A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD)...

10
Journal of Consulting Psychology 1965, Vol. 29, No. 1, 27-36 A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE 1 VIRGINIA C. CRANDALL, VAUGHN J. CRANDALL Pels Research Institute jor the Study oj Human Development AND WALTER KATKOVSKY Fordham University A Children's Social Desirability (CSD) questionnaire was constructed and administered to 956 Ss in Grades 3, 4, 5, 6, 8, 10, and 12. A direct question ("yes-no") form of the scale was presented to children in the 3rd, 4th, and Sth grades in individual testing sessions, and a true-false form was given older Ss in group sessions. For both forms of the questionnaire, split-half reliabilities and test-retest (1-month interval) reliabilities were high. Socially desirable responses were more frequently given by younger children than by older children, by dull Ss than by bright Ss, by girls than by boys, and by Negro children than by white children. Social class, size of family, and ordinal position were not associated with CSD responses. Questions were raised pertaining to the generality of this response and pertaining to the premises on which this measure of social desirability, as well as that of Crowne and Marlowe, is based. Recent research has demonstrated clearly that responses to questionnaires and other subjective report measures are often influ- enced by factors other than the manifest con- tent of the stimuli presented to the subject (e.g., Cronbach, 1946, 1950; Edwards, 1957; Fordyce, 1956; Taylor, 1961). Awareness of this phenomenon resulted in the more care- ful construction of self-report instruments in an attempt to control for such "irrelevant" influences as acquiescence and social desira- bility. Social Desirability scales (Crowne & Marlowe, 1960; Edwards, 1957; Hanley, 1956; Wiggins & Rumrill, 1959) have also been created to directly assess the differing tendencies of subjects to give socially desira- ble responses. The use of these scales now J This study was partially supported by Grant M-2238 from the United States Public Health Service. The following persons assisted in various aspects of data gathering and/or data analysis: Barbara Abrams, Suzanne Good, Jean Lindholm, Elinor Waters, Margo Wilson, and Shelley Wing. Grateful appreciation is also expressed to the following mem- bers of the Greene County (Ohio) School System and the Greenview, Tecumseh, and Yellow Springs Schools, whose cooperation greatly facilitated this study: E. Clayton Wisemann, guidance supervisor; Louis King, principal, Antioch School; Paul Coulter, principal, Tecumseh School; Harold Thomas, prin- cipal, Greenview Central; Hugh H. Haines, principal, Greenview South; Howard E. Swaim, principal, and Kenneth McDonough, guidance director, Greenview North; John Malone, principal, Bryan High School. makes it possible to investigate the extent to which the social desirability factor is operative on a specific self-report instrument, and also to test whether the propensity to appear socially desirable influences behavior in other, non-test-taking situations. In the main, research on social desirability has been limited to studies of adult subjects with few reports on the social desirability factor among children. 2 The present study is addressed to the investigation of this response in children. A major portion of current research in personality development using children as subjects relies on self-report techniques for the measurement of attitudes, motivations, expectations, and beliefs. Thus, the investi- gator of child behavior is also plagued with the problem of social desirability response distortion and the resultant reduction in the predictive utility of self-report instruments. Edwards (1957) has suggested two methods of eliminating this unwanted source of vari- ance: (a) One strategy is to word items in personality inventories in a subtle or neutral fashion as to their social desirability so that the subject has difficulty in deciding which answer he should give if he wishes to appear socially acceptable. His response must, then, 2 After this study was done, Lunneborg and Lunneborg (1963) reported the construction of another SD measure for children and the correlation of scores on this measure to certain "lie" and defensiveness scales. 27

Transcript of A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD)...

Page 1: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

Journal of Consulting Psychology1965, Vol. 29, No. 1, 27-36

A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE1

VIRGINIA C. CRANDALL, VAUGHN J. CRANDALL

Pels Research Institute jor the Study oj Human DevelopmentAND WALTER KATKOVSKY

Fordham University

A Children's Social Desirability (CSD) questionnaire was constructed andadministered to 956 Ss in Grades 3, 4, 5, 6, 8, 10, and 12. A direct question("yes-no") form of the scale was presented to children in the 3rd, 4th, andSth grades in individual testing sessions, and a true-false form was given olderSs in group sessions. For both forms of the questionnaire, split-half reliabilitiesand test-retest (1-month interval) reliabilities were high. Socially desirableresponses were more frequently given by younger children than by olderchildren, by dull Ss than by bright Ss, by girls than by boys, and by Negrochildren than by white children. Social class, size of family, and ordinal positionwere not associated with CSD responses. Questions were raised pertaining tothe generality of this response and pertaining to the premises on which thismeasure of social desirability, as well as that of Crowne and Marlowe, is based.

Recent research has demonstrated clearlythat responses to questionnaires and othersubjective report measures are often influ-enced by factors other than the manifest con-tent of the stimuli presented to the subject(e.g., Cronbach, 1946, 1950; Edwards, 1957;Fordyce, 1956; Taylor, 1961). Awarenessof this phenomenon resulted in the more care-ful construction of self-report instruments inan attempt to control for such "irrelevant"influences as acquiescence and social desira-bility. Social Desirability scales (Crowne &Marlowe, 1960; Edwards, 1957; Hanley,1956; Wiggins & Rumrill, 1959) have alsobeen created to directly assess the differingtendencies of subjects to give socially desira-ble responses. The use of these scales now

JThis study was partially supported by GrantM-2238 from the United States Public HealthService.

The following persons assisted in various aspectsof data gathering and/or data analysis: BarbaraAbrams, Suzanne Good, Jean Lindholm, ElinorWaters, Margo Wilson, and Shelley Wing. Gratefulappreciation is also expressed to the following mem-bers of the Greene County (Ohio) School Systemand the Greenview, Tecumseh, and Yellow SpringsSchools, whose cooperation greatly facilitated thisstudy: E. Clayton Wisemann, guidance supervisor;Louis King, principal, Antioch School; Paul Coulter,principal, Tecumseh School; Harold Thomas, prin-cipal, Greenview Central; Hugh H. Haines, principal,Greenview South; Howard E. Swaim, principal, andKenneth McDonough, guidance director, GreenviewNorth; John Malone, principal, Bryan High School.

makes it possible to investigate the extentto which the social desirability factor isoperative on a specific self-report instrument,and also to test whether the propensity toappear socially desirable influences behaviorin other, non-test-taking situations. In themain, research on social desirability has beenlimited to studies of adult subjects with fewreports on the social desirability factor amongchildren.2 The present study is addressed tothe investigation of this response in children.

A major portion of current research inpersonality development using children assubjects relies on self-report techniques forthe measurement of attitudes, motivations,expectations, and beliefs. Thus, the investi-gator of child behavior is also plagued withthe problem of social desirability responsedistortion and the resultant reduction in thepredictive utility of self-report instruments.

Edwards (1957) has suggested two methodsof eliminating this unwanted source of vari-ance: (a) One strategy is to word items inpersonality inventories in a subtle or neutralfashion as to their social desirability so thatthe subject has difficulty in deciding whichanswer he should give if he wishes to appearsocially acceptable. His response must, then,

2 After this study was done, Lunneborg andLunneborg (1963) reported the construction ofanother SD measure for children and the correlationof scores on this measure to certain "lie" anddefensiveness scales.

27

Page 2: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

28 V. CEANDALL, V. J. CEANDALL, AND W. KATKOVSKY

be based on the content of the item. However,Edwards (1957, p. 47) cites data collectedby Fordyce and Rozynko indicating thatMMPI items stated in a "subtle" fashion arestill affected by social desirability factors.(b) The amount of social desirability "pull"may be equated in the alternate choices ofitems in a forced-choice questionnaire, andEdwards used this latter approach in hisPersonal Preference Schedule. Nevertheless, anumber of other researchers have pointed outthat this method, too, does not completelycontrol for social desirability (Corah, Feld-man, Cohen, Gruen, Meadow, & Ringwall,1958; Dicken, 1959; Feldman & Corah,1960; Heilbrun & Goodstein, 1961). In addi-tion, this second, "equating" method limitsthe test constructor to a forced-choice formatwhich may be inappropriate for his specificpurpose.

In the final analysis, after controlling forsocial desirability "pull" as well as possible,it would seem best to measure directly sub-jects' tendencies to give socially desirableresponses. The relationship of their social de-sirability scale scores to their scores on thepersonality instrument in question might thenbe examined. If a significant degree of thevariance on the personality inventory wereshown to be accounted for by social desira-bility tendencies, then "correction" factorsmight be worked out to remove the remainderof this influence on the inventory scores.This is the approach taken by McKinley,Hathaway, and Meehl in using the K scoreas a correction factor for selected MMPIscales (McKinley, Hathaway, & Meehl, 1948;Meehl & Hathaway, 1946). In order to ac-complish this goal in research with children,however, a social desirability scale had tobe developed which was appropriate tochildhood experience.

It may be, however, that the desire to ap-pear socially acceptable is not merely a test-taking response but reflects a more pervasivedeterminant of individual behavior in manysituations. This likelihood has been sug-gested by Heilbrun and Goodstein (1961),Jackson and Messick (1958), Marlowe andCrowne (1961), and Allison and Hunt(1959). If this phenomenon is demonstratedwith children, then scores from a social de-

sirability scale might be related to theirbehavior in various other experimental andnatural situations to determine the generalityof the social desirability response. For boththese reasons the Children's Social Desira-bility (CSD) scale was constructed.

The basic aims of this particular study weretwo: First, it was necessary to determine thestrengths and limitations of the Children'sSocial Desirability scale for the kinds ofsubjects for whom it was intended. This re-quired investigation of group versus indi-vidual administration at various ages, theinternal consistency of responses to the meas-ure, test-retest reliability, etc. A second aimhad substantive considerations, i.e., to deter-mine the correlates of social desirability. Aninitial attempt was made here by examininga variety of common demographic variablessuch as age, sex, IQ, socioeconomic back-ground, ethnic origin, ordinal position in thefamily, and size of family, as these mightbe associated with the number of sociallydesirable responses given.

METHOD

Questionnaire and Administration ProceduresThe Children's Social Desirability (CSD) Question-

naire was patterned after the technique developedby Crowne and Marlowe in their adult scale(Crowne & Marlowe, 1960). The CSD presents achild with 48 tr-ue-false items stated so that thesubject can only answer them in a socially desirablemanner by dissembling. That is, some of the itemsask a child if he always behaves in some particularfashion which is prescribed by middle-class Amer-ican mores or always holds such attitudes and beliefs,e.g., "I am always respectful of older people."Other items ask him if he never does, or thinks ofdoing, those things which are disapproved in hisculture. "I have never felt like saying unkindthings to a person" is representative of this kindof item. Still other items pose an unacceptable be-havior or attitude and ask whether he sometimesacts or thinks in this fashion, e.g., "I sometimes feelangry when I don't get my way." If the subjectwishes to appear socially acceptable on this latterkind of item, he must maintain that he never actsin this unacceptable manner. Examples of otheritems are: "Sometimes I don't like to share mythings with my friends"; "When I make a mistake,I always admit I am wrong"; "I never forget tosay 'please' and 'thank you' "; "I always wash myhands before every meal." The operation to deter-mine the amount of social desirability is the numberof items on which the child maintains that he has

Page 3: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE 29

such an undeviating socially desirable attitude orbehavior.

Twenty CSD items from the Crowne-Marlowescale for adults were rephrased in simpler languagefor children. The remaining items were especiallyconstructed to sample a wide range of social experi-ences common to school-age children. The pos-sibility of acquiescence response sets was minimizedby keying 26 of the 48 items so that the subjectmust respond "true" and 22 so that he must respond"false" to appear socially acceptable.3

The CSD was administered to sixth-, eighth-, tenth-,and twelfth-grade students in group sessions. Thefollowing instructions were printed at the beginningof the questionnaire: "This questionnaire lists anumber of experiences that most children have atone time or another. Read each of these carefully.After you have read one, decide whether it doesor does not fit you. If it does, put a T (for true)in front of the statement; if it doesn't, put an F(for false) in front of the statement. If you haveany questions at any time, raise your hand andone of the persons who passed out these question-naires will come and explain it to you." The childrenwere reassured that their responses would not beshown to, or discussed with, anyone at their schools.

Before the study was undertaken, individual inter-views with a younger pretest sample indicated thatchildren below the sixth grade often have difficultydealing with the true-false form of the CSD ques-tionnaire, probably due to their lack of familiaritywith tests using a true-false format. Some of theseyounger children also did not know the meaningof several words appearing in the items and somewere unable to read well. As a result, a specialform of the CSD was devised for individual oraladministration to children below the sixth-gradelevel. In this questionnaire, true-false items werechanged to direct question form so that they couldbe answered "yes" or "no," and simpler words weresubstituted, e.g., "Have you ever felt like sayingunkind things to a person?"; "Are you alwayspolite to older people?" (see Footnote 3). However,in converting the true-false items to direct questions,it was impossible to retain a near-equal divisionof items keyed "yes" and "no" without awkwardsentence constructions, double negatives, etc. Thus,the direct question form of the CSD contains 13items keyed "yes" and 34 "no." (One item, Number12, appearing on the true-false form was droppedfrom this form because some schools do not givehomework in the earlier grades.) There is, therefore,some possibility that high CSD scores on this form

8 A full list of the items for this form of the CSDscale, as well as the direct question form of thescale to be described later, and a full table ofsplit-half (odd-even) reliability coefficients havebeen deposited with the American DocumentationInstitute. Order Document No. 8232, remitting$1.25 for microfilm or $1.25 for photocopies, andmake checks payable to: Chief, PhotoduplicationService, Library of Congress, Washington, D. C.20540.

may be positively associated with a dissent or"naysaying" response set. Wiggins and Rumrill(1959) and Jackson and Messick (1958) maintainthat they are related in the Edwards SD scale,although Couch and Keniston (1961) and Edwardsand Walker (1962) find acquiescence set to beindependent of social desirability in the Edwardsscale. Whether a dissent response set and socialdesirability are orthogonal or related in children'sresponses on the CSD scale is as yet undetermined.

Pretesting also showed that it was inadvisable toadminister even the direct question form to childrenbelow the third grade. The scale was inappropriatebecause of the short memory span common to youngchildren. For example, a small child could say withhonesty that he "never gets angry" because he canonly remember what he has done during the pastfew days or so. In such cases, the items would failto "pit" his honesty against his desire to appearsocially acceptable over a broad enough time spanto make it necessary for him to dissemble in orderto answer in the socially desirable direction.

The direct question form of the CSD was admin-istered to third-, fourth-, and fifth-grade childrenby using audograph recordings of the questions. Therecord was played individually to each subject inthe presence of an examiner who checked the child'sresponses on a test form. The questionnaire itemswere recorded in order to standardize tone, inflection,and rate of presentation. Individual oral administra-tion was found necessary because some children inthe third, fourth, and fifth grades still do not readwell enough to take the written form of the CSD.The following instructions were given these youngerchildren: "Here are some questions about thingsthat happen to all children your age. All the ques-tions have been put on the record (experimenterpoints). When the person on the record asks youa question, you tell me your answer. If you can'thear a question or if you don't understand anyof them, be sure to tell me and I'll have the recordsay it again." As with the older subjects, the chil-dren were reassured that their responses would notbe disclosed to anyone at their school.

SampleThe sample was composed of 956 students. Sub-

samples in various grades were: Grade 3, N — 115;Grade 4, #=115; Grade 5, #=106; Grade 6,# = 166; Grade 8, #=162; Grade 10, # = 183;and Grade 12, #=109. All children who were intheir respective schools at the time of testing wereused. The sample was drawn from five differentschools in order to make it as representative aspossible. The children who participated as subjectscame from a consolidated country school, a villageschool, a small-city school, a medium-city school,and a college-lab school, all located in southernOhio. None came from a large, metropolitan schoolsystem, however.

The socioeconomic status (SES) of the childrenin Grades 6, 8, 10, and 12 was determined byHollingshead's Two Factor Index of Social Position

Page 4: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

30 V. CRANDALL, V. J. CEANDALL, AND W. KATKOVSKY

(Hollingshead, 1957), an index based on the typeof occupation and amount of education of thehead of the household. Scores on these two factorsare differentially weighted and summed. The norma-tive sample of Hollingshead and Redlich (1958)obtained in New Haven, and those studied by manyother sociologists, indicate that all measures of SESfind lower social class families much more prevalentthan those of higher status. The present sixth-,eighth-, tenth-, and twelfth-grade sample is com-parable to that of Hollingshead and Redlich's norma-tive sample, except that this sample is not as heavilyskewed (as determined by nonsignificant Fisher gs).Thus, the distribution of older children in this studyis relatively "top heavy" with middle and high SESscores as compared with the general American popu-lation. (Complete SES information was availablefor all tenth-grade subjects used in the study, butcould be obtained for only parts of the sixth, eighth,and twelfth grades. However, subjects in these lattergrades all came from the same schools as those inthe tenth grade, and there was no reason to believethat the SES distribution of the remaining sixth,eighth, and twelfth graders would be dissimilar tothat of the tenth grade.)

For Grades 3, 4, and 5, an estimate of the chil-dren's SES was obtained from their fathers' occupa-tions only since information was not availableconcerning fathers' educations. However, father'soccupation has been shown by Hollingshead andRedlich (1958) to be the more predictive of thetwo indexes of SES. Because the fathers' occupationscould not be combined with their education forthis sample, the occupational scores were used inunweighted raw score form. The mean father'soccupation score for this younger sample was 4.3in the range of 1 to 7. The distribution showed asignificant skewness (Fisher gi significant at the.05 level) with scores piling up at the lower end ofthe range. However, again, this distribution is stillnot as skewed as the SES of the general Americanpopulation and indicates that higher status familieswere also somewhat overrepresented in our youngersample.

The California Test of Mental Maturity, theintelligence test used by all schools for Grades 6,8, 10, and 12 yielded a mean of 103.4 and an SD

of 14.15. The intelligence test which all but oneof the schools had used for Grades 3, 4, and 5 wasthe Lorge-Thorndike. The mean Lorge-Thorndikescore for the subjects who had had the test was 103.0with an SD of 12.51.

RESULTS AND DISCUSSIONData Analyses

Distribution characteristics of all variableswere tested with Fisher g statistics, thennonnormal distributions were normalized byMcCall T score transformations. Product-moment correlations were used for measuresof association, and tests of difference weret tests. Two-tailed tests were used to deter-mine all p values.

Characteristics of the CSD Scale .Uncorrected split-half (odd-even) reliabil-

ity coefficients range from .69 to .90 forsubsamples of boys and girls at various gradelevels. Corrected by the Spearman-Brownprophecy formula, correlations are .82 to .95(see Footnote 3).

Test-retest reliability was examined in onlypart of the total sample. The direct questionform of the CSD was readministered aftera 1-month interval to 63 of the youngerchildren. The correlation between first andsecond administration was .90. The true-falseform of the questionnaire was readministeredto 98 tenth graders after a similar 1-monthinterval and produced a correlation of .85.Thus, children's social desirability responsetendencies as measured by the CSD question-naire are demonstrably consistent over amonth interval as well as consistent from itemto item within the test itself.

Although results are presented in the fol-

TABLE lCSD MEANS AND STANDARD DEVIATIONS

Grade

34S68

1012

N

54665393699052

Boys

M

27.6618.7121.7517.6815.9813.0710.67

SD

10.8210.888.848.339.096.406.80

A'

61495373939357

Girls

M

30.6725.7623.0022.1217.5314.5514.47

SD

9.869.689.338.828.637.307.83

Both sexes

M

29.2621.7122.3819.6316.8713.8212.66

SD

10.4310.959.118.838.866.917.60

Page 5: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE 31

lowing discussion for children from thirdthrough twelfth grade, it should be remem-bered that the two forms of the CSD, theones for younger and for older children, varyin both format and method of administration.Thus, trends covering the whole range ofsubjects studied must be accepted only tenta-tively. Table 1 gives the descriptive statisticsfor the CSD scale. Scores ranged from 1 to47 for the 47-item direct question form ofthe questionnaire, and ranged from 0 to44 on the 48-item true-false form. The dataindicate that girls' CSD scores are higherthan boys' at all grade levels and that thereis a general tendency for socially desirableresponses to decrease with age (grade) forboth sexes. The smaller standard deviationsin the upper grades are evidence of increasingwithin-sample homogeneity of response, es-pecially among the male subjects.

Grade (Age) DifferencesIn order to examine more definitively the

tendency noted above for CSD scores to de-crease with age, t tests were run betweenCSD scores of subjects who were two gradesapart, as well as between the lowest andhighest grades of the older sample. (No com-parison was made between any of the lowergrades and any of the upper grades becauseof the difference in test forms and adminis-tration.) Relevant data are reported inTable 2. It will be seen that there was atendency for older children to give signifi-cantly fewer socially desirable responses withonly a few exceptions. The propensity to givesocially desirable responses apparently de-velops early in life since it is shown in thesedata to be well developed by grade three, theearliest grade tested here. It is possible thatthe maximum desire to appear socially ac-ceptable occurs at an even earlier age. Wecan only tell from these data, however,that after Grade 3 this response steadilydecreases.

The wider variability of scores at theyounger ages reported in Table 1 might sug-gest that either the culture is not as clear inits demands for some children as it is forothers and/or that children at these youngerages vary more in their wish to appearsocially acceptable. However, the larger means

TABLE 2TESTS or DIFFERENCE or CSD

SCORES BETWEEN GRADES

Grades compared

Grades 3 versus SBoysGirlsBoth sexes

Grades 6 versus 8BoysGirlsBoth sexes

t

2.956***4.280****5.200****

1.4043.398****2.921***

Direction

3 >3 >3>

6>6>6 >

555

888

Grades 8 versus 10Boys 1.826Girls 2.389**Both sexes 3.121***

Grades 10 versus 12BoysGirlsBoth sexes

2.235*.198

1.609

Grades 6 versus 12Boys 5.263****GirlsBoth sexes

5.179****6.849****

8> 108> 108> 10

10 > 1210 > 1210 > 12

6> 126> 126 > 12

Note.—For N, means, and standard deviations, see Table 1* p < .05.

**p <.02.*** p < .01.

****p <.001.

at the earlier grade levels would seem toindicate that, in general, the culture makesits expectations sufficiently clear by earlygrade-school age for them to be able torespond appropriately and consistently in asocially desirable direction. But why doesthe young child choose to, or need to, appearso socially acceptable? It may be that thisresponse reflects his greater dependency onthe approval of adults during the early yearswhen he feels neither instrumentally ade-quate nor emotionally independent. Since hecannot yet "do for himself" he must maintainthe good will of his caretakers by being, orattempting to appear, a "good boy" or a"nice girl." As the child develops greaterinstrumental competence he is then able toobtain rewards for behaviors other thansimply for appearing "nice" or "good."

In addition, the decrease in the olderchildren's CSD responses probably reflects

Page 6: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

32 V. CRANDALL, V. J. CRANDALL, AND W. KATKOVSKY

TABLE 3DIFFERENCES IN CSD SCORES OF

BOYS AND GlHLS

Grade Direction

345

3,4,5681012

6, 8, 10, 12

115115106336166162183109620

1.543.50***.70

3.55****3.19***1.181.362.71***3.47****

G> BG > BG > BG > BG >BG>BG>BG>BG > B

***p****p

(.heir desire to attain independence fromadults, and perhaps may even be evidence ofthe rebelliousness against social stricturesoften attributed to adolescent conflict withauthority figures. Nevertheless, it may be thatthe adolescent, like the young child, stillneeds to appear socially acceptable. Perhapsat this age, however, he no longer wishes toobtain acceptance from adults, but prefers toobtain it from his own gang, clique, or theadolescent subculture in general. In thisrespect, it should be noted that some itemsof the CSD scale probably represent morenearly the demands of the adult world and,in fact, may be antithetical to what teenagersconsider required social behavior to gainpeer-group acceptance.

Sex DifferencesTable 3 summarizes the results of the

tests of difference in CSD responses for boysand girls at each grade level. Significant sexdifferences are evident in the fourth, sixth,and twelfth grades. Furthermore, when allgirls in Grades 3, 4, and 5 are comparedwith all boys in these grades, the t ishighly significant. This difference also occurswhen CSD responses of girls in the sixth,eighth, tenth, and twelfth grades are com-pared with those of boys in these moreadvanced grades.

One might question whether this sex dif-ference is a reflection of the greater actualconformity of girls as compared with boys.While it is true that girls tend to be moreconforming than boys in actual behavior, it

is apparent from the undeviating behaviorrequired to give each socially desirable re-sponse that few girls could be truly as con-forming in overt behavior and attitudes asthe items demand for such a response. Thus,even though girls may be behaviorally moreconforming, the CSD items are so wordedthat girls' SD responses must still representan attempt to appear socially desirable ratherthan a true reflection of actual behavior. Thesex difference in CSD responses suggests thatit may be more important for American girlsto appear socially acceptable than for Amer-ican boys to do so.

Both Edwards (1957) and Marlowe andCrowne (1962) fail to find significant differ-ences between college men and women ontheir social desirability scales, althoughEdwards' difference approaches significance.It is difficult to determine why this sampleof girls attained significantly higher scoresthan the boys while Edwards and Marloweand Crowne did not find this difference intheir adult samples. These disparate resultsmay be due to the fact that the CSD scaleitems differ from those in the Marlowe-Crowne and the Edwards SD scales, or be-cause the present subjects are younger, orbecause the samples used by these other in-vestigators differed from our sample in an-other important respect. That is, the men andwomen used as subjects by Marlowe andCrowne and by Edwards were all collegestudents. It may be that college women havea lesser desire or need to present a sociallyacceptable facade than do women in general.If so, they would achieve lower SD scoreswhich are more similar to those of men. Inorder to pursue the possibility that girls withan intellectual-academic orientation might beas low on social desirability as boys withsimilar backgrounds, the CSD scores of thegirls and boys from the college laboratoryschool sample of the current study wereanalyzed separately. No significant differencebetween the two sexes was found (t — 1.67,p > .10). Thus, a tentative conclusion mightbe that females in the general Americanculture are more consistently taught to be,and expected to be, "nice" than are males,unless they come from academically orientedbackgrounds where parents of girls may be

Page 7: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE 33

less concerned with this aspect of behaviorand development. An alternative, but notmutually exclusive, possibility is that the girlwho is able to obtain satisfaction from intel-lectual competence, when it is valued, hasless need to get approval by appearing"sweet" and "good."

CSD Responses and IQFor the younger group of third-, fourth-, and

fifth-grade children the correlation betweenCSD responses and IQ was —.19, significantat the .01 level. For the older group of sixth,eighth, tenth, and twelfth graders, the associ-ation was —.20, significant at the .001 level.4While these correlations are moderate, theyindicate that, in general, the less intelligenta child is, the more he attempts to present asocially desirable facade. This negative rela-tionship seems to suggest that even dullerchildren, like younger children, must havebeen able to "pick up the cues" by whichthe culture communicates its expectations forchildren's attitudes and behaviors. The factthat less intelligent children give more sociallydesirable responses may be due to theirinability to get satisfaction from behaviorssuch as intellectual competence and socialleadership which are so highly valued inpresent-day American culture. Perhaps, then,these children fall back on those rewardswhich are still open to them for being respect-ful, polite, obedient, and cooperative, therewards reflected in the CSD scale. Such acompensatory explanation is, at present,highly speculative but could be tested infuture research.

CSD and Social ClassThe correlation of the children's SD scores

and Father's Occupation (the SES measureused for the lower grades) was a nonsignifi-cant — .09. For the older children, using thefull Hollingshead Two Factor Index, the rela-tionship between family socioeconomic status

*In addition to assessing associations of CSDscores and IQs for the total group of younger chil-dren and the total group of older children asreported above, correlations were run separately forboys and girls at each of the grade levels studied.The significant correlations did not form any ob-servable pattern, although all subsample correlationswere negative in direction.

and the subjects' CSD scores was —.16, sig-nificant beyond the .01 level.6 It would appearat first glance that there is some tendencyfor the older children who come from lower-class families to try to present a more sociallydesirable front than do children from higherstatus homes. However, because of the con-sistent relationship often reported betweensocial class and IQ, analyses of variance werecomputed on the children's CSD scores onthe basis of these two dimensions. For boththe older and the younger children, socialclass did not account for a significant amountof the variance, once the influence of intel-ligence was removed. Variance in CSD scoresdue to IQ did produce a significant effectfor both younger and older boys (F — 5.54,p < .05 and F = 8.15, p < .01) and for thesexes combined in the older group (F = 7.00,p< .01). Why IQ was influential only forthe boys and not for the girls is unknown.At any rate, the small but significant cor-relation of —.16 between CSD scores andsocial class is probably spurious and due onlyto the correspondingly higher intelligence ofthe upper-class child as compared with hislower-class peer.

CSD and EthnicityNegro children's CSD scores were compared

with those of white subjects. The test of dif-ference for the younger children yielded a t of2.13 (p < .05) with Negro subjects attainingthe higher scores. For the older children, twas 2.01 (p < .05), and again the Negrochildren gave the greater number of SDresponses. However, only 38 of the youngergroup and only 62 of the older children wereNegroes. These children constituted the verysmall Negro enrollment in those grades of theschools used in the present study, and itshould be remembered that these southwest-ern Ohio schools have a primarily white stu-dent body. It is possible, therefore, that thissample is not typical of most American

6 In addition to assessing associations of CSDscores with SES for the total group of youngerchildren and the total group of older children asreported above, correlations were run separately forboys and girls at each of the grade levels studied.The significant correlations did not form any ob-servable pattern, although 12 of the 14 subsamplecorrelations were negative in direction.

Page 8: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

34 V. CRANDALL, V. J. CRANDALL, AND W. KATKOVSKY

Negroes and that the differences found heremight not obtain between white childrenand more representative samples of Negrochildren.

In regard to the relationship of socialdesirability to IQ, social class, and ethnicgroup, another body of data was made avail-able to the present authors.6 These were theCSD scores of 470 children who were in thesixth, seventh, and eighth grades of threeSyracuse, New York, schools. They are de-scribed as "more than 90% Negro," comingfrom a "culturally deprived" section of thecity, and having parents who "are almostall in the lower class." The mean IQs ofthese children were 8S.O for the boys and88.2 for the girls (California Test of MentalMaturity). The boys of this sample hada mean CSD score of 20.92 with a standarddeviation of 8.05. The girls' mean was 22.48,standard deviation 9.14. When both sexeswere combined, the mean was 21.66, standarddeviation 8.61.

The CSD scores of these children werecompared with those of the sixth- and eighth-grade subjects of the current study who makeup a more representative sample of Americanchildren. The test of difference between boys'scores resulted in a t = 4.92, p < .001. Be-tween girls' scores the / was 3.23, p < .005,and for both sexes it was 5.56, p < .001. Ineach case, the Syracuse sample had signifi-cantly higher scores than the more representa-tive sample of the present study. It is impos-sible, of course, to ascribe their greater tend-ency to give socially desirable responses totheir lower status on one or another of thedeterminants, IQ, social class, or ethnicity,and this will not be attempted. This compari-son is only meant to illustrate that subjectsin the less privileged position on all thesedemographic variables do, in fact, demon-strate greater social desirability of response.

CSD scores were also examined in respectto size of the child's family and whether hewas first born or later born, but no significantresults were found.

In addition to the differences noted betweenthe Syracuse sample and the present one, itis interesting to observe that among the more

6 Grateful appreciation is expressed to David E.Hunt, Syracuse University, for these data.

representative sample of the present studyas well, those children in the less advan-tageous positions on each of the demographicvariables where significant results were ob-tained are those children who attained thehigher CSD scores; i.e., younger children,children with low IQs, Negro children, andgirls. It is tempting to generalize that whena child feels at a disadvantage for any oneor a number of reasons, one readily availableresponse is to attempt to appear "nice." Whilesupport for this hypothesis is not fully de-veloped, it seems to have some demonstrationin certain relationships found with CaliforniaPsychological Inventory variables. High CSDscores (subjects from the tenth-grade sampleused here) were negatively and significantlyrelated to the CPI variables called Sociability,Social Presence, and Self-Acceptance andpositively and significantly related to Self-Control and Good Impression. (The remain-ing CPI scales did not relate significantlyto CSD scores). These results will be morefully discussed in a future report. In themeantime, however, they seem to imply thatthe child who has a strong wish to appearsocially desirable is the child who is shy,withdrawn, unsure of himself in social situa-tions, lacking in self-confidence concerning hisown social skills, has a low sense of his ownpersonal worth, and has "little capacity forindependent thinking and action." He isrestrained and able to inhibit his antisocialimpulses. He tries to create a good impressionof himself and is concerned with how othersreact to him.

In addition to the foregoing individual dif-ferences in the personality correlates of socialdesirability, it may be that this response,whether verbally given or otherwise evi-denced, may also be partially determined byaspects of the situation in which the childfinds himself. To the degree that this is thecase, and our "compensatory" or "disadvan-taged" hypothesis is correct, then the sociallydesirable response should be especially calledforth when the child is interacting with some-one who is a member of some or all of themore favored groups on the demographicdimensions examined in this study; i.e., peoplewho are older, white, more intelligent, male,etc. All of the testing in the present study was

Page 9: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE 35

done by adult examiners introduced to thechildren as psychologists from the Fels Re-search Institute, and it is likely that thechildren expected their questionnaires wouldbe evaluated by similar persons. It would beinteresting in future research to find outwhether lower CSD scores would be obtainedfrom a similar sample of children if they weretested, for example, by an apparently lowerclass, Negro, nonprofessional girl—if such asituation could be arranged to seem reason-ably realistic. In the present study a begin-ning attempt was made along these lines inthe individual testing of the fourth-grade sub-sample. Half of the boys and half of the girlswere tested by a female examiner, and theremaining subjects by a male. The examinersdid not differ from each other on any of theother aforementioned dimensions, however,and the results showed no significant examinerdifferences in either the boys' or the girls'CSD scores.

In future research, it might also be possibleto compare high CSD scorers' behavior in twosocial situations, one in which they were ininteraction with high-prestige persons and theother where they were in situations withindividuals like themselves. Until studies ofthis nature are done, it is impossible to sayhow situation-specific or how generalizedsocial desirability tendencies are.

It should be pointed out that a possiblealternate interpretation of high scores on thepresent scale might be made. It might beargued that children who are of dull normalintelligence or are so young that they are in-capable of making fine semantic distinctions,are not bound by the exact "always" and"never" wording of the CSD items. Thus,they would be freer to endorse more itemsthan are the older, brighter children whorecognize these restrictions. This alternateinterpretation of CSD responses would alsopredict the negative relationships found be-tween intelligence and CSD scores and be-tween age and these scores. However, itwould not account for the sex differencesfound in the present study. The girls of thecurrent sample, who have higher CSD scoresthan the boys, are not only equal to the boysin age and intelligence, but the means oftheir intelligence test scores are slightly

higher than those of the boys (younger girls'mean IQ = 104.4, boys' mean = 101.5; oldergirls' mean IQ = 103.7, boys' mean = 103.3.The t tests between these means do not reachsignificance).

The interpretation that high-scale scoresrepresent an attempt to gain social approvalis supported by the results of other investiga-tions using the same operation for determin-ing social desirability tendencies (the Mar-lowe-Crowne SD scale). High M-C SD scalescores predict greater conformity undergroup pressure (Strickland & Crowne, 1962),the expression of approbation of a boring taskto the examiner (Marlowe & Crowne, 1961),and greater sensitivity to minimal social re-inforcements (Crowne & Strickland, 1961;Marlowe, 1962). These findings are mostplausibly interpreted as the product of adesire to obtain social approval. Furthermore,it will be remembered that CSD scores in thepresent study related (r = .51, p < .001)to the Good Impression scale of the CPI.

A word of caution should be made here.The present scale and the Marlowe-Crownescale are both based on an operation which"pits" the respondent's honesty and integrityagainst his need to present a socially accept-able picture of himself. This method ofmeasuring social desirability rests on twopremises, both of which may be faulty. Onthe one hand, it must be assumed that in-dividuals do not differ in their honesty andthat only their need to appear socially ac-ceptable causes the differences in socialdesirability scores. However, if subjects arenot homogeneous in the strength of theirhonesty, then honesty, as well as social desir-ability, will influence their scale scores.

The second premise must maintain thatsubjects do not differ in their need to appearhonest, and this assumption may also befaulty. Honesty is clearly an approved valuein our culture and a low scale score couldsimply reflect the subject's desire to impressthe examiner with his honesty. Thus, subjects'true honesty, as well as their need to appearhonest, detracts from high social desirabilityscale scores. However, the latter need actuallyreflects one form of the desire to appearsocially acceptable. The measurement paradoxis that this need leads to lower social desir-

Page 10: A CHILDREN'S SOCIAL DESIRABILITY QUESTIONNAIRE...The Children's Social Desirability (CSD) Question-nair e was patterne d after th techniqu develope by Crowne and Marlowe in their adult

36 V. CRANDALL, V. J. CRANDALL, AND W. KATKOVSKY

ability scale scores. These difficulties arguefor the development of other methods ofmeasuring social desirability.

Despite the limitations cited above, theCSD scale has thus far been shown to havepredictive utility. It correlates negatively andsignificantly with: achievement themes inchildren's stories to TAT-like pictures; vari-ous free-play behaviors of grade-school chil-dren, such as the frequency with which theyinstigated verbal and physical aggression,their recognition-approval seeking for theiraccomplishments, their concern with master-ing fine motor skills, the quality of theirlanguage, the time they spent alone involvedin tasks, the amount of independent achieve-ment efforts they displayed, etc. The strengthof these correlations varies greatly, however,depending on the sex of the subjects. Inaddition, boys' standardized achievement-testscores and their grade-point averages are alsonegatively correlated (in the .30s and .40s)with their CSD scores from fifth gradethrough twelfth. Such empirical findings willbe more fully described in future reports,but are mentioned here to indicate that theCSD scale, although not operationally pure,is somehow predicting individual differencesin a variety of behaviors in a systematicfashion.

REFERENCES

ALLISON, J., & HUNT, D. E. Social desirability andthe expression of aggression under varying con-ditions of frustration. Journal of Consulting Psy-chology, 1959, 23, 528-532.

CORAH, N. L., FELDMAN, J., COHEN, I. S., GRUEN,M., MEADOW, A., & RINGWALL, E. A. Social desir-ability as a variable in the Edwards PersonalPreference Schedule. Journal of Consulting Psy-chology, 1958, 22, 70-72.

COUCH, A., & KENISTON, K. Agreeing response setand social desirability. Journal of Abnormal andSocial Psychology, 1961, 62, 175-179.

CRONBACH, L. J. Response sets and test validity.Educational and Psychological Measurement, 1946,6, 475-494.

CRONBACH, L. J. Further evidence on response setsand test design. Educational and PsychologicalMeasurement, 1950, 10, 3-31.

CROWNE, B. P., & MARLOWE, D. A new scale ofsocial desirability independent of psychopathology.Journal of Consulting Psychology, 1960, 24, 349-354.

CROWNE, D. P., & STRICKLAND, BONNIE R. Theconditioning of verbal behavior as a function of

the need for social approval. Journal of Abnormaland Social Psychology, 1961, 63, 395-401.

DICKEN, C. F. Simulated patterns on the EdwardsPersonal Preference Schedule. Journal of AppliedPsychology 1959, 43, 372-378.

EDWARDS, A. L. The social desirability variable inpersonality assessment and research. New York:Dryden Press, 1957.

EDWARDS, A. L., & WALKER, J. N. Relationship be-tween probability of item endorsement and socialdesirability scale value for high and low groupson Edward's SD scale. Journal of Abnormal andSocial Psychology, 1962, 64, 458-460.

FELDMAN, M. J., & CORAH, N. L. Social desirabilityand the forced choice method. Journal of Con-sulting Psychology, 1960, 24, 480-482.

FORDYCE, W. E. Social desirability in the MMPI.Journal of Consulting Psychology, 1956, 20, 171-175.

HANLEY, C. Deriving a measure of test-taking de-fensiveness. Journal of Consulting Psychology,1957, 21, 391-397.

HEILBRUN, A. B., JR., & GOODSTEIN, L. D. Socialdesirability response set: Error or predictor vari-able. Journal of Psychology, 1961, 321-329.

HOLLINGSHEAD, A. The two factor index of socialposition. New Haven: Privately printed, 1957.

HOLLINGSHEAD, A., & REDLICH, F. Social class andmental illness. New York: Wiley, 1958.

JACKSON, D. N., & MESSICK, S. Content and stylein personality assessment. Psychological Bulletin,1958, 55, 243-252.

LUNNEBORG, PATRICIA W., & LUNNEBORG, C. E. Therelationship of social desirability to other test-taking attributes in children. Paper read at Amer-ican Psychological Association, Philadelphia,August 1963.

MCKINLEY, J. C., HATHAWAY, S. R., & MEEHL, P. E.The Minnesota Multiphasic Personality Inventory:VI. The K scale. Journal of Consulting Psychology,1948, 12, 20-31.

MARLOWE, D. Need for social approval and theoperant conditioning of meaningful verbal be-havior. Journal of Consulting Psychology, 1962,26, 79-83.

MARLOWE, D., & CROWNE, D. P. Social desirabilityand response to perceived situational demands.Journal of Consulting Psychology, 1961, 25,109-115.

MEEHL, P. E., & HATHAWAY, S. R. The K factoras a suppressor variable in the MMPI. Journal ofApplied Psychology, 1946, 30, 255-264.

STRICKLAND, BONNIE, & CROWNE, D. P. Conformityunder conditions of simulated group pressure asa function of the need for social approval. Journalof Social Psychology, 1962, 58, 171-181.

TAYLOR, J. B. What do attitude scales measure: Theproblem of social desirability. Journal of Abnormaland Social Psychology, 1961, 62, 386-390.

WIGGINS, J. S., & RUMRILL, C. Social desirability inthe MMPI and Welsh's factor scales A and R.Journal of Consulting Psychology, 1959, 23, 100-106.

(Received January 24, 1964)