top-rated free essay


By BrianBarr01 Apr 10, 2015 14819 Words
Self-Concept: Validation of Construct Interpretations
Author(s): Richard J. Shavelson, Judith J. Hubner and George C. Stanton Source: Review of Educational Research, Vol. 46, No. 3 (Summer, 1976), pp. 407-441 Published by: American Educational Research Association

Stable URL:
Accessed: 04-04-2015 14:12 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact

American Educational Research Association is collaborating with JSTOR to digitize, preserve and extend access to Review of Educational Research.

This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions

Review of Educational Research
Summer 1976, Vol. 46, No. 3, Pp. 407-441

Self-Concept: Validation of Construct

Richard J. Shavelson
Judith J. Hubner
George C. Stanton
Stanford University

Historically, education goals have tended to fluctuate from
emphasis solely on cognitive outcomes to major concern with
social and affective ones. The emphasis on achievement and the "cult of efficiency" (Callahan, 1962) early in this century was followed by a shift in the 1930's to the comprehensive high school with its social and affective concerns (cf. the Eight Year Study, Aikin, 1942). Then Sputnik initiated a rapid and dramatic reemphasis on cognitive outcomes (Bruner, 1960) from which the current trend seems to be moving in its emphasis on "humanistic" aspects of education. The sharp increase in the number of studies on self-concept is one reflection of the reemphasis on noncognitive

outcomes of
education. (For references to current educational studies, see reviews by Coller, 1971; Purkey, 1970; Yamamoto, 1972; Zirkel, 1971.) Another symptom of this shift has taken the form of
increased concern with enhancing the child's self-concept, espeThe authors wish to express their appreciation to Professors Lee J. Cronbach and Nathaniel L. Gage for their valuable, critical comments on earlier drafts of this paper.

The research reported herein was conducted at the Stanford Center for Research and Development in Teaching, which is supported in part by the National Institute of Education, Department of Health, Education, and Welfare. The opinions expressed in this publication do not necessarily reflect the position, policy, or endorsement of the National Institute of Education. (Contract No. NE-C-00-3-0061.)


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

cially on the part of Head Start teachers (e.g., Hoepfner, Stern, & Nummedal, 1971). According to Zirkel (1971, p. 211),
It has become increasingly
clear in the light of the
that the
schools' attempt to serve the disadvantaged
schools have a fundamental responsibility to enhance the
of their students [Clark, 1963; Marston,
1968; Tannenbaum, 1967].
This objective has been prescribed and described for virtually all (Fantini & Weinstein, 1968;
programs for the disadvantaged
Smiley, 1967; Gordon & Wilkerson, Note 1). Thus, improvement of a student's self-concept seems to be valued as an educational outcome in its own right.
But even if self-concept were not so valued, the construct has potential scientific importance for interpreting achievement outcomes. Most definitions link this construct to achievement, and there is some empirical evidence to support this theoretical linkage (Brookover, Le Pere, Hamachek, Thomas, & Erickson,

Note 2; Torshen, Note 3). Self-concept, then, whether used as an outcome itself or as a moderator variable that helps explain achievement outcomes, is a critical variable in education and in educational evaluation and research.

between a
Most self-concept
studies examine correlations
measure of self-concept and measures of other constructs (e.g., Coopersmith, 1967; Sears, Adenubi, Bloch, Hubner, Gamble, &
Crist, 1972; Brookover et al., Note 2; Torshen, Note 3; Bledsoe & Garrison, Note 4; Sears, Note 5); differences in mean self-concept scores among different populations of students (e.g., Hishiki, 1969; Soares & Soares, 1969; Zirkel, 1971); and changes in selfconcept attributable to some treatment (e.g., Herbart, Gelfand, & Hartman, 1969; Long, Ziller, & Henderson, 1968; Ludwig &

Maehr, 1967; Zirkel, 1971, 1972). Taken individually, they often provide important insights into the factors that motivate students in and out of school and into alternative courses of action that may enhance students' self-concepts

(e.g., Purkey, 1970;
Sears et al., 1972; Yamamoto, 1972; Spaulding, Note 6).
Yet, considered as a body of research, self-concept studies
today may be criticized in the same way as Crowne and Stephens (1961) and Wylie (1961) did almost twelve years ago, in that the of their measurements
may not be
self-concept interpretations
valid. First, definitions of self-concept are imprecise and vary from one study to the next. The imprecision makes it extremely difficult to specify (a) the population of self-concept items from which a representative

sample would be drawn for the instrument, or (b) the population of subjects for which the measurewould be appropriate. And, ment techniques and interpretations


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



as for variability in definitions, a review of such definitions (e.g., 1952; Bruner,
1958; Combs & Soper,
Coopersmith, 1967; Hamachek, 1965; James, 1963; Jersild, 1952; McDonald, 1965; Mischel, 1968; Mote, 1967; Piers & Harris, 1964; Rogers & Dymond, 1954; Sears & Sherman, 1964; Sherif & Cantril, 1947; Snygg & Combs, 1949; Sullivan, 1953; Symonds, 1951; Wylie, 1961; Brookover et al., Note 2; Bledsoe & Garrison, Note 4; Gordon, Note 7; Purkey, Note 8) revealed seventeen different conceptual dimensions on which they could be classified. Some of these dimensions were (a) emphasis on a stable or changing self-concept; (b) methods for changing self-concept-learning/reinforcement, creation of dissonance, or arousal of needs and defenses; (c) of self-concept-situational,

or inphenomenal,
ternal; (d) types of evaluation-normative
standard, absolute
and (e) dimensionality
personal standard, or nonevaluative;
or multidimensional.
A second difficulty in interpreting
measures of self-concept
arises because data are not readily available on the equivalence of various self-concept
In many
cases, researchers develop their own instruments for their own particular problem so that the number of different measurement techniques is increasing almost as rapidly as the number of
self-concept studies. Given the imprecision and variability in definitions, we have no reason to believe, a priori, that the measurement techniques are equivalent. The lack of empirically demonstrated

among self-concept measurements
makes it impossible to generalize across studies using different instruments.
data suggest that
And, for a given instrument,
of construct
across different
populations of subjects may be hazardous (Dyer, 1964; Gordon, Note 7; see also Zirkel's 1971 review of self-concept studies of culturally different students).
to test rival counterdata are not available
any self-report measure of
a personality variable, the self-concept interpretation
may be
challenged on the grounds that students may: (a) select responses they know to be socially desirable rather than responses that are self-descriptive (cf. Edwards, 1957a), or (b) be unable (cf. Snygg & Combs, 1949) or unwilling (cf. Cronbach, 1970) to report their "private" self-concepts. Crowne and Stephens' (1961) conclusion on this matter is consistent with ours: While studies of the effect of the social desirability variable on many of the commonly employed tests of selfacceptance have not been done, the results of some ... investigations ... would suggest that self-evaluative tests


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

are particularly susceptible to criticism on social desirability grounds. A common denominator in research findings on self-acceptance may well be the variable of social desirability. (p. 117)
In summary, then, it appears that self-concept research has
addressed itself to substantive
problems before problems of
definition, measurement, and interpretation have been resolved. Until these problems have been dealt with in a manner made
possible by advances in construct validation methodology, the of self-concept findings will be severely limited,
and data on students' self-concepts will continue to be ambiguous. The importance of the construct and the paucity of studies that examine self-concept
of their measurements
make construct validation research vitally needed. Unlike previous methodological critiques (e.g., Crowne & Stephens, 1961;
Gordon, C., 1969; Wylie, 1961), our approach is constructive that we (a) develop a definition of self-concept from existing definitions, (b) review some steps in validating a construct interpretation of a test score, and (c) apply these steps in examining five popularly used self-concept instruments. Perhaps this

approach will stimulate further construct validation research. Definition
of Self-Concept
Although construct validation can proceed with an informal,
intuitive definition, a mature construct definition should be formal and explicit. The ideal would be to define the construct with a network of associations or propositions that relate the construct to: (a) observable properties or quantities of the construct (the within-construct portion of the construct definition;

cf. Loevinger's, 1957, "structural component"), and (b) other constructs (the between-construct portion of the construct definition; cf. Loevinger's, 1957, "external component") which are, themselves, related to observables. This network of interrelationships, called a nomological network (cf. Cronbach & Meehl, 1955), locates a construct in a conceptual space. The withinportion of the definition specifies the features of the construct and links them to each other and to observable attributes of the locates the construct in a concepperson. The between-portion tual space that includes many other constructs related to or independent of the construct. For example, many definitions of feature (cf. Coopersmith,

self-concept include a multifaceted
1967; Piers & Harris, 1964; Purkey, 1970; Sears & Sherman, 1964; Brookover, Erickson, & Joiner, Note 9; Gordon, Note 10). The within-construct
portion of a definition may identify academic,


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



social, and physical self-concept facets and their interrelations. The between-construct
portion may relate each of these facets to
other constructs.
Thus, academic self-concept
may be more
closely related to achievement than physical self-concept.
Currently, in spite of the imprecision and variability of the informal and intuitive self-concept definitions, many self-concept definitions overlap in various ways. By integrating various features that are common to the definitions, and by extending the definitions when necessary, it is possible to construct a working definition of self-concept that is consistent with some current research and can be used to begin to integrate empirical evidence on the validity of self-concept interpretations.

In very broad terms, self-concept is a person's perception of himself. These perceptions are formed through his experience with his environment, perhaps in the manner suggested by Kelly rein(1973), and are influenced especially by environmental

forcements and significant others. We do not claim an entity within a person called "self-concept." Rather, we claim that the construct is potentially important and useful in explaining and predicting how one acts. One's perceptions of himself are thought to influence the ways in which he acts, and his acts in turn influence the ways in which he perceives himself. The exact

nature and direction of the influence of perceptions and behavior are important parts of the definition, but as yet are unclear and are an important focus of current self-concept
Self-concept is inferred from a person's responses to situations. The situations and the responses may be physical or symbolic. Since self-concept must be inferred, issues arise as to what are considered admissible observations. In most educational examinations of self-concept, a distinction is made between self-concept and inferred self-concept. Self-concept is restricted to a person's report of self (cf. Combs, Soper, & Courson, 1963; Parker, 1966). Inferred self-concept is another's attribution of a person's selfconcept. For the purposes of this paper, we will maintain the distinction between self-concept and inferred self-concept, and focus on the former.

Seven features can be identified as critical to the construct definition. Self-concept may be described as: organized, multifaceted, hierarchical, stable, developmental, evaluative, differentiable. Each of these features is considered below. An individual's experiences, in all their great diversity, constitute the data on which he bases his perceptions of himself. To a person recodes

reduce the complexity of these experiences,
them into simpler forms, or categories (Bruner, 1958). The particular category systems adopted by an individual are to some


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

extent a reflection of his particular culture. For example, a child's experience may revolve around his family, friends, and school. This may account for these categories in children's deabout themselves scriptive statements

(Jersild, 1952; Sears,
1963). The categories represent a way of organizing experiences and giving them meaning. One feature of self-concept, then, is that it is organized or structured.
A second feature of self-concept is that it is multifaceted; the particular facets reflect the category system adopted by a particular individual and/or shared by groups. At least in the white, middle-class population of students studied by Jersild (1952) and Sears (Note 5) the category system appears to include such areas as the school, social acceptance, physical attractiveness,

A third feature is that the multifaceted
structure of selfbe
concept may
generality (Super,
1963; Brookover et al., Note 9). That is, facets of self-concept may form a hierarchy from individual experiences in particular situations at the base of the hierarchy to general self-concept at the of this hierarchy is shown in

apex. One possible representation
Figure 1. This formulation is in some ways similar to the British psychologists' hierarchical model of intellectual abilities (cf. Vernon, 1950). At the apex of the hierarchy is general self-concept (cf. Spearman's "g"). General self-concept may be divided into two components: academic self-concept and nonacademic self-concept and practical abilities in the Vernon

(cf. verbal-educational
model). Academic self-concept
may be divided into subjectmatter areas (cf. specific group factors in the Vernon model) and then into specific areas within a subject matter (cf. specific factors). Nonacademic self-concept may be divided into social and physical self-concepts and then divided into more specific facets in a manner similar to academic self-concept.

If this line of reasoning is pursued to the base of the hierarof self-concept as situation-specific is
chy, a conceptualization
consistent with our definition. In extremely limited situations (such as those represented by laboratory experiments), alternative interpretations of a person's experience are reduced considerably. At this level, then, an observer's perception of a person's self-concept may correspond with the person's report of his self-concept. Nevertheless,

the distinction
between selfconcept and inferred self-concept is important. The correspondence between observer and self decreases as one moves up the self-concept hierarchy.
A fourth feature of self-concept is that general self-concept is stable. However, as one descends the self-concept
on specific situa(Figure 1), self-concept depends increasingly


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions




One possible representation

Figure 1
of the hierarchic

This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


of sel


Vol. 46, No. 3

tions and thus becomes less stable. At the base of the hierarchy, Furself-concept varies greatly with variation in situations. thermore, changes at the lower levels of the hierarchy are
at higher levels, makprobably attenuated by conceptualizations ing self-concept resistant to change (cf. Ludwig & Maehr, 1967; see Bem, 1972, for a similar notion from a different perspective). To change general

many situation-specific
with general self-concept,
would be restances, inconsistent
quired. For example, Ludwig and Maehr (1967) showed that
success and failure in an athletic task changed subjects' selfconcepts of specific physical ability but did not change their general self-concepts.
A fifth feature of self-concept is its developmental aspect (cf. Engle, 1959; Long, Henderson, & Ziller, 1967; Long et al., 1968; themselves
Sears, 1964). Infants tend not to differentiate
their environment. As they mature and learn from their increasof self from environment ing store of experiences, differentiation
begins. The self-concepts of young children are global, undifferenAs children begin to build concepts, tiated, and situation-specific.
as represented by the words I and me, they also begin to build concepts for categorizing events and situations. Young children have not started to coordinate the separate subparts of experience to integrate them within one conceptual self-framework. At different times during development, it seems likely that "as the child grows, different parts of him will become more important to him and different parts of his world will assume changing

significance" (Gordon, Note 7, p. 4). With increasing age and experience (especially acquisition of verbal labels), self-concept becomes increasingly differentiated. As the child coordinates and the parts of his self-concept,

we can speak of a
multifaceted, structured self-concept.
A sixth feature of self-concept is its evaluative character. Not only does the individual develop a description of himself in a particular situation or class of situations, he also forms evaluations of himself in these situations. Evaluations can be made

against absolute standards, such as the "ideal," and they can be made against relative standards such as "peers" or perceived evaluations of "significant
others." The evaluative
can vary in importance for different individuals and also for different situations. This differential weighting of the importance of the various evaluative dimensions probably depends
upon the individual's past experience in a particular culture, in a particular society, and so on. As far as we know, the distinction between
and self-evaluation
has not been
clarified either conceptually
or empirically.


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



terms self-concept and self-esteem have been used interchangeably in the literature. A seventh feature of self-concept is that it is differentiable from the other constructs with which it is theoretically related. It is the

beyond the scope of this paper to examine systematically
relationships between self-concept and other constructs, i.e., the between-constructs
of the nomological
it is possible to indicate the direction one could
take in specifying how self-concept is differentiable
from, and
related to, other constructs. For example, self-concept is influenced by specific experiences. Therefore, the more closely selfconcept is linked with specific situations, the closer is the relationship between self-concept and behavior in the situation. If one were to focus on the academic side of the hierarchy in Figure 1, one could hypothesize that (a) self-concept of mental ability should be more closely related to academic achievement than to ability in social and physical situations, and (b) self-concept of academic ability in science should be more closely related to achievement in science than to achievement in, say, English or overall grade-point-average

(cf. Brookover, Paterson, & Thomas,
Note 10).
In a similar manner, the other side of the hierarchy can be
between self-concept and other conexplored and relationships structs explicated.

of Self-Concept
of Test Scores:
of a test score involves
Validating a construct interpretation
an interplay of construct definition, instrument
and data collection. In this context, the construct definition sets the boundaries for potential measurement
techniques. It operIt
ates like a test plan for the development of an instrument.
specifies content areas (e.g., academic, social, and physical selfconcept), the type of stimuli (e.g., items referring to self), the observer (e.g., self as observer of self versus "other" as observer of self), the response (e.g., comparison with others or comparison with an absolute criterion), and so on. From this plan, an instrument is developed and data are collected. These data bear on the construct interpretation of the scores.

Interpretations of these data may be considered hypotheses to be challenged over and over with counterhypotheses
1971; Cronbach & Meehl, 1955). Initial construct validation
studies should examine the empirical and logical evidence in portion of the nomological netsupport of the within-construct of facets such as
work; e.g., whether or not measurements


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

academic, social, and physical self-concept warrant separate Later studies should examine evidence in supinterpretations. portion of the nomological netport of the between-constructs warrant the
work; whether the self-concept facet measurements
interpretations that this construct differs from other constructs, such as intelligence, locus of control, anxiety, and social desirability. If the empirical evidence is congruent with the construct

If the
definitions, test scores are given construct interpretations. data are incongruent with the definition, the definition or the instrument, or both, require revision. Note that empirical evidence cannot directly void the construct definition. Rather, it can reflect upon the measurement

technique, which, in turn, can
reflect upon the definition. If subsequent instrument revisions continue to produce empirical evidence incongruent
with the
definition, it may be that certain aspects of the construct cannot be measured with existing techniques. In this case, the nature of the warranted interpretations
should be specified, and the construct definition should be subjected to critical, logical analyses. Procedures for investigating
rival interpretations
fall into
three categories: logical, correlational, and experimental.

Logical Analyses
between the
The logical analysis examines the consistency
to subjects, instrument
construct definition and instructions
format, item content, and scoring procedures. It draws upon the conand upon psychometric
past experience
The function of the logical analysis is to generate
counterhypotheses as to the construct interpretation of the test score. For example, Sears and Sherman define self-concept as composed of ten traits (see Sears and Sherman, 1964, Table 5, p. 47); the Sears Self-Concept Inventory (Sears, Note 5) contains items

purported to measure each of these traits. A logical analysis of the inventory might lead to a counterhypothesis
that items
linked, say, to the facets of work habits, school subjects, mental ability, and social relations with teacher do not warrant separate interpretations, but relate rather to a single trait, academic self-concept (cf. Torshen, Note 3).

But it should be noted that "... the logical analysis of content cannot disprove a validity claim. The analysis puts forth a
whose pertinence can be verified only empiricounterhypothesis cally" (Cronbach, 1971, p. 475).


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions





between facets of a construct-such
as correlations between academic, social, and physical self-concept evidence on whether the facets deserve to be
between measures of
interpreted separately. Intercorrelations
one construct and other different constructs-such
as the correbetween
and social
evidence on whether scores on a construct
warrant the interpretation
that it is separate from other constructs. In a similar manner, correlations may be used to
examine other features of the construct definition such as its character, and hierarchical organizastability, developmental tion.
A variation on the correlational approach is to identify two populations expected to differ on the construct in question and to determine whether their scores on the construct measure differ (cf. Piers & Harris, 1964; Trowbridge, 1972; Zirkel, 1972). For example, Piers and Harris (1964) compared self-concept scores of retarded

public school children and adolescent, institutionalized,
children earned significantly higher self-concept scores.
Three correlational techniques can be useful in deciding how to interpret test scores. The first, factor analysis, arranges a matrix of correlations into convergences
or clusters among tests or
among items on a test. If the test operates as its design suggests, items measuring, say, "academic self-concept" should cluster together, and this cluster should be distinct from a cluster of items on, say, "physical self-concept." When this occurs, we gain some of test scores. If unanticiconfidence in our facet interpretation pated clusters are found in the test, some revision is called for. In some cases, factor analysis has been used to validate self-concept interpretations of subtest scores (e.g., Gordon, I. J., 1966; Piers & Harris, 1964; Torshen, Note 3; Sears, Note 5) and in some cases it has not (e.g., Coopersmith, 1967).

A second correlational
method, the multitrait-multimethod
matrix (Campbell & Fiske, 1959), examines patterns of intercorrelations between different traits (e.g., academic, social, and physical self-concept) measured by maximally different methods (say, self-report and peer-report of a student's self-concept; e.g., Bixler, 1965; Trickett, 1969). If, for example, factor analysis demonstrates that, for one instrument, items group into certain self-concept facets (say, academic self-concept) that are distinct from others (say, physical self-concept), this distinction should be maintained when different methods are used to measure the

matrix is constructed from
same traits. A multitrait-multimethod


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

correlations between (a) scores on different traits obtained by the same measurement method, (b) scores on the same trait obtained methods, and (c) scores on different
by different measurement
traits obtained by different measurement methods" (Campbell & Fiske, 1959, p. 81). (The distinction between factor analysis and matrix is made for the sake of clarthe multitrait-multimethod matrix to

ity. For the relation of the multitrait-multimethod
factor analysis and the analysis of variance, see Boruch, Larkin, Wolins, & MacKinney, 1970; Boruch & Wolins, 1970.) In this
matrix, "reliability is the agreement between two efforts to measure the same trait through maximally similar methods.
Validity is represented in the agreement between two attempts to measure the same trait through maximally different methods" (Campbell & Fiske, 1959, p. 83). Traits are isolated when the following convergent and discriminant validity criteria are satisfied (refer to Table 1; for a critique of these criteria, see Althauser & Heberlein, 1971):

1. Convergent criterion
a. A validity coefficient should be significantly greater than zero and of practical significance
(validity diagonal
2. Discriminant criteria
a. A validity coefficient should be higher than the correlations obtained between that variable and any other variable having neither trait nor method in common
b. A validity coefficient should be higher than the correlations among scores on different traits obtained by the same measurement
method (heterotrait-monomethod
c. The same pattern of interrelations
among traits should
be observed in correlations obtained with the same or
different methods
To our knowledge, the multitrait-multimethod
matrix has been
used to validate self-concept facet interpretations
of test scores
in only the studies by Dyer (1964), Gordon (1966), and Shulman (Note 11).
So far, the discussion of two correlational
focused primarily on the examination
of the within-construct
portion of the nomological network. These techniques are applicable to the between-construct portion as well. For example, the
matrix has been used to examine the
interpretation of a test score as measuring "humor" (Koppel & Sechrest, 1965). Rival hypotheses to this interpretation were that the score measured intelligence or extroversion. Three different


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



Table 1
Schematic Representation of Areas of a Multitrait-multimethod Matrix Method 2

Method 1
Traits A,





Method 3
C2 A3



Method 1

Method 2

Method 3

measurement methods (self-ratings, peer ratings, and objective responses) were used to measure humor appreciation, humor
creation, intelligence, and extroversion. The results were used to and exdetermine the degree to which humor, intelligence,
as constructs.
troversion could be distinguished
Consider the case in which one construct is measured by
methods A and B, and a second construct is measured by
methods A and C. For example, some theorists
argue that
(on this
only by self-report
issue, see Combs, Soper, & Courson, 1963; Parker, 1966; Sears & say
Sherman, 1964; Wylie, 1961). But some other constructs,
anxiety, might be measured by self-report, observation ratings, heart rate, and so on. To examine validity claims here, the
matrix may not be applicable, but factor


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

analysis is. Different measures of the same trait should cluster and perhaps two or more "measurement method" clusters might
be found.
Finally, if the network specifies a causal relationship, other correlational techniques such as path analysis might be used to examine causality (see, for example, Blalock, 1964; Crano, Kenny, & Campbell, 1972; Yee & Gage, 1968). Bixler (1965) attempted to use the cross-lagged panel analysis to examine causal effects of teachers' and peers' influence on changes in students' selfconcepts. The data suggest that neither teachers' nor peers' influence caused changes in students' self-concepts.

(See also
Wattenberg & Clifford, 1963; Torshen, Note 3; for attempts to examine causality in correlation studies of self-concept.)
Experiments can test some portion of a nomological network
Such experiments can be used to
against a counterhypothesis.
identify influences to which test scores are sensitive. Although most self-concept experiments
are not designed to test counof
as to the proposed self-concept interpretation
their measurements, the outcomes often bear on construct validity. Thus, experimental studies in which specific treatments are used to change only one aspect of self-concept, say physical self-concept, provide an indirect test of the within-construct portion of the nomological network. For example, Ludwig and

Maehr (1967) examined the effect of success and failure in
athletic tasks on physical and general self-concept. Subjects were randomly assigned to positive or negative feedback groups or a control group, regardless of athletic ability. In the feedback groups, feedback was either consistent or inconsistent with the subject's ability. If Ludwig and Maehr's instruments measured self-concept as specified in their nomological network, scores on measures of physical self-concept would be expected to increase with positive feedback, to decrease with negative feedback that was consistent with ability, to increase with positive, inconsistent information, to decrease initially with negative inconsistent information, and to remain unchanged in the control group. A similar but less distinct pattern of scores would be expected on the general self-concept measure. In general, the results of the study were consistent with these expectations.

are designed to
studies in which treatments
change subjects' test scores examine the construct interpretation against counter hypotheses (Cronbach, 1971). For example, do self-concept scores depend upon the subject's motivation, upon knowledge of socially desirable responses, or upon strategy



This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



for attacking the task? Parker (1966) examined the influence of students' expectations as to who would see their self-concept test scores on self-concept measurements. Data were collected with a self-report and an inferred self-concept test, first with the expectation of anonymity and then with the expectation that the teacher would see the scores. The expectancy variable did not influence mean scores on the self-report or inferred self-concept measures. But it did influence correlations between self-report and inferred self-concept measurements.

correlaThere are, then, a number of techniques-logical,
tional, experimental-for examining the validity of self-concept interpretations of test scores. (Our discussion has used the multifaceted feature of self-concept as an example, but other features may be examined as well.) Each contributes its own kind of

evidence to these interpretations. Most self-concept measurements have not been subjected to such examinations of their validity claims. In the following sections, we examine the validity of five widely used instruments. The discussion focuses on definitions and on the validity of the interpretations of the self-concept features (the within-construct portion of the nomological network). This focus seems desirable because, until we are confident about definitions and within-construct interpretations, it is difficult to interpret evidence bearing on the relation of self-concept to other constructs (the between-constructs portion).

Review of Five Major Self-Concept Instruments
Three aspects of construct validity-definition,
instrumentation, empirical data-are examined for five commonly used selfconcept instruments: (a) the Michigan State Self-Concept of Ability Scale, (b) the Self-Esteem Inventory, (c) the How I See Myself Scale, (d) the Piers-Harris Children's Self-Concept Scale, and (e) the Self-Concept Inventory. These instruments were

selected for several reasons. All authors presented some data bearing on construct interpretations of scores from their selfconcept instruments. All instruments are being used in educational research, evaluation, or both. All relate, then, in some way to students' educational experiences. All are representative of self-report measures of self-concept.

Michigan State Self-Concept of Ability Scale
The Michigan State Self-Concept of Ability Scale (SCA), a
measure of self-concept of academic ability, has been used in research by Brookover (Bilby, Brookover, & Erickson, 1972;
Brookover et al., Note 2; Brookover, Erickson, & Joiner, Note 7;


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

Brookover, Paterson, & Thomas, Note 10) to examine relationships between SCA and measures of antecedent and consequent variables, such as intelligence, school achievement, and perceptions of the evaluations of significant others. Construct Definition. "Self-concept of academic ability refers to behavior in which one indicates to himself [publicly or

privately] his ability to achieve in academic tasks as compared with others engaged in the same task [Note 7, p. 8]." Self-concept of ability can be characterized as one of the many concepts of self, hierarchical in nature (academic portion of Figure 1), dependent on a particular role, area of experience, situation, time, etc., relatively stable under similar stimulus situations, and primarily evaluative in nature.

Description of the Instrument. The SCA-General Scale (Form
A) consists of eight items, selected to differentiate students on achievement (Patterson, Note 12), which form a Guttman scale (reproducibility coefficient = 0.91). The items are self-evaluative questions about academic ability such as "What kind of grades do you think you are capable of getting in the following subjects?" For each of four subjects (mathematics, English, social studies, science), there are five response alternatives ranging from

"Mostly A's" to "Mostly F's." The eight items have been divided into two conceptual dimensions, each composed of two logical subsets (no examples were provided): (a) future-oriented and present-oriented items, and (b) comparative and absolute evaluations of ability. The SCA-Specific Subjects Scale (Form B) was constructed for the same eight items with separate responses requested for

science, mathematics, English, and social studies. The mathematics scale had a reproducibility coefficient of 0.90; the other scales were not analyzed (Paterson, Note 12).
Reliability. For samples of subjects drawn from grades seven to ten, internal consistency reliability of Form A ranged from 0.82 to 0.92 for boys and from 0.77 to 0.84 for girls (Brookover et al., Note 9; Brookover et al., Note 10). Form B internal consistency coefficients on a tenth-grade sample were above 0.89 for boys and girls (Brookover et al., Note 4).

Test-retest reliability coefficients, reported for one-year intervals, ranged from 0.69 to 0.72 for boys and 0.69 to 0.77 for girls in grade levels 8 to 12 (Brookover, et al., Note 4; Brookover, et al., Note 9). For the subject-matter areas of Form B, they ranged from 0.70 to 0.80 for boys and from 0.63 to 0.77 for girls

(Brookover, et al., Note 4).
These data, then, suggest a dependable rank ordering of persons at one point in time which remains fairly stable over one


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



year. In general, the reliabilities are high enough to permit study of individual differences.
Empirical Evidence in Support of Construct Interpretation.
Several analyses have examined the structure or dimensionality of the SCA-Form A. Scalogram analyses have yielded coefficients of reproducibility ranging from 0.86 to 0.97. A factor analysis (Paterson, Note 12) revealed two factors. The first factor was interpreted as self-concept of ability; the second factor, barely distinguishable from error, was interpreted as a time dimension and three presentdifferentiating between four future-oriented oriented items. These findings, then, support the interpretation of SCA as a unidimensional

measure perhaps of self-concept of
Studies by Brookover et al. (Note 2, Note 10) provide data which bear on whether scores on Forms A and B warrant separate
Correlations between Forms A and B ranged
from 0.62 to 0.73 for boys and 0.54 to 0.71 for girls. When these correlations were corrected for attenuation,
they ranged from
middle 0.70 to middle 0.90. These correlations along with Paterson's (Note 12, p. 163) report that "The general SCA mean scores are higher than any of the specific SCA scores but the general score is the closest to the specific score in the subject in which the student had his highest achievement" suggest that differential of Forms A and B should be made quite cauinterpretation

Data reported by Brookover et al. (Note 4) and Paterson (Note of the four
12) bear on whether
boys, the
correlation between scales ranged from 0.63 to 0.88; for girls, the range was 0.52 to 0.68. These correlations, when corrected for attenuation,
ranged from the mid-60s to the mid-80s. Even
are fairly high, differential prethese
dictions of grades in corresponding
subject matters from the
scales is possible. In all but one instance, the correlations between grades and scores on corresponding areas of the Form B scale (ranging from 0.43 to 0.61) were higher than the correlations between grades and scores in different areas. Interpretation of Scores from the SCA. Scores on the general SCA (Form A) probably warrant an interpretation

separate from
scores on the subject-matter
specific scales (Form B), if made
of the differquite cautiously. Likewise, cautious interpretation scales seems warranted. However, the high
ent subject-matter
correlations between Form A and the scales of Form B, the high correlations between the scales of Form B, and data on means reported by Paterson (Note 12) suggest that additional validation


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

work, such as a multitrait-multimethod
analysis, is needed to
clarify interpretations.
Finally, the issue needs to be raised as to
whether the SCA is a measure of self-concept,
an alternate
method of reporting grades, or an alternate method of reporting general mental ability. We suspect the self-concept interpretation will be shown to be warranted, but systematic examination is needed.

of these and other counterhypotheses
In regard to our definition of self-concept, the data reported on the SCA provides some support to a multifaceted, hierarchical view of self-concept.
Support for a multifaceted
comes from Brookover's definition of academic self-concept as one of many self-concepts and the empirical evidence-internal factor analysiscoefficients,
consistency and reproducibility
suggesting that the SCA is a unidimensional measure. Thus, one of many possible facets of self-concept may be tapped by the SCA. Support for a hierarchical view of self-concept comes from data on the relation of scores on SCA Forms A and B. The correlations suggest a strong general feature of academic self-concept that into subject-matter

can be differentiated
specific self-concepts
(see Figure 1).


The Self-Esteem Inventory (SEI: Coopersmith, 1967) has been
used as a measure of general self-concept (e.g., Coopersmith, 1967; Dyer, 1964; Smith, I. J., 1973; Epstein & Komorita, 1971). Most of Coopersmith's (1959, 1967) own empirical work examined relationships between SEI scores and measures of other constructs, such as popularity, role in group discussions, conformity and creativity.

Construct Definition.
Coopersmith (1967) defined self-esteem
... the evaluation which the individual makes and customarily maintains with regard to himself: it expresses an attitude of approval or disapproval, and indicates the
extent to which the individual believes himself to be
successful, and worthy. In short,
capable, significant,
self-esteem is a personal judgment of worthiness that is
expressed in the attitudes the individual holds toward
himself. It is a subjective experience which the individual
conveys to others by verbal reports and other overt
expressive behavior. (pp. 4-5)
Coopersmith elaborated especially on three features of this
definition: stability, structure, and process. Stability "centers upon the relatively enduring estimate of general self-esteem


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



rather than upon more specific and transitory changes in evaluation" (p. 5). Structure focuses on general self-esteem with the with
that self-esteem
may be multifaceted
sex, age,
Process focuses on "a
and other role-defining
judgmental process in which the individual examines his performance, capabilities, and attributes according to his personal standards and values, and arrives at a decision (evaluation) of his own worthiness."

Description of the Instrument. The SEI was developed from a
pool of original and reworded items from a scale reported by Rogers and Dymond (1954). Form A contains 58 items; eight
items form a "lie" scale, and 50 items form the self-esteem
measure. The items are simple, self-descriptive statements such as "I give in very easily." There are two response alternatives for each item: "Like me" or "Unlike me." In addition to a total

self-esteem score, four subscale scores have been identified conceptually: General Self (26 items), Social Self-Peers (8 items), Home Parents (8 items), School Academic (8 items).
Form B includes "those 25 items which showed the highest
item-total score relationships (Instructions, undated, p. 1)" in an item analysis of Form A. Total scores of Forms A and B correlated 0.86. This level of correlation has been approximated with four different samples (Coopersmith, 1967). Since this was a part-whole correlation, it is spuriously high.

Reliability. Internal consistency reliability for the four subscales of the SEI (Dyer, 1964) ranged from 0.28 to 0.82 for boys and girls at third, fifth, seventh, ninth, and eleventh grades. The total score test-retest

reliability coefficient was 0.88 over a
five-week interval with a sample of 30 fifth-grade students, and a coefficient of 0.70 was obtained over a three-year period with a sample of 56 public school students. These data suggest that a dependable measure can be obtained with total scores but, by using fewer items on subscales, the consistency of the measurement is greatly reduced. Although individual differences may be examined with total scores, some of the subscales appear too unreliable for such work.

Empirical Evidence in Support of Construct Interpretations.
Although a priori subscales for the SEI have been proposed,
Coopersmith has not examined the validity of separate interpretations of those subscales. He did report that differences in self-appraisals in distinct areas of experience were not found in test scores of 56 children aged 10 to 12. He reasoned that either these distinctions were not made by preadolescents

or, if they
were, they were made in the context of general self-esteem. This lack of differentiation between subscales would also be expected


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

from the low internal consistency coefficients and may suggest heterogeneity of items within subscales.
examined the subscale
Only one study has systematically
of the SEI. Dyer (1964) conducted a multitraitinterpretation multimethod validation study of the SEI using four subscale
(trait) scores: Self, Family, Friends, and School. An Attitude (AQ), with a semantic differential format, was
alternative measurement method. The AQ condeveloped
tained the concepts of "Myself," "With My Family," "With My
Friends," and "With My Schoolwork." For each concept, subjects rated themselves on eight bipolar adjectives (Dyer, 1964, p. 29; were adadjectives were not reported). The two instruments

ministered to 500 public school students in Flint, Michigan. Ten matrices were constructed for
separate multitrait-multimethod
boys and girls from grades three, five, seven, nine, and eleven. The ten matrices were examined according to the four validity criteria proposed by Campbell and Fiske (1959). The convergent validity coefficients ranged from 0.02 to 0.64. Twelve of the 20 entries for boys and 19 of the 20 entries for girls were statistically and thus met the convergent

validity criterion.
(Alpha is not stated, but correlations significant at p < 0.01 are indicated.)
When the convergent validity coefficients were compared with heterotrait-heteromethod
coefficients, only 22 of 40 validity coefficients met this first discriminant validity criterion. When the convergent validity coefficients were compared with the correcoefficients, only two met this sponding heterotrait-monomethod

criterion. They were the coefficients for "Family" for eleventhgrade boys and "School" for eleventh-grade girls. Finally, when
the patterns of relationships among the traits were examined for each matrix, only the eleventh-grade
data showed evidence of a
consistent pattern. In sum, about half of the validity coefficients met the convergent validity criteria but could not meet the three more stringent discriminant validity criteria.
Finally, Dyer examined all the correlations between different and the
traits (i.e., entries in the heterotrait-monomethod
triangles) by averaging over four coefficients for each of the six possible pairs of traits computed. These data indicated that the greatest differentiation (low correlations) between the traits was found between Friends and School (for girls and boys) and Family and School (for girls only). The mean correlations ranged from 0.18 to 0.45 across the five grade levels. The least differentiation

(high correlations) was
found for Self and School (for boys and girls) and Self and Friends (for girls only). Mean correlations ranged from 0.41 to 0.59 across grade levels.


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


In summarizing his multitrait-multimethod


analysis, Dyer

Strong method variance that was both shared, as indicated in the heteromethod triangles, and specific to each instrument, as indicated in the monomethod triangles,
contributed to the scores in most of the samples below the
eleventh grade, and thus resulted in lack of construct
validity. (p. 70)
Interpretation of Scores from the SEI. Although separate areas of self-esteem have been identified conceptually, there appears to be no empirical support for their separate interpretation. Data reported by Coopersmith (1967) bearing on the between-portion of the nomological network suggest that interpretation of the SEI separate from other variables (e.g., popularity, conformity) may be warranted. However, a decision as to whether the SEI

deserves a general self-concept interpretation must await
further studies in which the between-portion of the nomological network is examined and studies in which plausible counterhypotheses, such as the SEI is a measure of social desirability, are eliminated.

With respect to our definition of self-concept, the developmental (cross-sectional) data reported by Dyer (1964) seem to be supportive. These data hint that self-concept may become more differentiated with increasing age. In particular, some justification can be offered for a separate "Family" trait for eleventh-grade boys and a separate "School" trait for eleventhgrade girls. In the absence of factor analysis data, we cannot tell whether the traits themselves are not differentiable or whether the subscales have been inadequately formed. Finally, the testretest correlations suggest that general self-concept is fairly stable over a year's time.

How I See Myself Scale
Elementary- and secondary-school forms of the How I See
Myself Scale (HISM) have been developed in a series of research projects by Gordon (Note 7). At present, the scale is most often used to measure changes in certain aspects of self-concept resulting from intervention programs. Construct Definition. Gordon (1966, p. 53; Note 7, p. 34) defined self-concept as the organization of all of the child's biological and environmental experiences as he has interpreted them into one highly organized, highly integrated, multifaceted system. Selfconcept, that portion of the self-system of which the child is most aware, is the product of all his transactions at a particular point


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

in time and determines his behavior. As the child grows, different parts of self-concept and experience change in relative importance. According to Gordon, since behavior is determined by the
can be made from behavior to selfself-concept, inferences
self have been identified along with
corresponding behaviors from which inferences might be made: self as revealed from self-report, self as inferred from observed behavior, and self as inferred from projective techniques. The relationships between self-report and inferences from overt behavior and self-report and inferences from projective techniques are expected to be small (Gordon, Note 7, see pp. 35-36; see also Combs, Soper, & Courson, 1963).

Description of the Instrument. The elementary (ages 3-6 years) and secondary (ages 7-12 years) forms of the HISM are primarily statements developed from
comprised of simple, self-evaluative
Jersild's (1952) categories of children's responses in compositions are arranged on five-point,
about themselves. The statements
bipolar scales such as "I do well in school work... I do not do well in school work" (elementary form) or "I do very well in school... I do not do well in school" (secondary form).

Combs and Gordon (Gordon, Note 7) tested 80
slow-learning high school pupils over a two-week interval and obtained retest coefficients for four subscales (Teacher, Appearance, Body-build, Academic Achievement) ranging from 0.62 to 0.82. Yeatts (Note 13) reported total score stability coefficients ranging from 0.78 to 0.89 for third, fifth, eighth, and eleventh graders from a heterogeneous

public school population, after an
interval of nine days. No stability coefficients for the current subscale scores have been reported for children. Internal consistency coefficients have not been reported. Empirical Evidence in Support of Construct Interpretation.

The structure of the items on the HISM has been examined with factor analysis. In data from 960 laboratory school children, a number of factors were identified and remained consistent over grades three, five, eight, and eleven. Yeatts (Note 13), using data from boys and girls in public school grades three through twelve, found that the items loaded on five factors: physical appearance (e.g., "I like the way I look"), interpersonal adequacy (e.g., "my clothes are very nice"), teacher-student

(e.g., "teachers like me"),
academic adequacy (e.g., "I learn new things easily"), and autonomy ("I enjoy individual projects"). These labels are, however, only suggestive since, for example, items such as "My skin is nice looking" loaded on the factor, academic adequacy.(!) Depending on grade level and sex, additional factors emerged. According to Gordon (Note 7), changes in factors ". .. fit well with


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



developmental theory and self-concept theory" (p. 41). There is reasonable consistency in the factors found over grades and
sexes, but the composition of items within factors is sometimes curious. Since Gordon's definition of self-concept gives no clear direction as to which items should cluster with which others, interpretation of scores based on these factors seems hazardous. And, the fact that the same item may be incorporated into

several factors only creates additional problems of interpretation. Gordon (Note 7) reported correlations using three different
measurement methods-behavior
observation, projective technique, and self-report (HISM)-and a variety of traits. Although these data were supposed to form a multitrait-multimethod
matrix, the traits measured with the observational and projective techniques differed from those measured by the self-report method. Also, correlations were not reported for most of the traits which seemed to be related to the HISM factors. Gordon summarized the data as indicating that the corrrelations, although positive and significant, were generally low. He concluded that HISM measured self-concept somewhat differently from the inference approaches. We found the data uninterpretable.

Interpretation of HISM Scores. From the data available, interpretation of subscale scores-the scores used to report data hazardous. Further definitional
from the instrument-seems
and empirical work is needed in order to determine how to
interpret the subscales. Work also needs to be done in specifying the between-portion of the nomological network and in examining relations between scores on the HISM and scores on other constructs to determine whether self-concept interpretations are warranted.

With regard to our definition of self-concept, it is difficult to draw inferences from the HISM data until further work is done. The following, then, is speculative. First, the definition of selfconcept and factor analyses suggest a multifaceted self-concept, although it is not clear what those facets are. Second, some changes in factors with data from children of different ages suggest a developmental aspect to self-concept. However, these data do not support our contention that self-concept becomes more differentiated with age; they support Gordon's notion of change in emphasis at different ages.

The Way I Feel About Myself (WIFAM)
The Piers-Harris Children's Self-Concept Scale (The Way I Feel About Myself) was developed as a measure of general self-concept (Piers & Harris, 1964). The authors claim the instrument can be 429

This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

in clinical and counseling settings and in
used diagnostically
classrooms for psychological referral, but its primary use has been in research on the development and correlates (e.g., anxiety, intelligence) of self-concept (Piers, Note 14). of selfConstruct Definition.

While an explicit definition
Harris, 1964; Piers,
14), we
infer from their writings that self-concept is multifaceted, relaas age increases-with distinct detively stable-especially
Finally, Piers (Note 14) distinvelopmental
guishes between self-concept which is reported by the self from inferred self-concept which is inferred by others from the selfs (individual's) behavior.
The 80 items, originally deDescription of the Instrument.
categories, are simple descriptive
such as "I am a happy person," with a yes/no
response. These items were selected to discriminate
students with extremely low and high total scores. A total score or "cluster" scores can be obtained, the latter from items included in each of six clusters identified by factor analysis (Piers & Harris, 1964). The factors are labeled: Behavior, Intellectual and School Status, Physical Appearance and Attributes,

Popularity, Happiness and Satisfaction.
indicate stable
Reliability. Internal consistency
rank orderings of students on total scores: (a) KR21 coefficients for 95 items used in a standardization
study ranged from 0.78 to
0.93 for boys and girls in grades three, six, and ten; (b) corrected split-half coefficients for the total score were 0.90 and 0.87 for grade six and ten students, respectively. These rank orderings on total scores remain fairly stable over a period of several months: retest coefficients of 0.71 to 0.72 were found for the 95-item version and a coefficient of 0.77 was found for the current 80-item version (Piers, Note 14). Reliability

for "cluster
scores" have not been reported.
Empirical Evidence in Support of Construct Definition.
structure of the 80 items on the WIFAM has been examined by
factor analysis (Piers & Harris, 1964). Ten factors accounted for 42 percent of the total test-score variance (Piers & Harris, 1964). Six factors were judged large enough to be interpretable:

Behavior ("I do many bad things"), (b) Intellectual and School Status ("I am good in my school work"), (c) Physical Appearance and Attributes ("I am good looking"), (d) Anxiety ("I cry easily"), (e) Popularity ("People pick on me"), (f) Happiness and Satisfaction ("I am a happy person"). Although these factors suggest a multifaceted self-concept, in the absence of some conceptual structure, we do not know if these observed factors are the ones to be expected. (There is some


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



correspondence with the facets identified by Sears; see below.) Furthermore, there are insufficient data to determine whether the scores on separate factors warrant separate interpretation. For example, 13 items did not load on any of the factors, whereas 11 items loaded on more than one factor. Neither correlations between subscale scores nor subscale correlations with measures of other constructs are reported.

Piers (Note 14) reported some data that bear on the selfof the total score. For example, Mayer concept interpretation
(1965 reported by Piers, Note 14) found a correlation of 0.68 (p <.05) between the WIFAM and Lipsitt's Children's SelfConcept Scale for a sample of 98 males and females ranging in age from 12 to 16 years. Piers (Note 14) found correlations

between WIFAM and teacher ratings of fourth and sixth graders on self-concept ranged from 0.06 to 0.41. Correlations with peer ratings ranged from 0.26 to 0.49.
Interpretation of WIFAM Scores. There is some evidence to
suggest that total scores on the WIFAM warrant self-concept
interpretations. However, this statement is based on convergent of scores on WIFAM with
validity coefficients-correlations
other measures of self-concept. In some cases, measures of other constructs have correlated as highly with WIFAM total scores as have other measures of self-concept. In order to have reasonable additional conceptual work on

confidence in this interpretation,
the between-portion of the nomologial network accompanied with empirical data is required. At present, there are no data availof "cluster" scores. able on the interpretability
With respect to our definition of self-concept, data on the
WIFAM support a multifaceted view of self-concept, with general of
self-concept (total scores on WIFAM) a stable characteristic the individual. Data from the WIFAM bearing on the hierarchiof self-concept have not been reported.
cal characteristic
Likewise, data that would bear on the developmental
aspect of
Self-Concept Inventory


The Self-Concept Inventory provides a measure of general and ten facets of self-concept. The original form (Sears, Note 5) has been used as a research tool (e.g., Bixler, 1965; Gelfand, 1962; Trickett, 1969; Torshen, Note 3; Sears, Note 5; Spaulding, Note 6). A revised form is currently being used (Sears et al., Note 8; Sears, Note 15) to evaluate changes in self-concept resulting from

intervention programs.
For Sears, self-concept represents the
Construct Definition.
child's expectation of success in solving problems arid carrying


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

out tasks. The expectancies
are learned and can be changed
according to principles of learning. Furthermore, self-concept is multifaceted with each facet differing in importance or reward value. Since expectancies
are acquired within the context of a
facet (e.g., "school subjects"), an individual can predict success or failure of his behavior within that facet. Finally, various aspects of self-concept have properties similar to drive; to protect the self-concept, a person will select or strive toward those behaviors that maintain and enhance self-concept (see Sears & Sherman, 1964, p. 10).

Description of the Instrument. The original form of the SCI
contains 100 items, ten items for each of ten facets of self-concept. The ten facets were identified from compositions written by sixth graders and the items within each facet are short, descriptive phrases. The facets and example items are: (a) physical ability ("Being built for sports"), (b) mental ability ("Learning things rapidly"), (c) school relations with same sex ("Making friends easily with boys"), (d) school relations with opposite sex ("Making friends easily with girls"), (e) school relations with teacher ("Getting along well with teachers"), (f) work habits ("Getting my school work in on time"), (g) social virtues ("Being willing for others to have their way sometimes"), (h) happy qualities ("Getting a lot of fun out of life"), and (i) school subjects ("Getting good grades in school"). The student responds to each item on three questions: (a) satisfied with self, yes or no; (b) expect improvement, maybe or probably won't; and (c) compared with others, rate yourself.

The revised form of the SCI has 48 items representing
facets, most of which correspond to the original scale. The
revisions were: (a) Mental Ability was divided into Divergent and Convergent Mental Ability; (b) Social Relations with the Opposite Sex was dropped; (c) School Subjects was reduced in size; (d) the "personality" subareas (e.g., Happy Qualities, Attractive Appearance, Social Virtues) were reduced in size; and (e) the satisfaction and improvement responses were dropped; the student only rates himself as compared with others. Reliability. The reliability of both forms of the SCI, in general, appears to be adequate for research on individual differences. For the 100-item scale, Sears (Note 5) reported a corrected

split-half coefficient of 0.95 for the total score, self-rating. Similar coefficients for the subareas ranged from 0.79 to 0.93 (Bixler, 1965; Trickett, 1969; Sears, Note 5). Retest coefficients from fall to spring ranged from 0.29 to 0.67 for the total score. For the revised scale, Sears (Note 15) reported internal consistency coefficients of 0.90, 0.88, and 0.92 for the facet scores combined into total, personality, and mental ability scores, respectively. For the


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



nine separate facets, the coefficients ranged from 0.56 to 0.89. A retest (six-month interval) coefficient of 0.50 was reported for the total score (Sears et al., 1972). The revised form will not be discussed further since its manual is still in preparation.



in Support

of Construct


The structure of the SCI has been examined by factor analysis (Torshen, Note 3; Sears, Note 5). In Sears' analyses, 195 fifth graders were divided into four subgroups based on sex and their composite scores from the Primary Mental Abilities test

(superior or average). For the four separate factor analyses, items clustered on one factor: "generalization." Sears (Note 5) concluded that, although the ten subscales may measure different self-concept referents, "self-concept is a more or less unitary aspect of personality" (p. 81).

Torshen (Note 3), using data from 402 fifth-grade students,
found separate but correlated factors corresponding to each of the ten subareas of the SCI. In addition, two other factors, "self-improvement" and "self-rating," corresponded to two of the three rating scales on the SCI. The third type of ratingsatisfaction with self-did not appear as a separate factor since "it was obtainable from the other factors" (Torshen, Note 3, p. 47). Torshen's results, then, give some support to the interpretation of subarea scores as representing different subareas of self-concept.

The difference in the Sears (Note 5) and Torshen (Note 3)
findings might be explained in two ways. One explanation is that Sears divided students into homogeneous subgroups (by sex and aptitude), whereas Torshen designed her study to maximize
variability among students. Another explanation might be the differences in factor analytic method used. Sears used the principal components method with varimax rotation (Kaiser, 1958). In contrast, Torshen (Note 3; cf. JSreskog, 1969) imposed Sears' ten-subarea structure on the data and tested the goodness of fit. Although data have not been collected in a multitraitmultimethod matrix, parts of the matrix can be built from data reported by Bixler (1965), Trickett (1969), and Torshen (Note 3). In the Bixler study, heterotrait-heteromethod data are missing. Also, only six of the ten self-concept subareas are represented. In the Torshen study, only the heterotrait-monomethod data are

available. In the Trickett study, three methods were used: selfreport (SCI), teacher ratings, and peer ratings. However, data are available for only four of the ten traits and most reliability data are missing. In discussing the four criteria for convergent and discriminant validity of the SCI, data from all three studies will be used.


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

Reliability. With a few exceptions (Bixler data), the square root of the reliability
is the largest
entry in the
Convergent Validity. The validity coefficients are significantly greater than zero (except for two coefficients from Trickett). Discriminant Validity. Data are available in the heterotraitheteromethod triangles from only one study (Trickett, 1969). With this criterion, data from the SCI and peer ratings on four facets (happy qualities, school subjects, physical ability, appearance) suggest that three of the four (except happy qualities) can be discriminated. In comparing teacher ratings with the other two methods, the four facets cannot be discriminated.

data from Bixler (1965) and Trickett (1969) indicate that, in most cases, the correlations in the validity diagonal are not higher than corresponding
entries in the heterotrait-monomethod
triangle. These data, then, suggest that the facets may not be distinguishable from one another and that the type of measurement method employed strongly influences the scores obtained. These findings cast doubt on interpretations

of subscale scores as
separate facets of self-concept.
Because of missing data, patterns of intercorrelations
be examined thoroughly. A pattern of high intercorrelations
the heterotrait-monomethod
is found
"academic" traits: (a) School Subjects, (b) Mental Ability, (c) Work Habits, and (d) Social Relations with Teachers. Correlations among other facets do not show additional patterns. For example, a "social" trait using the Social Virtues, Social RelaBut the tions, and Happy subscales might be hypothesized.

do not support this hypothesis.
Likewise, a
"physical" self-concept facet might be hypothesized, but the data do not support this either.
Interpretation of Scores from the SCI. Most of the data presented above suggest that separate interpretation of the ten
subscales of facets of the SCI is not warranted. Nevertheless, certain subsets of facets, such as a subset called "academic self-concept," might warrant an interpretation
separate from
some other subset, such as "nonacademic
self-concept." This
observation is based on correlational data examined above and data presented by Torshen (Note 3) in her factor analysis and other data analyses. A decision as to whether the total score warrants a self-concept interpretation

must await thorough
examination of the between-portion
of the nomological network.
However, preliminary data from Sears (Note 5) and Torshen
(Note 3) suggest that such an interpretation may be warranted. With regard to our definition of self-concept, these data support to some degree, a hierarchical, multifaceted construct. This


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



observation is supported by data suggesting a strong general into, at least, academic and
self-concept discriminable
nonacademic facets (see Figure 1). The retest correlations lend some support to a stable, general self-concept; data are not available to evaluate the developmental aspect of the definition. Summary and Recommendations for Construct Validation

The enhancement of each student's self-concept is important in today's educational Zeitgeist either as an educational outcome or as a moderator of achievement. Most self-concept studies
examine intercorrelations between self-concept and other constructs, or differences in mean self-concept scores between different populations of students, or changes in self-concept due to some treatment. Taken individually, they often provide important insights into factors that motivate students in and out of school and into alternative courses of action that may enhance students' self-concepts. However, considered as a body of research, self-concept studies lack the focus that would result from an agreed-upon definition of self-concept, lack adequate validation of interpretations of self-concept measures, and lack empirical data on the equivalence of the many self-concept measures currently being used.

The methodology of construct validation represents one approach that may help draw together the diverse studies of self-concept. This methodology examines the ingredients of a construct definition and suggests logical and empirical methods for examining the validity of interpretations of self-concept measurements.

Self-concept was defined as an individual's perception of himself and a number of distinctive features were identified: (a) organized, (b) multifaceted, (c) hierarchical, (d) stable (general self-concept)/unstable (situational), (e) developmental, (f) descriptive and evaluative, and (g) differentiable from other constructs. Data from five instruments, commonly used as measures of

self-concept, tentatively support the interpretations of scores according to features of the proposed definition. Data from each of these instruments suggested a general self-concept interpretation of scores. This interpretation needs to be validated against alternative construct interpretations of these total scores. Although the subscores of the various instruments are moderately correlated, data from four of the five instruments suggested that general self-concept might be divided into different dimensions (multifaceted feature). There was not perfect agreement on the clustering of items from different instruments, but an examina-


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


Vol. 46, No. 3

tion of items within factors indicated sufficient consistency across instruments and subject populations (mainly Caucasian and middle class) to suggest that self-concept scores can be related to four general areas of experience: academic, social, emotional, and physical. In the absence of data comparing one inis tentative. strument with another, this interpretation

There is a paucity of data bearing on the hierarchical interpretation of self-concept scores. The data suggested a hierarchy in which general self-concept (at the apex) could be divided into academic self-concept and nonacademic self-concept (Ludwig & Maehr, 1967; Torshen, Note 3; Brookover et al., Note 9). Academic self-concept, in turn, may be divided into specific subject-matter self-concepts (Torshen, Note 3; Brookover et al., Note 9). The other facets may include, at least, a physical self-concept (Ludwig & Maehr, 1967). However, without additional information interrelating self-concept facets at various levels of the hierarsuch as self-concept divides into many chy, a counterhypothesis,

situation-specific responses all of which are positively correlated due to a verbal report factor, cannot be eliminated. Until the hierarchical feature of self-concept has been examined more
adequately, then, it must be considered extremely tentative. The hierarchical feature led to a hypothesis that general selfself-concept is unconcept is stable, whereas situation-specific stable. The data suggested a stable general self-concept. A few studies have examined the hypothesis that self-concept becomes increasingly stable moving from the base of the hierarchy to the apex. Ludwig and Maehr's (1967) data gave tentative support to Data reported by Brookover et al. (Note 9)

this interpretation.
suggested that self-concept of achievement in academic subjects is just as stable as self-concept of academic ability. Until stability is estimated at each level of the hierarchy, this feature must be considered tentative.

The data reviewed here have not examined the developmental
hypotheses directly. Data from Gordon (Note 7) showed that
different factors emerged from students at different ages. Some data from Dyer (1964) hinted at increased differentiation
age. But, further work needs to be done before this feature of the definition is adequately tested.
A distinction between descriptive and evaluative aspects of
self-concept is not supported empirically. For example, Torshen (Note 5) reported that a self-descriptive dimension of self-concept was subsumed by two evaluative dimensions.
The intent of the review has been to pull together some of the more significant, individual studies of self-concept into a coherent framework whose evidence
could be systematically
examined. It has showed a need to limit the number of selfconcept definitions, to examine the validity of the proposed 436

This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



of the definiconstructs, to examine the rival interpretations tions, and to ascertain the equivalence of self-concept instruments, so that data collected with one instrument can be related to data collected with another. Once some degree of standardization of definition, instrumentation, and interpretation

has been
and correlates of selfachieved, the determiners, consequences, concept, along with their implications for educational practices, can be investigated systematically.


1. Gordon, E. W., & Wilkerson, D. A. Compensatory education for the disadvantaged: Programs and practices for preschool through college. New York:

College Entrance Examination Board, 1966.
2. Brookover, W. B., LePere, J. M., Hamachek, D. E., Thomas, S., & Erickson, E. L. Self-concept

of ability

and school achievement,

II. USOE Cooperative


search Report, Project No. 1636. East Lansing: Michigan State University, 1965.
3. Torshen, K. P. The relation of classroom evaluation to students' self-concepts and mental health. Unpublished doctoral dissertation, University of Chicago, 1969.
4. Bledsoe, J. C., & Garrison, K. C. The self-concepts of elementary school children in relation to their academic achievement, intelligence, interests, and manifest

anxiety. USOE Cooperative Research Report, Project No. 1008, Athens: University of Georgia, 1962.

5. Sears, P. S. The effect of classroom conditions on the strength of achievement motive and work output on elementary school children. USOE Cooperative

Research Report, Project No. OE-873, Stanford, California: Stanford University, 1963.

6. Spaulding, R. Achievement, creativity, and self-concept correlates of teacherpupil transactions in elementary schools. USOE Cooperative Research Re-

port, Project No. 1352, Urbana: University of Illinois, 1963.

7. Gordon, I. J. A test manual for the how I see myself scale. Gainesville:


Educational Research and Development Council, 1968.
8. Purkey, W. W. The search for self. Gainesville, Florida: Florida Educational Research and Development Council Research Bulletin, 1968, 4 (2), whole. 9. Brookover, W. B., Erickson, E. L., & Joiner, L. M. Self-concept of ability and school achievement, III. USOE Cooperative Research Report, Project No. 2831, East Lansing: Michigan State University, 1967.

10. Brookover, W. B., Paterson

A., & Thomas, S. Self-concept

of ability and school

& L. M. Joiner, Self-concept

of ability and school

achievement. USOE Cooperative Research Report, Project No. 845, East Lansing: Michigan State University, 1962.
11. Shulman, L. S. Multiple measurement of self-concept. Paper presented at the annual meeting of the Educational Research Association, Chicago, 1968. 12. Paterson, A. Reliability and validity of self-concept of ability scale. In W. B. Brookover, E. L. Erickson,

achievement, III, USOE Cooperative Research Report, Project No. 2831, East Lansing: Michigan State University, 1967, 155-172.
13. Yeatts, P. P. Developmental change in the self-concept of children grades 3-12. Gainesville, Florida: Florida Educational Research and Development Council Research Bulletin, 1967, 3 (2), whole.

14. Piers,

E. V. Manual


the Piers-Harris,




Nashville, Tennessee: Counselor Recordings and Tests, 1969.
15. Sears, P. S. Memorandum with respect to use of the Sears Self-Concept Inventory. Stanford University, 1966. (mimeographed)

This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



Vol. 46, No. 3

Aiken, W. M. Story of the eight-year study. New York: Harper, 1942. Althauser, R. P., & Heberlein, T. A. Validity and the multitrait-multimethod matrix. In E. F. Borgetta (Ed.), Sociological methodology. San Francisco: Jossey-Bass, 1971.

Bem, D. J. Self-perception theory. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 6). New York: Academic Press, 1972. of self and
Bilby, R. W., Brookover, W. B., & Erickson, E. L. Characterizations student decision making. Review of Educational Research, 1972, 42, 505-524. Bixler, P. A. Changing self-concept in sixth-grade class groups. (Doctoral disser1965). Dissertation Abstracts, 1965, 26 3750.

tation, Stanford University,
(University Microfilms No. 65-12, 730)
research. Chapel Hill:
Blalock, H. M. Causal inferences in nonexperimental
University of North Carolina Press, 1964.
Boruch, R. F., Larkin, J. D., Wolins, L., & MacKinney, A. C. Alternative methods and Psychological
of analysis: Multitrait-multimethod
data. Educational
Measurement, 1970, 30, 833-853.
Boruch, R. F., & Wolins, L. A procedure for estimation of trait, method, and error to a measure. Educational
and Psychological
Measvariance attributable
urement, 1970, 30, 547-574.
Brownfain, J. J. Stability of the self-concept as a dimension of personality. Journal of Abnormal and Social Psychology, 1952, 47, 597-606. Bruner, J. S. Social psychology and perception. In E. E. Maccoby, T. M. Newcomb, & E. L. Hartley (Eds.), Readings in social psychology (3rd ed.). New York: Holt, Rinehart, & Winston, 1958, 85-94.

Bruner, J. S. The process of education. Cambridge: Harvard University Press, 1960.
Bruner, J. S. Social psychology and perception. In D. E. Hamacek (Ed.), The self in growth, teaching, and learning. Englewood Cliffs, New Jersey: Prentice-Hall, 1965.
Callahan, R. E. Education and the cult of efficiency: A study of the social forces that have shaped the administration
of the public schools. Chicago: University
of Chicago Press, 1962.
Campbell, D. T., & Fiske, D. W. Validation by the multitrait-multimethod Psychological Bulletin, 1959, 56, 81-105.
Clark, K. B. Educational stimulation of racially disadvantaged children. In H. A. Passow (Ed.), Education in depressed areas. New York: Bureau of Publications, Teachers College, Columbia University, 1963. of "self-concept" in early childhood education.

Coller, A. R. The assessment
Urbana, Illinois: ERIC Clearinghouse on Early Childhood Education, 1971. (ERIC Document Reproduction Service No. ED 057 910)
Combs, A. W., & Soper, D. W. The self, its derivative terms, and research. Journal of Individual Psychology, 1957, 13, 134-145.
Combs, A. W., Soper, D. W., & Courson, C. C. The measurement of self-concept and self-report. Educational and Psychological Measurement, 1963, 23, 493-500. types of self-esteem. Journal of
Coopersmith, S. A method for determining
Abnormal and Social Psychology, 1959, 59, 87-94.
Coopersmith, S. The antecedents of self-esteem. San Francisco: Freeman, 1967. Crano, W. D., Kenny, D. A., & Campbell, D. T. Does intelligence cause achievement?: A cross-lagged panel analysis. Journal of Educational Psychology, 1972, 63, 258-275.

Cronbach, L. J. Essentials of psychological testing (2nd & 3rd eds.). New York: Harper and Row, 1960 and 1970.
Cronbach, L. J. Test validation. In R. L. Thorndike (Ed.), Educational measurement. Washington, D.C.: American Council on Education, 1971.


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions


et al.


Cronbach, L. J., Gleser, G. C., Nanda, H., & Rajaratnam, N. The dependability of behavioral measurements:
Theory of generalizability for scores and profiles.
New York: Wiley, 1972.
Cronbach, L. J., & Meehl, P. E. Construct validity in psychological tests. Psychological Bulletin, 1955, 52, 281-302. behavior: A
Crowne, D. P., & Stephens, M. W. Self-acceptance and self-evaluating critique of methodology. Psychological Bulletin, 1961, 58, 104-121. Duval, S., & Wickland, R. A. The effects of objective self-awareness on attribution of causality. Journal of Experimental Social Psychology, 1973, 9, 17-31. by a multitrait-multimethod

Dyer, C. O. Construct validity of self-concept
analysis. (Doctoral dissertation, University of Michigan, 1963). Dissertation Abstracts, 1964, 25, 8154. (University Microfilms No. 64-8154) Edwards, A. L. The social desirability variable in personality assessment and research. New York: Dryden, 1957. (a)

Edwards, A. L. Techniques of attitude scale construction. New York: AppletonCentury-Crofts, 1957. (b) Engle, M. The stability of self-concept in adolescence. Journal of Abnormal and Social Psychology, 1959, 58, 211-215.

and locus of control in
Epstein, R., & Komorita, S. S. Self-esteem, success-failure, Negro children. Developmental Psychology, 1971, 4, 2-8.
New York: Harper & Row,
Fantini, M. D., & Weinstein, G. The disadvantaged.
Foa, U. G. New developments in facet design and analysis. Psychological Review, 1965, 72, 262-274.
Gelfand, D. The influence of self-esteem on rate of verbal conditioning. Journal of Abnormal Social Psychology, 1962, 65, 259-265.
Gergen, K. The effects of interaction goals and personalistic feedback on the presentation of self. Journal of Personality and Social Psychology, 1965, 1, 413-424.
Goode, W. J., & Hatt, P. K. Methods in social research. New York: McGraw-Hill, 1952.
methodologies. The Journal of Nervous and Mental
Gordon, C. Self-conceptions
Disease, 1969, 148, 328-364.
Gordon, I. J. Studying the child in school. New York: John Wiley, 1966. Hamachek, D. E. (Ed.). The self in growth, teaching, and learning. Englewood Cliffs, New Jersey: Prentice-Hall, 1965.
Herbart, E., Gelfand, D. M., & Hartman, D. P. Imitation and self-esteem determinants of self-critical behavior. Child Development, 1969, 40, 421-430. descent.
of 6th-grade girls of Mexican-American
Hishiki, P. C. Self-concepts
California Journal of Educational Research, 1969, 20, 56-62. Hoepfner, R., Stern, C., & Nummedal, S. G. CSE-ECRC preschool/kindergarten test evaluations. Los Angeles: UCLA Graduate School of Education, 1971. (ERIC Document Reproduction Service No. ED 055 123)

James, W. Psychology. Greenwich, Conn.: Fawcett, 1963.
Jersild, A. T. In search of self. New York: Teachers College Bureau of Publications, 1952. Joreskog, K. G. A general approach to confirmatory maximum likelihood factor 1969, 34, 183-202.
analysis. Psychometrika,
Kaiser, H. F. The varimax criterion for analytic rotation in factor analysis. Psychometrika, 1958,23, 187-200.
Kelley, H. H. The process of causal attribution. American Psychologist, 1973, 28, 107-128.
matrix analysis of sense
Koppel, M. A., & Sechrest, L. A multitrait-multimethod
of humor. Educational and Psychological Measurement, 1965, 25, 77-85. Loevinger, J. Objective tests as instruments of psychological theory. Psychological Reports, 1957, 3, 635-694.


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



Vol. 46, No. 3

changes in the
Long, B., Henderson, E. H., & Ziller, R. C. Developmental
self-concept during middle childhood. Merrill-Palmer Quarterly of Behavior and Development, 1967, 13, 201-215.
changes in the
Long, B., Ziller, R. C., & Henderson, E. H. Developmental
self-concept during adolescence. School Review, 1968, 76, 210-230. Ludwig, D. J., & Maehr, M. L. Changes in self-concept and stated behavioral preferences. Child Development, 1967, 38, 453-467.

Marston, A. R. Dealing with low self-confidence. Educational Research, 1968, 10, 134-138.
McDonald, F. J. Educational psychology, Belmont, Calif.: Wadsworth, 1965. Mischel, W. Personality and assessment. New York: John Wiley, 1968. in its relationship to children's
Mote, F. A study of the home environment
self-concepts. (Doctoral dissertation, Stanford University, 1966). Dissertation Abstracts International,
Microfilms No. 671967, 27, 3319A. (University
Parker, J. The relationship of self-report to inferred self-concept. Educational and Psychological Measurement, 1966, 26, 691-700.
Piers, E. V., & Harris, D. A. Age and other correlates of self-concept in children. Journal of Educational Psychology, 1964, 55, 91-95.
Purkey, W. W. Self-concept and school achievement.
Englewood Cliffs, N.J.:
Prentice-Hall, 1970.
Rogers, C. R., & Dymond, R. F. Psychotherapy and personality change. Chicago: University of Chicago Press, 1954.
Sears, P. S. Self-concept in the service of educational goals. California Journal for Instructional Improvement, 1963, 6, 3-12.
Sears, P. S., Adenubi, M., Block, M., Crist, J., Gamble, J., & Hubner, J. Effective reinforcement for achievement
behaviors in disadvantaged
children: The
first year. Technical Report No. 30, Stanford Center for Research and Development in Teaching, 1972. (ERIC Document Reproduction Service No. ED 067 442)
Sears, P. S., & Sherman, V. In pursuit of self-esteem. Belmont, Calif.: Wadsworth, 1964.
Sherif, M., & Cantril, H. The psychology of ego involvements, social attitudes, and New York: John Wiley, 1947.
Smiley, M. B. Objectives of educational programs for the educationally retarded and disadvantaged.
In P. A. Witty (Ed.), The educationally
retarded and
disadvantaged. Sixty-Sixth Yearbook of the National Society for the Study of Education, Part I. Chicago: University of Chicago Press, 1967. instruction on student attitudes. JourSmith, I. D. Impact of computer-assisted nal of Educational Psychology, 1973, 64, 366-372.

Snygg, D., & Combs, A. W. Individual behavior. New York: Harper & Row, 1949. Soares, A. T., & Soares, L. Self-perceptions of culturally disadvantaged children. American Educational Research Journal, 1969, 6, 31-46.

Sullivan, H. S. The interpersonal theory of psychiatry. New York: W. W. Norton, 1953.
Super, D. E. Toward making self-concept theory operational. In D. E. Super, J. P. Jordan, N. Maitlin, & R. Starishevsky
(Eds.), Career development: Selfconcept theory. New York: College Entrance Examination Board, 1963. 1951.
Symonds, P. M. The ego and the self. New York: Appleton-Century-Crofts, in the study of the
Tannenbaum, A. J. Social and psychological considerations
In P. A. Witty (Ed.), The educationally retarded and
socially disadvantaged.
disadvantaged. Sixty-Sixth Yearbook of the National Society for the Study of Education, Part I. Chicago: Chicago University Press, 1967.
Trickett, H. Stability and predictability of children's self-concept and perceptions by others: A developmental
study. (Doctoral dissertation. Stanford University, 1968). Dissertation Abstracts International, 1969,29, 2557A. (University
Microfilms No. 69-306)


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions



Trowbridge, N. Self-concept and socioeconomic status in elementary school children. American Educational Research Journal, 1972, 9, 525-538. Vernon, P. E. The structure of human abilities. London: Methuen, 1950. Wattenberg, W. W., & Clifford, C. Relation of self-concepts to beginning achievement in reading. Child Development, 1964, 35, 461-467. Wylie, R. The self-concept: A critical survey of pertinent research literature. Lincoln: University of Nebraska Press, 1961.

Yamamoto, K. (Ed.). The child and his image. Boston: Houghton Mifflin, 1972. Yee, A. H, & Gage, N. L. Techniques for estimating the source and direction of causal influence in panel data. Psychological Bulletin, 1968, 70, 115-126. Zirkel, P. A. Self-concept and the "disadvantage" of ethnic group membership and mixture. Review of Educational Research, 1971, 41, 211-225.

Zirkel, P. A. Enhancing the self-concept of disadvantaged students. California Journal of Educational Research, 1972, 23, 125-137.
RICHARD J. SHAVELSON Address: Graduate School of Education, University of California, 405 Hilgard Ave., Los Angeles, CA 90024. Title: Associate Professor. Degrees: B.A., University of Oregon; M.A., San Jose State College; Ph.D., Stanford University. Specialization: Learning and individual differences, research design and measurement. JUDITH J. HUBNER Address: School of Education, Stanford University, Stanford, CA 93405. Title: Doctoral Candidate. Degrees: B.A., University of California at Berkeley; M.A., San Francisco State College. Specialization: Child development.

GEORGE C. STANTON Address: School of Education, Stanford University, Stanford, CA 94305. Title: Doctoral Candidate. Degrees: B.A., Lake Forest College; M.A., Cornell University. Specialization: Learning and individual differences.


This content downloaded from on Sat, 04 Apr 2015 14:12:25 UTC All use subject to JSTOR Terms and Conditions

Cite This Document

Related Documents

Discover the Best Free Essays on StudyMode

Conquer writer's block once and for all.

High Quality Essays

Our library contains thousands of carefully selected free research papers and essays.

Popular Topics

No matter the topic you're researching, chances are we have it covered.