American Educational Research Journal

http://aerj.aera.net

Self-acceptance: The Evaluative Component of the Self-concept Construct

Lorrie A. Shepard
Am Educ Res J 1979 16: 139
DOI: 10.3102/00028312016002139

The online version of this article can be found at:

http://aer.sagepub.com/content/16/2/139

Published on behalf of

American Educational Research Association

and

http://www.sagepublications.com

Additional services and information for American Educational Research Journal can be found at:

Email Alerts: http://aerj.aera.net/alerts

Subscriptions: http://aerj.aera.net/subscriptions

Reprints: http://www.aera.net/reprints

Permissions: http://www.aera.net/permissions

Citations: http://aer.sagepub.com/content/16/2/139.refs.html

>> Version of Record - Jan 1, 1979

What is This?

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 American Educational Research Journal
Spring, 1979, Vol 16, No. 2, pp. 139-160

Self-acceptance: The Evaluative Component of

the Self-concept Construct

LORRIE A. S H E P A R D
University of Colorado

The validity of the construct self-acceptance was tested to map the evaluative
dimension of self concept Three constructs—self acceptance, self description,
and acceptance of others—were each measured by seven methods (checklist,
rating scale, sentence completion, forced-choice questionnaire, semantic differ-
ential, TA T, and Q-sort) to create a multitrait-multimethod matrix of inter-
correlations Instruments were either selectedfrom the literature or developed
by the author and were administered in counterbalanced order to 137 individ-
uals ranging in agefrom 14 to 82. The average convergent validity values were
.55 for self-acceptance, .42 for self-description, and .41 for acceptance of
others. Both of the self-constructs had strong discriminant validity from
acceptance of others; the average correlation of .22 between self-acceptance
and acceptance of others was consistent with their theoretical relationship.
Self-acceptance was only slightly discriminated from self-description (conver-
gent validity coefficient of .55 compared to average discriminant correlation of
.41). Although some additional evidence of discriminant validity exists in the
patterns of correlation with age and other constructs, self-acceptance and self-
description were not as distinct as the semantic definitions of the constructs
imply.

Self-concept is an important construct in education and psychology. It is studied

as an outcome of schooling and counseling and as a mediating variable that influences
academic achievement. Substantive investigations of the causes or consequences of
changes in self-concept are hindered, however, by inadequate construct definition
and operationalization (Shavelson, Hubner, & Stanton, 1976; Wylie, 1974; Zirkel,
1971). The significance of improving self-concept as a goal in educational programs
is reflected in major programs for the disadvantaged (Graham & Hess, 1965; Smiley,
1967; Stebbins, St. Pierre, Proper, Anderson, & Cerva, 1977); the concomitant
difficulty in measuring its accomplishment is seen in criticisms of the evaluations of

(All values referred to as average correlations were obtained by converting all of the
correlations in a category using Fisher's Z transformation, computing the arithmetic average,
and reconverting the mean transformed value to a correlation coefficient.)

139

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

those same programs (House, Glass, McLean, & Walker, 1978; Huron Institute, 1973;
Wisler, Burns, & Iwamoto, 1978).
In a recent review, Shavelson et al. (1976) tried to bring some order to the self-
concept literature. The major contributions of their article were a formal definition
of self-concept and a first step in examining the congruence of definition and
empirical results. From the theoretical literature, Shavelson et al. (1976) identified
seven essential features of the self-concept construct: it is organized, multifaceted,
hierarchical, stable, developmental, evaluative, and differentiable from other con-
structs. The body of research available allowed them to conclude that general self-
concept is indeed stable and multifaceted, having differential aspects such as aca-
demic, social, emotional and physical self. However, they found virtually no evidence,
positive or negative, that could be used to test the evaluative dimension of self-
concept. They concluded that "the distinction between self-description and self-
evaluation has not been clarified either conceptually or empirically" (p. 414).
The lack of research on the evaluative aspect of self-concept is disconcerting
because this ought to be the most salient feature of the construct; it is the dimension
that corresponds to good or bad mental health. From the time that self-concept was
first formulated in early psychological theories, positive self-concept was thought to
be an attribute of an effectively functioning person. It was taken to mean, "I see
myself; and I like what I see." Positive self-evaluation was so synonomous with good
mental health that Rogers (1949) referred to actual-self-ideal-self-correlations, now
called self-concept measures, as measures of adjustment. Perhaps it is precisely
because the evaluative component is assumed to be so axiomatically a part of the
construct that it has not been investigated empirically.
The distinction between self-description and self-evaluation is important to edu-
cational practice as well as to theoretical formulation of self-concept. The countless
educational programs that have the goal of enhancing the child's self-concept espouse
it either as a desired end in itself or as a means to foster subsequent achievement
gains (Bloom, 1977; Rubin, Dorle, & Sandidge, 1977). It is assumed that a child's
self-concept is free to vary, that it can be enhanced, and that it is not automatically
and permanently determined by the child's self-descriptions. This assumption is
emphasized when educators refer to the desired goal as being a realistic and positive
self-concept (cf. Purkey, 1970). It is believed that a low ability child can have a
positive self-concept. He may realistically recognize his relative position among
classmates but will be optimistic and enthusiastic about his own ability to make
progress in each arena, academically, socially, and physically. Ideally, self-valuing
will be high even when realistic self-descriptions are average or negative.
Self-acceptance is the one of several self-concept variables which best corresponds
to the evaluative component of the self-concept. Self-acceptance refers to an individ-
ual's satisfaction or happiness with himself, and is thought to be necessary for good
mental health (Rogers, 1951; Scott, 1968). Self-acceptance involves self-understand-
ing, a realistic, albeit subjective, awareness of one's strengths and weaknesses. It
results in an individual's feeling about himself that he is of "unique worth" (English
& English, 1958; Jersild, 1960). As the construct is defined, it is implied that self-
acceptance is distinguishable from self-description; it is the value or affect associated
with that self-description. The self-accepting person, aware of both his strengths and

140

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

shortcomings, values himself. The self-rejecting person considers himself of little

worth and is likely to have other symptoms of maladjustment.
Operationalizations of self-acceptance in well-known self-concept measures illus-
trate that the unique feature of the construct is positive affect attached to assessment
of personal strengths and weaknesses. In Shostrom's Personal Orientation Inventory
(1966), the Self-Acceptance subscale is intended to measure "affirmation or accept-
ance of self in spite of weaknesses or deficiencies" (p. 6). Similarly, the Tennessee
Self Concept Scale has multiple scales, one of these is the Self-Satisfaction scale that
reflects "how he feels about the self he perceives" (Fitts, 1965, p.2) or his degree of
self-acceptance.
The purpose of this study is to test the validity of the construct self-acceptance,
especially to determine whether self-valuing can be differentiated from self-descrip-
tion. The process of construct validation follows the usual procedures for developing
and testing scientific theories. It requires specifying a pattern of interrelationships
among constructs and observables, which Cronbach and Meehl (1955) called a
nomological network. By comparing postulated relationships among observables with
obtained correlations, it is possible to test the validity of a measure or of a construct.
An additional relevant and presumably separate variable, acceptance of others, is
included in the study. Acceptance of others has been theoretically designated as
related to, but distinct from, self-acceptance (Rogers, 1949; Sheerer, 1949; Stock,
1949). Currently, self-acceptance and acceptance of others are paired in explaining
classroom interactions (Gold, 1958; Lippitt, 1962; Schmuck & Schmuck, 1971). Both
constructs use the word acceptance which denotes the feeling of worth or merit for
both self and others. The self-accepting person is likely to be accepting of others,
which means respecting their rights and responding positively to them despite
recognition of specific faults.
The multitrait-multimethod approach (Campbell & Fiske, 1959) used in the
present study is a systematic empirical procedure for examining the agreement
between obtained and anticipated correlations. Following this conceptual scheme,
the researcher uses a theoretical rationale to identify a trait or construct and other
relevant variables that should be distinguishable from it. Then, each trait is measured
by two or more methods. In this study, self-acceptance, self-description, and accept-
ance of others are each measured by several methods to test the soundness of the
constructs.

METHOD
Sample
Participants were 137 high-school and college students, parents of high-school
students, and residents of a retirement community, all from a middle-class, university
community. Participants were not asked to put their names on the questionnaires.
Eighty-one of the subjects were female and 54 were men. Two students failed to
report their sex.

Instruments
Construct validation requires the simultaneous observation of relationships among
several measures of each of the constructs of interest. To determine accurately the

141

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

relative importance of overlapping concepts or test-specific correlations, each of the

three constructs—self-acceptance, self-description, and acceptance of others—was
measured by each of seven methods. Several instruments were included in the study
initially because they were exemplary measures of the construct from previous studies
or because they were developed by the author to parallel directly the definition of
self-acceptance. Additional instruments were developed to complete the method-by-
trait matrix.

Five-Point Rating Scale. The Berger scale of Expressed Acceptance of Self and Expressed
Acceptance of Others (1952) was selected because the evidence for its validity (criterion-related)
is more extensive than for most other measures. The title was changed to Five-Point Rating
Scale to avoid direct identification Of attitudes being assessed. This inventory contains two
scales measuring two constructs by a single method. The 36 items measuring self-acceptance
and 28 items measuring acceptance of others are interspersed and administered as one test. The
response mode is a modified five-point Likert scale anchored by "not at all true of myself
(scored 1) and "true of myself (scored 5).
Five-Point Self-Rating Scale. Items to measure the third construct, self-description in the 5-
poinl Likert, mode were developed by the author to form a "Five-Point Self-Rating Scale." In
addition, self-acceptance items were written which strictly followed the theoretical model,
stating the affect associated with a self-assessment. Phrases such as "I am happy with," "I am
bothered by," "It upsets me," and "I am embarrassed by" were coupled with statements which
appeared in other items as straightforward self-descriptions. After the appropriate items had
been reverse-scored, two self-acceptance scores, a self-description score, and an acceptance of
others score were obtained by summing appropriate item scores from the two Likert question-
naires.
FIRO-B. The modified Likert format was duplicated in the acceptance of others category on
the Fundamental Interpersonal Relations Orientation—Behavior {FIRO-B). The FIRO-B instru-
ment, developed by Schutz (1967), comprises six subscales corresponding to Expressed and
Wanted behavior in the areas of social inclusion, control, and affection. The Expressed subscales
describe the respondent's own behavior and the Wanted subscales describe the behavior desired
of others. The Expressed Inclusion and Expressed Affection subscales were expected to be
measures of acceptance of others. The words the author uses to typify positive inclusion describe
behaviors that would be expected of one who is accepting of others: associate, interact, attend
to, togetherness, pay attention to, interested. Expressed Affection even more directly parallel
the acceptance or rejection of others. Terms used to exemplify positive affection are love, like,
emotionally close, personal, intimate, friend (Schutz, 1967, p. 5).
Each item was answered either as to frequency of occurrence on a scale from 1 to 6, or using
a scale indicating how many people the respondent felt were characterized by the item statement,
from 1 if the participant felt that way toward "most people" to 6 if the participant felt that way
toward "nobody."
The scores used in this study differed from Schutz's in that, rather than dichotomize the
responses (e.g., 1 and 2 equal 0; 3, 4, 5, and 6 equal 1), the full score-continuum was retained.
Scale scores were a simple sum of the numbers recorded by the participant after the reverse
scoring of positively worded items.
Judgments About Self A Judgments About Self checklist was developed to parallel one aspect
of the semantic description of self-acceptance, that is, the lack of negative affect associated with
negative self-descriptions. The participant read through the inventory checking faults. After
finishing, the respondent was asked to go through the instrument a second time and indicate
how bothered he or she was to exhibit each negative characteristic checked.
Based on pilot work (Shepard, 1972), a self-description score was obtained from the Judgments
About Self instrument as a tally of the number of negative attributes checked without attention

142

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

to the associated degree of concern. A high score was an indication of very negative self-
descriptions. The acceptance score was computed by summing the concern weights assigned to
each of the items. Statements which the participant said were not applicable were scored zero.
A high scorer was someone who was self-rejecting.
Q-Sort. The Q-Sort technique was included as an assessment method because of its popular-
ity. As a measurement mode, the Q-sort requires that the participant order statements, in this
case self-descriptions, along a continuum from "most like me" to "least like me." To improve
the psychometric properties of the instrument and facilitate the comparison of two Q-sorts, the
participant is required to arrange the statements along the continuum so that they approximate
a normal distribution. The sorting of self-statements results in a profile that can be summarized
in a single statistic: the sum of the products of scale positions and statement weights.
The most important use of the Q-sort as a measure of self-acceptance has been to determine
the agreement between actual and ideal self-perceptions. Such a coefficient, the correlation
between self and ideal sorts, was introduced by Butler and Haigh (1954) as a measure of self-
esteem. The self-ideal discrepancy score indicates not only the individual's perception of himself
as possessing each characteristic, but also the degree to which he values this state.
The original Butler-Haigh items could be classified into self-acceptance stems, self-description
stems, and stems that were unclassifiable. The first two sets of items were the basis for the
development of new Q-sorts to measure each of the two constructs. Additional items were
written to make the number of positive and negative stems about equal and to add areas of self-
evaluation which would not be assessed otherwise. After 26 items had been selected for the self-
acceptance scale, 26 parallel items were written to form an acceptance of others scale.
Each participant completed four Q-sorts: acceptance of self, self-description (actual), self-
description (ideal), and acceptance of others. One measure of self-acceptance was the correlation
between an actual and an ideal self-sort of the self-description items. This agreement score had
no counterpart for the other two constructs. A product-sum score was computed for self-
acceptance, self-description (actual), and acceptance of others sorts. In anticipation of this
summative scoring procedure, weights were assigned to each self-statement based on the average
of rankings assigned by six trained judges. Product-sum scores were the sum of all the item
weights multiplied by the position assigned to each by the respondent.
Word Rating Scale. A self-acceptance measure using a semantic differential format was
developed in pilot work (Shepard, 1972) and modified for this study. The participant was given
four semantic differentials and asked to rate each of four individuals known to him on a set of
15 bipolar adjectives. Person A was someone whom the subject thought highly of, Person B was
regarded positively but not as highly, Person C was an individual moderately rejected by the
subject, and Person D was someone strongly rejected. Following these four sets of responses,
the participant was asked to rate himself on the same 15 scales. A self-acceptance score was the
participants' location on the continuum from Person A to Person D who represented varying
degrees of acceptability to him.
The adjective pairs had been used by Osgood, Suci and Tannenbaum (1957, p. 242). The
scales included nine from the evaluative dimension (happy-sad, beautiful-ugly, clean-dirty,
honest-dishonest, valuable-worthless, good-bad, pleasant-unpleasant, fair-unfair, and healthy-
sick); three from the potency dimension (large-small, deep-shallow, and strong-weak); and four
from the activity factor (active-passive, sharp-dull, hot-cold, and fast-slow). Other scales were
added which the authors thought were particularly relevant: calm-excitable, adaptable-inflexi-
ble, self-assertive-submissive, and tense-relaxed. For this study, self-assertive-submissive was
arbitrarily replaced by black-white because the second pair is less specific. A random half of the
bipolar adjectives placed the positive descriptor first to avoid a response set induced by the
format of the scale.
To construct a self-acceptance score, the participant received four scores descriptive of the
discrepancy between the self rating and each of the four stimulus persons. Each score was the
sum of the squared discrepancies for each comparison of adjective pairs. The square roots of

143

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

these total scores were also computed; the four square roots were considered more appropriate
distance measures and could be interpreted so as to allow a comparison of the relative distance
between ME and Person A and ME and Person D. An additional score was computed as the
ratio of the ME-A distance to the ME-A plus ME-D distance; the denominator represented the
entire continuum from A to D. This final score assumed, of course, that the ME value was
between A and D and not more extreme than either of them.
Two short self-description instruments were constructed in the semantic differential mode.
The first contained the evaluative items from the self-acceptance measure, simply scored by
adding across adjective scales rather than comparing the item-scale ratings to the four persons.
A second scale was composed of more specific adjective pairs representing the participant's
physical, intellectual, or social self (e.g., effective-ineffective, coordinated-uncoordinated, and
sociable-unsociable). The final score was the sum across 14 adjective pairs.
The acceptance of others measure was the same set of 20 adjective pairs used in the self-
acceptance device but with "other people" as the stimulus. The use of "other people" as the
stimulus for a semantic differential has many potential difficulties. The task of considering all
other people as an entity may be so difficult as to introduce many variables other than the
person's acceptance or rejection of others into the response pattern. If this were the case, the
semantic differential instrument would not correlate with other measures of acceptance of
others. Since the success of the stimulus stem could not be predicted a priori, it was decided to
use the same rating scales with more specific stimuli.
Three short video tapes were put together to serve as a stimulus for semantic differential
ratings. The first two segments were excerpts of two different women talking about their
personal problems taken from a longer tape made by Harrington (1971). The third tape segment
was not a performance but was taken from a tape made for a study by Remer (1973) about
modeling in counseling.
The Video Tape Word Rating represented the same set of twenty rating scales as in the self-
description and acceptance of others semantic differentials with "First Woman," "Second
Woman," and "The Two Women Talking" as the three objects to be rated.
Forced-Choice Questionnaire. Edwards (1954) used a forced-choice format to control for a
social desirability response set. Although there is no standard procedure which can take into
account very different individual views as to the social desirability of a given response, the
forced-choice format does control for the response set for most individuals if there is fairly good
concensus among respondents (as well as among judges) as to the relative social acceptability
of responses. In a forced-choice test, the participant is presented with a series of item pairs,
matched as to their judged desirability, and told to choose the statement which is more
personally true.
The Personality Orientation Inventory (Shostrom, 1966) was the source of most of the item
stems used to construct the forced-choice instrument. The POI was selected because it contains
scales that claim to differentiate self-acceptance from general positive self-regard. More careful
content analysis, however, revealed that Shostrom's operationalization of the two constructs at
the item level failed to maintain the distinction between self-description and self-evaluation
sufficiently to meet the criteria for this study.
A /W-based, forced-choice instrument was constructed by the author to assess each of the
three constructs after the 150 items in the POI had been classified as a self-accepting statement,
a straight self-description, an acceptance of others statement, or an irrelevant stem by two
judges. The POI was not designed to measure self-description; therefore, 14 new items were
written.
The classifiable items from the POI and the newly constructed statements were submitted to
two groups of graduate and undergraduate students in education to be rated on social
desirability. These 50 "judges" were given short training materials in which social desirability
was defined and examples given of different degrees of social desirability. Items with negative
social desirability were not included because in the forced-choice format the participant's failure

144

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

to select a negatively worded self-acceptance statement could not necessarily be given the same
value toward a final score as the selection of a positive self-acceptance item.
The final Forced-Choice Questionnaire was composed of item pairs in which a construct-
relevant statement is matched with an irrelevant stem to yield equal mean social desirability
scores and approximately equal variability. Statements referring to self-acceptance, self-descrip-
tion, and acceptance of others are always paired with irrelevant foils to avoid the problems of
ipsative measurement.
Incomplete Sentences. Although the tests discussed above represent several methods of
assessing the constructs, if the term method were construed in a larger sense, the aforementioned
tests would all be examples of a single modality—data derived from the participant's self-report.
The selection of the last two tests included in this study was made in an effort to gather
additional data that would not be dependent on self-report.
The sentence completion test is a quasi-projective means of attitude assessment. Although
the person may still edit and control personal responses, there is some significance attached to
the answer he selects from the universe of all possible responses. The Rotter Incomplete
Sentences Blank (Rotter & Rafferty, 1950) was used as a model for the development of a new
sentence completion questionnaire which was likely to elicit responses scoreable as to self-
acceptance, self-description, or acceptance of others.
The scoring examples in the Rotter and Rafferty manual (1950) were read to determine
which items were frequently responded to with sentences which reflected self-description, extent
of self-acceptance, or acceptance of others. Ten sentence fragments were retained from the
Rotter instrument and were expected to most frequently represent the constructs according to
the following scheme:
Self-Description: I c a n ' t . . . , I am very . . .
Self-Acceptance: I f e e l . . . , I regret..., I failed . . .
Acceptance of Others: Men . . . , People . . . , Most women . . . , Other people . . .
Additional stems were written which were expected to give each of the three constructs about
equal representation. Of course, no special instructions were given to the participants to suggest
that each response should be relevant to a particular construct.
The Incomplete Sentences test was scored by six judges working independently. The scoring
instructions used to train the judges, who were doctoral students in educational research,
included thorough definitions of the constructs, a scoring format for items as well as a total-
score procedure, and examples along a five-point continuum for each of the three constructs.
Participants were assigned to judges using a Greco-Roman design. Every judge read every
participant's tests, but scored each third of the tests according to a different construct. Every
subject was scored twice. After the computation of interjudge reliability, the two scores obtained
for each person on each construct were averaged to arrive at the score used in the multitrait-
multimethod analysis.
The Thematic Apperception Test (Murray, 1943) is a true objective device. Although it does
not meet high psychometric standards of good interjudge reliability and criterion validity
(Adcock, 1965; Kleinmuntz, 1967), it is still a popular instrument in research and clinical
practice. The TA T was not designed originally to assess the specific constructs considered by
this study; rather it was intended to be a broader measure of personality, general adjustment,
and motivating needs. The TA T is, however, a set of relatively ambiguous stimuli to which
researchers frequently attach their own set of interpretations. It was hoped that a more specific
set of scoring procedures developed for this study would not only increase the relevance of the
TAT to the study, but also would so structure the interpretations that interjudge agreement
would be enhanced.
Six cards were selected from the TAT for administration. Twenty TA T protocols gathered as
part of a projective tests course and cited in textbook case studies were read in an attempt to
identify those cards that would most likely produce material scoreable for the three constructs.
The TA T cards were administered by graduate students in education who may or may not

145

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

have had previous experience with the TA T Each administrator was given an orientation so
that problems of time limits and prompting (when a participant failed to tell a story or left out
what lead up to or what came after the scene in the picture) would be handled consistently. The
standard instructions from the Murray (1943) manual were read to the participant. The TAT
administrators took notes on the stories told by each participant. Tape recordings were also
made of each test. The stories were transcribed from the examiner's notes to be scored by
trained judges. The tapes were used whenever notes were unclear or incomplete.
The judges who scored the TA T responses were doctoral students who had each had at least
one course in projective techniques and had administered and interpreted the TA T under the
supervision of an experienced user of the test. In addition, the judges were given extensive
instructions about the constructs and examples of stories which were relevant to particular
constructs. Three pairs of judges working independently read random thirds of the participant's
protocols. The judges determined final scores by sorting participants for a given construct until
their relative position could be indicated on a one to seven continuum, a high score representing
the positive end of the scale. The scores given by two judges for a particular participant and
construct were averaged for use in the analysis.

Procedures
The tests were administered in three-hour sessions on Saturdays or weeknights
over a two-week period at the counseling facility at the University of Colorado. Each
participant attended only one session. To encourage the participants to take the task
seriously, every person was given a separate office with a desk to work in.
Each test packet began with a page of instructions that stressed participation in a
research study. The purpose of the study was to gather information about the validity
of the tests, not to evaluate the participants; they were urged to answer as honestly
as they could. Each participant was told there would be two interruptions, once to
take an individual oral test (the TAT) and once to see a video tape in a different
room. Residents of the retirement community were tested at the residence and in two
sittings rather than one.
Each of the test packets had instructions in front, with space provided to indicate
age and sex, followed by the Forced-Choice Questionnaire. This test was administered
first in all cases to avoid any effect of the other tests in sensitizing subjects to the
specific personal attitudes being measured, thereby unbalancing the matched social
desirability of item pairs. All other tests in the battery were randomly ordered in an
attempt to prevent a systematic order or fatigue effect.

RESULTS A N D DISCUSSION
The results of the study consist of a large correlation matrix depicting the obtained
relationships between every pair of variables measured. The multitrait-multimethod
analysis of the matrix involves the appraisal of the relative magnitude of coefficients
in subsets of the matrix.

Elimination of Inadequate Measures

Although one may not legitimately discard tests because they do not confirm one's
expectations about convergent and discriminant validity, it is appropriate to eliminate
tests that appear to measure nothing but error, that measure factors other than the
intended construct, or that are a very poor measure of a construct, before drawing
conclusions about relationships among constructs. As a general rule, if many instru-

146

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

ments designed to measure a particular construct do not correlate well with one
another, the construct lacks validity. If, however, one instruments fails to correlate
with other measures of the same construct that are strongly related to each other,
that particular instrument is said to lack construct validity, especially if it also has
poor internal properties. The latter determination is made in this section.
Table 1 contains statistical data on all those instruments eventually eliminated
from consideration. The instruments are correlated with the tests retained in the
study and among themselves. The rationale for omitting each of these tests is
presented in the following paragraphs.
All three scales of the Forced-Choice Questionnaire were ineffective measures of
each of the three constructs. The correlations reported in Table 1 for the forced-
choice subtests were computed after the measure had been improved by eliminating
items which were negatively correlated with the total scale scores.
The forced-choice format had been developed in an attempt to control for a social
desirability response set. Therefore, it cannot be set aside lightly because it does not
correlate well with other measures of the construct. The forced-choice instrument
could be the most accurate measure of self-acceptance, for example, with the
intercorrelations of the other tests attributable to social desirability. However, the
forced-choice tests also had very poor internal properties. Even after removing the
items which had slight negative correlations with the total score, the internal
consistency coefficient for the self-acceptances subtest was improved only from .20
to .50. Such an instrument cannot be expected to locate persons reliably along a
continuum representing the construct. The other two subtests were even weaker in
this respect, making it likely that they are poor measures of the constructs rather
than valid measures uncontaminated by social desirability. The issue of social
desirability as the possible cause of correlations for other measures will be dealt with
as part of the discussion of discriminant validity.
The Thematic Apperception Test failed in general to correlate with any other tests.
Only six coefficients were significantly non-zero considering all three of the TA T
scores crossed with all other tests. Given the total number of correlations, this is
about what one would expect by chance (with a = .05). The only pattern in the
correlations was a consistent, small relationship between the TAT for all three
constructs and the Incomplete Sentences measure of acceptance of others.
The poor performance of the TA Tcan plausibly be attributed to internal properties
of the measure. Mean interjudge reliabilities averaged across three pairs of judges
were .36, .46, and .57 for self-acceptance, self-description, and acceptance of others
respectively. The three TA T scores correlated .51 and .50 with each other, indicating
that whatever is being measured is probably one factor, not three.
Although the FIRO-B was highly internally consistent, it was omitted from the
discussion of construct validity because it measures some things or things other than
the constructs of interest. Particularly, it does not measure acceptance of others as
was originally thought possible. The correlations reported in Table 1 are for three
subscale scores rather than six. The intercorrelation matrix of subscales was inspected
in an early analysis. Several of the subtests were not distinguishable; therefore,
subtests 1, 2, 5, and 6 were combined as one homogenous variable or factor, FIRO-
B(\). Subtests 3 and 4 were treated as specific factors.
The FIRO-B(l) scale failed to correlate with measures of acceptance of others

147

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

00
TABLE 1
Instruments Eventually Eliminated from the Evaluation of Construct Validity
Instruments ; Eventually Eliminated

-Me)
II Q

ffect
iclus
hers
«T ? CA < +
4> (X
'oT c" O <& J3 'oT

d Ratinj 1 (Mean Sq. Disc

Foreed Choi e (Self-Acceptanc

d Ratinj 9 ([A-Me]/[ ( A M
Foreed Choi e (Acceptance of i
Foreed Choi e (Self-Descriptio

em. Dif Self-Accept ance)

em. Dif. Self-Accept ance)
ressed & W;anted
ressed & W;anted
3 5<*- 3
? t en O

' (Accep ance of Oth<

FIR*O-B (W; nted Contro

FIR<0-B (Ex ressed Cont

em. Dif. Acceptance

o Tape Lating Scale
em. Dif. Acceptance
em. Dif. Acceptance

o Tape .ating Scale

o Tape ] ating Scale
TAT (Self-D scription)
TAT (Self-A ceptance)
o o o o <u CL Cv, Cu a: i OS 1 » i *t% i an '
X X «!
WW
«

FIR<
< 6 5£
Instruments Retained in the Analysis > > >
Judgments About Self .25 .04 -.13 .08 -.02 -.05 -.03 -.16 -.37 .02 -.09 -.05 -.31 -.10
(Checklist-Self-Acceptance)
Five Point Rating Scale (Berger) .23 .04 -.11 .18 -.04 .08 .25 -.08 -.35 -.04 .00 .03 -.10 -.15
(Likert-type - Self-Acceptance)
Five Point Self Rating Scale .29 .13 -.13 .08 -.05 -.01 .10 -.02 -.30 -.02 -.06 -.11 -.01 -.08
(Likert-type - Self-Acceptance)
Self-Ideal Correlation .18 .04 -.06 .12 -.09 .00 .10 .02 -.31 .05 .06 -.01 -.25 -.17
(Q-sort - Self-Acceptance)
Self-Acceptance Product Sum .22 .19 -.09 .18 -.08 .06 .19 .07 -.43 .04 .07 -.04 -.19 -.17
(Q-sort)
Incomplete Sentences .12 .08 -.19 .14 -.05 .08 .17 .00 -.27 -.07 -.07 .04 -.16 -.10
(Self-Acceptance)
Judgments About Self .18 -.05 -.07 .00 -.10 -.01 .02 -.06 -.14 .05 -.09 -.01 -.06 .06
(Checklist - Self-Description)

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

Five-Point Self Rating Scale .23 .25 -.09 .06 -.15 -.05 .34 .22 -.27 -.02 .01 .15 -.12 -.22
(Likert-type - Self-Description)
Self-Description Product Sum .21 .06 -.08 .13 -.04 .09 .16 .01 -.33 .01 .04 -.03 -.30 -.15
(Q-sort)
Me (1) .16 .13 -.07 .07 -.20 -.02 .14 .06 -.42 -.07 .15 .12 -.40 -.48
(Semantic Dif. - Self-Description)
Me (2) .08 .30 -.09 .07 -.19 -.02 .24 .12 -.43 -.01 .06 .11 -.43 -.36
(Semantic Dif. - Self-Description)
Incomplete Sentences .08 .05 -.17 .15 -.03 .10 .22 .16 -.22 -.15 -.05 -.05 -.13 -.10
(Self-Description)
Five-Point Rating Scale (Berger) .17 -.07 .17 -.04 -.04 .02 .20 -.37 -.05 .01 .14 .23 .03 .01
(Likert-type - Acceptance of Others)
Acceptance of Others Product Sum .18 -.08 .05 .18 .23 .18 .13 -.14 -.12 .18 -.05 .06 .09 .18
(Q-sort)
Others .10 -.05 .19 .07 .06 .07 .12 -.07 -.06 .20 .05 .21 .05 .03
(Sem. Dif. - Acceptance of Others)
Incomplete Sentences .11 -.07 .06 .24 .23 .17 .12 -.08 -.11 -.01 -.11 .05 .01 .07
(Acceptance of Others)
Forced Choice (.50) .15 -.04 .01 .01 .07 .03 -.20 -.12 .07 -.03 .14 -.06 .00
(Self-Acceptance)
Forced Choice (.43) .04 -.02 -.09 -.09 .26 .05 -.10 .14 .15 .23 .01 -.12
(Self-Description)
Forced Choice (.15) -.06 .01 .11 .01 -.07 .06 -.05 -.17 .07 .03 .03
(Acceptance of Others)
TAT (.38)* .51 .50 .00 .16 -.01 .08 .03 -.03 .08 .07
(Self-Acceptance)
TAT (.46)* .50 -.22 -.05 .18 .07 -.02 -.01 .14 .19
(Self-Description)
TAT (.57)* -.07 .04 .05 -.01 -.01 -.02 .07 .23
(Acceptance of Others)
FIRO-B (Expressed & Wanted Affection) (.94) .32 .04 .09 .03 .24 -.03 -.14
(Expressed & Wanted Inclusion)
FIRO-B (.92) .09 .00 -.15 -.12 -.06 -.10
(Expressed Control)
Note. Those numbers appearing in parentheses are Cronbach's a internal consistency coefficients. Those numbers appearing in asterisked parentheses are interjudge reliability
coefficients.

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

o
TABLE 1 (Continued)
Instruments Eventually Eliminated

'ord R ating 1 (Me;an Sq. Discrep. A-Me)

(Expressed & Wanted Incl usion)
action)
orced <Choice (Ace eptance of Oth ers)

(Sem. Dif.-Accep tance of Others

(Sem. Dif.-Accep]tance of Others
M
+
'«' '5s /
5T

orced <Choice (Self -Descriptio

(Sem. Dif.-Accep tance of Ot

orced (Choice (Self -Acceptanc

IRO-B (Expressed & Wanted

(Sem. Dif.-Self-A cceptance)

IRO-B1 (Expressed Control)
AT (Acceptance oif Others)

ideo Tape Rating Scale 3

ideo Tape Rating Scale 2
IRO-B (Wanted Control)

ideo Tape Rating!Scale 1

9os

AT(S«:lf-Acceptan ce)

'ord Riating 9 flA-]

AT(Se jlf-Descripti

(Sem. Dif.-Self-A
Instruments Retained in the Analysis fc u. fc H H H u. u- u. > > > ^ *
FIRO-B (.83) -.07 -.09 .05 .23 .24
(Wanted Control)
Video Tape Rating Scale 1 (.82) .25 .11 .04 .21
(Sem. Dif.-Acceptance of Others)
Video Tape Rating Scale 2 (.84) .28 -.10 -.09
(Sem. Dif.-Acceptance of Others)
Video Tape Rating Scale 3 (.88) .09 -.12
(Sem. Dif.-Acceptance of Others)
Word Rating 1
(Mean Sq. Discrep. A-Me) .57
(Sem. Dif.-Self-Acceptance)
Word Rating 9
([A-Me]/[(A-Me) + (D-Me)])
(Sem. Dif.-Self-Acceptance)

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

possibly because it was an undifferentiable collection of statements about both the

participant's treatment of others and behavior wanted from others. The "other-
directed" behaviors described by the author as affection and inclusion apparently
have only small overlap with "acceptance." It had not been Schutz's purpose to
measure acceptance of others. Considering the potency of some of the descriptors,
e.g., love and intimacy, it may be that the FIRO-B reflects the nature of the
participants closest relationships rather than a more global response to "others."
Some unexpected correlations did obtain for the FIRO-B scales not intended as
measures of the constructs of interest. The consistent, moderate, negative correlations
between FIRO~B(3), "Wanted Control," and self-acceptance and self-description are
relevant to the validation of the constructs by demonstrating theoretically appropriate
relationships with other variables. These will be considered later in the discussion.
The apparent failure of the FIRO-B subscales to function as expected in relation
to measures of acceptance of others causes one to question the scoring procedure
adopted in this study. However, when each participant's test was rescored using the
answer keys provided by the author, correlations between the Shepard (1973) and
Schutz (1967) scorings of each of the six subscales ranged from .87 to .94. Thus,
substituting scores obtained from the Schutz key would not alter the relationship of
the FIRO-B subscales to other variables.
The Video Tape Word Rating scales were omitted from subsequent analyses
because they appeared to fail to assess a general acceptance of others. Perhaps the
Video Tape Word Rating scales were effective measures of the participants' "accept-
ance" of each of the three persons rated. The relatively high internal consistency
coefficients suggest that the items' scores reflect at least a consistent impression of or
response to the person viewed rather than the evaluation of specific, unrelated traits.
But, it is inappropriate to call this response acceptance of others since low correlations
among the three ratings clearly indicate that a generalizable response to others is not
being assessed.
The final instrument to be eliminated from the analysis was the semantic differ-
ential measure of self-acceptance. The Word Rating score No. 1, referred to in Table
1, is the sum of the squared discrepancies between the participant's rating of himself
on the semantic differential and his rating of Person A, someone he is very accepting
of and has a positive affective reaction toward. Score 9 is the square root of the A-
ME discrepancy sum. The two scores are representative of the nine scores computed
for the instrument. None of these scores was strongly related to other measures of
self-acceptance. The largest correlations were predictably with the semantic differ-
ential measures of self-description; the first (ME 1) has snared items and the second
is still assessment by a common method.
Using different adjective pairs, a semantic differential measure of self-acceptance
had functioned moderately well in a pilot study. A clue as to why the instrument was
ineffective in this study is provided by the intercorrelation of the four "person" scores
in Table 2. It had been expected that the self-accepting participant would rate himself
very similar to Person A and very unlike Person D. The positive correlations between
all of the scores rather than inverse relationships between A and B with C or D
indicate that the instrument did not function as intended. The A through D ratings
certainly did not form a continuum. Participants apparently developed a response
mode whereby they rated themselves similar to all four persons rated or generally

151

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SHEPARD

TABLE 2
Intercorrelation of Four Person Ratings on Word Rating Scale
(A-ME)2 (B-ME)2 (C-ME)2 (D-ME)2
(A-ME)2 .63 .45 .12
(B-ME)2 .49 .41
(C-ME)2 .55

unlike all four persons. The final scores were more a function of this response mode
than of the individual's self-acceptance.
Elimination of several instruments from the analysis causes the multitrait-multi-
method matrix to be incomplete, but the analysis is not jeopardized by the deletions
for two reasons. First, most of the instruments eliminated represent a common
method across three constructs; therefore, the matrix is not badly unbalanced.
Secondly, the relationship between constructs is made clearer by the elimination of
the instruments with poor internal properties. The forced-choice and TAT methods
were eliminated for all three constructs. The exclusion of the FIRO-B and Video
Tape Rating Scale did not create a gap in the matrix since these instruments had
duplicate method types within the construct. The only gap caused by discarding
inadequate measuring devices is for a semantic differential method of assessing self-
acceptance. The multitrait-multimethod matrix remains intact after removing the
instruments discussed above; each construct is assessed by several measures and
every method is used to measure at least two constructs.

Multitrait-Multimethod Analysis
Convergent validity is the first condition necessary for construct validation. The
correlations reported in Table 3 indicate a large degree of convergent validity among
the several measures of each of the constructs. Separate measures of self-acceptance
were on the average correlated .55 with each other. The agreement among measures
of self-description and acceptance of others was not as great; average correlations
among tests were .42 and .41, respectively.
The strengths of the relationships among tests should always be evaluated in light
of the reliability of each measure. Two constructs, acceptance of self and acceptance
of others, had the highest internal consistency coefficients. This outcome could be
anticipated on a theoretical basis since these two constructs are more unitary traits
than is self-description (as treated in this study). Self-description is measured as an
aggregate of evaluations of different parts of oneself. The assessments of intellectual,
physical, social, and emotional self may not be highly correlated or consistent. A
similar effect, caused by nonparallel parts of an instrument, accounts for the two
lowest internal consistency values among measures of self-acceptance. Judgments
About Selfimd the Five-Point Self Rating are composed of items expressing affect
associated with a number of different self-descriptions. Although the rationale for
the instruments assumes that the self-accepting person will have a final score which
is high in the total amount of positive affect expressed, it does not require that all
items be answered consistently.
In general, this study was more effective in assessing self-acceptance than in
measuring the two related constructs. Measures of self-acceptance had internal

152

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 TABLE 3
Multitrait-Multimethod Matrix
Self-acceptance Self-description Acceptance of Others

1 2a 2b 3 4 6 1 2b 4 5a 5b 6 2a 4 5 6 r-H

Self-Acceptance
1. Judgments About Selfa (.74)1 .66 .55 .55 .57 .58 .58*h .43 .57 .40 .48 .50 .36 .33 .11 .36
(Checklist-Self-Acceptance)
2a. Five Point Rating Scale (Berger) (.92) .65 .53 .72 .59 .40 ,56 .54 .48 .41 .45 .52 .37 .07 .32
(Likert-type-Self-Acceptance)
2b. Five Point Self Rating Scale (.78) .40 .50 .42 .57 JO .37 .33 .36 .45 .36 .28 -.03 .31
(Likert-type-Self-Acceptance)
3. Self-Ideal Correlation .50 .39 .24 .35 .76* .39 .28 .31 .22 .21 .10 .34
(Q-sort-Self-Acceptance)
4. Self-Acceptance Product Sum (.71) .56 .31 .47 M .47 .41 .11 •29 .08 .21
(Q-sort)
d2
6. Incomplete Sentences (56) .28 .40 .42 .32 .41 .69* .17 .24 .07 .28*
(Self-Acceptance)
Self-Assessment
1. Judgments About SelP (.87) .46 .33 .30 .27 .34 .35 .20 .05 .31
(Checklist-Self-Description)
2b. Five Point Self Rating Scale (.67) .40 .41 .59 .45 .28 .18 -.10 .12
(Likert-type-Self-Description)
4. Self Description Product Sum (.44) .48 .43 .44 .21 .29 -.06 .15
(Q-sort)
5a.Me(l) (.70) .60 .27 .27 .07 .05 .10
(Semantic Dif.-Self-Description)
5b. Me (2) (.77) .41 .12 .01 -.12 .04
(Semantic Dif.-Self-Description)
6. Incomplete Sentences (.59) .14 .14 -.05 .28*
(Self-Description)
Acceptance of Others
2a. Five-Point Rating Scale (Berger) (.78) .50 .25 .31
(Likert-type-Acceptance of Others)
4. Acceptance of Others Product Sum (.65) .42 .53
(Q-sort)
5. Others (.87) .41
(Semantic Dif.-Acceptance of Others)
6. Incomplete Sentences (.79)
(Acceptance of Others)
SEX .04 .03 -.02 -.06 -.06 .02 -.06 -.06 .05 .05 .01 .02 .29 .13 .03 .10
AGE .31 .34 .14 .17 .17 .13 .21 -.08 .27 .01 -.03 .14 .22 .30 .14 .31
Note. Internal consistency values, reported in parentheses, were obtained for most instruments using Cronbach's a. For sentence completion tests inteijudge reliabilities
are reported instead. Q-sort reliabilities were computed using Hoyt's formula; Judgments About Self-self-description internal consistency value was obtained using KR2i.
a
Negatively scored tests were reversed scaled before including in matrix.
b
Underlined coefficients indicate correlations between different constructs using the same method; asterisked values include shared errors of measurement.

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

properties which were as good as or better than those for acceptance of others
instruments. There was greater convergent validity among measures of self-accept-
ance. The more effective assessment of the principal construct perhaps reflects the
greater attention given in this study to the specification of the construct and the
selection of its operationalization in distinct measurement modes.
The acceptance of others measures had good internal consistency and the highest
agreement among subjective judgments made in scoring the sentence completion
tests. The limited overlap among the several measures (.41) is more attributable to
specific operationalizations of the construct by different measurement modes than to
the unreliability of the instruments.
To attach meaning to the magnitude of convergent validity correlations, they must
be compared to appropriate discriminant validity coefficients. As a second basic
requirement of the multitrait-multimethod approach, convergent validity values must
exceed their respective heterotrait-heteromethod correlations. The relative size of
average convergent and discriminant validity values is illustrated in Figure 1. The
conditions for distinquishing between constructs are dramatically met for both self-
acceptance and self-description when compared with acceptance of others. There is
some evidence that self-acceptance was discriminated slightly from self-description.
Self-acceptance measures had an average correlation of .55 with one another across
methods, but correlated on the average only .41 with self-description measured by
the same array of different methods. Although there is a strong relationship between
self-acceptance and self-description, the difference in these average intercorrelations
suggests that there is a reliable portion of self-acceptance that cannot be accounted
for by self-descriptions. There is no discriminant validity, however, for self-descrip-
tion from self-acceptance. The correlations among measures of self-description are
not greater than the correlations between self-acceptance and self-description.
A further, stricter requirement for discriminant validity, outlined by Campbell and
Fiske (1959), is that the correlations between different measures of the same construct
be greater than the correlations between different constructs measured by the same
method. If instruments are adequately measuring a construct, the object of assessment
will be a single concept, not several unique specifications of the construct associated
with each mode of measurement.
For both self-acceptance and self-description correlated with acceptance of others,
the final criterion for discriminant validity is met. The correlation between self-
acceptance and self-description measured by the same methods is very high however.
Measured by different methods, the two constructs are correlated .41 on the average.
When common methods are used, the average correlation increases to .64, which is
higher than the convergent validity values for either of the two constructs. Even
when one eliminates the tests that involve two scorings of the same data (such tests
will inflate the correlations because of shared errors of measurement), the heterotrait-
monomethod correlation for self-acceptance and self-description is too high to meet
the Campbell-Fiske specification. Of course, the Campbell-Fiske criterion is based
on the comparison of constructs that are not highly related (i.e., that show strong
discriminant validity). When the multitrait-multimethod analysis is used to determine
if two constructs are really one or are merely strongly related, it is likely that the
coefficient representing the correlation between the constructs, plus the influence of
method specific variance, will exceed the values for convergent validity.

154

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

MONOTRAIT MONOTRAIT HETEROTRAIT HETEROTRAIT

MONOMETHOD HETEROMETHOD MONOMETHOD HETEROMETHOD
(reliability) (convergent v a l i d i t y ) (discriminant v a l i d i t y )

1.00

.90

.80

o .70
MEAN CORRELAT]

.60

.50

.40

.30

.20

.10

.00
SA SD AO SA SD AO SA SA SD SA SA SD
SD AO AO SD AO AO

FIGURE 1.

The presence of method-specific variance does not influence the size of either the
convergent validity coefficients (correlations among different methods) or the het-
erotrait-heteromethod discriminant validity correlations. These are the values to be
examined in evaluating the validity of the constructs. The magnitude of the hetero-
trait-monomethod correlations, however, answers a different and more practical
question about the interchangeability of the methods for assessing each of the
constructs. The monomethod correlation between self-acceptance and self-description
is high because the constructs are strongly related (or are one construct) and because
of the unique contribution each method makes to the reliable portion of subjects'
scores. The presence of method specific variance indicates that despite the level of
convergent validity, the several measures of self-acceptance are not measuring exactly
the same thing but are assessing several subconstructs (Fiske, 1971) associated with
each method.
The final desideratum proposed by Campbell and Fiske is that the same pattern of
correlations among traits (constructs) be maintained whether the same or different
methods are used. This requirement is satisfied. It may be observed in Figure 1 that
the same relative correlations between constructs are maintained in the heterotrait-
monomethod column as in the heterotrait-heteromethod column. The pattern also
may be observed for each individual method.
The observation of discriminant validity between the self-constructs and accept-
ance of others is the best evidence to discount the possibility that a social desirability
response set caused the intercorrelations of methods within constructs. If participants'
tendencies to represent themselves in a favorable light were determining their
responses, then there would necessarily be much higher correlations between both
self-acceptance and self-description with acceptance of others. "Everything would

155

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

correlate with everything else" if social desirability were the one underlying "con-
struct" being assessed.
The overall interpretation of the multitrait-multimethod matrix supports the
construct validity of self-acceptance. But this is a verification of the construct in its
broadest terms. In popular usage, the construct is frequently associated with high
self-esteem or with good self-concept. Treated in this way, the concept of self-
acceptance is substantiated when there are large correlations with self-description.
Much more care must be taken, however, when self-acceptance is used to mean a
person's feelings about himself independent of the socially determined value of self-
description. The evidence is only slight that the affective reaction of an individual to
self-descriptions will be any different from the normative a priori value attached by
the person's culture to those assessments. The theoretical specification of self-accept-
ance would lead one to predict a strong relationship between self-acceptance and
self-description. This postulated relationship is supported by the high correlation
(.41); but given the only modest convergent validity value (.55), this presents a
problem for discriminant validity at the same time.

The Influence of Measurement Artifacts on the Magnitude of Correlations

The evaluation of the multitrait-multimethod matrix must take into account any
factors other than the constructs that influence the size of the correlations. The
appraisal of convergent and discriminant validities has proceeded uninterrupted
because potential artifacts of scaling or imperfect measurement techniques actually
had little effect on the correlations.
A number of instruments resulted in skewed distributions. A concomitant problem,
often causing the first, was a ceiling or floor effect whereby an instrument failed to
discriminate at one end of the continuum. The impact of these problems on the
magnitude of correlation coefficients was judged to be small, however, after careful
examination of scatter plots for every pair of variables.
Validity coefficients must also be assessed, of course, in light of the reliabilities of
the measures involved. The reliability values reported in the diagonal of Table 2
were computed using various estimates of internal consistency. In the case of sentence
completion tests, the correlations between judges' ratings were used. A second
multitrait-multimethod matrix was constructed after the validity coefficients had
been corrected for attenuation. The pattern of correlations did not differ from those
already reported.

Corroborating Evidence of Construct Validity

The validity of the construct self-acceptance depends on whether it exists and is
distinguishable from other constructs and on whether it functions in meaningful
relation to other personality concepts. The subject of this study is self-acceptance;
other constructs were introduced to satisfy the multitrait requirements and to gather
data that would allow the observation of the construct in relation to other relevant
variables (Cronbach & Meehl, 1955). The scientific worth of a concept can be
documented if the theoretical relationship between the construct and other person-
ality variables is observable between measurement approximation of the construct
and quantifiable forms of the variables.

156

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

The correlation of participants' age with each of the measures in the study offers
further evidence for a slight distinction between self-acceptance and self-description.
Measures of self-acceptance, on the average, correlated .22 with age. The correlation
between age and self-description was .12. The self-assessment instrument with the
largest relationship to age was also much more closely correlated with self-acceptance
measures than were other self-description tests. Self-acceptance is expected in theory
to increase with age, just as maturity is expected to increase with age. The mature
individual, whose actual traits may not have improved, will be better able to accept
his strengths and weaknesses. The age-self-acceptance correlations do not, however,
necessarily distinguish self-acceptance from other positive personality constructs
since acceptance of others correlated with age to the same degree (.24).
The FIRO-B(3) data presented in Table 1 also support the construct validity of
self-acceptance. Scale 3 of the FIRO-B is a measure of the individual's Wanted
Control, i.e., how much he wants others to determine his actions for him. The self-
accepting person would think well enough of himself to wish to govern his own
behavior. The postulated relationship between self-acceptance and Wanted Control
was verified by a consistent negative correlation between the two variables. Self-
acceptance measures correlated —.34 on the average with FIRO-BQ). For measures
of self-description with FIRO-B(3), the average correlation was —.29. The unique
properties of the self-acceptance construct are corroborated by the negligible corre-
lation between acceptance of others and the FIRO-B(3) scale (—.09).
The final construct that can be used to reflect on the operation of self-acceptance,
as theorized, is acceptance of others. From the review of the literature of both
empirical studies and less rigorous discussions of the constructs, one expects that
there will be a positive relationship between self-acceptance and acceptance of others.
The heterotrait correlation of .22 confirms this expectation. Another small piece of
evidence for the slight difference between self-acceptance and self-description is that
the average correlation of self-description with acceptance of others is only .13.

CONCLUSIONS
The study produced modest evidence that self-acceptance has construct validity.
Despite method-specific operationalizations of the construct, there was substantial
agreement (average correlation of .55) among the several measures of self-acceptance.
Self-acceptance had significant discriminant validity from acceptance of others
regardless of whether the same or different methods were used. The small correlation
between acceptance of self and acceptance of others is consistent with the theoretical
relationship expected between the constructs. The correlation between measures of
self-acceptance and age gives further evidence that the construct functions as postu-
lated.
The positive correlation between self-acceptance and self-description, .41, parallels
the theoretical conceptualization of the constructs. Theoretically, the two constructs
are expected to be highly related. However, given a convergent validity coefficient
for the self-acceptance measures of only .55, the magnitude of the correlation—
regardless of how much it was expected—is too large to allow confidence in
discriminant validity between the constructs. The evidence for discriminant validity
is tenuous. Unless more sensitive measurement techniques eventually show otherwise,
studies should not proceed on the basis that self-acceptance can be measured

157

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

distinctly from self-description as the semantic distinctions between the constructs

suggest. The results yield no explanation as to why self-acceptance and self-descrip-
tion should be so highly related.
It might be that the social norms for judging self-worth or self-satisfaction are so
uniformly imparted to each member of society that self-acceptance is predictable
from self-description. For example, high intelligence, athletic prowess, and physical
attractiveness are so uniformly valued that positive self-descriptions in these respects
are invariably linked with self-satisfaction and positive emotions. A broader cultural
sampling would be needed to encompass those persons who apply a different set of
norms to their self-descriptions and can truthfully report satisfaction with their low
intelligence or dissatisfaction with their physical beauty. Some evidence for the
uniform norms conjecture exists in the correlations among all of the subjects' ideal-
self Q-sorts which averaged .96.
Apart from the psychometric conclusion that self-acceptance and self-description
cannot be clearly distinquished, the tangling of these two constructs may have
implications for classroom practices aimed at effective goals. If values and affect
follow nearly automatically from most self-descriptions, it makes little sense to try to
alter self-esteem directly without changing the "facts" upon which that self-image
may be based. Specifically, this would mean that a child's self-evaluation about his
or her ability to learn would more likely increase because of a noticeable improvement
in learning rather than from receiving "unearned" praise or a pep talk about his or
her capabilities.

REFERENCES
ADCOCK, C. J. Review of the Thematic Apperception Test. In O. K. Buros (Ed.), The sixth
mental measurements yearbook. Highland Park, New Jersey: Gryphon Press, 1965.
BERGER, E. M. The relation between expressed acceptance of self and expressed acceptance of
others. Journal of Abnormal and Social Psychology, 1952, 47, 778-782.
BLOOM, B. S. Affective outcomes of school learning. Phi Delta Kappan, 1977, 59, 193-198.
BUTLER, J. M., & HAIGH, G. V. Changes in the relation between self-concepts and ideal-
concepts. In C. R. Rogers & R. F. Dymond (Eds.), Psychotherapy and personality changes.
Chicago: University of Chicago Press, 1954.
CAMPBELL, D. T., & FISKE, D. W. Convergent and discriminate validation by the multitrait-
multimethod matrix. Psychological Bulletin, 1959, 56, 81-105.
CRONBACH, L. J., & MEEHL, P. E. Construct validity in psychological tests. Psychological
Bulletin, 1955,52,281-302.
EDWARDS, A. L. Manual: Edwards Personal Preference Schedule. N. Y.: Psychological Corpo-
ration, 1954.
ENGLISH, H. B., & ENGLISH, A. C. A comprehensive dictionary of psychological and psychoan-
alytical terms. New York: Longmans, Green and Co., 1958.
FISKE, D. W. Measuring the concepts of personality. Chicago: Aldine Publishing Co., 1971.
FITTS, W. H. Manual for the Tennessee Self-Concept Scale. Nashville: Counselor Recording and
Tests, 1965.
GOLD, M. Power in the classroom. Sociometry, 1958, 25, 50-60.
GRAHAM, J., & HESS, R. Handbook for Project Head Start. Chicago: University of Chicago
Press, 1965.
HARRINGTON, S. A. The effects of basic encounter group experience with prospective counselors.
Unpublished doctoral thesis, University of Colorado, 1971.

158

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

 SELF-ACCEPTANCE

HOUSE, E. R., GLASS, G. V., MCLEAN, L. D., & WALKER, D. F. N O simple answer: Critique of
the Follow Through evaluation. Harvard Educational Review, 1978, 48, 128-160.
Huron Institute. The quality of the Head Start planned variation data. Cambridge, Massachusetts:
Huron Institute, August 30, 1973.
JERSILD, A. T. Child psychology. Englewood Cliffs, New Jersey: Prentice-Hall, 1960.
KLEINMUNTZ, B. Personality measurement. Homewood, Illinois: The Dorsey Press, 1967.
LIPPITT, R. Unplanned maintenance and planned change in the group work process. Social
Work Practice. New York: Columbia University Press, 1962.
MURRAY, H. A. Thematic Apperception Test manual. Cambridge, Massachusetts: Harvard
University Press, 1943.
OSGOOD, C. E., SUCI, G. J., & TANNENBAUM, P. H. The measurement of meaning. Urbana,
Illinois: University of Illinois Press, 1957.
PURKEY, W. W. Self-concept and school achievement. Englewood Cliffs, New Jersey: Prentice-
Hall, 1970.
REMER, R. A performance criterion for counselor selection (Doctoral dissertation, University of
Colorado, 1972). Dissertation Abstracts, 1973, 33, 4099-A. (University Microfilms No. 73-
1819,264)
ROGERS, C. R. A coordinated research in psychotherapy: A non-objective introduction. Journal
of Consulting Psychology, 1949, 13, 149-153.
ROGERS, C. R. Client-centered therapy. Boston: Houghton Mifflin, 1951.
ROTTER, J. B., & RAFFERTY, J. E. Manual: Rotter Incomplete Sentences Blank. New York:
Psychological Corporation, 1950.
RUBIN, R. A., DORLE, J., & SANDIDGE, S. Self-esteem and school performance. Psychology in
the Schools, 1977, 14, 503-506.
SCHMUCK, R. A., <& SCHMUCK, P. A. Group processes in the classroom. Dubuque, Iowa: William
C. Brown Company Publishers, 1971.
SCHUTZ, W. C. The FIRO scales. Palo Alto, California: Consulting Psychologists Press, 1967.
SCOTT, W. A. Conceptions of normality. In E. F. Borgatta & W. W. Lambert (Eds.), Handbook
ofpersonality theory and research. New York: Rand McNally, 1968.
SHAVELSON, R. J., HUBNER, J. J., & STANTON, G. C. Self-concept: Validation of construct
interpretations. Review of Educational Research, 1976, 46, 407-441.
SHEERER, E. T. An analysis of the relationship between acceptance of and respect for self and
acceptance of and respect for others in ten counseling cases. Journal of Consulting Psychology,
1949, 13, 169-175.
SHEPARD, L. A pilot study of three measures of self-acceptance and one measure of ego strength.
Boulder, Colorado: Laboratory of Educational Research, University of Colorado, 1972.
SHEPARD, L. A multitrait-multimethod approach to the construct validity of self-acceptance
(Doctoral dissertation, University of Colorado, 1972). Dissertation Abstracts, 1973, 33, 4189-
A. (University Microfilms No. 73-1828, 232)
SHOSTROM, E. L. Personality Orientation Inventory. San Diego, California: Educational and
Industrial Testing Service, 1966.
SMILEY, M. B. Objectives of educational programs for the educationally retarded and disadvan-
taged. 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.
STEBBINS, L. B., ST. PIERRE, R. G., PROPER, E. C , ANDERSON, R. B., & CERVA, T. R. Education
as experimentation: A planned variation model, Volume IV-A, An evaluation of Follow Through.
Cambridge, Massachusetts: Abt Associates, Inc., 1977. (Also issued by the U. S. Office of
Education as National evaluation: Patterns of effects, Volume II-A of the Follow Through
Planned Variation Experiment series.)
STOCK, D. An investigation into the interrelations between the self-concept and feelings directed
toward other persons. Journal of Consulting Psychology, 1949, 13, 176-180.

159

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014

SHEPARD

WISLER, C. E., BURNS, G. P., & IWAMOTO, D. Follow Through redux: A response to the critique
by House, Glass, McLean, and Walker. Harvard Educational Review, 1978, 48, 171-185.
WYLIE, R. C. The self-concept (Vol. 1). Lincoln, Nebraska: University of Nebraska Press, 1974.
ZIRKEL, P. A. Self-concept and the "disadvantage" of ethnic group membership and mixture.
Review of Educational Research, 1971, 41, 211-225.

LORRIE SHEPARD. Address: Laboratory of Education Research, University of

Colorado, Boulder, CO 80309

160

Downloaded from http://aerj.aera.net at UNIVERSITY OF WINDSOR on July 2, 2014