Ballard 1992

Psychological Reports, 1992, 71, 1155-1160.
O Psychological Reports 1992
SHORT FORMS O F T H E MARLOWE-CROWNE

SOCIAL DESIRABILITY SCALE '
REBECCA BALLARD
L n d e r University
Summary.-Three short forms of the Marlowe-Crowne Social Desirability Scale
were constructed from the results of principal components analysis (N= 399). Those
subscales were compared with short forms developed by previous researchers who used
the same methodology. Examination of the subscales indicated that 13 of the scale's
33 items were isolated by at least two of the three reported studies. Those items were
used to construct a composite subscale, which appeared to offer a useful alternative to
the full scale. Further analysis of the subscale's contents, however, raised questions
about the dimensionality of the Marlowe-Crowne scale. Caution was urged in the use
and interpretation of both the full inventory and the short form until the meaning of
scale scores can be clarified.
For decades, researchers have been intrigued by the possibility that

self-report measures might be contaminated by social desirability response
bias. The Marlowe-Crowne Social Desirability Scale, developed in 1960
(Crowne & Marlowe, 1960), initially seemed to be a promising measure of
that bias; its test-retest and internal consistency reliabilities were relatively
high, and its correlation with the Edwards Social Desirability Scale (Edwards, 1957) was low, indicating, according to the authors, that it was devoid of psychopathological content. Subsequent examination of other aspects
of the scale, however,-led some researchers to question its usefulness. Goldfried (1964), for example, suggested that different items on the scale may
discriminate between high and low scorers among males better than among
females. Strahan and Gerbasi (1972) found no sex differences but suggested
that, in general, some of the items on the scale might not be useful. This
concern was echoed by Ballard, Crino, and Rubenfeld (1988), who found
that only 16 of the items consistently key in the original socially desirable
direction. They felt that some of these items' discriminatory ability might be
solutidn to the perceived problems with this scale, these
weak. One
authors concluded, would be to construct subscales. Retaining only those
items which appear to measure social-desirabllty response bias effectively
could restore some meaning to scale scores.
Previous researchers have isolated similar subsets of items by constructing Marlowe-Crowne short forms through principal components analysis.
This study also performs a principal components analysis to estimate whether
'Address correspondence to R. Ballard, School of Business Administration, Lander University,

Greenwood, SC 29649.
1156
R. BALLARD
previous research results can be replicated. Only if similar results are

obtained across studies could a subscale be used with confidence.
Previous Research on Subscales

Strahan and Gerbasi (1972) performed principal components analysis on
Marlowe-Crowne data from 361 (176 men, 185 women) students in introductory psychology. The items were interspersed with 40 other items of interest
to the researchers. Their analysis yielded a dominant component, which accounted for about 13% of the total variance, and a second component, which
accounted for less than half that of the first. Sex and class were included in
the analysis as dichotomous variables but proved nonsignificant.
The loading on the dominant component was the initial criterion for
inclusion on Strahan and Gerbasi's (1972) short forms; items were then
added according to the criterion that each form contain equal numbers of
positively keyed and negatively keyed items. Two 10-item subscales were
constructed, which were combined to form a third, 20-item subscale (see Table 1 below). Although individual factor loadings were not reported, Strahan
and Gerbasi (1972) acknowledged that the range of absolute loadings on the
first factor was from .11 to .54, with a mean of .35. For their selected items,
the range was from .28 to .54, with a mean of .42. The means and standard deviations of the full scale and subscales are listed in Table 2 below.
Reynolds (1982) also used principal components analysis to create a
short form of the original inventory. His sample of 608 undergraduate students were tested for sex differences by comparing mean scale scores; no
differences were apparent. His principal factor accounted for 15.9% of the
total variance, and a second factor accounted for approximately 5 % of the
variance.
Reynolds (1982) chose .40 as the minimum acceptable factor loading for
inclusion on the short form, based on a suggestion by Lindeman, Merenda,
and Gold (1980). The range of loadings on his dominant factor was from
.15 to .54, with a mean of .38. For the chosen items, the range was from
.40 to .54, with a median loading of .46. O n the basis of item-to-total correlations, to increase internal consistency reliability, he added two items, one
at a time to form two additional subscales; see Table 1. The means and standard deviations of the f
d scale and subscales are listed in Table 2.
METHOD
Subjects
Subjects were 399 students at a middle-sized southern university who
were enrolled in courses on principles of management and personnel management. They were predominantly middle class, 19 (5%) were black, 361 were
white (70%) and 19 ( 5 % ) were other; 263 were men (66'70) and 136 were
women. Participation in t h s research was voluntary and was not related to
receipt of course credit.
1157
MARLOWE-CROWNE: SHORT FORMS
Procedure
The Marlowe-Crowne Social Desirability Scale was administered with
standard instructions. Anonymity was assured, and subjects were told that
they would be (and were) briefed on the purposes and results of the research
at a later date.
RESULTS
The principal components analysis yielded a primary factor which accounted for 12% of the variance in scale scores. A second factor, which accounted for 5.4% of the variance, was isolated. I n all, 13 factors accounted
for more than 3% of the variance in scale scores (i.e., had eigenvalues of 1
or greater).
SHORTFORMSOF
Item Strahan & Gerbasi
XX
X1
PO)*
(10)
1
2
X
3
4
X
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
THE
(1972)
X2
(10)
TABLE 1
MARLOWE-CROWNE
SOCIALDESIRABILITY
SCALE
Reynolds (1982)
C
B
A
(13)
(12)
(11)
X
X
X
X
X
X
X
X
X
X
X
X
Present Study
2
1
(12)
(11)
t
(13)
X
X
X
X
X
X
X
3
(13)
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
(continued on next page)

'Number of items. tcomposite.
1158
R. BALLARD
SHORTFORMSOF
THE
Item Strahan & Gerbasi (1972)

XX
X1
X2
(20)*
(10)
(10)
33
X
X
*Number of items. tComposite.
TABLE 1 (Com'o)
&OWE-CROWNE SOCM DESIRABILITYSCALE
Reynolds (1982)
C
B
A
(13)
(12)
(11)
3
(13)
Present Study
2
1
(12)
(11)
t
(13)
The mean loading on the first factor was .33; the absolute loadings
ranged from .083 to ,508. Consistent with Reynolds1 method, a primary subscale was constructed from the 11 items that had factor loadings of .40 and
above. Two other items, those which had loadings on the first factor of
,399 and .390, were added to form second and third subscales of 12 and
13 items, respectively; see Table 1. Reliabilities for these scales, Reynolds'
(1982) and Strahan and Gerbasi's (1972) subscales, and the full MarloweCrowne scale are reported in Table 3 . The means and standard deviations of
the full scale and subscales are listed in Table 2. The effect of sex on scale
scores was not tested.
TABLE 2
MEANSAND STANDARD
DEVIATIONS
OF SU HE MARLOWE-CROWNE
SOCIAL D E s m n ~ SCALE
n
AND SUBSCALES
Scale
(items)
Strahan & Gerbasi (1972)
SD
Reynolds (1982)
SD
Full Scale (33)"

(20)
X l (10)
X2 (10)
C (13)
B (12)
A (11)
3 (13)
2 (12)
1 (11)
Composite (13)
*Number of items.
xx
'These items also had the highest remaining item-to-total correlations.
Present Research
SD
1159
MARLOWE-CROWNE: SHORT FORMS

TABLE 3
REUABILITY ESTIMATES
FOR MAFCLOWE-CROWNE
SOCIALD E S I R A B L U ~
SCALEAND SHORTFORMS
Scale
Strahan & Gerbasi
(1972)
Full Marlowe-Crowne Scale

(33 items)
Composite Subscale
(13 items)
Strahan & Gerbasi (1972)
X1 (10 items)
X2 (10 items)
XX (20 items)
Reynolds (1982)
A (11 items)
B (12 items)
C (13 items)
Present Study
3 (13 items)
2 (12 items)
1 (11 items)
.73-.83
Reliabilities
Reynolds
(1982)
.82
Present Study
.75
.70
.59-.70
.49-.75
.73-.87
.63
.66
.79
.50
.54
.7 1
.74
.75
.76
.64
.67
.68
.70
.69
.69
DISCUSSION
This research was initiated to answer one question: could a common
subset of Marlowe-Crowne items be identified across all three reported investigations? In fact, nine of the 33 items were identified by the three
studies, and four more were isolated by two of the three. Those items were
combined to form a 13-item composite subscale; see Table 1. As reported in
Table 3, this short form has a reliability of .70-only .05 less than that for
the full scale. In addition, it is nearly as well-balanced as the full scale; eight
of the items contained in this subscale have a socially desirable keying direction of false, while five are true.
Although the composite subscale seems to provide a useful alternative to
the f d inventory, analysis of its content raises questions about the &mensionality of the Marlowe-Crowne scale. First, is avoidance (avoiding disapproval) the dominant component of the full scale? This interpretation would
be consistent with the conclusions of other researchers (Robinson & Shaver,
1773; Milham & Jacobson, 1978) and appears to be supported by the evidence at hand. Of the 13 items on the composite subscale, ten can be classified as measuring avoidance.' Since the subscale was constructed on the basis of primary factor loadings, its content should be representative of the
dominant component of the full scale.
'The classification scheme for the Marlowe-Crowne items is contained in Robinson and Shaver
(1973, p. 729) and is attributed co Jack French.
1160
R. BALLARD
Second, since the scale clearly has a dominant component, is it unidimensional? Evidence from all three studies suggests that it is not. I n none
of the studies does the amount of variance accounted for by the primary factor exceed 16%; so, even in the best case, more than 84% of the variance in
scale scores remains unaccounted for by the first factor.
Third, is the scale bidimensional, as other researchers (Robinson & Shaver, 1973; Millham & Jacobson, 1978) have theorized? Again, the evidence
at hand does not appear to support their contention. If the scale were bidimensional, one would expect the two primary factors to account for the majority of the variance in scale scores. At best (Reynolds, 1982), the first two
factors account for approximately 20% of variance in scale scores. I n this
study, in fact, the 13 factors with eigenvalues of 1 or greater together only
account for 69% of the variance in scale scores; in order to account for
80%, 22 factors would have to be included. While the scale does have a
dominant component, then, it does not appear to be either unidimensional or
bidimensional.
I n conclusion, the reported research on the Marlowe-Crowne Social Desirability Scale raises consistent, critical concerns about the meaning of scale
scores. If the inventory is to be interpreted with confidence or accuracy, researchers must concentrate on clarifying the meaning of those scores. Pending such clarification, caution is urged in the use and interpretation of both
the f d scale and the subscale.
REFERENCES
BALLARD,
R., CRINO,M. D., & RUBENFEU),S. (1988) Social desirability response bias and the
Marlowe-Crowne Social Desirability Scale Psychological Reports, 63, 227-237.
CROWNE,
D. P., & MARLOWE,D. (1960) A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354.
EDWARDS,
A. L. (1957) The social desirability variable in personality assessment and research.
New York: Dryden.
GOLDFRIED,
M. R. (1964) A cross-validation of the Marlowe-Crowne Social Desirability Scale
items. Journal of Social Psychology, 64, 137-145.
LINDEMAN,
R. H., MERENDA,
P. F., & GOLD,R. Z. (1980) Infroduction to bivariate and multivariate analysis. Glenview, IL: Scott, Foresman.
MILLHAM,
J., &JACOBSON,
L. I. (1978) The need for a proval In H London & J . E. Exner, JL
(Eds.), Dimensions of personalily. New York: ~ f e y pp:
. 365-j90.
REYNOLDS,
W. M. (1982) Development of reliable and valid short forms of the MarloweCrowne Social Desirability Scale. lournu1 of Clinical Psychology, 38, 119-125.
ROBINSON,
J. P., & SHAVER,P. R. (1973) Measures of social psychological attitudes. Ann Arbor,
MI: The Institute for Social Research.
STRAHAN,
R., & GERBASI,K. C. (1972) Short, homogeneous versions of the Marlow (sic)Crowne Social Desirabhty Scale. lourno1 of Clinical Psychology, 28, 191-193.
Accepted October 19, 1992.

Ballard 1992

Uploaded by

Copyright:

Available Formats

Ballard 1992

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ballard 1992

Uploaded by

Copyright:

Available Formats

Psychological Reports, 1992, 71, 1155-1160.

O Psychological Reports 1992

SHORT FORMS O F T H E MARLOWE-CROWNE

For decades, researchers have been intrigued by the possibility that

'Address correspondence to R. Ballard, School of Business Administration, Lander University,

previous research results can be replicated. Only if similar results are

Previous Research on Subscales

MARLOWE-CROWNE: SHORT FORMS

(continued on next page)

Item Strahan & Gerbasi (1972)

Strahan & Gerbasi (1972)

Full Scale (33)"

'These items also had the highest remaining item-to-total correlations.

MARLOWE-CROWNE: SHORT FORMS

Full Marlowe-Crowne Scale

You might also like