Académique Documents
Professionnel Documents
Culture Documents
With the resurgence of interest in the self-system in the late 1970s and early
1980s (see Harter, 1983; Leahy, 1985; Suls & Greenwald, 1985) came a
notable shift from emphasis on the self as a generalized, unitary construct to
increased recognition that self-perceptions vary across the domains of one's
life (Damon & Hart, 1982; Harter, 1982, 1983; Marsh, Smith, & Barnes, 1983;
Shavelson, Hubner, & Stanton, 1976). Consistent with a multidimensional
view of self, researchers devised a new wave of domain-specific self-report
instruments to assess children's self-concept within particular content
areas, in addition to general self-worth (e.g., Harter, 1982; 1985; Marsh et
al., 1983; see Byrne, 1996; Wylie, 1989, for reviews). The psychometric
quality of these multidimensional measures has been well documented
(Byrne, 1996; Wylie, 1989), but the area is still plagued with concerns about
how to best tap children's self-perceptions and what different measures tell
us about children's self-knowledge.
To address these issues, two studies were conducted to evaluate the utility
and comparability of domain-specific evaluations of self among
preadolescent, elementary school-age children. In the first study, two of the
most widely used multidimensional measures of children's self-concept were
compared in terms of reliability, validity, and stability, as well as
correspondence across common domains. In the second study, children's
self-perceptions in four major areas (peer relations, schoolwork, appearance,
and physical/athletic ability) were further examined to determine (a) the
correspondence among various types of domain-specific, self-report
approaches (questionnaires, ratings, interview data), (b) the correspondence
between self and others' perceptions of the child's performance in those
domains, and (c) the information children report using to evaluate
themselves in each domain.
STUDY 1
In at least three studies children's responses to the SDQ-1 and the SPPC have
been compared (Byrne & Schneider, 1988; Marsh, 1990; Marsh & MacDonald-
Holmes, 1990). Across studies, the construct validity of both measures was
verified by confirmatory factor analyses. Marsh and MacDonald-Holmes
(1990) also provide support for the convergent and discriminant validity of
both instruments using multitrait, multimethod analyses and found the scales
to be comparable in terms of internal consistency (coefficient alpha = .82 to
.93 for SDQ-1 subscales, and .81 to .86 for SPPC subscales). With regard to
concurrent validity, significant correlations were found between social self-
concept on the SPPC and peer assessments of sociability/leadership (Byrne
& Schneider, 1988) and between academic self-concept on both the SDQ-1
and the SPPC and teacher evaluations of achievement (Marsh & MacDonald--
Holmes, 1990). Importantly, significant, positive correlations were observed
across the SPPC and SDQ-1 for scores in comparable domains, ranging from
.54 to .86 for Grade 5 to 8 students (Byrne & Schneider, 1988) and from .56
to .68 for Grade 5 students (Marsh & MacDonald-- Holmes, 1990).
These data support previous reports of the psychometric quality of the two
scales and further suggest that, despite format and item variations, these two
measures yield similar estimates of self-concept in specific (comparable)
domains. Study 1 is a replication and extension of these findings regarding
the comparability of the scales in a Canadian sample of fifth and sixth
graders. As in prior studies, the two instruments were compared in terms of
(a) subscale interrelations, (b) internal reliability (Cronbach a), and (c)
correspondence between scores obtained on comparable subscales.
Extending previous research, we examined the stability or test-retest
reliability of subscale scores over a 1-week period, and evaluated the
correspondence of self-reports with teacher and peer assessments in each
domain. Although peer and teacher evaluations are often used as relevant
validity criteria, such comparisons generally have not been conducted across
numerous domains.
Method
Results
Comparisons of the SDQ-1 and the SPPC were made on the basis of (a)
reliability (internal consistency, test-retest stability), (b) subscale
interrelations, and (c) validity (correlations with teacher and peer ratings).
The correspondence of the two instruments was examined through
correlations between scores obtained in comparable domains.
Discussion
Data from over 200 fifth- and sixth-grade students revealed an impressive
pattern of consistency across two of the most widely used multidimensional
measures of self-concept. Comparable estimates of internal consistency and
test-retest reliability were obtained, along with highly similar patterns of
correlations across subscales, and across evaluations of self and other. Not
surprisingly, then, self-concept scores in comparable domains were highly
correlated across the two measures. Consistent with prior studies, these data
do not favor one measure over the other, and instead support the conclusion
that the SDQ-1 and SPPC provide comparable and psychometrically similar
assessments of self-concept.
Finally, the children were asked to describe the "cues" they used to develop
their own self-appraisals, by asking them to explain "how they knew" how
competent they were in each domain. Our focus on children's views of the
"data" used to determine self-assessments constitutes a unique focus within
the literature. Previous efforts have been largely attempts to validate
statistically multidimensional and hierarchical models of the self on the basis
of responses to questionnaires. Although useful, such procedures reveal little
about how individuals (especially children) themselves form their constructs
of self within specific domains.
Method
The children participated in one group testing session and one individual
interview over a 2-3 week period (order counterbalanced across classrooms),
and were assured of the confidentiality of their responses in both sessions.
During group testing, children completed the SPPC as well as a series of
rating scales on which they evaluated participating classmates and
themselves along several dimensions (described later). Teachers were asked
to complete a similar series of ratings on participating children. During the
interview, children were asked to (a) respond to a series of open-ended
questions about themselves, (b) rate how well they perceived themselves to
perform in each of four domains, and (c) explain how they determined their
own competencies in each domain.
Children also completed a series of five peer rating scales, providing peer
assessments of overall liking or popularity (sociometric measure) and of each
participant's competence in each of four domains (schoolwork, athletic
ability, peer relations, appearance), identical to those completed for Study 1.
The sociometric measure was administered first, with the order of the
remaining four scales counterbalanced across classrooms. Children's ratings
of their own competence in each of the four domains (completed along with
peer evaluations) provided an additional index of children's self-perceptions
of competence in each domain, one embedded in a context that likely
enhanced social comparisons.
2. Tell me some things about yourself that you think are good.
3. Tell me some things about yourself that you think are not so good.
5. Tell me the things about yourself that you are not really proud of.
Next, children rated on a 10-point response wheel how well they perceived
themselves to perform in each domain. Response wheels were made from
two circles of interlocking cardboard, each a different color. On the face of
each circle were lines that create 10 equal "pie slices" and the interlocking
circles could be rotated to reveal any number of "slices" of either color. Each
colored circle was labeled with either a positive or negative self-description
for a particular domain (e.g., "I get along really well with other kids" and "I
don't get along well at all with other kids" for the peer relations domain; "I do
really well in my schoolwork" and "I don't do well at all in my schoolwork" for
the academic domain). These response wheels, easily understood by the
children, corresponded to 10-point rating scales, allowing for greater
response variability, with higher scores indicating more positive self-
perceptions. They differed from other self-ratings in that (a) they provided a
visual representation of one's self-assessments that may have been more
meaningful for some students and (b) they did not highlight social
comparison (as did self-ratings embedded in the peer evaluations). The
response wheels provided a final set of self-evaluations, completed on a
different day.
Finally, children were asked "how they could tell" if they were doing well or
not doing well in each domain, in an attempt to identify the sources of
information the children used to evaluate their performance in each domain.
Responses were recorded verbatim and later categorized in terms of the type
of "cues" used to determine their self-assessments. Six major types of cues
were described by the participants, although many of the categories only
occurred in particular domains: objective outcomes (grades, test scores,
goals, wins/losses, etc.); direct feedback from others ("__ tells me I am
good"); direct affect or liking (e.g., "They tell me that they like me");
friendship (e.g., "because I have friends"); self-observations (interpretations
of one's own behavior); and performance inferred from the behavior of others
(evaluations of self based on how one is treated by others). The latter two
categories were further divided into several subcategories. Specifically, self-
observations included (a) general self-observations (e.g., "I look at myself in
the mirror," "When I get all dressed up"); (b) self-descriptions of
psychological states (e.g., "I feel good/pretty"); (c) work-related behavior
(references to completion of work, speed of performance, neatness, etc.); (d)
ability/inability to perform tasks (e.g., "I could do it perfectly the first time I
tried"); (e) interpretations of one's own performance (references to ready
understanding, task ease/difficulty, enjoyment, etc.). Performance inferred
from the behavior of others included the subcategories of (a) general
statements (e.g., "The way teachers/coaches treat you"); (b) positive social
behaviors or absence of negative behaviors (e.g., "They invite me to parties,"
"They don't tease me," or "They choose me as captain"); (c) negative social
behaviors or absence of positive behaviors (e.g., "They never include me
when they play"); (d) shared or mutual affect/cooperation (e.g., "We laugh a
lot when we get together," "We have fun together"); (e) communication (e.g.,
"We talk to each other"); (f) intimacy/loyalty/acceptance (e.g., "I can tell my
friends anything," "I can trust her not to tell secrets"); (g) similarity/ shared
values (e.g., "We like the same things"). Unclassifiable and "don't know"
responses were coded separately (see Table 8 for a list of categories).
Interrater reliability, obtained for a random sample of 25% of the children,
was computed across categories and was consistently high across domains
(agreement = 89% for academics; 90%, peer relations; 91%, athletics; and
95%, appearance).
Most surprising were results obtained in the peer relations domain. There was
little correspondence among self-report questionnaire and selfratings and no
relationship between interview measures and self-assessments derived from
questionnaires or rating scales (see Table 5).
In clear contrast, relations between self- and other evaluations were virtually
nonexistent in the domains of peer relations and, especially, appearance. In
the social domain, the correlations between self and peer or teacher
evaluations were sometimes significant but very small in magnitude, and
almost nonexistent when parent evaluations or spontaneous self-descriptions
were considered. In the appearance domain, there were no significant
relations obtained between self- and other assessments, suggesting perhaps
that beauty is indeed in the eye of the beholder.
These results are highly similar to those reported by Marsh and his colleagues
(Marsh, 1988; Marsh & Craven, 1991). Across eight studies in which the SDQ-
1 was used to assess children's self-perceptions and a single-item rating
scale was used to assess teacher perceptions of these children, Marsh (1988)
found average self-other agreement correlations for academic, athletic, peer,
and appearance domains to be .37, .38, .29, and .16, respectively. Marsh and
Craven (1991) evaluated agreement between children's SDQ-1 subscale
scores and perceptions of teachers, mothers, and fathers. Self-other
correlations tended to be somewhat higher than those reported in Marsh
(1988) but the pattern of results was similar, with the highest agreement
observed in the academic and athletic domains and somewhat weaker
agreement observed in the peer domain. The consistency of our results with
those of previous studies supports the robustness of this differential pattern
of self-other agreement across domains.
In both the academic and athletic domains, children most often relied on
rather direct sources of information, such as that obtained from academic or
athletic outcomes including references to goals, scores, winning, losing, etc.,
in the athletic domain (mentioned by 55% of the children) and references to
grades, test scores, marks, etc., in the academic domain (mentioned by 82%
of the children). In addition, a substantial number of children relied on direct
feedback from others, with 42% and 33% of the children mentioning this
source of information in the athletic and academic domains, respectively. It is
important to recall here that in these two domains in particular, direct
feedback from others is likely to be a consistent source of information, given
our earlier findings of a rather high correspondence across teachers, parents,
and peers with regard to children's performance in these two domains.
In summary, the results of the present study suggest that the measurement
of the self remains a difficult task even when the focus is on domain-specific
assessment rather than a global or unitary construct of self. As we have
demonstrated, the correspondence among different self-perception
measures varies considerably as a function of the domain being tapped, with
a reasonable degree of coherence across self-measures observed in the
athletic domain but virtually no relationship across self-measures in the
social domain.
Further, the present results indicated that the correspondence between self
and others' perceptions also varied across domains, with significant
correlations observed between self and other assessments in the athletic and
academic domains, but little or no correlation between self and other
perceptions observed in the social and appearance domains. This pattern of
differential correspondence across domains is not readily attributable to
variations in the consistency of others' perceptions of one's competence
within domains, although the correlations obtained among others'
perceptions were slightly higher in the athletic and academic domains than in
the appearance and social domains.
In conclusion, the results of the present study suggest that the recent shift
within the self literature from more general to domain-specific
selfassessments has led to the finding that children may process information
about the self quite differently across domains. Thinking of these domains as
parallel has been primarily a matter of convenience in previous research, but
on closer inspection such an assumption breaks down. In the present study,
for example, we have demonstrated that the consistency of self-
assessments, the correspondence between self and other evaluations, and to
some extent the consistency of others' evaluation varies considerably across
domains, as do the sources of information children utilize as feedback
concerning their own performance or competence. It would be beneficial to
consider in the future the unique ways in which children process information
regarding themselves as a function of the domain assessed and the
implications of such processing differences for subsequent behavior, as well
as self-evaluation.
[Reference]
REFERENCES
[Reference]
BYRNE, B. M. (1996). Measuring self concept across the life span: Issues and
instrumentation. Washington, DC: American Psychological Association.
BYRNE, B. M., & SCHNEIDER, B. H. (1988). Perceived competence scale for
children: Testing for factorial validity and invariance across age and ability.
Applied Measurement in Education, 1, 171-187.
[Reference]
[Author Affiliation]
This research was supported by grants from the University of Waterloo and
the Social Sciences and Humanities Research Council of Canada. Portions of
this paper were presented at the University of Waterloo Conference on Child
Development, Waterloo, ON, May 1988. Some of the data were collected as
part of an unpublished undergraduate honors thesis by Vanessa Ploc, and as
a pilot study for an unpublished doctoral thesis by Dr. Annie Steinhauer. We
thank the participating staff, parents and students at Alpine, Crestview, and
Southridge public schools in Kitchener-Waterloo, ON, for their cooperation in
this research, and Vanessa Ploc and Annie Steinhauer for their assistance in
data collection.
Merrill-Palmer Quarterly, October 1999, Vol. 45, No. 4, pp. 602-623. Copyright
1999 by Wayne State University Press, Detroit, MI 48201
Copyright Wayne State University Press Oct 1999. Provided by ProQuest LLC. For permission to reuse this
article, contact Copyright Clearance Center.