IJT logo



Friedemann Pfäfflin,
Ulm University, Germany

Walter O. Bockting,
University of Minnesota, USA

Eli Coleman,
University of Minnesota, USA

Richard Ekins,
University of Ulster at Coleraine, UK

Dave King,
University of Liverpool, UK

Managing Editor:
Noelle N Gray,
University of Minnesota, USA

Editorial Assistant:
Erin Pellett,
University of Minnesota, USA

Editorial Board


book Historic Papers


© Copyright

Published by
Symposion Publishing

ISSN 1434-4599

Volume 6, Number 2, 2002

Gender Stereotype and Self among Transgenders: Underlying Elements

Dr. Sam Winter,
Department of Education
University of Hong Kong, 
Pokfulam Road,
Hong Kong, 
People’s Republic of China
Fax: (852) 28585649

Nuttawut Udomsak,
University of the Thai Chamber of Commerce,

Citation:  http://www.symposion.com/ijt/ijtvo06no02_02.htm


In an article published earlier in The International Journal of Transgenderism (Winter and Udomsak, 2002) showed that while Thai MtF transgenders displayed actual self-concepts that were strongly female-stereotyped (that is, consistent with their own beliefs about femaleness) their ideal self concepts, and aspirations for change were distinctly less female-stereotyped.

This finding raised the following question: what underlying considerations, if not the simple pursuit of stereotyped femaleness, governed their ideal self and aspirations for personal growth? To answer this question, the Adjective Checklist (ACL) data from the original study (Winter and Udomsak, 2002) were further analysed in a three-step procedure.

Firstly, an attempt was made to identify the underlying essence of the traits we employed in the ACL. Using findings from earlier ACL research by Williams and co-workers (Williams and Best, 1990; Williams et al., 1998, 1999), we ascribed to each ACL trait-item a set of 14 scores, each of which reflected the degree to which that trait reflected an important psychological feature. These features represented (a) affective meaning (three scores: favourability, strength and activity), (b) ego-state (five scores: critical parent, nurturing parent, adult, free child and adapted child), (c) higher-order personality factors (five scores: extraversion, agreeableness, conscientiousness, emotional stability and openness), and (d) psychological importance (one score: indicating the degree to which the trait is a ‘core’ element of personality).

Secondly, in order to reduce the data somewhat, these scores were factor analysed. The 14 scores were loaded onto four factors. On the basis of the loadings, these factors were labelled: ‘resourceful / dependable’ (factor I), ‘intrusive / controlling’ (factor II), ‘risk-taking / stimulation-seeking’ (factor III), and ‘caring / harmonious’ (factor IV).

Thirdly, multiple regression analyses were employed to identify which if any of these factors appeared to underlie participants’ (i) gender-trait stereotypes, (ii) actual-self, (iii) ideal-self, (iv) aspirations to acquire traits (traits desired but not possessed) and (v) aspirations to lose traits (traits possessed but not desired).

The analysis revealed that gender-trait stereotypes were predicted by factors I, II and III (all underlying male-stereotyped traits) as well as by IV (underlying those that were female-stereotyped). Factors I, II and III could therefore be considered ‘male’ factors, while factor IV was ‘female’. As one might expect, factor IV (the ‘female’ factor) predicted those traits endorsed for actual self, while factor II (a ‘male’ factor) acted in a counter-predictive way.

Our main interest focused on finding factors that might shed light on our earlier findings on ideal self and aspirations for change. We found that ideal self was predicted by a gender-inconsistent mix of factor IV (the ‘female’ factor) and factor I (a ‘male’ factor). Factor I was also important in predicting those traits which participants aspired to acquire, and in counter-predicting those (unwanted) traits, which they aspired to lose. Beyond this, factor IV (the ‘female’ factor?) somewhat paradoxically predicted aspirations to lose traits.

In short, participants’ ideals for self seemed to embody qualities of care and harmony (‘female’ qualities), but also resourcefulness and dependability (‘male’ qualities). Indeed, participants wished not only to retain whatever ‘male’ qualities of resourcefulness and dependability they had, but also to acquire more of these qualities. Furthermore, while they valued many ‘female’ qualities like care and harmony, they also aspired to lose some of these qualities.

These findings, which are, at face value, gender-anomalous, attest to personal growth goals that transcend (or indeed run counter to) gender stereotype; instead they conform to notions of maturity and personal efficacy.


Keywords: transgender, Thailand, self-concept, personality traits, gender-trait stereotypes.




Western research into self-concept indicates that, while transgender persons typically display many characteristics stereotypically associated with their chosen gender (Fleming et al., 1981; Skrapec and MacKenzie, 1981; Kuiper and Cohen-Kettenis, 1988; Cole et al., 1997) they retain other traits associated with their biological gender (Skrapec and MacKenzie, 1981; Fleming et al., 1984; Cole et al., 1997).

Similar findings come from a recent study with 204 Thai MtF transgenders, published in The International Journal of Transgenderism (Winter and Udomsak, 2002). A shortened version of the Adjective Checklist (ACL) from Gough and Heilbrun (1980) was used to examine participants’ actual and ideal self-concepts, as well as aspirations for change, in terms of their gender-trait stereotypes (beliefs about traits associated with men and women). For each personality trait, gender-trait stereotyping was expressed using an M% figure (a percentage figure indicating the number of persons thinking the trait described men more than women, divided by the number expressing a clear view one way or the other). A high M% value indicated high male stereotyping for the trait in question. Actual and ideal self-concepts were expressed by endorsement rates (with a high figure for a trait indicating that many participants felt it was an aspect of themselves as they currently were (in the case of actual self-concept) or wished to be (in the case of ideal self-concept). Aspirations for change (i.e., either to acquire or lose traits) were indicated by the discrepancy between endorsement rates for actual and ideal self-concepts.

Inter alia, it was found that actual and ideal self concepts tended to be discrepant, with actual self stereotypically being female, and ideal self being distinctly less stereotyped. The authors concluded that, despite their transition towards a female gender role, participants overall avoided a slavish pursuit of stereotyped ‘femaleness’. Instead, like many of their counterparts (male and female) in the wider (non-transgender) community, they constructed ideals and pursued growth goals for self that were governed by other considerations.

The question remained: what considerations, if not the pursuit of stereotypical femaleness, were they governed by? To answer this question we examined our data again, this time focusing on underlying features of the ACL traits other than their gender-trait stereotypy, and employing multiple regression analysis in order to see how well these features explained (in a statistical – i.e., predictive sense) our original data.


Research Steps

  1. In order to extract the underlying essence of the traits we were employing in the ACL, we drew upon earlier ACL research by Williams and co-workers (Williams and Best, 1990; Williams et al., 1998, 1999) which provides, for each ACL trait, 14 scores on important psychological features. These scores were called ‘psychological features’ scores.
  2. In order to reduce the number of variables to be employed in the subsequent multiple regression, these scores were factor analysed.
  3. Multiple regression analyses were employed to identify which, if any, of these factors appeared to underlie the participants’ (i) gender-trait stereotypes, (ii) actual-self, (iii) ideal-self, (iv) aspirations to acquire traits (traits desired but not possessed) and (v) aspirations to lose traits (traits possessed but not desired).

It is important to note that, at all of these steps, ACL traits were our unit of analysis.


Methodology: Original Collection of Data

First of all, it may be useful to reprise important information on participants, as well as the instrumentation and procedure by which the original data were collected. For further information on these aspects of methodology see Winter and Udomsak (2002). 



There were 204 participants in this study. They were aged 17 to 42 years (mean 23.0 years) and at the time of the study were all living full-time in a cross-gendered role. All reported having experienced early feelings of female identity. The vast majority had undergone some sort of medical treatment (pharmacological or surgical) to feminise their physical appearance. A few had undergone sex reassignment surgery. In every case their speech habitually exhibited the pronouns and particles employed by females in the Thai language. All described themselves as ‘kathoey’ (a Thai word commonly used to describe those we might call MtF transgenders) or its related terms. To the extent that the particpants displayed all these qualities they were, for the purpose of this study, defined by the authors as MtF transgenders.

Participants identified themselves as having the following work: 120 cabaret performers, 41 bar workers, 36 university students, and 7 others engaged in shop work etc. In fact these categories were all somewhat arbitrary, as some participants fell into two or more categories. As one might therefore expect, significant group differences in response patterns were relatively few in number, and small in scale.



The instrument was a shortened ACL consisting of 81 adjectives drawn from the original pool of 300. Each checklist, printed in Thai, was given to the participants three times, with instructions (a) to check any adjective that described participants ‘as they would like to be’, (b) to check any adjective describing them ‘as they currently were’, and (c) to evaluate each adjective according to ‘whether they believed men displayed the trait more often than women, women more often than men, or both sexes equally.’ Items and instructions had been translated by a qualified professional and back-translated to check for adequacy. The 81 adjectives are displayed in Table 1, arranged in terms of the gender-trait stereotypes actually expressed by the transgender sample. 

Table 1: List of ACL trait-items, arranged according to MtFs’ gender-trait stereotypes

Stereotypically Female   Stereotypically Male
Strongly* Moderately*  Mildly* Neutral* Mildly* Moderately*  Strongly*
fearful affected  curious  ambitious  active  adventurous  (none)
feminine  affectionate  meek  assertive  cruel  aggressive   
fussy  anxious nagging  boastful  energetic  autocratic   
sensitive  appreciative  stern  changeable  flirtatious  coarse   
shy  attractive  submissive  confident  hard-headed courageous  
soft-hearted charming  talkative  determined  high-strung daring   
superstitious  complaining  timid  dominant  humorous  disorderly   
sympathetic  dependant  whiny  frivolous  independant  egotisical   
weak  dreamy    gentle  reckless  enterprising   
  emotional    initiative  sophisticated  forceful   
  excitable    inventive  stolid  lazy   
  mild    poised  unemotional  loud   
  prudish    rational    masculine   
  sentimental    realistic    progressive   
  sexy    self-confident   robust   
  worrying    warm    rude   
      wise    severe   


*Strongly Female (M% less than or equal to 10)
Moderately Female (M% over 10 but less than or equal to 20)
Mildly Female (M% over 20 but less than or equal to 33)
Neutral (M% over 33 but less than 67)
Mildly Male (M% 67 or more but less than 80)
Moderately Male (M% 80 ore more but less than 90)
Strongly Male (M% 90 or more)



Potential participants were approached at their places of work or study and invited to take part in the research. It was presented as an attempt to attempt to understand transgenders as people, and to communicate that understanding as widely as possible. Participants were told that there were three questionnaires (each of which would take about ten minutes to complete), that there were no right or wrong answers, and that participants should answer in a way that reflected their feelings. Two potential participants refused to take part, leaving 204 who actually did. Participants completed all three questionnaires at one sitting. Completed questionnaires were checked for thoroughness, and for any responses indicating the need for further instructions or questioning. Any questions from the participants were answered by the research assistant, a native Thai speaker. 

For actual and ideal self-concept 204 and 201 questionnaires were usable respectively. For the gender-trait stereotype questionnaire the mean number of usable responses was 194.7 with a range from 189 to 199, depending on the item.

We can now turn to the main topic of this report: the re-analysis of our original data.


Methodology: Re-analysis of Data

In view of the small and largely non-significant differences between the study/employment sub-groups, data were pooled for all participants. Inferential statistical operations were performed by way of SPSS-PC Plus Version 10.1.

Step one: ascription of scores indicating underlying psychological features

Each ACL trait-item was ascribed 14 scores each of which reflected the degree to which that trait reflected important psychological features. The scores were drawn from research by Williams and colleagues in a series of studies, both American and international, employing the ACL. 
The scores represented:

  1. affective meaning (three scores: favourability (FAV), strength (STR) and activity (ACT)), 
  2. ego-state (five scores, for critical parent (CP), nurturing parent (NP), adult (A), free child (FC) and adapted child (AC)),
  3. higher-order personality factors (five scores: extraversion (EXT), agreeableness (AGR), conscientiousness (CON), emotional stability (EMS) and openness (OPN)) and
  4. psychological importance (one score: (PI) indicating the degree to which the trait is a ‘core’ element of personality). 

A couple of examples may be useful here. 
The item ‘coarse’ was ascribed the following scores: 
FAV=398,STR=517, ACT=512, CP=2.5, NP=0.2, A=0.2, FC=1.8, AC=2.5, EXT=2.09, AGR=2.03, CON=2.69, EMS=2.24, OPN=2.56, PI=269. 
By contrast the item ‘gentle’ was ascribed these scores: 
FAV=635, STR=492, ACT=362, CP=0.3, NP=3.9, A=0.5, FC=1.7, AC=1.2, EXT=3.69, AGR=4.46, CON=3.19, EMS=3.83, OPN=3.35, PI=356.

More information about the 14 variables, and about the research by which scores were derived, can be found in the Appendix.

Step two: factor analysis

As indicated earlier, the 14 scores for psychological features were considered too many to be placed into multiple regression analyses as predictor variables. Accordingly factor analysis (principal components with oblique rotation by way of the direct oblimin method) was used to tease out underlying elements, and reduce the data to a more manageable level. Any factors with eigenvalues above 1.0 were to be used as independent variables in the subsequent multiple regression procedure.

Step Three: multiple regression analysis

Factors identified at step two were employed as independent variables in a set of stepwise regression procedures, each with a different dependent variable. 

The dependent variables were:

  1. gender-trait stereotype scores for each trait (operationalised as an M% figure, high M% values indicating traits that were viewed as stereotypically male),
  2. actual self-concept scores for each trait (or ASC%, operationalised as the percentage of participants endorsing each trait as a description of current self),
  3. ideal self-concept scores for each trait (or ISC%, operationalised as the percentage endorsing each trait as a description of ideal self),
  4. aspirations to acquire traits (for each trait, the percentage of respondents who did not endorse the trait for actual self, but did for ideal; here abbreviated as AA%) and,
  5. aspirations to lose traits (for each trait, the percentage who endorsed the trait for actual self, but did not for ideal; here abbreviated to AL%). 

The last two dependent variables were chosen instead of the single variable employed in the earlier article by Winter and Udomsak (2002) (‘aspirations for change’) in view of evidence from that article that there were qualitative differences between traits desired and those eschewed.

Once again, we stress that during all three steps ACL items were employed as the unit of analysis. In step one, the ACL traits were ascribed scores, in step two, their scores were subjected to a factor analysis, and in step three, the M% scores (for gender-trait stereotypy) and endorsement rates (for actual and ideal self and aspirations for change) were employed as dependent variables. 



Factor Analysis

The 14 psychological features were themselves underlain by four factors with eigenvalues above 1.0. These factors, and the associated loadings, are displayed in Table 2.

Table 2: Factor analysis based on 14 scores for psychological

Factor Loadings
  Factor I Factor II Factor III Factor IV
FAV .403 -.134 .149 .603
STR .444 .416 .425 .189
ACT .032 .542 .740 -.210
CP -.114 .958 -.156 -.021
NP -.171 .052 -.076 1.035
A .968 .072 -.109 -.070
FC -.147 -.234 .980 .015
AC -.868 .085 -.120 .065
EXT .135 .047 .641 .456
AGR .109 -.248 .124 .808
CON .539 .437 .098 .356
EMS .757 -.249 -.058 .252
OPN .296 -.020 .710 .182
PI .265 .146 .109 .674
Eigenvalue 7.581 2.233 1.496 1.016



  Factor I Factor II Factor III Factor IV
M% .336** .365** .336** -.222*
Factor I - - - -
Factor II .087 - - -
Factor III .356** .111 - -
Factor IV .495** -.094 .297** -


Factor I: ‘resourceful / dependable’
Factor II: ‘intrusive / controlling’
Factor III: ‘risk-taking / stimulation-seeking’
Factor IV: ‘caring / harmonious’
Significance of correlations: *: p<0.05, **:p<0.01

As often happens in this sort of research, the naming of factors presented a challenge. The difficulty was exacerbated by the fact that the psychological features entered into the analysis were themselves of an abstract and higher-order nature. In arriving at factor labels, the authors found it helpful to consider not only the 14 psychological features but also the ACL items that generated the highest loadings for the factor in question. Furthermore, in doing so it was useful to consider highest positive as well as highest negative loadings, so that one might ascertain not only what a factor represented, but also what it counter-represented. 

Factor I expressed highest loadings from A, AC (this one negative) and EMS, with a substantial loading from CON. This factor accounted for 54.15% of variance. It drew especially high loadings from the traits ‘rational’, ‘realistic’, ‘independent’, ‘inventive’, ‘unemotional’ and ‘wise’ (all positive), and ‘dependent’, ‘disorderly’, ‘emotional’ and ‘whiny’ (all negative). The label ‘resourceful / dependable’ was chosen for this factor.

Factor II expressed a high loading from CP, with a substantial one also coming from ACT. This factor accounted for 15.95% of variance. It drew especially high loadings from the traits ‘aggressive’, ‘dominant’, ‘determined’, ‘forceful’, ‘nagging’, ‘severe’ and ‘stern’ (all positive), and ‘dreamy’ and ‘lazy’ (both negative). The label ‘intrusive / controlling’ was chosen for this factor. 

Factor III expressed highest loadings from FC, ACT, and OPN, and a substantial loading from EXT. The factor accounted for 10.67% of variance. It drew especially high loadings from the traits ‘adventurous’, ‘daring’, ‘energetic’ and ‘excitable’ (all positive), and ‘unemotional’, ‘meek’, ‘mild’, ‘prudish’, ‘shy’, ‘submissive’, ‘timid’ and ‘weak’ (all negative). The label ‘risk-taking / stimulation-seeking’ was chosen for this factor.

Factor IV expressed highest loadings from NP and AGR and PI, with a substantial additional loading from FAV. The factor accounted for 7.26% of variance. It drew especially high loadings from the traits ‘affectionate’, ‘gentle’, ‘sympathetic’, ‘warm’, ‘soft-hearted’, ‘sensitive’, ‘appreciative’ (all positive), and ‘rude’, ‘disorderly’, ‘whiny’ and ‘reckless’ (all negative). The label ‘caring / harmonious’ was chosen for this factor. 

Again, a couple of examples may be useful here. The ACL trait-item ‘coarse’ displayed loadings as follows: factor I = 2.56, factor II = -0.74, factor III = 0.60, factor IV = -1.22. By contrast the item ‘gentle’ displayed the following loadings: factor I = 0.26, factor II = -1.19, factor III = -0.57, factor IV = 2.23.

Factors I to III correlated positively with participants’ gender-trait stereotypes, and therefore appeared to be male-associated. The reverse was true for factor IV. Simple bivariate correlations therefore indicated that intrusiveness and control, resourcefulness and dependability, as well as the taking of risks and seeking of stimulation were all perceived as stereotypically male traits, while caring and social harmony were seen as stereotypically female. The relevant correlation coefficients are shown in Table 2, along with the intercorrelations between the factors themselves. 


Multiple Regression Analysis 

(a) Regression of factors upon gender-trait stereotype

Results for the regression upon M% are displayed in Table 3a. All four factors had a significant predictive effect on M%. Strongest was factor II (‘intrusive / controlling’), accounting for 13.3% of variance. Then came factor I (‘resourceful / dependable’), accounting for a further 9.3% of variance. In each case the Beta values were positive. factor IV (‘caring / harmonious’) came next, predicting another 15.9% of variance. In this case the Beta value was negative. Finally, factor III (‘risk-taking / stimulation-seeking’) predicted another 7.4% of variance, yielding a positive Beta value. To summarise, the regression analysis confirmed the picture painted by bivariate correlation analysis; participants’ stereotypes regarding male traits were underlain by intrusiveness and control, resourcefulness and dependability, as well as the taking of risks and the seeking of stimulation. In contrast, caring and social harmony underlay a female stereotype.


Table 3a: Gender-trait stereotype (M%) 

  R square change  Sig.  Beta  t-value  Sig.
Factor II .133 .001 .243 2.815 .006
Factor I .093 .003 .466 4.615 .000
Factor IV .159 .000 -.518 -.5.204 .000
Factor III .074 .002 .297 3.231 .002


Factor I ‘resourceful / dependable’ 
Factor II ‘intrusive / controlling’
Factor III ‘risk-taking / stimulation-seeking’
Factor IV: ‘caring / harmonious’

(b) Regression of factors upon actual self-concept

Results for the regression upon ASC% are displayed in Table 3b. Only two factors appeared to predict ASC%. Factor IV (‘caring / harmonious’) accounted for a substantial 45.3% of variance, yielding a positive Beta value. Beyond this, factor II (‘intrusive / controlling’) accounted for another 6.1%, with a negative Beta value. In short, participants’ actual self-concepts expressed underlying qualities of caring and social harmony, as well as rejecting qualities of intrusiveness and control.

Table 3b: Actual self-concept (ACS%)

  R square change  Sig.  Beta  t-value  Sig.
Factor IV .453 .000 .650 8.188 .000
Factor II .097 .003 -.248 -3.120 .003

(c) Regression of factors upon ideal self-concept

Regression results for ISC% are displayed in Table 3c. Again, two factors appeared to underlie ISC%. First factor IV (‘caring / harmonious’) accounted for a substantial 48.4% of variance. Beyond this, factor I (‘resourceful / dependable’) accounting for a further 9.7%. Both factors yielded positive Beta values. In short, participants’ ideal self-concepts expressed underlying qualities of caring and harmony, as well as of resourcefulness and dependability.

Table 3c: Ideal self-concept (ICS%)

  R square change  Sig.  Beta  t-value  Sig.
Factor IV .484 .000 .518 6.141 .000
Factor I .061 .003 .359 4.263 .000

(d) Regression of factors upon aspirations to acquire traits

Results for the regression upon AA% scores are displayed in Table 3d. Only one factor appeared to underlie AA%. This was factor I (‘resourceful / dependable’), accounting for 49% of variance, and yielding a positive Beta value. The participants’ aspirations for change therefore seemed to be towards greater resourcefulness and dependability, and little else.

Table 3d: Aspirations-to-acquire traits (AA%)

  R square change  Sig.  Beta  t-value  Sig.
Factor I .490 .000 .700 8.714 .000

(e) Regression of factors upon aspirations to lose traits

Results for the regression upon AL% scores are displayed in Table 3e. Two factors underlay AL%. First, factor I (‘resourceful / dependable’) had an effect, accounting for 17.9% of variance and yielding a negative Beta value. Beyond this, factor IV (‘caring / harmonious’) had an effect, accounting for a further 7.8% and yielding a positive Beta value. In short, participants’ aspirations for change seemed to be away from underlying qualities of non-resourcefulness and non-dependability, as well as those that indicate caring and social harmony.

Table 3e: Aspirations-to-lose traits (AL%)

  R square change  Sig.  Beta  t-value  Sig.
Factor I .179 .000 -.528 -5.182 .000
Factor IV .078 .005 .322 2.862 .005



Summary of findings

  1. Participants’ beliefs about gender-trait stereotypy were predicted by factors I, II and III (all expressing the male stereotype) and by factor IV (expressing the female stereotype).
  2. Participants’ own actual self-concepts were predicted by factor IV (a female-stereotypic factor) and counter-predicted by factor II (a male-stereotypic factor).
  3. On the other hand, ideal self-concepts were predicted by factor IV (female-stereotypic) and factor I (male-stereotypic).
  4. Aspirations to acquire desired traits were predicted by factor I (male-stereotypic).
  5. The same factor counter-predicted aspirations to lose traits. Beyond this effect, there was a predictive effect for factor IV (female-stereotypic) revealing a paradoxical tendency to cast aside traits underlain by this factor, even though this same factor underpinned ideal self.

Regression for M%

Turning first to the results of the regression for M%, one might query how the typical view is that intrusiveness and control, resourcefulness and dependability, risk-taking and stimulation-seeking are stereotypically male qualities, while caring and social harmony are stereotypically female qualities. To answer this question the authors have conducted an analysis identical to the one here, using the M% data from Williams and Best’s pooled international (non-transgender) sample (1990). Full details of the analysis are beyond the scope of this paper, and in any case are available from this paper’s first author. For the present, the findings can be summarised thus: (a) all four factors predicted gender-trait stereotype, (b) the factors entered the regression equation in the order II, I, IV, III, (c) factors I, II and III underlay male-stereotypical traits, and (d) factor IV underlay female stereotypy. In short, the regressions were in all important respects closely similar to our own. We may conclude that the ideas of the MtF transgenders in our study about what constitutes ‘maleness’ are very similar to those of men and women worldwide. In retrospect this should not surprise us. It may be that we are observing some historical and pervasive pan-cultural gender beliefs. Alternatively, we may be seeing an effect of increasingly globalised (Westernised?) culture in which many of us, at least those who are urbanised, tend more and more to form similar beliefs about the world around us, including about the qualities of men and women. 

Actual self-concept

We now turn to the findings on actual self-concept. Beyond endorsing traits underlain by caring and social harmony, participants resolutely avoided endorsing any underlain by intrusiveness or control. At one and the same time they appeared to embrace a set of highly female-stereotypic qualities, and reject a set of male ones. Together then, these two forces appeared to underlie the highly female-stereotypic concept of current self reported in our earlier paper.

Interestingly, actual self-concept seemed not to be defined by ‘resourcefulness / dependability’ (factor I) or ‘risk-taking / stimulation-seeking’ (factor III) – whether by the presence or absence of those qualities. We may conclude that these underlying qualities have little part to play in the way our participants defined their current selves. Of course, the possibility exists that these two factors were very much a part of their actual self, but that in the collectivist and Buddhist culture of the Thais it is not considered seemly, at least among MtF transgenders, to report that one has them. Interesting though this possibility is, it is beyond the interest of this study, which is to identify the underlying features for the endorsements they did make. 

In passing, we should note the very great predictive role shown by factor IV (‘caring / harmonious’) upon actual self-concept, an effect much greater than upon gender-trait stereotype. The contrast was clearest when a forced entry regression was performed, employing that factor as the only predictor. Entered alone, it accounted for 45% of variance in ASC%, but only 4.9% of M%. We may conclude that, though our participants were living in a female gender role, for them the qualities of caring and social harmony constituted even more powerful forces for defining self than for defining femaleness itself.

Finally in our discussion of actual self-concept, we should note that the most important factor underlying ASC% (factor IV) incorporated moderate loadings from FAV and PI. To the extent that participants judged the traits they possessed as favourable and important, and on the premise that FAV and PI scores might tell us anything about self-esteem (Le Verrier, 1987), we might conclude that self-esteem was reasonably high for this sample.

Ideal self-concept

We now turn to the findings on ideal self-concept. Participants again strongly endorsed qualities underlain by factor IV (‘caring and social harmony’). As was the case for actual self-concept, this factor appeared to be very important indeed. It accounted for 48% of variance in ISC%. We may conclude again that, though our participants’ ideal was to live in a female gender role, for them the qualities of caring and social harmony constituted even more powerful forces for defining ideal self than for defining femaleness. As was noted earlier, this factor incorporated moderate loadings from FAV and PI. This is unsurprising, as we would expect people either to view favourably and important those traits which they have as ideals, or, alternatively, to incorporate favourable and important traits into their ideal for self. Williams and Best report a similar finding in their 1990 study with non-transgender males and females.

Beyond the effect of factor IV, there was an effect for factor I (‘resourceful / dependable’). It is interesting to note the contrast between the roles played by this factor in actual and ideal self. While resourcefulness and dependability had no part to play in defining current self, it clearly had an important role to play in defining the ideals of the participants, accounting for nearly 10% of additional variance. Indeed, when entered alone it accounted for 37.9% of variance in ISC%. This compares with only 11.3% of variance in M%. We may therefore speculate that resourcefulness and dependability play a more powerful part in defining participants’ ideals for self than in defining ‘maleness’. 

We should note that, in endorsing traits underlain by these two factors participants were at one and the same time embracing female- and a male-stereotypical qualities. It appears to be these two underlying elements, in particular the idealisation of resourcefulness and dependability, that lead to the gender-ambiguity in ideal self that was reported in the Winter and Udomsak (2002) report. 

Interestingly, ideal self-concept seemed neither to incorporate or exclude intrusiveness and control or risk-taking / stimulation-seeking. It therefore appears that these underlying qualities have little part to play in the ways our participants defined their ideal selves.

Aspirations to change

We turn last of all to participants’ aspirations to change. Participants aspired to acquire traits (i.e. acquire those they wanted that they did not currently possess) that loaded highly on factor I (‘resourceful / dependable’). Beta values were positive. In a rather symmetrical arrangement of affairs, they aspired to lose traits (i.e., lose those they possessed but did not want) that were not loaded on this factor.

To the extent that this factor was somewhat male-stereotyped (a bivariate correlation of 0.336 and explaining an additional 9.3% of M% variance beyond that explained by factor II), it is clear that, perhaps paradoxically, participants were in important ways aspiring away from a female-stereotyped self and towards a somewhat more male-stereotyped alternative.

Beyond the effect of factor I, participants aspired to lose traits that were loaded on factor IV - the ‘caring / harmonious’ factor. This factor, recall, was the ‘female’ factor (an important element in both actual and ideal self). The finding seems doubly paradoxical. Firstly. it again suggests a group tendency to abandon a ‘female’ element of self. Secondly, it suggests a group tendency to cast aside an important underlying element of ideal self! This second paradox is perhaps more apparent than real. We should also bear in mind how important a role this factor played in defining actual self. In seeking to cast aside traits loaded on this factor, participants – or some of them – may only be abandoning undesired traits (perhaps a small number) that lurk in an overall desired set. One might almost describe them as ‘surplus to requirements’ within that set which loads on factor IV.

What are these surplus traits? A search for outliers high on both AL% and factor IV loadings revealed three traits of note: ‘emotional’, ‘sensitive’ and ‘soft-hearted’. It is above all these that, while contributing to ‘femaleness’, were seen as unwanted baggage by a substantial portion (in each case around a third) of the sample. It would be interesting to try to identify what had aroused such aspirations for change. For the present, the mere existence of aspirations away from a stereotypically female self towards a more gender-ambiguous one is clear evidence of a set of personal growth goals unhampered by strict gender considerations.


To summarise our findings on gender trait stereotype: our Thai MtF sample appeared to view maleness and femaleness in much the same way as people might elsewhere, incorporating four underlying elements: 'resourcefulness / dependability', 'intrusiveness / control', 'risk-taking / stimulation-seeking', and 'care / harmony'.

To summarise our findings on transgendered self: actual self in our sample embraced an underlying female element of caring and social harmony, as well as rejecting a male element of intrusiveness and control. These forces therefore displayed consistency, each with the other, and each with the participants’ chosen gender roles. Ideal self similarly embraced the female element of caring and social harmony, but also revealed a group tendency towards an underlying male element of resourcefulness and dependability. To this extent there was a degree of inconsistency in the forces shaping ideal self.

The aspirations of the participants to change were towards traits expressing the male element of resourcefulness and dependability, and resolutely away from those that did not. Beyond this the group sought to cast aside a limited number of traits underlain by the female element of caring and social harmony. To this extent there was a degree of inconsistency in the forces shaping aspirations for change. 

The idealisation of aspirations towards stereotypically male qualities of resourcefulness and dependability, as well as the desire to cast aside certain stereotypically female qualities of caring and social harmony, all point to a sophisticated set of personal growth goals focused much more on perceptions of maturity and personal efficacy than upon narrow conformity to a female gender stereotype. One is reminded of the recent conclusion of Best and Williams (1998: 116) that:

…the use of the concepts of masculinity and femininity in the study of men’s and women’s self-concepts is not a particularly useful one. Men and women do differ in their self-perceptions, but these differences are best approached by concepts other than the overworked and often poorly defined concepts of masculinity and femininity.

The current study throws light on what concepts may be of use, at least in research with transgenders.


Closing comments

By way of closing comments, we should note that the research reported here displays limitations. Those that apply to the ACL methodology have been discussed in our earlier study (Winter and Udomsak, 2002). Additional criticisms arise out of the 14 psychological features employed as tools for analysing the data. The criticisms relate to concepts and metrics. Let’s take them in turn.

Firstly, notwithstanding cross-cultural research findings to the contrary (e.g. Osgood et al., 1975; McCrae and Costa, 1997; Williams et al., 1998) the three dimensions of affective meaning (FAV, STR and ACT), the five higher order personality factors (EXT, AGR, CON, EMS, OPN) and psychological importance (PI) may not be entirely relevant for non-Western cultures such as Thailand. The same may be true for the five ego states (CP, NP, A, FC, AC). All are derived from the writings of Western academics (Osgood et al. (1957), Tupes and Christal (1961), Berne (1961) and Allport (1937). The psychological features that these writers identified, and which we use here to differentiate among ACL traits, may not be the most salient ones for Thailand. Others may be more salient. 

Secondly, even if the concepts employed are relevant to Thai culture, there remains a question about the way in which the psychological feature scores used in this study were obtained. All were drawn from research with non-Thai samples. For two (FAV and PI) the scores came from a pooled international sample (Thailand not represented). The other twelve came from US samples. If Thai samples (especially MtF transgenders) had been used then the ACL traits may have been rated differently, and entirely different factor structures (and therefore regression equations) may have resulted. 

By way of example, PI seems to play a less important role in predicting M% for our Thai transgenders than was the case for Williams and Best’s 1990 international sample of men and women (nontransgendered). The bivariate correlations reflect this: 0.090 versus 0.226 (p=0.042) respectively. The difference may arise out of the special nature of Thai culture, since the correlation between PI and M% in the Thai men and women in Williams et al's (1998) study was similarly small (0.008). It therefore appears that, while there is a global tendency to view male traits as more important than female traits, Thai culture views them as equally important. On this last point note the finding by Arrindell et al. (1997) that Thailand is, in world terms, a relatively ‘feminine’ society. It would be interesting to know what role this aspect of Thai culture might play in accounting for the apparently large numbers of MtF transgenders in Thailand.



Psychological Features Employed in Step One of the Re-analysis

Affective meaning scores

Osgood et al. (1957) argued that words carry, along with their denotative meanings, a range of connotative (or affective) meanings, each of which can be characterised on three dimensions: Evaluation, Potency and Activity. Cross-cultural work has confirmed the relevance of these ideas worldwide (e.g. Osgood et al., 1975). More recently, Williams and Best (1990) asked US university students to rate each of the 300-ACL traits on each of the three affective meaning dimensions, two of them renamed Favourability (FAV) and Strength (STR). Scores were computed for each trait that represented affective meaning on each dimension. These scores had a mean of 500 and a standard deviation of 100.

A subsequent cross-national study by Williams et al., (1998) provided FAV scores for ten countries spanning five continents, this time presented in a range from 1 (very unfavourable) to 5 (very favourable) with a scale mid-point of 3 (neutral). As well as presenting scores for each country, the authors reported pooled data from all the countries involved. 

Given a choice between two sets of FAV scores, those from the international sample (Williams et al., 1998) were preferred to those from the US sample (Williams and Best, 1990). In any case preliminary analysis suggested that the latter scores were less closely correlated with the dependent variables of interest. Consequently, the STR and ACT scores of Williams and Best (1900) and the FAV scores of Williams et al. (1998) - providing a total of three scores for each ACL trait - were employed in the re-analysis of our Winter and Udomsak (2002) data.

Ego-state scores

Transactional analysis theory describes five ego states: Critical (or Controlling) Parent, Nurturing Parent, Adult, Free Child and Adapted Child. Williams and Best (1990:102–103) describe the five ego states thus:

The Critical Parent (CP) ego state designates a set of feelings, attitudes and behavior patterns that resemble those of parental figures and represents that part of personality which criticises, finds fault, and reflects the rules of society and the values of the individual. The Nurturing Parent (NP) ego state represents a parental figure that nurtures and promotes growth. The Adult (A) ego state represents patterns that are adapted to current reality and are used for logical reasoning and precise predictions. The Free Child (FC) ego state is a relic of an individual’s own childhood and is characterised by fun, frivolity, self-indulgence, and natural spontaneous feelings. The Adapted Child (AC) ego state is also a relic of childhood and is manifested by behaviors that are inferentially under the domination of parental influence and are characterised by conforming and compromising behaviors.

Berne (1961) and others have argued that, while all of us displays each ego state at some time when interacting with others, individuals differ in the degree to which any state assume primacy. Drawing on the opinions of expert raters in the US, Williams and Best (1990) provided, for each of the 300 ACL traits, scores indicating the degree to which they indicate each of the five ego states. Ratings ranged from 0 (not at all descriptive of the ego state in question) to 4 (highly descriptive of the ego state), with a notional mean of 2.

The Williams and Best CP, NP, A, FC, and AC scores (1990) (a total of five scores for each ACL trait) were employed in the re-analysis of our Winter and Udomsak (2002) data.

Higher-order personality factor scores

Tupes and Christal (1961/1992) and Norman (1963) proposed that major dimensions of personality can be conceived in terms of five factors – Extraversion, Agreeableness, Conscientiousness, Emotional Stability (sometimes viewed in terms of its opposite, Neuroticism) and Openness (EXT, AGR, CON, EMS and OPN respectively). Drawing on Formy-Duval (1993), we can describe the five factors thus. High EXT persons are sociable, prefer stimulating environments to relaxed ones, are active, outgoing and assertive. Talkative, frank, adventurous, energetic and enthusiastic persons may be regarded as high in EXT. High AGR persons are kind, loving and trusting. Affectionate, cooperative, sensitive, good-natured, gentle and warm people may be regarded by others as high in AGR. High CON individuals are organised and motivated towards achievement, self-disciplined and competent. Deliberate, dependable, responsible, thorough, efficient, persevering, scrupulous and reliable persons may be regarded as high in CON. High OPN people are curious and exploratory, trying the new and unfamiliar as opposed to the commonplace. Artistic, imaginative, insightful, intelligent, original, clever, polished, inventive, sophisticated and foresighted people may be regarded as high in OPN. High EMS persons are ‘cool, calm and collected’, contented and stable. Low EMS persons are anxious, hostile and impulsive, and display their emotions frequently. Nervous, tense, highly-strung, moody, temperamental, touchy and emotional persons may be regarded as low EMS.

Support for five-factor theory comes from a range of research, including a cross-cultural study by McCrae and Costa (1997). In a US study, Formy-Duval, working alone and with others (Formy-Duval, 1993; Formy-Duval et al., 1995) asked university students to rate each of the 300-ACL traits on each of the five factors. From these ratings scores were computed for each factor on each trait. Scores on the five-factor (ACL-FF) scoring scheme range from 1 (highly counter-indicative of the personality factor in question) to 5 (highly indicative), with a notional mean of 3. Williams et al. (1999) used this ACL-FF scoring scheme to re-analyse the Williams and Best (1990) cross-cultural data. The same scores (five for each ACL trait) were employed in the re-analysis of our Winter and Udomsak (2002) data.


PI scores

Building on Allport’s (1937) notion of central and secondary traits, Williams et al. (1998) examined, in 20 countries across 6 continents, the relative psychological importance (PI) of each of traits in the 300-ACL. More ‘important’ traits ‘provide more information concerning the individual’s psychological make-up and, hence, are more useful in understanding and predicting behavior. More important traits may tell us more about “what the person is really like” (Williams et al, 1998: 25). Psychologically important traits are in that sense more fundamental (or ‘core’) than less important ones.

For each country Williams et al. (1998) reported PI scores ranging from 1 (little or no importance) to 5 (critical or outstanding importance) with a notional mean of 3. They also reported cross-national PI scores based on the pooled data. Those scores (one for each ACL trait) were employed in the re-analysis of our data (Winter and Udomsak, 2002).



For advice and information: John Sachs of the University of Hong Kong, and John Williams of Georgia State University, USA; for questionnaire translation: Pattaranit ‘Poom’ Imampai; for help in distributing questionnaires: Werachat ‘Tor’ Suttipan, Paniti ‘Diana’ Thongtuk and Warachai ‘Ton’ Kuadkengkrai.



Allport, G.A. (1937) Personality: A psychological interpretation. New York, NY: Holt.

Arrindell, W.A., Hatzichristou, C., Wensink, J., Rosenberg, E., van Twillert, B., Stedema, J., and Meijer, D. (1997) Dimensions of national culture as predictors of cross-national differences in subjective well-being. Personality and Individual Differences, 21(1): 37–53.

Berne, E. (1961) Transactional Analysis in psychotherapy: a systematic individual and social psychiatry. New York: Ballantine Books.

Best, D.L. and Williams, J.E. (1998) Masculinity and femininity in the self and ideal self descriptions of university students in 14 countries. In G. Hofstede, (Ed.) Masculinity and femininity: The taboo dimension of national cultures. Thousand Oaks, CA: Sage.

Cole, C.M., O’Boyle, M., Emory, L.E., and Meyer, W.J. (1997) Comorbidity of gender dysphoria and other major psychiatric diagnoses. Archives of Sexual Behavior, 26: 13–26.

Fleming, M., Cohen, D., Salt, P., Jones, D., and Jenkins, S. (1981) A study of pre- and post-surgical transsexuals’ MMPI characteristics. Archives of Sexual Behavior, 10: 161–170.

Fleming, M.Z., MacGowan, B.R., and Salt, P. (1984) Female-to-male transsexualism and sex roles: self and spouse ratings on the PAQ. Archives of Sexual Behavior, 13: 51–57

Formy-Duval, D.L. (1993) Scaling the Adjective Check List for the Five-Factor Model of personality. Unpublished Master’s thesis. Wake Forest University.

Formy-Duval, D.L., Williams, J.E., Patterson, D.J., and Fogle, E.E. (1995) A “Big Five” scoring system for item pool of the Adjective Check List. Journal of Personality Assessment, 65: 59–76.

Gough, H.G. and Heilbrun, A.B. (1980) The Adjective Check List manual. Palo Alto, CA: Consulting Psychologists Press.

Kuiper, A.J., and Cohen-Kettenis, P.T. (1998) Gender role reversal among postoperative transsexuals. International Journal of Transgenderism, 2 (3): http://www.symposion.com/

Le Verrier, L.C. (1987) Theoretical and methodological issues related to masculinity and femininity. Unpublished Master’s thesis. Wake Forrest University

McCrae, R.R. and Costa, P.T. (1997) Personality trait structure as a human universal. American. Psychologist. 52: 509–516.

Norman, W.T. (1963) Toward an adequate taxonomy of personality attributes: replicated factor structure in peer nomination personality ratings. Journal of Abnormal and Social Psychology. 66: 574–583.

Osgood, C.E., Suci, G.J., and Tannenbaum, P.E. (1957) The measurement of meaning. Urbana: University of Illinois Press.

Osgood, C.E., May, W.H., and Miron, M.S. (1975) Cross-cultural universals of affective meaning. Urbana: University of Illinois Press.

Skrapec, C., and MacKenzie, K.R. (1981) Psychological self-perception in male transsexuals, homosexuals and heterosexuals. Archives of Sexual Behavior. 10: 357–370.

Tupes, E.C. and Cristal, R.E. (1961) Recurrent personality factors based on trait ratings. USAF ASD technical report no.61–97. (reprinted in Journal of Personality, 1992, 60: 225–251).

Williams, J.E. and Best, D.L. (1990) Measuring sex stereotypes: A multi-nation study. Newbury Park, CA: Sage Publications. 

Williams, J.E., Satterwhite, R.C., and Saiz, J.L. (1998) The importance of psychological traits: A cross-cultural study. New York: Plenum Press.

Williams, J.E., Satterwhite, R.C., and Best, D.L. (1999) Pan-cultural gender stereotypes revisited: the Five Factor Model. Sex Roles, 40, 513–525.

Winter, S. and Udomsak, N. (2002) Male, female and transgender: Stereotypes and self in Thailand. International Journal of Transgenderism, 6 (1): http://www.symposion.com/

Correspondence to Dr. Sam Winter