated.  The  
order   of   the   items   was   set   in   groups   of   three   in   the   following   order:  
Masculine,   Feminine   and   Neutral,   which   is   also   the   order   used   by   Bem.  
Note   that   Swedish   participants   were   encouraged   to   rate   themselves,  
which   is   contrary   to   Bem’s   instructions   to   the   American   sample.   Bem  
asked   that   participants   should   express   what   they   felt   was   typically  
American   in   a   general   sense   rather   than   using   the   BSRI   to   provide   a  
personal  profile.  
 
 
 
 
 
 
 

20
Table   6.   The   different   subgroups   of   the   Swedish   test   sample.   Note   that   the  
percentages  pertaining  to  each  sex  is  the  relative  frequency  of  the  subgroup,  whereas  
the  relative  frequency  of  n  relates  to  N.  
 
 
                           Age  
                  ____________________________  
 
Subgroup           n   M   F   Mean   Min   Max   SD  
 
 
Pre-­‐school  student  teachers   24   4    20   23.3     19   40   5.01  
  Relative  frequency  (%)     (20)   (17)   (83)      
 
Communication  Studies  students  49   11    38   24.7     20   44   5.21  
  Relative  frequency  (%)     (42)   (22)   (78)  
 
Amateur  football  players     12   12    -­‐-­‐   39.4     25   54   9.1  
  Relative  frequency  (%)     (10)   (100)  -­‐-­‐  
 
Pre-­‐school  teachers       16   4    12   48.9     24   61   9.55  
  Relative  frequency  (%)     (14)   (25)   (75)  
 
Comprehensive  student  teachers  12   12    -­‐-­‐   36.0     21   51   10.45  
  Relative  frequency  (%)     (10)   (100)  -­‐-­‐  
 
Various  university  staff     5   5    -­‐-­‐   43.0     18   60   16.07  
  Relative  frequency  (%)     (4)   (100)    -­‐-­‐      
 
All  subgroups  (N)       118   48      70   31.2     18   6   112.00  
  Relative  frequency  (%)     (100)  (41)  (59)  
 
 
 
  All   the   Swedish   participants   completed   the   answering   sheets  
impeccably.   At   all   occasions   students   were   provided   with   immediate  
feedback  as  everyone  had  completed  the  task.  That  is,  after  completion  
the   students   themselves   were   allowed   to   score   their   sheets   after   being  
informed   which   items   belonged   to   which   scale.   Scoring   was  
straightforward   and   students   were   able   to   estimate   themselves  
whether   they   were   high   -­‐   high,   low   -­‐   low   or   high   -­‐   low   on   either   the   M-­‐

21
scale   or   F-­‐scale.   The   neutral   Social   Desirability   Scale   was   ignored.  
Participants  only  had  access  to  their  own  score  so  as  to  keep  the  ethical  
integrity  of  the  testing  situation.  No  further  conclusions  were  drawn  by  
students   except   for   their   relative   standing   in   comparison   to   the  
American-­‐elicited  items.    
  However,   the   administration   tended   to   trigger   interesting  
discussions  on  gender  and  the  face  validity  of  the  instrument.  In  some  
cases,   particularly   with   regard   to   the   Communication   Studies   students,  
the   inventory   caused   some   merriment   in   the   class.   They   argued   that  
items   were   “out-­‐dated”   and   not   at   all   appropriate   for   outlining  
masculine  and  feminine  stereotypical  behaviour  in  Sweden.    
  In   evaluating   the   BSRI   in   the   following   estimations   are   made   on  
the   basis   of   three   types   of   scores:   judgements   by   the   men   of   the  
sample,  judgements  made  by  the  women  of  the  sample  and  both  men  
and   women   combined.   This   is   also   the   strategy   chosen   by   Bem   for  
producing  the  original  norms  for  BSRI.  
 
Reliability  of  the  inventory  scales  
Cronbach’s  Coefficient  α  was  used  as  a  measure  of  internal  consistency  
also  for  the  translated  version  of  BSRI  (Table  7).  It  would  appear  that  
reliability   from   this   perspective   largely   concurs   with   that   of   Bem’s  
normative  sample  with  one  exception.  The  Social  Desirability  Scale  as  
judged   by   the   Swedish   sample   shows   poor   consistency   (α   =   .56   for   the  
combined   sample),   whereas   the   same   scale   in   Bem’s   original   1973  
sample   is   acceptable   (α   =   .70   for   the   combined   Stanford   sample).  
However,   the   SD-­‐scale   is   nevertheless   less   reliable   than   the   M-­‐scale  
and   the   F-­‐scale   also   in   the   Bem   normative   sample.   After   criticism   from  

22
Walkup   and   Abbot   (1978),   Bem   proposed   that   the   Social   Desirability  
Scale  should  no  longer  play  a  part  in  the  inventory.  The  items  are  still  
included  in  the  inventory  but  serve  only  the  purpose  of  being  “fillers”;  
providing   a   general   environment.   They   are   rated   by   respondents   but  
not   considered   by   the   research   or   clinician   making   use   of   the  
inventory.   The   Social   Desirability   Scale   has   been   removed   completely  
in  the  Swedish  revision  of  BSRI.  
  The   fact   that   the   SD-­‐scale   is   of   questionable   reliability   in   the  
Swedish  setting  strongly  suggests  that  it  should  not  be  used  to  qualify  
or   verify   the   independence   of   androgyny   as   a   construct,   which   was  
originally   done   by   Bem   in   the   normative   sample.   She   showed   that  
androgyny   was   nearly   uncorrelated   with   androgyny   as   expressed   by  
Student’s  t  -­  ratio  (that  is  r  =  .03  for  males  and  r  =  -­  .10  for  females).  
 
 
Table   7.   Coefficient   alpha   for   both   the   Swedish   sample   and   Bem’s   1978   Stanford  
sample  
 
 
Scale             Males     Females     Combined  
 
 
Masculinity  
  Swedish  sample       .80       .82        .82  
  Stanford  sample       .86       .87        .86*  
   
Femininity  
  Swedish  sample       .84       .74        .81  
  Stanford  sample       .78       .78        .80*  
         
Social  Desirability  
  Swedish  sample       .63       .47       0.56  
  Stanford  sample       n/a       n/a          .70*  
 
*  These  values  are  from  Bem’s  1973  Stanford  sample  

23
 Either   an   indigenously   derived   scale   of   Social   Desirability   for   a  
Swedish   context   be   constructed   to   replicate   Bem’s   construct  
procedure,   or   the   averaged   F-­‐scale   and   M-­‐scale   means   are   correlated  
with   each   other   to   estimate   the   degree   to   which   they   correspond.   A  
high   degree   of   correspondence   would   suggest   that   Masculinity   and  
Femininity   as   separate   constructs   is   not   a   successful   one,   whereas   a  
low  degree  of  correspondence  would  suggest  the  opposite.  Bem  offers  
this  reliability  check  also  in  the  commercially  published  version  of  the  
inventory   and   thus   shows   that   the   two   remains   uncorrelated   (Females  
r  =  .00  and  Males  r  =  -­‐.05).      
  However,   this   is   not   the   case   for   the   Swedish   sample.   Whilst  
correspondence   between   the   scales   as   judged   by   the   female  
participants  is  low  (r  =  .16)  the  correspondence  as  judged  by  the  male  
participants  is  moderate  (r  =  .41).  The  same  procedure  applied  to  the  
total  Swedish  sample,  including  both  men  and  women,  also  signifies  a  
degree  of  correspondence  although  weak  (r  =  .30).    
  These  results  suggest  that  the  use  of  BSRI  in  a  Swedish  setting  in  
merely   a   translated   form   is   not   straightforward   in   spite   of   acceptable  
alpha   values   for   both   scales.   There   appears   to   be   some   overlap  
between   the   Masculinity   Scale   and   the   Femininity   Scale,   which   are  
likely  due  to  cultural  differences.  This  conceptual  discrepancy  between  
the  samples  requires  closer  scrutiny.      
  Performing   an   inter-­‐item   total   correlation   analysis   shows   that   a  
number  of  variables  are  either  weakly  correlated  to  the  scales  or  that  
men   and   women   strongly   disagree   on   certain   items   as   being   sex-­‐
appropriate   (Table   8).   For   example,   it   appears   that   being   athletic  
(M4SPORT),   analytical   (M8ANALYT),   competitive   (M19TAVL)   or  

24
ambitious  (M20AMBI)  is  considered  by  both  men  and  women  not  to  be  
typically  masculine.  Similarly  all  participants,  irrespective  of  sex,  seem  
to  agree  that  yielding  (F1UNDFAL),    
 
Table   8.   Inter-­item   total   correlation   as   based   on   female   and   male   participant   scores  
separately  and  combined.  To  decipher  item  labels  see  Table  1  &  2  above.  Items  which  
either  correlate  weakly  to  the  scale  or  sample  to  which  they  apply,  or  items  regarding  
which  men  and  women  appear  to  disagree  strongly,  are  boldened  and  enlarged.    
 
 
Scale  items  (M)     M   F   Comb   Scale  items  (F)   M   F   Comb  
 
 
M1SJLVFR       .51   .50   .52     F1UNDFAL   .17   -­.01  .06  
M2FORSV       .50   .54   .53     F2MUNTR   .23   .17   .25  
M3OBERO     .13   .61   .40     F3BLYG     .22   .21   .16  
M4SPORT     .01   -­.10  -­.03     F4TILLGV     .64   .48   .59  
M5BESTMD     .65   .52   .56     F5SMICKR   .31   .31   .25  
M6STPER       .63   .34   .43     F6LOJAL     .45   .32   .39  
M7KRFTFL     .53   .54   .56     F7FEMINI     .36   .51   .49  
M8ANALYT     .15   .32   .27     F8SYMPAT   .57   .40   .50  
M9LEDARE       .39   .58   .59     F9KNSLBH   .64   .39   .56  
M10RISK       .33   .53   .46     F10FORST     .48   .38   .48  
M11BSLUT     .42   .39   .44     F11MDLID  .61   .53   .58  
M12SJLVT       .53   .31   .35     F12LINDR     55   .42   .52  
M13DOMIN     .58   .56   .55     F13MILD     .41   .46   .41  
M14MASK       .39   .34   .38     F14VARM     .51   .54   .57  
M15BSLST     .39   .58   .53     F15OMSIN  .73   .61   .70  
M16AGGRS     .37   .09   .21     F16LTLUR   .40   -­.05  .17  
M17LEDAR     .39   .48   .48     F17BARNS   .25   .07   .15  
M18INDIV     .14   .55   .34     F18SVOR   .12   .22   .20  
M19TAVL     .21   .25   .29     F19BARKR   .41   .23   .38  
M20AMBI     .32   .23   .21     F20VARSM   .72   .42   .59  
 
 
 
cheerful   (F2MUNTR),   shy   (F3BLYG),   flatterable   (F5SMICKR)   and  
childlike   (F17BARNS)   do   not   describe   characteristics   that   are   typically  
feminine,  nor  is  use  or  no  use  of  harsh  language  (F18SVOR)  a  variable  

25
appropriate   to   describe   femininity.   Observe,   however,   that   men  
disagrees  with  women  on  the  status  of  being  independent  (M3OBERO),  
aggressive   (M16AGGRS)   and   individualistic   (M18INDIV).   The   women  
suggest   that   independence   and   individualism   are   typical   male  
characteristics  whereas  men  do  not.  On  the  other  hand,  men  feel  that  
aggression   to   some   degree   is   typically   male   while   women   appear   to  
disagree.   In   a   like   manner   men   and   women   do   not   agree   on   whether  
women   are   typically   gullible   (F16LTLUR).   Men   say   they   are   while  
women   object.   Also,   men   seem   to   think,   to   a   higher   degree   than   do  
women,  that  it  is  typically  feminine  to  love  children  (F19BARKR).  
  In   order   to   come   to   terms   with   the   response   patterns   of   the  
Swedish   sample—deviant   in   comparison   to   the   American   normative  
sample—and  increase  the  integrity  of  the  two  scales,  seven  items  were  
removed  from  the  Masculinity  Scale  (M3,  M4,  M8,  M16,  M18,  M19  and  
M20)   and   eight   items   from   the   Femininity   Scale   (F1,   F2,   F3,   F5,   F16,  
F17,   F18,   F19).   A   new   reliability   analysis   was   performed   on   the  
resulting  shortened  version  of  the  translated  inventory.    
  This   procedure   increased   internal   consistency   as   expressed   by  
Coefficent  Alpha  for  both  scales  and  with  regard  to  females,  males  and  
men   and   women   combined   (Table   9).   Inter-­‐item   correlations   have   also  
become   more   robust   (Table   10).   To   further   confirm   the   conceptual  
basis   of   the   BSRI   in   a   Swedish   setting   a   factor   analysis   (principal  
components   with   varimax   rotation)   was   performed   on   the   combined  
and   revised   scales.   A   first   run   revealed   a   solution   of   seven   factors  
explaining   66.5   %   of   total   variance.   However,   in   this   solution   two  
factors   dominate   and   together   explain   41.4   %   of   total   variance.   The  
variance   of   the   remaining   five   factors   ranges   from   4.0   to   7.3   %.   In   a  

26
second  run  the  principal  component  analysis  was  restricted  to  a  two-­‐
factor   solution   only,   which   would   seem   to   corroborate   the   existence   of  
two  separate  constructs  (Table  11)  explained  by  approximately  equal  
proportions   of   the   total   variance   (22.0   %   and   19.3   %   respectively).  
However,   there   is   still   some   overlap   between   the   two   scales.  
Masculinity   (M14MASK)   loads   moderately   on   both   factors   and   the  
same  is  true  of  Soft-­‐spoken  (F13MILD).      
 
Table  9.  Coefficient  alpha  for  both  the  unrevised  and  the  revised  Swedish  version  of  
BSRI  
 
 
        Masculinity         Femininity  
        ______________________       _______________________  
 
Sample       Unrevised   Revised   t     Unrevised   Revised   t  
 
 
Males       .80     .85     33.0*     .84     .87     57.0*  
Females       .82     .84     83.0*     .74     .83     17.4  
Combined       .82     .84     83.0*     .81     .86     33.4*  
 
*  p<  .01  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

27
Table   10.   Inter-­item   total   correlation   as   based   on   the   corrected   Femininity   and  
Masculinity  Scales.  To  decipher  item  labels  see  Table  1  &  2  above  
 
 
Scale  items  (M)  M   F   Combined     Scale  items  (F)   M   F   Combined  
 
 
M1SJLVFR     .43   .54   .51       F4TILLGV     .59   .58   .60    
M2FORSV     .51   .46   .48       F6LOJAL     .43   .34   .39  
M5BESTMD   .66   .52   .57       F7FEMINI     .39   .44   .49  
M6STPER     .64   .40   .47       F8SYMPAT   .58   .46   .52  
M7KRFTFL   .52   .56   .57       F9KNSLBH   .65   .56   .65  
M9LEDARE     .59   .63   .64       F10FORST     .58   .49   .57  
M10RISK     .40   .47   .44       F11MDLID     .61   .57   .59  
M11BSLUT   .47   .45   .49       F12LINDR     .65   .46   .58  
M12SJLVT     .48   .30   .32       F13MILD     .32   .37   .32  
M13DOMIN   .58   .58   .56       F14VARM     .55   .58   .61  
M14MASK     .42   .30   .37       F15OMSIN     .77   .71   .76  
M15BSLST     .43   .54   .51       F20VARSM   .73   .47   .61  
M17LEDAR   .43   .53   .52  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

28
  
 
Table  11.   Factor  analysis  (principal  components,  varimax  rotation)  of  the  revised  M  
and  F-­scales  combined.  Note  that  the  cut-­off  point  is  set  to  >.35.  Loadings  included  on  
this  criterion  are  boldened  and  enlarged  in  the  table.  
 
 
      Factors               Factors  
      _________________           __________________  
 
Items     I     II       Items     I     II  
 
 
M1SJLVFR      .08     .58       F4TILLGV     .70      .02    
M2FORSV      .14     .58       F6LOJAL     .48      .19  
M5BESTMD    .11     .66       F7FEMINI     .62     -­‐.17  
M6STPER      .21     .60       F8SYMPAT   .61      .12  
M7KRFTFL    .01     .68       F9KNSLBH   .74      .00  
M9LEDARE     -­‐.08     .75       F10FORST     .65      .03  
M10RISK      .04     .53       F11MDLID     .67     -­‐.08  
M11BSLUT   -­‐.22     .55       F12LINDR     .68     -­‐.06  
M12SJLVT      .14     .41       F13MILD     .37     -­.39  
M13DOMIN   -­‐.19     .66       F14VARM     .71      .19  
M14MASK     -­.40     .44       F15OMSIN     .82      .03  
M15BSLST     -­‐.06     .56       F20VARSM   .67     -­‐.10  
M17LEDAR      .00     .64    
 
 
 
 
  Cut-­‐off   point   for   factor   loadings   were   set   to   >.35,   which   is   in  
accordance  with  Bem’s  (1981)  factor  analysis  of  the  original  BSRI  in  an  
effort  to  arrive  at  a  shorter  version  of  the  inventory.  This  cut-­‐off  point  
is   more   stringent   than   recommended   for   the   current   number   of  
variables   and   sample   size.   Child   (1990),   for   example,   recommends   a  
cut-­‐off  point  of  >.20  at  p<  .05  and  >.26  at  p<  .01  when  N  =  100  and  the  
analysis  involves  no  more  than  30  variables  (see  also  Burt,  1952).  
 

29
  
  In  summary,  it  seems  not  advisable  to  use  the  BSRI  in  it’s  original  
American   rendition   in   a   Swedish   context,   since   it   contains   items  
describing   gender   characteristics   that   appear   not   similarly   understood  
in  Sweden.  However,  in  view  of  the  fact  the  integrity  of  the  inventory  
seems   to   hold   in   the   revised   version,   this   could   be   used   in   a   Swedish  
research   setting   where   it   is   desirable   to   compare   gender   to   other  
variables.   In   so   doing,   however,   one   must   bear   in   mind   that   the  
Swedish   revised   version   (which   will   be   termed   BSRI-­‐SE   in   the  
following)   only   accounts   for   41.4   %   of   total   variance,   which   suggests  
there  is  more  to  the  understanding  of  gender  roles  in  Sweden  than  is  
expressed   by   the   BSRI.   Ideally   a   new   item   pool   should   be   derived   from  
a  larger  and  more  representative  Swedish  sample  than  has  been  used  
for   this   study,   and   a   new   and   entirely   indigenous   set   of   scales   be  
constructed.  
 
Procedure  for  classification  and  standardisation  
The   classification   procedure   suggested   by   Bem   (1981)   to   replace   the  
dismissed  Student’s  t  -­‐  ratio  as  a  basis  for  deciding  to  what  category  a  
respondent   might   belong   is   to   some   extent   straight-­‐forward.   The  
median   split   method   is   simple   and   quick   given   there   are   norms  
available   by   which   to   check   arrived-­‐at   scores.   The   estimation   of   the  
strength   of   sex-­‐typing,   however,   is   more   complex   and   the   BSRI   manual  
is   somewhat   paradoxical   and   far   from   straight   forward   in   explaining  
how   the   four-­‐way   classification   procedure   relates   to   estimating   the  
degree   to   which   a   person   may   be   regarded   as   typically   masculine,  
feminine,   androgynous   or   undifferentiated.   For   the   standardising  

30
process   of   the   SE-­‐version   of   BSRI   the   following   procedures   were  
employed,   which   are   based   on   Bem’s   recommendation   as   far   as   the  
four-­‐way   classification   by   means   of   median   splits   go.   The   use   and  
relevance   of   standard   scores,   however,   differs   from   what   Bem  
suggests.   This   will   be   discussed   separately   below.   The   procedure  
below   describes   the   classification   process   with   reference   to   already  
obtained  norms.  
 
The  median  split  method  
  As   answering   sheets   have   been   collected,   coded   and   fed   into   a  
suitable  computer  statistics  package  (such  as  the  SPSS  8.0,  which  was  
used   for   this   study),   the   first   step   is   to   compute   each   individual’s  
arithmetic   average   raw   score   on   both   the   Femininity   Scale   and   the  
Masculinity  Scale.  
  These  average  scores  are  then  compared  to  the   median  scores  of  
both   scales   arrived   at   by   finding   the   average   medians   for   the   whole  
sample   tested.   The   medians   for   the   Swedish   sample   are   4.92   for   the  
Masculinity   Scale   and   5.19   for   the   Femininity   Scale   (see   Table   12  
below).   These   are   the   normative   medians   by   which   an   individual’s  
score   is   split   from   the   total   sample   into   a   particular   category.   For  
example,   if   one   respondent’s   averaged   raw   scores   on   the   F-­‐scale   and  
the   M-­‐scale   is   4.69   and   4.75   respectively.   He   or   she   would   be   classified  
as  undifferentiated  because  4.69  is  below  the  normative  median  for  the  
F-­‐scale  (i.e.  below  4.92)  and  4.75  is  also  below  the  normative  median  
for  the  M-­‐scale  (i.e.  below  5.19).  Similarly,  if  an  individual  on  average  
has  scored  3.85  of  the  F-­‐scale  and  6.08  on  the  M-­‐scale,  he  or  she  should  
be  classified  as  masculine,  since  the  F-­‐score  falls  below  the  norm  and  

31
the  M-­‐score  falls  above  the  norm  (see  Figure  1).  If  a  score  falls  on  the  
norm  precisely  it  is  best  regarded  as  falling  above  the  norm.  
  Note  that  depending  on  context  and  purpose  of  research  it  might  
be   advisable   to   create   other   norms   specific   for   the   sample   or  
population   involved   in   the   research.   This   is   also   Bem’s   (1981)  
recommendation.  

____________________________________________________________

MASCULINITY SCORE

Below median Above median

Undifferentiated Masculine Below median

(low F - low M) (low F - high M) FEMININITY SCORE
Feminine Androgynous Above median
(high F - low M) (high F - high M)

______________________________________________________________________
Figure  1.  The  BSRI  four-­way  classification  as  suggested  by  Bem  (1981).  

 
  Note   that   Bem   recommends   a   slightly   different   procedure.   She  
recommends   that   the   averaged   individual   scores   from   both   scales   be  
standardised   into   T-­   scores   (that   is,   first   obtain   z   -­   scores,   which   is  
easily  done  with  for  example  SPSS.  Then  multiply  the  z  -­  score  with  10  
and  add  50.  The  formula  for  converting  z  into  T  is  simple:  T  =  z  (10)  +  
50.   The   standardised   T   -­‐   scale   has   a   mean   of   50   and   a   standard  
deviation  of  10).  However,  Bem  then  suggests  that  the  M-­‐F  Difference  
Score   be   calculated   from   the   standardised   F-­‐score   and   M-­‐score,   and  
that  this  difference  score  in  turn  is  standardised  in  the  same  way.  This  
spells   problems,   however,   and   is   discussed   in   the   following.   The   M-­‐F  
Difference  Score  (i.e.  the  result  of  subtracting  the  averaged  masculine  

32
score   from   the   averaged   feminine   score)   is   important   and   should   be  
accompany  the  classification.  It  tells  the  strength  and  direction  of  sex-­‐
typing.  
 
The  M-­F  Difference  score  
  The   median   split   categorises   a   subject   only   with   reference   to   the  
norms   of   a   certain   sample.   It   does   not   consider   the   difference   between  
an   individual’s   scores   on   the   M-­‐scale   and   the   F-­‐scale.   The   result   is  
inevitably  that  some  individuals  are  classified  as  masculine  or  feminine  
in  spite  of  the  fact  that  their  scores  are  fairly  equal.  Conversely,  some  
may  be  categorised  as  androgynous  or  undifferentiated  although  they  
score  very  differently  on  masculinity  and  femininity.  This  is  a  problem  
for   individuals   who   score   near   the   cut-­‐off   points   (i.e.   the   median  
splits).   In   spite   of   this,   the   median   split   is   the   method   that   Sandra   Bem  
recommends  for  most  purposes  because  it  is  a  fairly  straight-­‐  forward  
and  simple  procedure  if  norms  are  available.    
  However,   M-­‐F   Difference   score   also   has   bearing   on   classification  
and   could   perhaps   at   times   serve   as   a   more   precise   basis   for  
classification   if   appropriate   ratios   were   decided   within   which   scores  
may  be  classified.  While  Bem  (1981)  dismisses  the  Student  t  -­  ratio  as  
the   basis   of   classification   replacing   it   by   the   median   split,   the  
importance   of   a   ratio   remains   in   the   classification   procedure  
nevertheless.    
  Bem  proposes,  that  every  score  should  be  standardised—and  Bem  
favours  T  -­  scores—and  that  the  M-­‐F  Difference  score  be  calculated  and  
standardised  as  T  -­‐  scores  also.    This  makes  little  sense,  however,  since  
Bem  is  keen  on  pointing  out  the  importance  of  sign  (plus  or  minus)  of  

33
the   M-­‐F   Difference   score.   “High   scores   in   either   direction”,   Bem   argues  
(1981),   “indicate   a   tendency   to   be   strongly   sex-­‐typed   (or   sex-­‐
reversed),   positive   scores   indicate   femininity   and   negative   scores  
indicate  masculinity”  (p.  7).  In  the  light  of  this  statement  observe  that  
the   whole   point   of   applying   a   T   -­‐   transformation   is   to   avoid   negative  
scores  (Cohen,  Swerdlik  &  Phillips,  1995).  The  problem  with  the  BSRI  
Manual   recommendation   on   how   to   deal   with   the   M-­‐F   Difference  
scores  therefore  becomes  twofold:    
  First,   if   the   difference   between   the   two   already   standardised  
scores   is   subject   to   a   T-­‐score   transformation   also   the   values   become  
inordinately   large.   In   a   sense   the   procedure   entails   standardising   the  
standardisation.    
  Second,   the   T   -­   score   transformation   also   removes   the   negative  
signs,  which  makes  void  the  notion  of  a  ratio  in  which  the  sign  of  any  
value  provides  a  clue  as  to  its  direction.  Such  a  ratio  could  be  expressed  
on   the   basis   of   T   -­‐   scores   also,   of   course,   but   the   immediate   appeal   of   a  
positive  or  negative  sign  signifying  direction  is  lost.  In  other  words,  it  is  
easier   to   understand   the   significance   of   the   M-­‐F   Difference   score   if   this  
quality  is  retained.    
  For  the  purpose  of  this  study,  therefore,  the  M-­‐F  Difference  score  
was  calculated  from  the  raw  scores  of  the  F-­‐scale  and  the  M-­‐scale,  and  
the   resulting   difference   was   transformed   into   z   -­   scores   rather   than   T   -­  
scores   (Figure   2).   This   transformation   allows   for   both   negative   and  
positive   signs,   since   the   mean   of   z   -­   scale   is   0.00   and   its   standard  
deviation   1.00.   For   anyone   desiring   to   try   other   transformations   the  
raw  data  for  BSRI-­‐SE  are  appended  to  this  research  report  (Appendix  
II).  

34
 16

14

12

10

4
Std. Dev = 1,00
2
Mean = 0,00
0 N = 118,00
-2,75 -2,25 -1,75 -1,25 -,75 -,25 ,25 ,75 1,25 1,75 2,25 2,75 3,25

Figure  2.  The  distribution  of  z  -­  standardised  M-­F  Difference  scores  as  compared  to  
normal  distribution.  

 
  The   value   of   each   standardised   M-­‐F   Difference   score   tells   not   only  
how   the   individual   relates   to   the   normative   sample   but   retaining   the  
negative   and   positive   signs   through   using   z   -­   scores   also   says  
something   about   the   strength   of   sex-­‐typing.   A   large   negative  
discrepancy  between  the  M  and  F  scores  signifies  that  an  individual  is  
strongly   sex-­‐typed   in   a   masculine   direction,   whereas   a   large   positive  
difference  means  that  an  individual  is  strongly  sex-­‐typed  in  a  feminine  
direction.   Small   differences   indicate   that   a   person   is   either  
undifferentiated  (i.e.  low  mean  scores  on  both  scales)  or  androgynous  
(i.e.   high   mean   scores   on   both   scales).   For   classifying   any   individual  
according   to   the   BSRI   or   the   BSRI-­‐SE   a   standard   score   indicating   the  
distribution   and   strength   of   sex-­‐typing   should   be   given   together   with  
the  classification.  

35
 Are  Swedes  masculine,  feminine,  androgynous  or  
undifferentiated?  
  An   interesting   aspect   of   evaluating   the   BSRI-­‐SE   on   a   Swedish  
sample   is   how   Swedes   fare   in   terms   of   classification   distribution   as  
compared   to   Bem’s   normative   sample.   The   revised   procedure   of  
classifying   participants   on   the   basis   of   a   median   split—as   outlined  
above—was  followed,  and  the  average  medians  of  the  Swedish  sample  
were   employed   as   norms   (Table   12).   Observe   that   these   normative  
values   have   been   weighted   in   order   to   compensate   for   the   lesser  
number  of  men  in  the  sample.1      
  Differences   were   observed   between   the   two   samples   (Table   13),  
and  note  the  clerical  error  in  the  American  sample:  it  is  made  up  of  101  
%  males  and  99  percent  femals  [sic!].  However,  consider  the  fact  that  
in  both  samples  roughly  half  fall  in  either  a  category  which  is  not  sex-­‐
typed   (undifferentiated   or   androgynous)   or   a   category   which   is   sex-­‐
typed  as  either  masculine  or  feminine.  
 
 
 
 
 
 
 
 
 
 
 
 

1   For   future   users   of   the   BSRI-­‐SE   it   might   be   useful   to   have   the   weighting   equation  

readily  at  hand  for    calculating  weighted  means    by  hand  in  a  smaller  sample:  The  
weighted   X   =   ΣXF   +   ΣXM/nF   +   nm   that   is,   the   number   of   males   multiplied   by   the  
means     for   males.   The   same   procedure   is   performed   for   the   female   sample.   The  
resulting  sums  are  added  and  divided  by  the  total  number    of  the  sample  (i.e.  males  
and  females  combined).  

36
  
Table  12.  The  male  and  female  average  judgments  of  the  Masculinity  and  Femininity  
scales   as   based   on   the   revised   BSRI.   Values   for   females   and   males   combined   have  
been  weighted  to  compensate  for  the  difference  in  numbers  between  the  two  sample  
groups.   Boldened   and   enlarged   values   in   the   table   serve   as   norms   for   classifying  
participants  in  the  Swedish  sample  according  to  a  median  split.  
 
 
      Masculinity     Femininity   Difference  (M  -­  F  scale)  
      ______________     _______________   __________________  
   
Sample     M   MD   SD     M   MD   SD     M   MD   SD  
 
 
Males     5.07   5.15   1.26     4.71   4.79   1.31     .36   .36   -­‐.05  
Females     4.60   4.77   1.33     5.32   5.46   1.17     -­‐.72   -­‐.69   .16  
Combined     4.79   4.92  1.30     5.07   5.19  1.23     -­‐.28   -­‐.26   .08  
 
 
 
 
Table   13.   Classification   distribution   in   samples   (in   percentages)   for   Bem’s   original  
BSRI  and  the  revised  BSRI-­SE.  Cp  denotes  cumulative  percentage.  
 
 
          Original  BSRI  sample   Revised  BRSI-­SE  sample  
          __________________       _______________________  
 
Sex-­type  classes     M   Cp   F   Cp     M   Cp   F   Cp    
 
 
Undifferentiated       27   27   18   18     31   31   27   27  
Androgynous       20   47   30   48     21   52   20   47  
Masculine         42   89   12   60     40   92   17   64  
Feminine         12   101   39   99     8   100   36   100  
 
 
 
 
 
 
 
 
 
 
 

37
Table  14.  Sex-­typing  in  the  different  subgroups  of  the  Swedish  test  sample.  
 
 
Sample  subgroup       n     Undiff   Andr     Masc     Fem  
 
 
Pre-­school    student  teachers      
      Males     4     1     3     0     1  
      Females     20     3     4     1     11  
      n       24     4     7     1     12  
 
      Percentage       17     29     4     50  
 
Communication  Studies  students  
      Males     11     4     1     5     1  
      Females     38     11     7     9     11  
      n       49     15     8     14     12
     
      Percentage       31     16     29     24
   
   
Amateur  football  players    
      Males     12     7     1     3     1  
      n       12     7     1     3     1  
 
      Percentage       59     8     25     8  
     
Pre-­school  teachers  
      Males     4     2     0     2     0  
      Females     12     4     3     2     3  
      n       16     6     3     4     3  
 
      Percentage       37     19     25     19  
               
Comprehensive  student  teachers    
 
      Males     12     0     4     7     1  
      n       12     0     4     7     1  
 
      Percentage       0     33     59     8  
     
Various  university  staff  
      Males     5     3     1     1     0  
      n       5     3     1     1     0  
 
      Percentage       60     20     20     0  
 

38
  
A   paired-­‐samples   t   -­   test   (pairs   consisting   of   American  
males/Swedish   males   and   American   females/Swedish   females)   was  
employed   to   test   whether   differences   observed   were   statistically  
significant.   They   were   not   (t   =   .143,   n.s.   and   t   =   -­‐.059,   n.s.   respectively),  
which   suggests   either   that   samples   are   indeed   similar   in   terms   of   the  
distribution  of  gender  stereotypes  in  the  two  cultures,  or  alternatively  
that   the   median   split   is   not   sensitive   enough   a   method   by   which   to  
classify  a  sample  of  the  size  used  for  BSRI-­‐SE  (i.e.  N  =  118).  
 
Sex-­typing  in  the  sample  sub-­groups  
  Although  the  Swedish  sample  is  small  in  comparison  to  that  used  
by   Bem,   it   adds   to   the   understanding   of   gender-­‐roles   in   Sweden   to  
categorise  participants  within  sample  groups  also  (Table  14).  At  least  it  
may   raise   a   few   questions   as   to   the   relationship   between   chosen  
profession   or   activity   and   professed   gender   stereotype.   The   two   larger  
sample   groups:   Pre-­‐school   students   teachers   and   Communication  
studies  students  are  interesting  in  this  respect.  
  Students   of   Communication   Studies,   who   will   graduate   as   Public  
Relation   Officers   with   a   employment   opportunities   in   a   wide   variety   of  
contexts,  seem  to  a  large  extent  (70  %)  to  be  either  undifferentiated  or  
masculine   in   their   sex-­‐type   orientation.   Pre-­‐school   student   teachers,  
on   the   other   hand,   are   largely   directed   towards   either   feminine   or  
androgynous   behaviour   (79   %).     The   latter   should   perhaps   be  
compared   to   full-­‐time   working   pre-­‐school   teachers;   a   group   in   which  
stereotypical  gender  roles  are  more  evenly  distributed:  56  %  are  either  
undifferentiated   or   androgynous   whereas   44   %   are   sex-­‐typed   as  

39
typically  feminine  or  masculine.  
 
 
Concluding  remarks  
It   is   clear   from   this   research   that   the   Bem   Sex-­‐Role   Inventory   (BSRI)  
should   not   be   used   in   a   Swedish   context   in   its   original   form.   It   has  
limited   face   validity,   as   spontaneously   argued   by   many   of   the  
respondents.   It   would   also   appear   that   while   some   scale   items   of   the  
American   Femininity   Scale   and   the   Masculinity   Scales   are   indeed  
shared   between   the   two   cultures   others   are   not,   which   weakens   the  
statistical   reliability   of   the   inventory.   Rather,   the   BSRI-­‐SE   should   be  
used,   which   has   been   revised   to   better   accommodate   the   cultural  
differences.   However,   it   must   be   remembered   that   although   the  
revised   version   has   better   psychometric   properties,   and   does   indeed  
confirm   Bem’s   theoretical   stance   of   regarding   femininity   and  
masculinity   as   separate   constructs   rather   than   extreme   ends   of   the  
same   dimension,   the   BSRI-­‐SE   only   accounts   for   41.4   %   of   total  
variance.   There   are   obviously   other   aspects   of   gender   in   the   Swedish  
sample,  which  are  not  accounted  for  in  the  inventory  and  that  deserves  
further  research.  
  The  evaluation  and  revision  of  BSRI  interestingly  showed  that  on  
the   basis   of   BSRI-­‐SE   there   is   no   statistically   significant   difference  
between  the  distribution  of  gender  role  categories  in  the  United  States  
and   in   Sweden.   This   is   different   from,   for   example,   the   findings   of  
Maloney,  Wilkof  and  Dambrot  (1981)  who—using  the  BSRI  in  Israel—
found  that  while  American  and  Israeli  women  did  not  differ,  American  
and  Israeli  men  did.  Israeli  males  were  significantly  less  androgynous  

40
than  were  American  men  (6  %  versus  20  %).  
  However,  the  apparent  similarity  between  American  and  Swedish  
cultures  in  this  respect  is  nevertheless  interesting.  Sweden,  unlike  the  
United   States   has   been   described   as   a   “feminine”   country   in   terms   of  
what  Hofstede  (1984)  designates  as  “masculine  and  feminine  goals”  in  
his   exceedingly   extensive   research   into   international   differences   in  
work-­‐related  values.  In  this    
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

41
Table  15.  The  Masculinity  Societal  Norm  (Hofstede,  1984).  
 
 
                             Low  MAS              High  MAS  
 
 
•  People  orientation         •  Money  and  things  orientation  
 
•  Quality  of  life  and  environment     •  Performance  and  growth  are  important  
are  important  
 
•  Work  to  live           •  Live  to  work  
 
•  Service  ideal           •  Achievement  ideal  
 
•  Inter-­‐dependence  ideal       •  Independence  ideal  
 
•  Intuition             •  Decisiveness  
 
•  Sympathy  for  the  unfortunate     •  Sympathy  for  the  successful  achiever  
 
•  Levelling:  don’t  try  to  be  better     •  Excelling:  try  to  be  the  best  
than  others  
 
•  Small  and  slow  are  beautiful     •  Big  and  fast  are  beautiful  
 
•  Men  need  to  be  assertive  but     •  Men  should  behave  assertively  
can  also  take  caring  roles       and  women  should  care  
 
•  Sex  roles  in  society  should  be  fluid   •  Sex  roles  in  society  should  be  clearly  
              differentiated  
 
•  Differences  in  sex  roles  should  not   •  Men  should  dominate  all  settings  
mean  differences  in  power  
 
•  Unisex  and  androgyny  ideal     •  Machismoideal  (ostentative  manliness)  
 
 
 
 
research   Hofstede   has   calculated   a   Masculinity   Index   (MAS)   for   each  
country   participating.   The   content   on   which   the   MAS   is   based   is  
outlined   below   (Table  15).  The  United  States  receives  an  MAS  Index  of  
62,  whereas  Sweden  receives  MAS  6,  which  is  the  lowest  index  value  of  

42
all   the   39   countries   participating   in   the   study.   Japan   has   the   highest  
MAS   index   at   87   and   Sweden   is   joined   at   the   lower   end   by   Norway  
(MAS  10),  The  Netherlands  (MAS  14)  and  Denmark  (MAS  22).  
  The   findings   of   the   present   research   seem   to   corroborate  
Hofstede’s   findings   as   far   as   the   Swedish   sample   is   concerned:   74   %   of  
the   participants   (men   and   women   together)   are   indeed   classified   as  
either   undifferentiated   or   androgynous,   thus   largely   conforming   to   the  
attributes   outlined   by   Hofstede   as   being   typical   of   a   low   MAS   Index  
(see   also,   Daun,   1996).   However,   this   is   more   or   less   the   case   for   the  
American   sample   also:   66   %   are   either   undifferentiated   or  
androgynous,   which   is   somewhat   contradictory   to   the   relatively   high  
MAS  Index  of  62.  
  No   conclusions   may   be   drawn   from   these   the   within-­‐group  
classifications   done   as   part   of   the   present   research.   The   groups   are   too  
small.  However,  future  research  should  for  example  look  into  whether  
the   over-­‐representation   of   women   in   teacher   training   (cf.   OECD,   1993)  
also   means   an   over-­‐representation   of   non-­‐masculine   gender   roles,  
which   would   perhaps   be   the   common   sense   assumption   to   make.  
Amongst   all   teachers   participating   in   the   present   study   (n   =     52,   of  
which  20  are  male  and  32  are  female)  the  four  gender  role  categories  
are   apparently   evenly   distributed—in   spite   of   female   over-­‐
representation:  19  %  are  undifferentiated,  27  %  are  androgynous,  23  
%   are   masculine   and   31   %   are   sex-­‐typed   as   feminine.   This   would   be  
interesting   research,   not   only   to   corroborate   the   usefulness   and  
reliability   of   the   BSRI-­‐SE   further,   but   also   to   bring   new   light   to   the  
understanding  of  much  of  education  as  sexist  and  gender  biased,  which  
is   often   seen   as   threat   to   the   increasing   equality   between   the   sexes   (cf.  

43
Houston,  1996;  Morgan,  1996).  
  The   division   of   humanity   into   different   roles,   gender   roles   and  
other,  will  continue.  However,  the  meaning  of  gender  is  changing  over  
time   and   across   cultures   with   increasing   speed   and   with   the  
sometimes   dubious   aid   of   media   and   commercial   promotion   (e.g.  
Sullivan   &   O’Connor,   1988;   Wagner   &   Banos,   1973).   Research  
instruments   such   as   the   BSRI   need   to   reflect   this   change   (Lonner,  
1990).  The  revision  of  the  BSRI  into  BSRI-­‐SE  is  such  an  effort.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

44
 References  

American   Psychological   Association   (1993,   January).   Call   for   book   proposals   for  
test  instruments.  APA  Monitor,  24,  12.  
 
Bailey,  M.  J.  (1996).  Gender  identity.  In  R.  C.  Savin-­‐Williams  &  K.  M.  Cohen  (Eds.).  
The  lives  of  lesbians,  gays,  and  bisexuals:  children  to  adults  (pp.  71-­‐88).  New  York:  
Harcourt  Brace  College  Publishers.  
 
Baumeister,   R.   F.   (1988).   Should   we   stop   studying   sex   differences   altogether?  
American  Psychologist,  42,  756-­‐757.  
 
Bem,   S.   L.   (1974).   The   measurement   of   psychological   androgyny.   Journal   of  
Consulting  and  Clinical  Psychology,  42(2).  155-­‐162.  
 
Bem,   S.   L.   (1975).   Sex-­‐role   adaptability:   one   consequence   of   psychological  
androgyny.  Journal  of  Personality  and  Social  Psychology,  31,  634-­‐643.  
 
Bem,   S.   L.   (1977).   On   the   utility   of   alternative   procedures   for   assessing  
psychological   androgyny.   Journal   of   Consulting   and   Clinical   Psychology,   45,   196-­‐
205.  
 
Bem,  S.  L.  (1981).  Bem  Sex-­Role  Inventory.  Palo  Alto,  CA:  Mind  Garden.  
 
Bjerrum-­‐Nielsen,   H.   &   Rudberg,   M.   (1994).   Psychological   gender   and   modernity.  
Oslo:  Scandinavian  University  Press.  
 
Broverman,  D.  M.,  Vogel,  S.  R.,  Broverman,  D.  M.,  Clarkson,  F.  E.  &  Rosenkrantz,  P.  S.  
(1972).  Sex-­‐role  stereotypes:  a  current  appraisal.  Journal  of  Social  Issues,  28,  59-­‐78.  
 
Burt,   C.   (1952).   Tests   of   significance   in   factor   studies.   British   Journal   of   Psychology.  
Statistics  Section,  5,  109-­‐133.  
 
Child,  D.  (1990).  The  essentials  of  factor  analysis  (2nd  edition).  London:  Cassell.  
 
Cohen,   R.   J.,   Swerdlik,   M.   E.   &   Phillips,   S.   M.   (1995).   Psychological   testing   and  
assessment:   an   introduction   to   tests   and   measurements   (3rd   edition).   London:  
Mayfield.  
 
Daun,   Å.   (1996).   Swedish   mentality   (Translated   by   Jan   Teeland).   University   Park,  
PA:  Pennsylvania  State  University  Press.  
 
Ellis,   L.   (1996).   Theories   of   homosexuality.   In   R.   C.   Savin-­‐Williams   &   K.   M.   Cohen  
(Eds.).   The   lives   of   lesbians,   gays,   and   bisexuals:   children   to   adults  (pp.   9-­‐34).   New  
York:  Harcourt  Brace  College  Publishers.  
 
 

45
Gough,   H.   G.   (1957).   Manual   for   the   California   Psychological   Inventory.   Palo   Alto,  
CA:  Consulting  Psychologists  Press.  
 
Hassler,   M.   (1991).   Androgynie   [Androgyny].   Göttingen,   Germany:   Verlag   für  
Psychologie  Hogrefe.  
 
Hoenig,   J.   &   Kenna,   J.   C.   (1974).   The   nosological   position   of   transsexualism.  
Archives  of  Sexual  Behavior,  3,  272-­‐287.  
 
Hofstede,   G.   (1984).   Culture’s   consequences.   International   differences   in   work-­
related  values.  London:  Sage.  
 
Houston,   B.   (1996).   Gender   freedom   and   the   subtelties   of   sexist   education.   A.  
Diller,  B.  Houston,  K.  P.  Morgan  &  M.  Ayim  (Eds.).  The  gender  question  in  education.  
Theory,  pedagogy  and  politics  (pp.  50-­‐63).  Oxford:  Westview  Press.  
 
Howells,  K.  (1987).  Sex  roles  and  sexual  behaviour.  D.  J.  Hargreaves  &  A.  M.  Colley    
(Eds.).  The  psychology  of  sex  roles  (pp.269-­‐286).  London:  Harper  &  Row.  
 
Kaschak,   E.   &   Sharratt,   S.   (1983).   A   Latin   American   sex   role   inventory.   Cross-­
Cultural  Psychology  Bulletin,  18(1),  3-­‐6.  
 
Kranau,  E.  J.,  Green,  V.  &  Valencia-­‐Weber,  G.  (1982).  Acculturation  and  the  Hispanic  
woman:   attitudes   toward   women,   sex-­‐role   attribution,   sex-­‐role   behavior,   and  
demographics.  Hispanic  Journal  of  Behavioral  Science,  4,  21-­‐40.  
 
Larsson,   S.   (1997).   Det   andra   jaget   vid   manlig   transvestism.   Ett   jag-­teoretiskt   och  
kognitionspsykologiskt   perspektiv   [The   second   self   and   male   transvetism.   A   self-­‐
theoretical   and   cognitive   perspective],   Doctoral   dissertation.   Uppsala,   Sweden:  
Department  of  psychology,  Uppsala  University.  
 
LeVay,  S.  (1993).  The  sexual  brain.  London:  MIT  Press.  
 
Lipman-­‐Blumen,   J.   (1989).   Female   leadership   in   formal   organizations:   Must   the  
female  leader  go  formal?  In  H.  J.  Leavitt,  L.  R.  Pondy  &  D.  M.  Boje  (Eds.).  Readings  in  
managerial  psychology  (pp.  325-­‐344).  Chicago,  IL:  Chicago  University  Press.  
 
Lonner,  W.  J.  (1990).  An  overview  of  cross-­‐cultural  testing  and  assessment.  R.  W.  
Brislin  (Ed.).  Applied  cross-­cultural  psychology  (pp.56-­‐76).  London:  Sage.  
 
Maloney,  P.,  Wilkof,  J.  &  Dambrot,  F.  (1981).  Androgyny  across  two  cultures:  United  
States  and  Israel.  Journal  of  Cross-­Cultural  Psychology,  21(1),  95-­‐102.  
 
McHugh,   M.   C.,   Koeske,   R.   D.   &   Frieze,   I.   H.   (1986).   Issues   to   consider   in   conducting  
nonsexist  research:  a  guide  for  researchers.  American  Psychologist,  41,  879-­‐890.  
 
 

46
McKnight,   J.   (1997).   Straight   science?   Homosexuality,   evolution   and   adaption.  
London:  Routledge.  
 
Money,   J.   &   Ehrhardt,   A.   A.   (1972).   Man   and   woman,   boy   and   girl.   Baltimore,   MD:  
Johns  Hopkins  University  Press.  
 
Morgan,  K.  P.  (1996).  The  androgynous  classroom:  liberation  or  tyranny?  A.  Diller,  
B.   Houston,   K.   P.   Morgan   &   M.   Ayim   (Eds.).   The   gender   question   in   education.  
Theory,  pedagogy  and  politics  (pp.  64-­‐74).  Oxford:  Westview  Press.  
 
Murphy,   T.   F.   (1990).   Homosexuality,   a   philosophical   analysis.   Journal   of  
Homosexuality,  19,  132-­‐137.  
 
OECD  (1993).  Education  in  OECD  countries.  A  compendium  of  statistical  information.  
Paris:  OECD.  
 
Ross,   M.   W.   (Ed.).   (1985).   Homosexuality,   masculinity   and   femininity.   New   York:  
Harrington  Park  Press.  
 
Rowland,   R.   (1977).   The   Bem   Sex-­‐Role   Inventory.   Australian   Psychologist,   12,   83-­‐
88.  
   
Segall,   M.   H.,   Dasen,   P.   R.,   Berry,   J.   W.   &   Poortinga,   Y.   H.   (1990).   Human   behavior   in  
a   global   perspective.   An   introduction   to   cross-­cultural   psychology.     Oxford:  
Pergamon.  
 
Spence,   J.   T.,   Helmreich,   R.     &   Stapp,   J.   (1975).   Ratings   of   self   and   peers   on   sex-­‐role    
attributes   and   their   relation   to   self-­‐esteem   and   conceptions   of   masculinity   and  
femininity.  Journal  of  Personality  and  Social  Psychology,  43,  568-­‐571.  
 
Sternberg,   R.   J.   (1993).   What   is   the   relation   of   gender   to   biology   and   environment?  
An  evolutionary  model  of  how  what  you  answer  depends  on  just  what  you  ask.  In  
A.   E.   Beall   &   R.   J.   Sternberg   (Eds.).   The   psychology   of   gender   (pp.   1-­‐8).   London:  
Guilford  Press.  
 
Sullivan,   G.   L.   &   O’Connor,   P.   J.   (1988).   Women’s   role   portrayals   in   magazine  
advertising:  1958  -­‐  1983.  Sex  Roles,  18,  181-­‐188.  
 
Wagner,   L.   C.   &   Banos,   J.   B.   (1973).   A   woman’s   place:   a   follow-­‐up   analysis   of   the  
roles   portrayed   by   women   in   magazine   advertisements.   Journal   of   Marketing  
Research,  10,  213-­‐214.  
 
Walkup,   H.   &   Abbott,   R.D.   (1978).   Cross-­‐validation   of   item   selection   on   the   Bem  
Sex-­‐Role  Inventory.  Applied  Psychological  Measurement,  2,  63-­‐71.  
 
 
 

47
Williams,   W.   L.   (1996).   Two-­‐spirit   persons:   gender   nonconformity   among   native  
American  and  native  Hawaiian  youths.  In  R.  C.  Savin-­‐Williams  &  K.  M.  Cohen  (Eds.).  
The  lives  of  lesbians,  gays,  and  bisexuals:  children  to  adults  (pp.  416-­‐432).  New  York:  
Harcourt  Brace  College  Publishers.  
 
Williams,   J.   E.   &   Best,   D.   L.   (1989).   Sex   and   psyche:   self-­concept   viewed   cross-­
culturally.  Newbury  Park,  CA:  Sage.  

48
49
 Appendix I

BSRI-­‐SE     The  Bem  Sex-­Role  Inventory  

Utvärdet  och  reviderat  för  en  svensk  kontext  

 
 
 
 
Roland  S.  Persson,  PhD  
 
Centrum  för  psykologi  
Högskolan  för  lärarutbildning  och  kommunikation  
Högskolan  i  Jönköping  
 
(Original  instrument  konstruerat  av  Sandra  Bem,  Stanforduniversitet,  USA,    
och  publicerat  1981  av  Mind  Garden,  Palo  Alto,  CA)  
 
 
 
 
 
 
 
 
 
 
 
Råpoäng  M   Råpoäng  F   M-­F  poäng   Klass   Standard    
         
 

50
BSRI-­‐SE   Kön:     Ålder:     Datum:     Kod:  

Ta  ställning  till  följande  påståenden  (1  till  25)  genom  att  i  boxen  framför  varje  påstående  skriva  den  
siffra,  som  bäst  motsvarar  hur  du  upplever  hur  dessa  utsagor  stämmer  på  dig.  Läs  och  bedöm  alla  
utsagor!    Du  kan  använda  dig  av  siffrorna  1,  2,  3,  4,  5,  6  eller  7  för  att  beskriva  hur  sanna  du  tycker  
att  dessa  påståenden  är.  Siffran  1  betyder  att  ett  påstående  är  aldrig  eller  nästan  aldrig  sant,  medan  
siffran   7   betyder   att   ett   påstående   är   alltid   eller   nästan   alltid   sant.   Värdena   däremellan   beskriver  
olika  grader  av  dessa  två  påståenden  (se  figuren  nedan).  
 
 
1    Det  är  aldrig  eller  nästan  aldrig  sant          3     Det  är  alltid  eller  nästan  alltid  sant      7  
 
 
 
1.      Jag  är  full  av  självförtroende     14.      Jag   är   angelägen   att   lindra  
                    sårade  känslor      
2.        Jag  är  tillgiven         15.      Jag  är  förstående  
3.      Jag  försvarar  mina  övertygelser     16.      Jag  har  lätt  för  att  ta  beslut  
4.      Jag  är  bestämd         17.      Jag  är  självtillräcklig  och    
                    klarar  mig  själv  
 
5.      Jag  är  mild  och  stillsam       18.      Jag  är  en  lojal  person  
6.      Jag  har  en  stark  personlighet     19.      Jag  är  dominant      
7.      Jag  är  sympatisk         20.      Jag  är  maskulin  
8.      Jag  är  en  varm  person       21.      Jag  är  ömsint  
 
9.      Jag  är  en  kraftfull  person       22.      Jag  är  varsam  
10.      Jag  är  en  ledartyp         23.      Jag  är  villig  att  fatta  beslut  
                    och  att  stå  för  dem  
11.      Jag  är  medlidsam         24.      Jag  agerar  ledare  
12.      Jag  är  känslig  för  andras  behov     25.      Jag  är  feminin    
13.      Jag  är  villig  att  ta  risker                  
 
 
            Tack   för   att   du   har   besvarat   alla   påståenden!
   
 
 
 
 
 
 
 
 
 
 
 
 
 

51
De  olika  påståendena  på  svarsblanketten  har  ordnats  slumpmässigt  enligt  följande.  
M   anger   att   påståendet   tillhör   maskulinitetsskalan   och   F   visar   att   påståendet  
tillhör   femininitetsskalan.   Siffrorna   efter   M   respektive   F   anger   variabelnummer  
och   kan   identifieras   i   ovanstående   utvärdering   av   BSRI.   M-­‐skalan   består   av   13  
påståenden   och   F-­‐skalan   av   12.   I   forskningsrapporten   beskrivet   hur   revision   av  
detta  inevntorium  gått  till  väga  samt  hur  resultat  skall  bedömas  och  räknas  ut.  
                       
                     
1.   Jag  är  full  av  självförtroende                   M1  
2.   Jag  är  tillgiven                       F4  
3.   Jag  försvarar  mina  övertygelser                   M2  
4.   Jag  är  bestämd                       M5  
5.   Jag  är  mild  och  stillsam                     F13  
 
6.   Jag  har  en  stark  personlighet                   M6  
7.   Jag  är  sympatisk                       F8  
8.   Jag  är  en  varm  person                     F14  
9   Jag  är  en  kraftfull  person                     M7  
10.   Jag  är  en  ledartyp                       M9  
 
11.   Jag  är  medlidsam                       F11  
12.   Jag  är  känslig  för  andras  behov                   F9  
13.   Jag  är  villig  att  ta  risker                     M10  
14.   Jag  är  angelägen  att  lindra  sårade  känslor               F12  
15.   Jag  är  förstående                       F10  
 
16.   Jag  har  lätt  för  att  ta  beslut                   M11
   
17.   Jag  är  självtillräcklig  och  klarar  mig  själv               M12  
18.   Jag  är  en  lojal  person                     F6  
19.   Jag  är  dominant                       M13  
20.   Jag  är  maskulin                       M14  
 
21.   Jag  är  ömsint                         F15  
22.   Jag  är  varsam                       F20  
23.   Jag  är  villig  att  fatta  beslut  och  att  stå  för  dem               M15  
24.   Jag  agerar  ledare                       M17  
25.   Jag  är  feminin                       F7  
 
 
 
 
 
VIKTIGT!   För   närmare   beskrivning   av   BSRI:s   ursprungliga   konstruktion   och   tolkning   se   Bem  
(1974;   1975;   1981).   Användaren   av   detta   reviderade   svenska   inventorium   bör   för   ändamålet   alltid  
införskaffa  det  amerikanska  och  kommersiellt  publicerade  materialet  från  Mind  Garden,  Palo  Alto,  
Kalifornien  med  tanke  på  lagen  om  upphovsrätt.  Detta  svenska  material  är   ett   forskningsmaterial  
av  begränsad  spridning  och  har  ett  icke-­‐kommersiellt  syfte.  
 
 
 
 

52
 Appendix  II  
Normative raw scores for BSRI-SE
Part 1:
Subject   M-­scale   F-­scale   MF  Diff   Diff  z   Subject   M-­scale   F-­scale   MF  Diff   Diff  z  
mean   mean   mean   mean  
1f   4.69   4.75   -­‐.06   .1904   34f   5.77   5.33   .44   .60383  
2f   5.77   5.33   .44   .6038   35f   5.08   3.92   1.16   1.19911  
3f   4.31   6.50   -­‐2.19   -­‐1.5706   36f   3.77   6.17   -­‐2.40   -­‐1.7442  
4m   5.15   4.58   .57   .7113   37f   6.46   5.67   .79   .89320  
5m   5.00   5.42   -­‐.42   -­‐.1072   38f   5.08   5.08   .00   .24005  
6f   5.38   5.92   -­‐.54   -­‐.2064   39f   4.54   5.17   -­‐.63   -­‐.28082  
7f   3.69   5.58   -­‐1.89   -­‐1.3225   40f   4.85   5.08   -­‐.23   .04989  
8m   4.46   5.42   -­‐.96   -­‐.5536   41f   3.85   6.08   -­‐2.23   -­‐1.6036  
9m   5.53   5.67   -­‐.14   .1243   42f   5.08   5.50   -­‐.42   -­‐.1072  
10f   4.00   6.08   -­‐2.08   -­‐1.4796   43f   5.54   6.00   -­‐.46   -­‐.1402  
11f   5.38   5.75   -­‐.37   -­‐.0.658   44f   4.31   4.83   -­‐.52   -­‐.1898  
12f   3.31   5.58   -­‐2.27   -­‐1.6367   45f   3.54   4.75   -­‐1.21   -­‐.7603  
13f   4.15   4.67   -­‐.52   -­‐.1898   46f   4.46   5.00   -­‐.54   -­‐.2064  
14f   4.54   5.92   -­‐1.38   -­‐.9009   47f   4.31   5.58   -­‐1.27   -­‐.8099  
15f   4.01   5.33   -­‐1.32   -­‐.8513   48m   6.08   3.08   3.00   2.7203  
16f   4.15   4.00   .15   .3640   49m   5.54   4.00   1.54   1.5132  
17f   4.31   5.92   -­‐1.61   -­‐1.0910   50m   3.62   5.08   -­‐1.46   -­‐.9670  
18f   5.92   5.00   .92   1.0006   51m   3.92   4.17   -­‐,25   .0333  
19f   3.54   5.33   -­‐1.79   -­‐1.2398   52m   5.08   4.25   .83   .9262  
20f   3.92   5.42   -­‐1.50   -­‐1.0001   53m   6.00   4.58   1.42   1.4140  
21f   3.00   6.75   -­‐3.75   -­‐2.8603   54f   4.38   5.58   -­‐1.20   -­‐.7520  
22f   4.00   6.08   -­‐2.08   -­‐1.4796   55f   3.69   5.75   -­‐2.06   -­‐1.4531  
23f   4.92   5.75   -­‐.83   -­‐.4461   56f   5.38   5.17   .21   .4136  
24f   4.00   5.17   -­‐1.17   -­‐.7272   57f   4.08   6.25   -­‐2.17   -­‐1.5540  
25f   4.54   4.92   -­‐.38   -­‐.0741   58f   5.23   6.83   -­‐1.60   -­‐1.0828  
26m   4.54   5.25   -­‐,71   -­‐.3469   59f   1.69   5.25   -­‐3.56   -­‐2.7032  
27m   5.38   4.42   .96   1.0337   60f   6.23   5.00   1.23   1.2569  

53
28m   4.46   3.17   1.29   1.3065   61m   3.38   4.92   -­‐1.54   -­‐1.0331  
29f   4.54   4.75   -­‐.21   .0664   62f   4.38   3.58   .80   .9014  
30f   4.46   4.33   .13   .3475   63f   5.31   4.42   .89   .9758  
31m   6.46   5.92   .54   .6865   64f   5.38   4.54   .84   .9345  
32f   5.23   5.42   -­‐.19   .0829   65f   4.62   5.00   -­‐.38   -­‐.0741  
33f   5.15   4.67   .48   .6369   66f   4.38   6.00   -­‐1.62   -­‐1.0993  
 
 

Part  2  
Subject   M-­scale   F-­scale   MF  Diff   Diff  z   Subject   M-­scale   F-­scale   MF  Diff   Diff  z  
mean   mean   mean   mean  
67f   4.69   5.83   -­‐1.14   -­‐.7024   93f   5.31   4.83   .48   .6369  
68f   4.15   5.50   -­‐1.35   -­‐.8761   94f   4.77   4.75   .02   .2565  
69f   4.15   5.50   -­‐1.35   -­‐.8761   95f   4.54   4.75   -­‐.21   .0664  
70f   5.38   6.25   -­‐.87   -­‐.4792   96f   5.00   6.17   -­‐1.17   -­‐.7272  
71f   5.08   4.83   .25   .4467   97f   4.85   6.25   -­‐1.40   -­‐.9174  
72f   5.15   3.92   1.23   1.2569   98f   5.23   5.42   -­‐.19   .0829  
73f   4.15   5.17   -­‐1.02   -­‐.6032   99f   4.08   4.08   .00   .2400  
74m   4.54   4.67   -­‐.13   .1325   100m   5.77   6.17   -­‐.40   -­‐.0906  
75m   4.15   3.83   .32   .5046   101f   4.85   4.75   .10   .3227  
76m   4.62   4.58   .04   .2731   102m   5.54   2.00   3.54   3.1668  
77m   4.54   4.67   -­‐.13   .1325   103m   5.34   3.92   1.46   1.4471  
78m   6.00   5.17   .83   .9262   104m   5.08   6.00   -­‐.92   -­‐.5205  
79m   4.69   4.33   .36   .5376   105m   6.31   5.75   .56   .7030  
80m   4.77   4.58   .19   .3971   106m   6.31   5.42   .89   .9758  
81m   5.00   5.83   -­‐.83   -­‐.4461   107m   5.46   4.42   1.04   1.0999  
82m   5.69   4.00   1.69   1.6373   108m   4.92   4.33   .59   .7278  
83m   5.08   4.33   .75   .8601   109m   4.54   5.67   -­‐1.13   -­‐.6942  
84m   3.93   4.58   -­‐.65   -­‐.2973   110m   5.15   6.50   -­‐1.35   -­‐.8761  
85m   6.00   5.92   .08   .3061   111m   5.00   4.83   .17   .3806  
86m   4.31   4.75   -­‐.44   -­‐.1237   112m   6.62   4.50   2.12   1.9928  
87f   4.92   5.25   -­‐.33   -­‐.0327   113m   5.62   4.75   .87   .9593  
88f   4.85   6.00   -­‐1.15   -­‐.7107   114m   5.23   4.75   .48   .6369  

54
89f   3.69   6.33   -­‐.2.64   -­‐1.9426   115m   4.15   4.75   -­‐.60   -­‐.2560  
90f   4.92   4.42   .50   .6534   116m   4.54   4.33   .21   .4136  
91m   5.15   3.92   1.23   1.2568   117m   4.08   3.75   .33   .51288  
92m   5.08   4.33   .75   .86013   118m   5.54   6.00   -­‐.46   -­‐.1402  
 
 
 
 
 
 
 

55