Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/266223536
Article
CITATION READS
1 2,031
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Louise A Ellis on 12 May 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Self-Concept Research: Driving International Research Agendas
Although the Self Description Questionnaire II (SDQ-II) has widespread applicability to many areas of research of
psychology, its length sometimes limits its utility. This article evaluates a brief, 51-item version of the SDQ-II in a
sample of 1319 secondary school students. The psychometric properties were good; the self-concept specific scales
were reliable (.77- .90; Md = .82), first and higher-order confirmatory factor models fit the data well, and
correlations among the scales were moderate (.04 - .60; Md = .22). Gender and age differences were suggestive of
developmental trends that are consistent with results based on responses on the longer version.
Over the past two decades, the general conception of Instrument: Self-Description Questionnaire II Short
self-concept has undergone a dramatic transformation.
Early researchers treated self-concept as a unidimensional Fifty-one of the strongest items from the original 102-
construct, arguing that given the heavy dominance of an item SDQII were selected, 31 of which are positively
overall, general self-concept factor; it is virtually worded (see Appendix 1). The SDQII-S was intended to
impossible to differentiate among its subcomponents. assess the same seven non-academic scales (physical
Owing to a substantial amount of construct validity ability, physical appearance, same sex relations, opposite
research, most of which was directed by Marsh and sex relations, parent relations, emotional stability and
colleagues (see Marsh & Hattie, 1996), contemporary honesty/truthfulness), three academic scales (mathematics,
theorists rely on a multifaceted, hierarchical construal, verbal and general school) and overall global self-concept.
which becomes increasingly differentiated as the A summary of the 11 scales is presented in Appendix 2.
individual moves from infancy to adulthood. The response format was not changed from the original
The strongest evidence for the multidimensionality of 1-False to 6-True response scale.
self-concept comes from the work with the Self
Description Questionnaires developed for primary school Procedure
students (SDQI), secondary school students (SDQII) and
young adults (SDQIII). These measures are frequently The SDQII-S was administered in group settings
used in self-concept research and have been evaluated to during class hours in the sixth week of the school year.
be among the best in terms of psychometric properties and The SDQII-S was the second measure in a battery of
construct validation (Hattie, 1992; 1996). In previous questionnaires, which were administered by one of the
research on the SDQ II, factor analysis clearly identified authors of study according to the standardised procedures
the 11 SDQII scales (target factor loadings were at least used in previous SDQ research.
.48; median = .68), each scale was reliable (median alpha
= .86), and correlations among the factors were small Statistical Analyses
(median r = .15).
Investigations on self-concept will often require that The statistical analyses described in this study were
the SDQ be administered to students in conjunction with a conduced with SPSS 10.0 and LISREL 8.2 (Joreskog &
number of other measures. However, the length of the Sorbom, 1993). Before any of the analyses were performed,
SDQ instruments may strain students patience and the responses to negatively worded items were reversed so
endurance, particularly if being administered with other that all varied along the one scale, where 6 represented the
measures. The purpose of this study was to evaluate a brief highest level of self-concept and 1 represented the lowest.
version of the original 102-item SDQII measure. The brief Results from the analyses on the SDQII-S were compared
version consists of 51 of the original items and is aimed to with those presented in the SDQII manual (Marsh, 1990).
improve the instruments utility without sacrificing its As there were much fewer students from Year 11
excellent psychometric properties. completing the SDQII-S in the present study, we combined
the results of Year 11 with those of Year 10 for analyses
Method comparing responses of students in different grades.
The confirmatory factor analyses (CFAs) were
Participants conducted with LISREL 8.2 (Joreskog & Sorbom, 2002).
For a detailed description of the CFA procedure using
A total of 1319 students from Grade 7 (n = 934), Grade LISREL refer to Byrne (1998), Joreskog and Sorbom
10 (n = 725) and Grade 11 (n = 65) participated in the study. (1996) or Kelloway (1998). Separate factor analyses were
The mean age of students within each of the grade levels conducted on responses from students from Year 7 and
was 11.95 (Grade 7), 14.93 (Grade 10) and 16.09 (Grade Year 10/11, as well as for the total sample. Following
11). The overall sample consisted of 925 females (53.8 %) Marsh, Balla, and Hau (1996), and Marsh, Balla, and
and 795 males (46.2%). The participants were drawn from 7 McDonald (1988), the Tucker-Lewis index (TLI), the
secondary Catholic schools in New South Wales, Australia. relative noncentrality index (RNI), and root mean square
error of approximation (RMSEA) were used to evaluate
1
Self-Concept Research: Driving International Research Agendas
goodness of fit, but we also present the 2 test statistic. Effects of Gender and Age
RMSEA values less than .05 indicate good fit and values
as high as .08 represent reasonable errors of approximation A series of one-way analysis of variance procedures
in the population (see Marsh et al., 1996; Schumacker & was carried out to detect gender and age differences on
Lomax, 1996). The TLI and CFI vary along a 0-to-1 each of the SDQII-S scales. Due to the sizeable number of
continuum in which values greater than .90 and .95 are tests and considerable sample size, statistical significance
typically taken to reflect acceptable and excellent fits to was set at the .01 level. Previous research on the SDQII has
the data (Bentler, 1990; Bentler & Bonett, 1980). An index consistently found sex differences in Physical Ability,
of .90 can be roughly interpreted as being able to explain Appearance and Math favouring males and in Verbal,
90% of the covariation among the measured variables. Honesty/Trustworthiness and Same-Sex Relations favouring
Correlations among the SDQII-S scales were also females (Marsh, 1987; Marsh, Parker, & Barned, 1985;
examined. The correlations were then analysed separately Marsh, Smith, Marsh, & Owens, 1988). Consistent with the
for Year 7 versus Year 10/11 to determine whether or not original SDQII, we found that Males scored significantly
the correlations are reasonably similar for different ages. higher on Math [F(1, 1718) = 38.64, p = .000] and
Finally, the effects of grade level and gender were Appearance [F(1, 1718) = 11.51, p = .00], whereas
determined by a series of univariate and multivariate females scored significantly higher on
analyses. Honesty/Trustworthiness [F(1, 1718) = 56.55, p = .00],
Same-Sex Relations [F(1, 1717) = 33.26, p = .00] and
Results and Discussion Verbal [F(1, 1717) = 6.94, p = .01]. For the three
combined scales (Academic, Nonacademic and Total
Reliability Estimates. Self), the effect of gender was not statistically significant.
We also sought to examine whether these effects varied
Internal consistency estimates for the SDQII-S scales with age. At an alpha level of .05, we found no significant
for the total sample reached an acceptable level at both the age X gender interactions, suggesting that the effects of
combined scale (range = .87 to .90) and subscale (range = gender may remain relative from Year 7 through Year
.77 to .90) levels, with a median alpha coefficient of .82. 10/11. This finding is concordant with previous research
As shown in Table 1, subscale coefficient alphas were with the SDQII (Marsh, 1990).
somewhat higher for Year 10/11 students (range = .81 to Marsh (1990) reported that there is a consistent
.92) than Year 7 students (range = .67 to .89). pattern of age effects across the range of early and middle
Interestingly, for the three combined scale scores, the adolescents responding to the SDQII. Self-concept
reliability estimates for Year 7 versus Year 10/11 students declines through about grade 8 or 9, levels out, and then
were more similar than the estimates for the individual starts to increase in grades 10 and 11. We found a decline
scales. Coefficient alphas for the SDQII scales in the on many of the SDQII-S scales with age. Students in Year
normative archive (median = .86) were higher than those 7 scored significantly higher than Year 10/11 students on
based on Year 7 responses (median = .80), but very similar Physical Ability [F(1, 1722) = 36.47, p = .00], Appearance
to those based on Year 10/11 responses in the present [F(1, 1715) = 14.78, p = .00], Parent Relations [F(1, 1721)
investigation (median = .85). = 101.61, p = .00], Opposite-Sex Relations [F(1, 1721) =
13.96, p = .00], Math [F(1, 1722) = 26.77, p = .00], Verbal
Factor Analysis [F(1, 1721) = 12.04, p = .00], General School [F(1, 1721)
= 35.04, p = .00] and General Self [F(1, 1722) = 43, p =
For the CFA each measured variable was allowed to .00].
load only on the factor that it was designed to measure and Shavelson, Hubner, and Stanton (1976) proposed that
all other factor loadings were constrained to be zero. This self-concept becomes more differentiated with age.
restrictive, a priori factor structure provided a fairly good Consistent with this proposal, Marsh, Barnes, Cairns, and
fit to the data (TLI = 0.94; CFI = 0.95; RMSEA = 0.07), Tidman (1984) found that the size of correlations
and all factor loadings were statistically significant substantially decrease in the preadolescent years. Marsh
(varying from.41 to .94; median = .71). The mean (1990) has also reported that there is a weak tendency for
factor loading for Year 10/11 students (.74) is higher than correlations to be smaller for the youngest and oldest
that for Year 7 students (.69). subjects completing the SDQII. The pattern of correlations
Correlations among the factors ranged from .04 to .60 for Year 7 versus Year 10/11 students is presented in
(median r = .22). Seventy percent of the correlations were Table 4. Consistent with the findings reported by Marsh
less than .30. These statistics provide support for the (1990), the size of the correlations was somewhat lower
distinctive nature of each factor. In addition, compared for Year 10/11 students (Range = .01 to .55; Median = .20)
with the average of all correlations, a) correlations were than Year 7 students (Range = .08 to .63; Median = .22).
generally higher among the 3 academic factors, b)
In conclusion, this preliminary study suggests that the
correlations were generally higher among the 7 non-
psychometric properties of the SDQII-S are almost as
academic factors (particularly the first 4), and c)
good as those of the full-length measure. Although
correlations between academic and non-academic factors
additional data needs to be gathered from a wider range of
tended to be lower.
secondary school year groups, and information is needed
regarding the validity of the SDQII-S, it appears that the
shorter measure will serve most research needs quite well
2
Self-Concept Research: Driving International Research Agendas
Contact Details
References
3
Self-Concept Research: Driving International Research Agendas
Table 1
Coefficient Alpha Estimates of Reliability at Time 1 for Year7 students, Year 10/11 students, and the
Total Sample, as well as the Number of Items within Each Scale
Coefficient Alpha
Scale Year 7 Year 10/11 Total Normative
(N = 691) (N = 628) (N = 1319) Data on
SDQ II
(N = 5494)
Subscales:
Physical Abilities .79 .86 .83 .85
Physical Appearance .87 .89 .88 .91
Same-Sex Relationships .75 .81 .77 .86
Opposite-Sex Relationships .80 .84 .82 .90
Honesty/Trustworthiness .79 .84 .81 .84
Parent Relations .67 .84 .78 .87
Emotional Stability .79 .78 .78 .83
Verbal .88 .92 .89 .86
Math .89 .90 .90 .90
School .80 .87 .84 .87
General .80 .85 .82 .88
Combined scores:
Nonacademic .87 .87 .87
Academic .88 .89 .89
Total .90 .89 .90
4
Self-Concept Research: Driving International Research Agendas
Table 2
Summary of Parameter Estimates for Confirmatory Factor Analysis of Different Grade Levels and the
Total Group for the Eleven-Factor A Priori Model
Scale Year 7 Year 10/11 Total Group
Physical:
1 .79 .81 .80
2 .92 .95 .94
3 .43 .57 .48
4 .64 .70 .67
Appearance:
1 .84 .86 .86
2 .89 .91 .91
3 .70 .73 .73
4 .73 .74 .74
Same Sex:
1 .71 .72 .67
2 .54 .81 .66
3 .65 .56 .59
4 .63 .65 .63
5 .51 .68 .61
Opp. Sex:
1 .53 .62 .57
2 .88 .88 .88
3 .84 .80 .83
4 .57 .70 .62
Hon/Trust:
1 .51 .68 .60
2 .74 .72 .72
3 .64 .63 .63
4 .59 .75 .67
5 .57 .57 .55
6 .67 .71 .70
Parent:
1 .74 .78 .77
2 .69 .78 .75
3 .74 .79 .79
4 .36 .64 .50
Emot. Stab:
1 .62 .64 .63
2 .59 .62 .59
3 .57 .57 .57
4 .59 .60 .60
5 .82 .83 .83
Verbal:
1 .61 .71 .65
2 .78 .82 .80
3 .75 .87 .81
4 .86 .91 .88
5 .83 .87 .85
Math:
1 .84 .86 .85
2 .88 .93 .90
3 .88 .87 .88
4 .65 .71 .68
School:
1 .51 .72 .60
2 .73 .76 .76
3 .83 .87 .85
4 .85 .87 .86
General:
1 .67 .73 .71
2 .74 .77 .77
5
Self-Concept Research: Driving International Research Agendas
Table 3
Correlations among the SDQII-S Scales
Scale Phys Appr Sm Op Hon Prnt Emt Ver Math Schl Genl
Sx Sx n
Physical 1.00
Appearance .33 1.00
Same. Sex .23 .24 1.00
Opp. Sex .28 .36 .46 1.00
Honesty .10 .08 .28 .12 1.00
Parent .18 .20 .23 .08 .32 1.00
Emot Stab .19 .20 .28 .25 .20 .14 1.00
English .10 .22 .16 .13 .22 .17 .13 1.00
Math .14 .19 .12 .04 .18 .17 .12 .15 1.00
School .17 .26 .21 .13 .32 .30 .17 .56 .58 1.00
General .36 .48 .38 .28 .36 .42 .28 .38 .38 .60 1.00
Note. Self-concept factors are: Phys = Physical, Appr = Appearance, Sm Sx = Same-sex relations, Op Sx =
Opposite-sex relations, Hon = Honesty/Trustworthiness, Prnt = Parent Relations, Emtn = Emotional Stability,
Ver = Verbal, Schl = School, and Genl = General.
6
Self-Concept Research: Driving International Research Agendas
Table 4
Correlations among the SDQII-S Scales for Year 7 and Year 10/11 students
Scale hys Appr Sm Op Hon Prnt Emtn Eng Math Schl Genl
Sx Sx
Year 7
Physical 1.00
Appearance .36 1.00
Same. Sex .21 .24 1.00
Opp. Sex .29 .37 .42 1.00
Honesty .11 .15 .34 .16 1.00
Parent .17 .25 .30 .10 .34 1.00
Emot Stab .18 .17 .24 .23 .18 .09 1.00
English .13 .21 .17 .13 .24 .17 .02 1.00
Math .16 .22 .16 .08 .20 .18 .12 .18 1.00
School .22 .29 .24 .15 .35 .31 .10 .59 .59 1.00
General .38 .47 .39 .28 .37 .38 .18 .40 .44 .63 1.00
Year 10/11
Physical 1.00
Appearance .28 1.00
Same. Sex .27 .23 1.00
Opp. Sex .32 .38 .52 1.00
Honesty .09 .00 .20 .07 1.00
Parent .15 .13 .20 .10 .31 1.00
Emot Stab .20 .24 .34 .28 .23 .19 1.00
English .04 .22 .15 .16 .19 .14 .05 1.00
Math .08 .13 .06 .01 .16 .13 .05 .10 1.00
School .10 .21 .17 .14 .29 .26 .13 .53 .56 1.00
General .32 .48 .40 .34 .34 .41 .32 .34 .30 .55 1.00
Note. Self-concept factors are: Phys = Physical, Appr = Appearance, Sm Sx = Same-sex relations, Op Sx =
Opposite-sex relations, Hon = Honesty/Trustworthiness, Prnt = Parent Relations, Emtn = Emotional Stability,
Eng = English, Schl = School, and Genl = General.
7
Self-Concept Research: Driving International Research Agendas
Appendix 1
8
Self-Concept Research: Driving International Research Agendas
Appendix 2