Reliability and validity of the self‐perception profile for adolescents: An Australian sample

Abstract

In this study, we explored the psychometric properties and factorial validity of Harter's Self‐Perception Profile for Adolescents (SPPA) with an Australian sample drawn from the Raine study 14‐year follow‐up. Participants, boys (n = 795) and girls (n = 758) from Grades 8, 9, and 10, completed the SPPA. Cronbach's alpha showed good internal reliabilities for seven of the eight subscales and global self‐worth. Boys had significantly higher athletic (p < .001), physical appearance (p < .001), and romantic (p < .05) self‐perceptions, while the girls perceived significantly higher behavioural conduct (p < .001) and close friendships (p < .001). Exploratory factor analysis yielded similar factors to those reported by Harter with North American adolescents, although cross loadings resulted in one additional factor. Our findings generally supported Harter's conceptualisation of the self as a multidimensional construct, and with minor modifications, the use of the SPPA with Australian adolescents.

Key words:

• adolescents
• Harter
• Raine study
• self‐perceptions

INTRODUCTION

Self‐esteem has long been recognised as playing a dynamic and important role in an individual's emotional experiences, future behaviour, and long‐term psychosocial well‐being. It is also well recognised that the formation of self‐esteem is a complex process, influenced by an individual's socialization experiences, achievement‐related behaviour, and person‐to‐person interaction, that in turn varies according to cultural context. While there is strong support for the importance of self‐esteem in traditional areas of psychosocial health, there is now focus on newer related areas. DuBois and Hirsch (2000) argue that by affording higher priority to linking the study of self‐esteem to other aspects of the development, researchers would be better positioned to understanding self‐processes in adolescence. Recent findings focus on the relationships between perceptions of the self and a range of variables: diet and appearance satisfaction (Barker & Bornstein, 2010), neural activity (Pfiefer et al., 2009), self‐deception and self‐liking (Mar, DeYoung, Higgins, & Peterson, 2006), overexcitability (Rinn, Mendaglio, Rudasill, & McQueen, 2010), body weight and obesity (McCullough, Muldoon, & Dempster, 2009; O'Dea, 2006), emotional expression (Polce‐Lynch, Myers, Kliewer, & Kilmartin, 2001), and smoking behaviour (Glendinning & Inglis, 1999).

As socio‐cultural and ethnic dimensions of self‐esteem continue to be examined in new areas and broader contexts (Akande, 2009; Greenberg, 2008; Michaels, Barr, & Roosa, 2007; Schmitt & Allik, 2005), researchers need to be confident that their choice of self‐perception measurement is appropriate for the sample of interest. While the Self‐Perception Profile for Adolescents (SPPA; Harter, 1988) was designed for use in the USA, it has been widely used in other countries. There is support for its use with French Canadian (Bouffard et al., 2002) and Norwegian adolescents (Wichstrøm, 1995), and one study has tested its validity and reliability in Australia (Trent, Russell, & Cooney, 1994). However, it is documented that self‐esteem is bound to culture, from within (Purdie & McCrindle, 2004) and between cultures (Hattie, 1992). Furthermore, the components of self‐concept can change over time and generations. The adolescent in 2010 is developing in different socio‐cultural contexts than those of 20 years ago. Lifestyles have changed significantly since Harter's conceptualisation of the self and the design and validation of SPPA in her country. Duda and Hayashi (1998) caution researchers when using instruments validated in one culture with another. For example, Moksnes, Moljord, Espnes, and Byrne (2010) found differences in the factor structure of an instrument developed in Australia when employed with Norwegian adolescents. In relation to measures of self‐perception, Harter (1999, p. 134) notes that ‘at minimum any meaningful interpretation requires that these instruments show comparable psychometric properties’. In sum, instruments that were conceptually and psychometrically sound for measuring self‐esteem in adolescent populations in one country several decades ago may not be appropriate for use with youth of today in another country. This article differs from an earlier Australian study by Trent et al. (1994), who worked with a younger sample, omitted a subscale from the SPPA, and added two subdomains (English and Mathematics) that were not a part of Harter's SPPA. We employed data for all eight SPPA subscales and global self‐worth (GSW) with a larger and older sample. To our knowledge, the SPPA in its complete form has not been examined for validity and reliability with an Australian sample of adolescents.

Based on factor analytic and other statistical models (Harter, 1982; Hattie & Marsh, 1996; Shavelson & Marsh, 1986), there is now general acceptance of the multidimensional nature of self‐esteem and related constructs (e.g., self‐concept, self‐perceptions), although there is some debate regarding the sensitivity of global measures of self‐esteem (Bodkin‐Andrews, O'Rourke, & Craven, 2010; Glendinning & Inglis, 1999; Supple & Plunkett, 2011). Currently used measures (DuBois, Tevendale, Burke‐Braxton, Swenson, & Hardesty, 2000; Harter, 1985, 1988; Marsh, 1988, 1991) were designed around this conceptual framework of the self in which there are domain‐specific self‐perceptions and GSW. To accommodate the transition from childhood to adolescence, Harter (1988) developed the SPPA, expanding and redesigning the Self‐Perception Profile for Children (SPPC). According to the changing roles and relationships that adolescents experience in a variety of contexts, Harter proposed that domain‐specific self‐perceptions become more diverse with development. Consequently, in developing the SPPA, Harter added three subscales to the SPPC: romantic, close friendship, and job competence. The SPPA and the SPPC measures reflect not only the multidimensional nature of the self but also the increasing ability of the individual to differentiate among aspects of the self.

Harter explains her model of self‐concept as an integration of two approaches; the need to consider both the multidimensional nature of self‐evaluative judgements, as well as the individual's overall sense of self‐worth. Within this framework, it is essential to recognise that GSW is a construct in its own right, representing the overall judgement about oneself as a person. According to Harter, GSW involves a complex combination of discrete judgements about the self and can be directly assessed. In the SPPA, GSW is not assessed by combining domain‐specific judgements. Thus, Harter conceptually and empirically separates the domain‐specific judgements of competence from the individual's global sense of self‐worth. Harter reports that GSW is qualitatively different from self‐descriptions in each of the specific domains. Although GSW is in part influenced by competence in these domains, it is also influenced by how important one deems the domain. Hence, the relationship between GSW and the other subdomains is not hierarchical.

Harter (1988, 1999) reported varying relationships between GSW and domain‐specific self‐perceptions, with physical appearance showing the strongest relationship. Researchers have consistently supported this finding (Seidah & Bouffard, 2007; Trent et al., 1994; Wichstrøm, 1995). Although the effect of physical appearance on GSW is generally considered a reflection of societal influence on the importance of good looks, Harter goes further to describe this domain of physical appearance as ‘the portable self’, something that an individual carries into every situation. As a result, there is a unique relationship between one's perception of physical appearance, ‘the outer self’ and GSW, ‘the inner self’. How closely this relationship emerges in Australian adolescent culture is of interest in the current study.

During the development of the SPPA, Harter found that gender impacted differently on specific domains of self‐esteem. Physical and biological changes occur within contexts that make adolescence a particularly unique period in which gender role expectations of significant others play a strong role. As boys and girls undergo the many physical and social changes of early adolescence, they must come to terms with the new definitions of their roles as females or males. Consistent with findings reported for gender differences in self‐perceptions across the lifespan, adolescent males have been shown to have higher self‐perceptions in athletic competence, physical appearance, and GSW, than their female counterparts, and these are generally reported in studies using the SPPA (Harter, 1988; Trent et al., 1994). Harter (1988) also reported that female adolescents rated their perceptions higher on the behavioural conduct subscale.

Our purpose in this study therefore was to add to the existing body of research examining the validity and reliability of the SPPA (Harter, 1988) with Australian adolescents. We asked the following questions.

1	Does the SPPA measure the different facets of self‐perception as conceptualised by Harter's (1985, 1988) multidimensional model of the construction of the self?
2	Does the SPPA have acceptable levels of internal consistency with Australian adolescents?
3	What are the relationships between the domain‐specific self‐perceptions and GSW?
4	Do Australian adolescent girls and boys differ in domain‐specific self‐perceptions and GSW?

METHOD

Participants

The adolescents were a subsample (N = 1,568) of a possible 1,860 who participated in the 14‐year‐old follow‐up in the longitudinal West Australian Pregnancy Cohort (Raine) Study. The original study began in 1989 as described by Newnham, Evans, Michael, Stanley, and Landau (1993). The socio‐demographic characteristics of the cohort are broadly representative of the Western Australian population (Li et al., 2008). The subsample participated in the data collection of Harter's (1988) SPPA. Of this subsample, four participants who were in the seventh grade were eliminated from the data analysis as the SPPA was developed on adolescents from the eighth to eleventh grade. Thirteen participants were excluded as information on their grade level was missing. The 1,551 participants remaining in this study included 757 females and 794 males from Grades 8 to 10, with an age range from 157 to 180 months (M = 168, standard deviation (SD) = 2.30).

Measures

Self‐perceptions were assessed using Harter's (1988) SPPA. The scale has 45 items which tap GSW and eight specific domains including: (1) athletic competence, (2) physical appearance, (3) social acceptance, (4) close friendship, (5) romantic appeal, (6) behavioural conduct, (7) job competence, and (8) scholastic competence. Each subscale consists of five statements. Of the 45 items, 20 are reversed to avoid acquiescence. The SPPA employs a structured alternate format designed to avoid socially desirable responses. The participant first decides which statement is most true for her or him, and then proceeds to decide whether the statement is ‘really true’ or ‘sort of true’. The score for each statement ranges from 1 (lowest) to 4 (highest). The items are presented as pairs of statements contrasting two types of adolescents on one characteristic. In our study, a few participants did not respond to some items. Where there were 4/5 responses in a subscale, the mean of the four responses was used. Where a participant answered fewer than four of the five subscale items, their subscale score was removed from the analysis (see Table-1).

Table 1 SPPA subscale and GSW mean (SD) scores for males and females

	N	Total	n	Males	n	Females	Gender difference
Scholastic	1,551	2.86 (0.60)	794	2.88 (0.60)	757	2.84 (0.60)	p = .21
Conduct	1,551	2.90 (0.52)	794	2.83 (0.52)	757	2.98 (0.54)	p < .01
Athletic	1,549	2.84 (0.67)	792	2.97 (0.63)	757	2.69 (0.68)	p < .01
Appearance	1,551	2.70 (0.65)	794	2.82 (0.58)	757	2.53 (0.68)	p < .01
Acceptance	1,550	3.18 (0.54)	793	3.16 (0.53)	757	3.20 (0.55)	p = .14
Friend	1,551	3.33 (0.59)	794	3.18 (0.57)	757	3.48 (0.57)	p < .01
Romantic	1,497	2.64 (0.50)	764	2.68 (0.50)	733	2.61 (0.49)	p < .01
Job	1,507	2.91 (0.52)	770	2.89 (0.51)	737	2.93 (0.53)	p = .11
Global self‐worth	1,551	3.15 (0.53)	794	3.20 (0.46)	757	3.09 (0.58)	p < .01

Procedures

The ethics committee of Princess Margaret Hospital provided approval to carry out the research. Volunteer participants completed the questionnaire individually under the supervision of trained researchers, either in their own home or at the Institute.

Design and statistics

All statistical analyses were undertaken using SPSS version 17 (SPSS Inc., Chicago, IL, USA). Descriptive statistics are provided for each of the subscales. Cronbach's alpha was used to indicate the internal consistency of each subscale. T‐tests explored the effect of gender on each of the self‐perception subscales and GSW.

Factorial validity was explored using confirmatory factor analysis (AMOS v17, SPSS Inc., Chicago, IL, USA). Because the model failed to fit, we used principal components analysis with oblique (promax) rotation to explore the factor patterns in the data. Harter (1988) recommended this approach as correlations were expected between some subscales. Factor loadings of 0.3 or more were considered when interpreting factors. All items from the eight subdomains were included in the analysis. GSW, the overall value placed on oneself by the individual, differs from a specific domain of competence (Harter, 1988). For this reason, Harter recommended that it should not be included in the factor analysis of the SPPA.

RESULTS

Factor analysis

The confirmatory factor analysis (CFA) testing Harter's eight subdomain model did not provide adequate fit statistics (χ² = 3,502, df = 712, p < .01; Comparative Fit Index (CFI) = 0.87; Adjusted Goodness‐of‐Fit Index (AGFI) = 0.85; Root‐mean‐square‐error of approximation (RMSEA) = 0.05). Subsequently, the exploratory principal component analysis yielded nine factors with an eigenvalue greater than 1. These factors accounted for 61.5% of the variance. Eight of the factors identified the same domains of self‐perception (see Table-2) as the SPPA (Harter, 1988); however, there were some cross loadings. The romantic subscale split into two resulting in nine rather than eight factors. Three of the factors paralleled Harter's subscales: athletic, scholastic, and job competence. All of the physical appearance items loaded on a single factor, with one of the items cross loading onto the romantic factor. All five of the social acceptance statements loaded on one factor; however, two statements from the close friend domain loaded on this factor. These two statements also cross loaded onto the close friend factor, which included all five close friend statements with all loadings above 0.3. Four of the five behavioural conduct statements loaded on a single factor, while the fifth statement failed to load substantially on any factor but had low loadings (<0.3) on three factors including behavioural conduct. With our Australian data, Harter's romantic subscale yielded two separate components with all five of the original items loading on one factor, and the two negatively worded statements also forming another factor. The correlations between factors (see Table-3) indicated that there were moderate correlations between the social acceptance and the close friend factors and between behavioural conduct and scholastic factors. The correlation between the two romantic factors was low (−0.11).

Table 2 Promax pattern matrix for the eight subdomains of the SPPA

Scale items	Component
	1	2	3	4	5	6	7	8	9
Athletic competence 3	0.86
Athletic competence 39	0.81
Athletic competence 12	0.79
Athletic competence 21	0.79
Athletic competence 30	0.70
Social acceptance 11		0.86
Social acceptance 2		0.84
Social acceptance 38		0.78
Social acceptance 29		0.70
Social acceptance 20		0.60
Close friendship 35		0.48				0.35
Close friendship 8		0.43				0.40
Physical appearance 13			0.91
Physical appearance 22			0.89
Physical appearance 4			0.82
Physical appearance 40			0.73
Physical appearance 31			0.54					0.48
Scholastic competence 28				0.85
Scholastic competence 37				0.76
Scholastic competence 10				0.75
Scholastic competence 1				0.74
Scholastic competence 19				0.71
Behavioural conduct 34					0.81
Behavioural conduct 7					0.80
Behavioural conduct 16					0.79
Behavioural conduct 43					0.78
Close friendship 17						0.84
Close friendship 44						0.80
Close friendship 26						0.79
Job competence 23							0.85
Job competence 5							0.83
Job competence 41							0.78
Job competence 14							0.50
Job competence 32							0.32
Romantic 24								0.69
Romantic 6								0.65
Romantic 33								0.63
Romantic 15								0.31	0.85
Romantic 42								0.32	0.79
Eigenvalue	8.34	3.63	3.13	2.30	1.92	1.58	1.31	1.22	1.16
% Variance	20.85	9.08	7.82	5.74	4.80	3.96	3.27	3.04	2.91

Table 3 Correlation matrix between self‐perception components

Subscale	1 A	2 SA	3 PA	4 SC	5 BC	6 CF	7 JC	8 R	9 R‐
1 A	1.00
2 SA	0.38	1.00
3 PA	0.38	0.36	1.00
4 SC	0.18	0.29	0.29	1.00
5 BC	0.02	0.25	0.29	0.46	1.00
6 CF	0.12	0.46	0.06	0.10	0.18	1.00
7 JC	0.17	0.32	0.16	0.28	0.17	0.15	1.00
8 R	0.24	0.26	0.20	0.12	−0.09	0.15	0.12	1.00
9 R‐	0.16	0.34	0.23	0.12	0.14	0.16	0.22	‐0.11	1.00

A = athletic competence; SA = social acceptance; PA = physical appearance; SC = scholastic competence; BC = behavioural conduct; CF = close friendship; JC = job competence; R = romantic.

Internal consistency

The Australian sample showed good internal consistency for all subscales except the romantic subscale. Cronbach's alpha ranged from a low of 0.68 for the romantic subscale (N = 1,443) through to a high of 0.87 for the athletic subscale (N = 1,536). In Table-4, internal consistency reliabilities from our study are compared with those reported for the American (Harter, 1988), French Canadian (Bouffard et al., 2002), Australian (Trent et al., 1994), and Norwegian (Wichstrøm, 1995) studies. The latter study used both a structured alternative format and a Likert scale. While some minor variations were evident, there was overall consistency across these studies.

Table 4 A comparison of studies of internal reliability for Harter's SPPA

Subscale	Raine	Harter	Bouffard	Trent et al.	Wichstrøm^a
Athletic	0.87	0.89	0.91	0.89	0.66/0.79
Appearance	0.86	0.86	0.91	0.85	0.76/0.87
Friend	0.82	0.82	0.84		0.75/0.78
Social acceptance	0.81	0.81	0.83	0.85^b	0.56/0.76
Scholastic	0.82	0.82	0.77	0.91^c	0.60/0.69
Conduct	0.77	0.78^d	0.76	0.59	–
Job	0.72	0.74^d	0.73	–	–
Romantic	0.68	0.81	0.63	0.70	0.65/0.75
GSW	0.81	0.85	0.85	0.82	0.68/0.77

a Structured alternative/Likert.

b ^b Combined close friend/social acceptance.

c ^c Combined scholastic competence/math competence.

^d Used Harter's (1988) revised D sample only.

Correlations between GSW and subscales

GSW was most highly correlated with physical appearance (see Table-5), and there were moderate correlations (≥0.4) between GSW and the social acceptance, behavioural conduct, scholastic, and athletic domains. Among the subdomains, there was a moderate correlation between social acceptance and best friend, and between scholastic competence and behavioural conduct.

Table 5 Correlations between global self‐worth (GSW) and specific subscales

Subscales	GSW	Scholastic	Behaviour	Job	Social	Athletic	PA	CF
Scholastic	0.41
Behaviour	0.46	0.48
Job	0.27	0.26	0.18
Social	0.49	0.23	0.19	0.29
Athletic	0.40	0.19	0.06	0.19	0.42
Physical appearance	0.65	0.24	0.24	0.14	0.34	0.43
Close friend	0.37	0.18	0.26	0.24	0.57	0.21	0.18
Romantic	0.32	0.13	0.04	0.27	0.43	0.32	0.33	0.31

Gender differences

As shown in Table-1, adolescent males had significantly higher self‐perceptions than females for GSW, athletic competence, physical appearance, and romantic appeal. The females had significantly higher self‐perceptions of their behavioural conduct and close friendships. There were no significant differences between genders for scholastic, social acceptance, and job competence domains.

DISCUSSION

The main purpose of this study was to examine the psychometric properties and the factorial structure of the SPPA (Harter, 1988) in its original form for use with Australian adolescents. From the results of our analyses and from comparisons made with Harter's (1988) findings for the SPPA, it appears that the psychometric properties and the factorial structure of the original version developed for American adolescents are generally consistent with our Australian sample. Although there were methodological differences between our studies, the French Canadian (Bouffard et al., 2002), Norwegian (Wichstrøm, 1995), and the previous Australian study (Trent et al., 1994) findings reported for the SPPA were generally comparable.

Harter's (1988) model of the construction of the adolescent self was generally supported as eight of the factors identified in our study measured the same facets of self‐perception that she had conceptualised. Three of the factors, athletic, scholastic, and job competence, mirrored the items used in Harter's subscales. Comparisons could not be made with the Trent et al. (1994) study, because they did not include job competence. They argued that the participants in their study, lacking experience in paid employment, would not have the work experiences to give consistent responses on perceived competence in the job domain. However, the age range was older in our sample. We also noted that in all but one of the job competence items, the statements relate to the potential to perform a paid job rather than being in actual current employment. Essentially, even if some respondents had no experience in paid work, they could respond about perceptions of competence as if they had a job. Based on our findings from the factor analysis and the internal consistency reliability, there seems little justification for omission of the job subscale from the SPPA when working with Australian adolescents older than those in the Trent study (Trent et al., 1994).

The physical appearance factor was also consistent with Harter's subdomain, although it had one item that cross loaded with romantic appeal. This item focused on perceptions of being good‐looking, whereas the other statements were more focused on satisfaction with one's looks. When Wichstrøm (1995) used the GSW subscale in the factor analysis, physical appearance loaded with GSW to form one factor, further evidence of the inextricable link between physical appearance and the overall value that a person places on the self as a worthy being.

Our findings for close friendship and social acceptance also were similar to those reported by Harter (1988), with each emerging as separate factors. However, in our sample, there were two close friendship items that cross loaded on social acceptance. These two items focused on the ability to make a close friend, whereas the other items in this subscale asked about having a close friend. Trent et al. (1994) found that although close friendships and social acceptance formed separate factors, they were highly correlated with each other. On this basis, they combined these factors to form one social factor, limiting the possibility for comparison with our study. In our study, the adolescents were older; the correlations between these factors lower; and the factors were better separated. This finding is consistent with Harter's view that there is increased ability to differentiate among domains of competence with age.

For romantic appeal, our findings differed from those reported by Harter (1988). Instead of one well‐defined factor, we found that the two negatively worded items had acceptable loadings with the remaining three positively worded romantic items. However, they also formed a separate factor. Thus, two separate romantic appeal factors emerged. The tendency for negatively worded items to form separate factors has been reported elsewhere (Schmitt & Allik, 2005; Supple & Plunkett, 2011). It was perplexing that in our study, this separation in loading of negatively worded items was isolated to the romantic subscale. A possible explanation may lie in the inability of adolescents in our sample to differentiate romantic appeal when statements are presented negatively. Adolescents might be cognitively aware of romantic attraction (the media makes that very clear), but may have little real‐life experience in the romantic domain. Bouffard et al. (2002) also reported difficulties in the factor structure in relation to romantic attraction and recommend more in‐depth research of this factor, especially when comparing age‐related responses and cultural differences in awareness of romantic attraction.

With behavioural conduct, we suspect that even a slight cultural difference in word interpretation might influence factor loadings. The use of the word ‘feel’ in item 25 could well have been the cause for this item having low loadings (<0.3) across three factors. The statement, Some teenagers feel really good about the way they act but other teenagers don't feel that good about the way they often act did not work with this Australian sample. The other items in this subscale directly ask about ‘doing’ rather than ‘feeling’. Item 25 did not load on any factor, and deleting it from the behavioural conduct subscale resulted in an increase of Cronbach's alpha from 0.77 to 0.80. In the Trent et al. (1994) study, this item loaded with social acceptance rather than with behavioural conduct. A rewording of the statement would be recommended for use with future Australian samples. Even if cultures appear to be very similar, careful attention needs to be given to wording in pencil and paper questionnaires.

There is good correspondence between Harter's (1988) and our findings for internal consistency reliabilities. All subscales were at an acceptable level given that there were only five items per subscale (Pallant, 2005). The most noticeable difference was for the romantic subscale (0.68), which was considerably lower than the coefficient (0.81) reported by Harter. The reliability of the romantic subscale is lower not just in our study, but also in the studies carried out by Bouffard et al. (0.63) in Canada, by Trent et al. (0.70) in Australia, and by Wichstrøm (0.65) in Norway using the structured alternate scale. It is possible that the media influences on the importance of romantic attraction occur at an earlier age in the USA.

For physical appearance and athletic competence, the internal reliabilities were high and consistent with earlier studies (Bouffard et al., 2002; Harter, 1988; Trent et al., 1994; Wichstrøm, 1995). Similarly, the alpha levels for job competence were at an acceptable level and varied little from those reported by Harter (1988) and Bouffard et al. (2002).

With respect to behavioural conduct, our findings for reliability were at an acceptable level and consistent with those reported from the American and French Canadian studies. Trent et al. (1994), with a younger Australian sample, reported much lower reliability for behavioural conduct (0.59). Again, consistent with the findings of Harter (1988) and Bouffard et al. (2002), we found acceptably high alpha levels for social acceptance and close friendships. Internal consistency for scholastic competence was also acceptable and comparable to those reported by Harter and Bouffard et al.

When relationships between specific domains and GSW were examined, Harter (1988) reported that physical appearance was the major contributor to GSW. Our results were consistent with Harter's (1999) conceptualisation of the unique relationship between one's perception of physical appearance, ‘the portable self’, and GSW, ‘the inner self’. Our results also reflected Harter's intercorrelation cluster of scholastic competence and behavioural conduct, in which adult support was considered most relevant. Harter also reported a second cluster comprised of correlations among more peer‐relevant domains; physical appearance, romantic appeal, and social acceptance. The relationships between these peer‐relevant domains were not as strong as those in the Harter study.

Previous studies (Harter, 1988, 1999; Marsh, 1989) have clearly shown that boys and girls differ in level of perceptions of competence, more notably in some domains than others. Our findings were similar to those of Harter (1988) and Trent et al. (1994). The boys in our study had higher perceptions in athletic competence, physical appearance, and GSW. Unlike their studies (Harter, 1988; Trent et al., 1994), we found that the boys also had higher perceptions on the romantic subscale. Physical appearance, athletic competence, and romantic appeal all fall within what Harter (1999) refers to as peer‐relevant domains. Similarly to Harter (1988), girls in our study were higher in perceptions of behavioural conduct, but they also had higher perceptions of close friendships. This too is not surprising, given that girls are more likely to seek out one best friend, than are boys (Harter, 1999). Trent reported no differences between boys and girls for behavioural conduct or with their combined subscale of social acceptance and close friendship. The French study (Bouffard et al., 2002) reported gender differences that were similar to those found in our study.

Cultural differences, language modifications of the SPPA, and differences in sample sizes and age ranges are all limiting factors when making comparisons between our findings and previous validation studies. The cross‐sectional nature of this study prevented examination of test re‐test reliability, an issue that needs to be addressed to pursue longitudinal studies (Shevlin, Adamson, & Collins, 2003). Important differences found by other researchers in the use of particular subscales, such as job competence, social acceptance, close friendships, and romantic attraction, highlight the need for the regular assessment of instruments developed previously, especially if developed in another culture where social and cultural influences may be different. Mitigating the effects of negative bias remains a challenge for future investigators. In our study, with the exception of the romantic subscale, we can conclude that there is general support for the factor validity and reliability of the SPPA, and its use with Australian adolescents.