dimensions and types of social status: a cross-age perspective

Developmental Psychology1982, Vol. 18, No. 4, 557-570

Copyright 1982 by the American Psychological Association, Inc.OOI2-1649/82/I804-0557$00.75

Dimensions and Types of Social Status:A Cross-Age Perspective

John D. CoieDuke University

Kenneth A. DodgeIndiana University

Heide CoppotelliDuke University

Children's sociometric status was conceptualized in terms of independent di-mensions of social preference and social impact. In Experiment 1, peer perceptualcorrelates of these dimensions were investigated with children in Grades 3, 5,and 8. Social preference was highly positively related to cooperativeness, sup-portiveness, and physical attractiveness and negatively related to disruptivenessand aggression. Social impact was related to active, salient behaviors of bothpositive and negative valence. Whereas the correlates were found to be similarat each grade level, greater proportions of the variance in these dimensions couldbe predicted at the younger ages than at the older ages. In Experiment 2, thesedimensions were used to select children into five sociometric status groups, calledpopular, rejected, neglected, controversial, and average. Peer perceptions of thebehavioral correlates of these groups were solicited and found to reveal distinctprofiles. A previously unidentified group of controversial children was perceivedas disruptive and aggressive (like the rejected group), but also as social leaders(like popular children). It is suggested that researchers consider controversialchildren as a distinct group in future behavioral and epidemiological studies.

Over the last several years, there has beena resurgence of interest in children's socialadjustment, particularly as this adjustmentis defined by the child's peer group (Asher& Hymel, 1981; Hartup, 1975, 1979). Therecent revival of interest in sociometric datahas been triggered by a more general interestin children's social cognition (Shantz, 1975),and by a growing body of evidence that peersare an important source of data for predict-ing a child's future adjustment (Cowen, Ped-erson, Babigian, Izzo, & Trost, 1973; Kohl-berg, Lacrosse, & Ricks, 1972; Roff, Sells,&,Golden, 1972). The 1940s and 1950s wereperiods of extensive study of children's peerrelationships. The sociometric interview pro-vided most of the data for these studies, andin the period between Moreno's (1943) orig-inal work and Gronlund's (1959) summaryof sociometric research, many of the majorquestions of children's social choices wereexplored. Sociometric data are now used to

Requests for reprints should be sent to John D. Coie,Department of Psychology, Duke University, Durham,North Carolina 27706.

select children for preventive interventionprograms (e.g., Oden & Asher, 1977).

In spite of its lengthy history, the litera-ture on sociometric choices has had a num-ber of serious shortcomings. One problem inthis literature has been the inconsistencywith which popularity or acceptance amongpeers is operationally defined/Sometimes itis defined simply by social acceptance (thenumber of peer nominations to the question,"Whom do you like most?") and sometimesthe acceptance score is combined with ameasure of social rejection (defined by manynominations to the question, "Whom do youlike least?"). Since these two dimensions—acceptance and rejection—are only slightlynegatively correlated, the kinds of social sta-tus distinctions that can be drawn from so-ciometric data vary greatly depending onwhether acceptance and rejection scores areused together to define types of status orwhether acceptance alone is used as the in-dex of status. For example, when negativestatus is defined solely in terms of the smallnumber of liking nominations received frompeers, the result is a confounding of two

557

558 J. COIE, K. DODGE, AND H. COPPOTELLI

types of negative status that can be seenwhen the combined use of acceptance andrejection scores is undertaken (Asher &Hymel, 1981; Gronlund & Anderson, 1957;Northway, 1944). Under this latter condi-tion a distinction must be made at the neg-ative status end of the acceptance dimensionbetween children who areliked Wy their peers and children who aresimply not nominated by their peers as likedor dishkedi. 1'he latter children are some-times referred to as isolates and sometimesas neglected children. Recently, Gottman(1977) and Hymel and Asher (1977) havedemonstrated that both positive and negativenominations must be used to avoid confound-ing rejected and neglected1 status groups.

A further confounding occurs at the pos-itive status end of the acceptance dimensionwhen only positive nomination questions areused. Children who have high liking scores

5w~3isliking scores have been calledpopularyby most researchers (e.g., Dun-

mBgtCn"71957). A second group, those chil-dren who receive high liking scores and highdisliking scores, has been ignored by mostresearchers. Roff et al. (1972) suggested thatthese children behave in ways different frompopular children. We prpposgTo calTthishighly nominated gr<know of no systematic resejirch that KaTbeenconducted with this group of children.

The omission of negative sociometric nom-inations from past research was often madeon ethical grounds by investigators who werereluctant to have children make what arethought to be pejorative decisions about theirpeers. Hymel and Asher (1977) suggested,however, that children react to such ques-tions easily and in matter-of-fact ways whenthey are presented properly. We concur.

Peery (1979) recently demonstrated theimportance of using both positive and neg-ative dimensions in predicting preschoolers'social comprehension skills. One questionunanswered by previous research concernsthe ways in which peers perceive these dif-ferent groups of children. The present re-search focuses on this question.

A second major problem with the litera-ture on sociometric choices has been the lackof studies of the developmental changes inthe bases for sociometric choices. Are the

reasons for choosing a peer as liked or dis-liked the same among third graders andamong eighth graders? Over 20 years agoGronlund (1959) observed that variations insociometric questionnaire items across stud-ies made it difficult to compare the reputa-tional correlates of positive and negative peerstatus at different age levels. The two studiesreported here represent a systematic attemptto determine the developmental changes (ifany) in the kinds of behavior peers observeamong children at different ends of the so-ciometric status dimension.

Experiment 1

Overview

Dunnington (1957) and Peery (1979)have suggested that peer nomination scoresthat indicate a child's like or dislike for apeer can be combined to create two new di-

' Peery (1979) has made this same distinction withthe introduction of a social impact dimension but dis-tinguishes between amiable and isolated children at thelow end of this social impact scale. His isolated groupcorresponds to our neglected group, since we add theclassificatory refinement of insisting that the neglectedgroup include only those who receive no positive nom-inations. We have followed Northway (1944) and Gron-lund and Anderson (1957) in using the term neglectedto refer to such children rather than the term isolated.As Asher and Hymel (1981) pointed out, the term iso-lated is often used to refer to children at the low endof a social acceptance scale based only on a positivenomination item. Gottman (1977) noted the dual useof the term isolated to refer to those who are observedto physically withdraw from others and to refer to thoseidentified sociometrically, Thus, we attempt to avoidthese confusions with the term neglected.

2 The distinctions among popular, controversial, re-jected, and isolated children are not merely a functionof using a peer nomination method and two questions("Who is liked most?" and "Who is liked least?").Three of our groups could be derived by using a peerrating method and a positive-to-negative scale. Thosechildren usually receiving positive ratings would becalled popular. Those children usually receiving nega-tive ratings would be called rejected. Those childrenusually receiving ratings near the mean of the scalemight prove to be either average or neglected. Thosechildren receiving a high variance in ratings (severalratings at each extreme) would be called controversial.The nomination method has an advantage over the rat-ing method in distinguishing between average and ne-glected children. The inclusion of a positive nominationitem with a rating scale (cf. Green, Forehand, Beck,& Vosk, 1980) would permit the differentiation of theselatter two groups.

SOCIOMETRIC STATUS 559

mensions of sociometric status. The sum ofa child's liking score plus his or her dislikingscore yields a new score that Peery called"social impact." A child's liking score minushis or her disliking score yields a score thatPeery called "social preference." The goalof the first experiment was to identify thesets of behavioral descriptions most closelyassociated with these two social status di-mensions in the minds of the child peer groupand to determine whether these correlateschange across age levels.

The set of behavior descriptions for whichchildren nominated their classmates wassimilar to items used in the Bower Class Play(Bower, 1960). The 24 items used in thepresent study were selected from items foundelsewhere in the sociometric literature. Itemsthat have been known to predict positive ornegative status at any age level were selectedfor this study. Unlike the Bower Class Play,the items in the present study were writtenin behavioral terms so that children couldmatch a peer to a behavioral description witha minimum of inference. For example, oneitem, called "starts rights," was written asfollows: "This person starts fights. He or shesays mean things to other kids and pushesthem and hits them." The only exceptionsto the use of behavioral descriptions weretwo items that asked a child to nominatepeers who were physically attractive or un-attractive (e.g., "Who are the best lookingboys and best looking girls?"). These itemswere included because of recent findings thatattractiveness is a strong correlate of liking(Kleck, Richardson, & Ronald, 1974; Lan-glois & Stephen, 1977).

In the first study, children at three agelevels (third, fifth, and eighth grades) wereadministered a sociometric nomination in-terview. In the interview, "like most" and"like least" nominations were solicited, sothat social impact and social preferencescores could be derived. Following these twoquestions, subjects were asked to nominatethree peers who best fit each of 24 behavioraldescriptions. Scores were calculated as thesum of all nominations received in a partic-ular category and then were standardizedwithin a grade level, so that cross-grade com-parisons could be made independently of thesize of classes. In a series of multiple regres-

sion analyses, standardized nomination scoreson the 24 behavioral descriptions were usedas predictor variables, and social impact andsocial preference scores were used as crite-rion variables. These analyses were con-ducted separately for each age level so thatcross-grade comparisons of the correlates ofthe sociometric dimensions could be made.

MethodSubjects. All of the children in the third and fifth

grades of a Durham County, North Carolina school,and all of the children in three eighth-grade classroomsof a junior high school in the same district as the ele-mentary school served as subjects in this study, followingthe receipt of written parental permission. Approxi-mately 20% of the sample were black, and 49% weremale. In all, 311 children, that is, 94 third graders (meanage = 8.9 years), 112 fifth graders (mean age =11.0years), and 105 eighth graders (mean age = 14.0 years),served as subjects, both as nominators and as nominees.

Procedure. Subjects were escorted individually orin small groups to a private research trailer located out-side of their school and were told of the strict confi-dentiality of their responses to an interview. In the in-terview, each child was asked to name, from the grade-level roster placed in front of them, three classmateswhom he or she liked most and then three whom he orshe liked least. Following this, each child was asked toname three children who best fit each of 24 standardizedbehavioral descriptions. The labels for each of thesedescriptions are listed in Table I.3 At the end of theinterview, each child was reminded of the confidentialityof the responses and was asked not to discuss the in-terview with peers. According to teacher reports, thisrequest appeared to be honored.

Approximately 12 weeks after the first interview(which had taken place in January of the school year),all children were administered the same interview againto establish the reliability of nomination scores. Thedata from the first interview were analyzed for thisstudy.

Results and Discussion

The total number of nominations receivedby each child on each of the 26 questions(liked most, liked least, and 24 behavioraldescriptions) were calculated and weretransformed into standardized scores withineach grade level. The 12-week test-retestreliabilities of these scores were calculatedby Pearson product-moment correlations,and ranged from .46 to .88, with a mediancorrelation of .65. The reliabilities for like

3 The first author will provide, upon request, actualdescriptions used in the sociometric interview.


Table 1Correlations of Standardized BehavioralDescription Scores to Like Most and LikeLeast Scores

Item label Like most Like least

Like mostCooperates with peersDisrupts the groupSupports peersAffiliates with peersActs shyActs snobbishGets rejected by peersDefends self in argumentsAggresses indirectlyGets teased by peersStarts fightsHas short temperRemains calmFeelings hurt easilyActs independentBragsSeeks help constantlyAchieves muchAchieves littjeAttractive physicallyUnattractive physicallyLeads peersGets into trouble with teacherTried to be teacher's favorite

_.51*

-.07.63*.41*

-.12-.04-.28*

.37*-.14-.18-.02

.01

.43*-.05

.36*-.01-.11

.39*-.19*

.57*-.30*

.51*-.03

.40*

-.21*-.31*

.78*-.24*

.11-.05

.66*

.30*

.03

.74*

.14

.70*

.67*-.28*

.06-.06

.65*

.41*-.22*

.41*-.25*

.52*-.08

.71*-.17

Note, n = 311.

most and like least scores were each .65.These reliabilities are similar to those re-ported by Roff et al. (1972). Also, the reli-abilities were similar across the three gradelevels.

To assess the correlates of the like mostand the like least scores, each of the 24 be-havioral description variables was correlatedto these two scores. These correlations arelisted in Table 1. As can be seen in the table,the major correlates of the like most scoreinclude the descriptions "supports peers,""attractive physically," "cooperates withpeers," and "leads peers." The major cor-relates of the like least score include the de-scriptions "disrupts the group," "aggressesindirectly," "starts fights," "gets into troublewith teacher," and "acts snobbish." Thesecorrelations were repeated within each gradelevel and were found to reveal similar, almostidentical, patterns at each grade level. Theonly differences in patterns were that the

correlation coefficients were lower, gener-ally, at the eighth grade than at the thirdand fifth grades. For example, the correla-tions of "disrupts the group" and "acts snob-bish" with the like least score decreased sig-nificantly as children got older (ps < .01, by

> Z scores). Overall, children seemed to haveclearer ideas about the correlates of rejectionthan of acceptance.

Two new sociometric variables, socialpreference and social impact, were derivedfrom the like most (LM) and like least (LL)scores. As suggested by Peery (1979), socialpreference was calculated by the formulaLM — LL. Social impact was calculated bythe formula LM + LL. The two scores de-rived by these formulas were then used ascriterion variables in several sets of multipleregression analyses, with the behavioral de-scription scores as predictor variables. Toassess the predictions of social preferenceand of social impact separately at each ofthree grade levels, six series of stepwise mul-tiple regression analyses were performed. Inthe stepwise analyses, a predictor variablehad to Increase the variance explained b"y~atleast 3% (calculated as squared multiple cor-jglation change) and had to contribute sig^mficantly to the prediction of "the criterion'variable (at p < .01) to be reported as a sig-nificant predictor variable.

The multiple regression analyses predict-ing social preference for the three grade lev-els are summarized in Table 2. The propor-tions of variance accounted for by thedescriptive items at a .01 level of significancewere high for all grade levels. The majorpredictors of high social preference weresimilar across all three grade levels.4 Theitems "supports peers" and "attractive phys-ically" were highly positive predictors of so-cial preference. The negative correlates of

4 When different predictor variables were entered inthe regression analyses for different grade-level regres-sions on the same criterion variable, tests of significancefor between-grade-level differences in the simple cor-relations between the predictor and the criterion vari-able were calculated. Unless there were significant dif-ferences in the correlations for the grade-level correlationsof predictor and criterion, the inclusion or exclusion ofa particular predictor variable in'the multiple regressionwas not deemed to reflect a grade-level difference in theregressions.


social preference were similar for third andfifth graders but were somewhat different forthe eighth graders. For the latter group, lowsocial preference was less characterized byovert, active misconduct (disruptive) andunattractiveness and more closely linked tonot fitting in with peers (not accepted). Moreof the variance in social preference was pre-dicted by fewer variables at Grade 3 thanat Grades 5 or 8. Given that the reliabilitiesof these variables were similar across gradelevels, it is reasonable to conclude that theolder children were more differentiated andless stereotyped in their perceptions of thecorrelates of social preference.

The fact that perceived physical attrac-tiveness or unattractiveness is related to so-cial preference has been well documentedamong preschoolers (Dion & Berschied,1974) and among elementary-school-ageand early adolescent males (Kleck et al.,1974; Langlois & Stephen, 1977). Althoughthe latter studies suggested that perceivedattractiveness predisposes people to prefermore attractive peers, we were interested inknowing whether the physical attractivenessitems carried predictive power that would

not be carried by items describing behaviorrather than attributes. To this end we reranthe multiple regressions with the two phys-ical attractiveness items deleted from the setof predictors. The proportion of variance in

. social preference accounted for by behav-ioral items was not different from that shownin Table 2 except for the fifth-grade regres-sion in which the squared multiple correla-tion dropped to .62.

The regression analyses predicting socialimpact are described in Table 3. The step-wise regression analyses yielded several pre-dictor variables at each grade level. Bothpositive items (e.g., "supports peers") andnegative items (e.g., "disrupts the group")contributed positively to social impact. In-terestingly, the "picked-on" item was a sig-nificant predictor of social impact for theeighth grade but not for the other grades.As with social preference, more of the vari-ance in social impact appeared to be pre-dicted, by fewer variables at Grade 3 thanat Grades 5 or 8.

The behaviors associated with high socialimpact included a combination of active,positive behaviors ("supports peers" and

Table 2Multiple Regressions of Behavior Descriptions on Social Preference for Third, Fifth,and Eighth Grades

Variable R df

Third grade

SupportiveDisruptiveUnattractive

.65

.81

.84

.42

.66

.71

67.08*61.89*15.65*

1,922,913, 90

.65-.62-.60

.14-.18-.21

Fifth grade

UnattractiveSupportiveIndirectly aggressiveAttractiveDefends self

.61

.71

.77

.80

.82

.36

.50

.59

.64

.68

62.48*30.33*24.68*13.79*12.52*

1, 1102, 1093, 1084, 1075, 106

-.60.56

-.56.56.27

-.23.13

-.14.26.19

Eighth grade

Not acceptedSupportiveDisruptiveAttractiveSnobbish

.51

.65

.71

.75

.77

.26

.43

.51

.56

.59

36.96*28.96*17.13*12.27*7.35*

1, 1032, 1023, 1014, 1005, 99

-.51.48

-.37.43

-.20

-.34.24

-.14.13

-.24

Note. F values are for R1 change estimates. For third grade, n = 94; fifth grade, n = 112; eighth grade, n*p< .01.

105.


Table 3Multiple Regressions of Behavior Descriptions on Social Impact for Third, Fifth,and Eighth Grades

Variable

Teacher troubleLeaderAffiliates

DisruptiveSupportiveAttractiveDefends self

FightsSupportivePicked-onAttractiveDisruptiveShy

R

.61

.74

.78

.54

.63

.66

.69

.45

.62

.67

.72

.75

.76

R2

.38

.55

.61

.29

.40

.43

.47

.20

.38

.45

.52

.56

.58

F

Third grade

55.64**33.84**15.89**

Fifth grade

45.54**19.72**6.41**7.28**

Eighth grade

26.05**29.40**12.29**14.22**10.30**3.69*

df

1, 922,913,90

1, 1102, 1093, 1084, 107

1, 1032, 1023, 1014, 1005, 996, 98

r

.61

.46

.47

.54

.19

.20

.25

.45

.40

.32

.32

.43-.26

ft

.19

.43

.26

.47

.23

.28

.18

.29

.29

.24

.24

.17-.20

Note. F values are for R2 change estimates. For third grade, « = 94; fifth grade, n •* p < .06. ** p< .01.

112; eighth grade, n = 105.

"leads peers") and salient, negative behav-iors ("disrupts the group," "gets into troublewith teacher," and "starts fights"). This wastrue for all age groups. Because of the waysocial impact scores were computed, it wasmathematically possible for popular, re-jected, and controversial types of children"to have high social impact scores. Therefore^it becomes necessary to discern whether thismixed pattern of positive and negative be-haviors is actually perceived to be charac-teristic of that group of children we refer toas controversial (i.e., that subset of high-im-pact children who receive many social ac-ceptance votes and many social rejectionvotes) or whether the results of the regres-sion analyses simply reflect the mixed com-position of the high-impact group. Since thenumber of controversial children was verysmall at any one grade level in this sample,it was not possible to resolve this questionin the present study.

The preceding analyses have been focusedon the dimensions of social status—prefer-ence and impact—that can be derived frompositive and negative sociometric choicemeasures. Instead of dimensions of status,

one could think in terms of the types of statusthat particular children may hold withintheir peer groups. This latter orientation ismore compatible with the kind of concernthat has given rise to the recent resurgenceof interest in sociometrics—the identifica-tion of specific children who are at risk forfuture problems and can be identified as suchon the basis of the social status they havein their peer group. The question raisedabove regarding the mixed correlates of highsocial impact is also more effectively resolvedby looking at the ways in which peers per-ceive groups of children who are identifiedby sociometric choice data as either popular,rejected, or controversial. This focus ontypes of social status marked the purpose ofa second study. (See the General Discussionsection for further discussion of the resultsof Experiment 1.)

Experiment 2

The design of the second experimentcalled for the identification of children whoqualified as fitting extreme types of socialstatus Within their peer group. These types


of social status can be thought of in termsof sectors within the two-dimensional gridformed by the social preference and socialimpact variables. Figure 1 illustrates the re-lationship between the dimensions of socialpreference and social impact and the positiveand negative sociometric measures fromwhich the former are derived. It also illus-trates the relationship between the dimen-sions of social preference and social impactand the five types of social status groups thatwere compared in the present study. An av-erage status group was identified to providea reference group with whom the more ex-treme groups could be compared.

Once these status types were identified,the different ways in which peers tended to

view these types of children could be assessedand the possibility examined that peer per-ceptions of the children who fit these statustypes change with the age of the peer group.The instrument for obtaining behavioral de-scriptions from peers was altered for Exper-iment 2. The 24-item pool was reduced to6 items on the basis of a hierarchical clusteranalysis (Johnson, 1967) of the data fromExperiment 1. These 6 behavioral descrip-tion items (or, more precisely, the numberof times a child was nominated for each itemby his or her grade level school peers) con-stituted the dependent variables for Exper-iment 2. Since the basic question to be ad-dressed in this study concerns potential dif-ferences (as perceived by peers) on behavior

Social Preference

Liked• Most

NEGLECTED AVERAGEM/

CONTROVERSIALI Social

Impact

LikedLeast

Figure 1. The relationships between positive and negative social choice measures, the dimensions ofsocial preference and social impact, and five types of social status. (sz = standard deviation for stan-dardized scores.)


patterns among the five social status groups,the design for the data analysis called for a5 (status group) X 2 (gender) X 2 (race) X 3(grade level) multivariate analysis of vari-ance, using a weighted least squares methodbecause of unequal cell sizes (Speed, Hock-ing, & Hackney, 1978). Planned contrastsbetween controversial and popular groupsand between controversial and rejectedgroups were performed.

Method

Procedure. Over the 2 years following the data col-lection for Experiment 1, children entering the third,fifth, and eighth grades (N = 537) of the same schoolssampled in Experiment 1 were administered a confi-dential sociometric interview in whieh they were askedto nominate three peers as liked most and three as likedleast. Following these questions, the children were askedto nominate three peers who best fit each of the followingsix behavioral descriptions:

1. Cooperates. "Here is someone who is really goodto have as part of your group, because this person isagreeable and cooperates—pitches in, shares, and giveseveryone a turn."

2. Disrupts. "This person has a way of upsetting ev-erything when he or she gets into a group—doesn't shareand tries to get everyone to do things their way."

3. Shy. "This person acts very shy with other kids,seems always to play or work by themselves. It's hardto get to know this person."

4. Fights. "This person starts fights. They say meanthings to other kids or push them, or hit them."

5. Seeks help. "This person is always looking for help,asks for help even before they've tried very hard."

6. Leader. "This person gets chosen by the others asthe leader. Other people like to have this person incharge."At the end of the interview, children were reminded ofthe confidentiality of the interview data and were askednot to discuss the interview with peers.

The number of times a child was named by same-grade peers was calculated lor eacn child. These scoreswere "standardized by grade level, and these standardscores for each of the six behavior descriptions wereused as the dependent variables in the multivariate andunivariate analyses.

Status group selection. The data for the 311 subjectsinterviewed in Experiment 1 were combined with thosedata obtained from the 537 children interviewed forExperiment 2. Of this total of 848 children, 52% weremale and 32% were black. The sample had a total of213 third graders, 300 fifth graders, and 335 eighthgraders.

The raw nominations for the liked most and likedleast categories were tallied, standardized, and trans-formed into social preference and social impact scores,as was the case in Experiment 1. Social preference andsocial impact scores were standardized with grade levelso that equivalent selection procedures were employedacross grade level. These scores were then used to iden-

tify children for the five distinct social status groupsaccording to the following criterion: (a) Thejaopulargroup consisted of all of those children who received asocial preference score of greater than 1.0, a like moststandardized score of greater than 0, and a like leaststandardized score of less than 0. (b) Thg rejected groupconsisted of all of those children who received a socialpreference score of less than -1.0, a like least stan-dardized score of greater than 0, and a like most stan-dardized score of less than 0. (c) The neglected groupconsisted of all of those children who received a socialimpact score of less than —1.0 and an absolute like mostscore of 0. The neglected children, therefore, had no oneidentifying them as among the three people they likedmost. They differed from the rejected children in that Ithe rejected children received many nominations as liked nleast, whereas the neglected children did not. (d) Thecontroversial group consisted of those children who re-ceived a "social impact score of greater than 1.0 and whoreceived like most and like least standardized scores thatwere each greater than 0. Thus, members of this con-troversial group were all above their class mean for bothpositive and negative sociometric nominations, (e) Theaverage group consisted of all of those children whoreceived a social preference score that was greater than—.5 and less than .5. Selection to a status group wasmade independently ofVchTia's sender and race! *

By following the preceding selection procedures, 486children were selected from the original pool of 848children as fitting one of the five social status types.Since the criteria for group identification were followedstrictly, the numbers selected to each status group var-ied. There were 104 in the popular group, 111 in therejected group, 112 in the neglected group, 62 in thecontroversial group, and 77 in the average group.

Results

Gender effects. The number of boys andgirls who were selected to each of the fivesociometric status groups were tallied. Fivechi-square tests were performed to examinewhether boys and girls were differentiallyselected to the five status groups. One chi-square test was significant: boys (15.4% oftotal) were more likely to be selected to therejected group than were girls (10.6%),X2(l) = 4.39, p < .05. This finding held ateach grade level, and it joins a large bodyof data demonstrating that boys experiencemore overt difficulties in peer relations thando girls (e.g., Rutter, 1976).

A significant multivariate main effect ofgender was found for the pool of six behav-ioral descriptions, F(6, 422)= 15.81, p<.001, as were several significant univariateeffects. Specifically, females received jmore,nominations than males for the items "co-^operates," F(\, 427) = 16.76,p< .001, and


;' F(l, 427) = 15.82,p * .001, whereasmales received more nominations than fe-males for the items ̂ 'fights" F(l, 427) =57.75, p < .001, and '̂ seeks help." F(l,

* 427) = 4.59, p < .04. No significant genderdifferences were found for the "disrupts"and "leader" descriptions. No significant in-teractions for gender and grade level wereobtained.

Race effects. The racial identities of chil-dren in the five status groups were compiled,and five chi-square tests were performed totest the significance of the differences in ra-cial composition for these groups. Propor-tionately fewer blacks (12.5%) were selectedas popular than were whites (25.4%),y (1) =10.41, p < .01; proportionately more blacks(18.4%) were selected as controversial thanwere whites (10.2%), X

2(D = 6.31, p < .05.Although these effects are labeled here asrace effects because race is the demographicfactor involved in the analysis, these effectsare more properly viewed as reflecting theimpact of majority versus minority status inthe population being surveyed. In this caseblack children were in the minority, and theresults should be interpreted as telling ussomething about minority standing and peerstatus rather than about race per se.

There was a significant multivariate maineffect for race, F(6, 422) = 7.65, p < .001,along with four significant univariate effectsfor race. Blacks received more nominationsfor the "disrupts" item, F(l, 427) = 26.53,/><.001; for "fights," F(l, 427) = 27.83,p < .001; and for "seeks help," F(l, 427) =6.54, p < .01. Black children received fewernominations for the "cooperates" item, F(l,427) = 15.18, p < .001. There were no sig-nificant race effects for the "shy" and"leader" items. The fact that fewer blacksqualified for the popular group serves to ac-count for much of these racial differences.That fewer blacks were popular is likely aconsequence of the relative numbers ofblacks and whites in this school populationand the tendency, as Singleton and Asher(1979) noted, for there not to be much cross-race "best-friend" nominating on sociomet-ric surveys of the sort used here.

Status group effects. A significant mul-tivariate main effect for social status groupwas found, F(24, 1682) = 25.78, p < .001.

All six univariate effects for status were alsosignificant. The status group means for thesix behavioral descriptions are displayed inTable 4, along with the F values for the maineffects. Since we were interested primarilyin the contrasts between the popular andcontroversial groups, and between the con-troversial and rejected groups, planned con-trasts of these differences were also ana-lyzed. The F values for these analyses arealso found in Table 4.

The most meaningful way to describe thedifferences among status groups is to con-struct profiles of each group. The popularchildren were viewed by peers in obviouslylprosocial terms. They received high scores)for the "cooperates" and "leader" descrip-\tions and low scores for the "disrupts,"/"fights," and "seeks help" descriptions. The*.rejected children displayed an opposite pro-file. They received low scores for the two,prosocial items ("cooperates" and "leader"),and high scores for the three negatively va-lenced descriptions ("disrupts the group,""fights," and "seeks help"). These two groupsdid not differ from each other in scores forthe "shy" description.

The ̂ controversia^children are best seenin contrast to the popular and rejected chil-dren and displayed a profile that combinedfeatures of these two groups. The contro-versial children were similar to the rejectedchildren in being perceived as djsruptivg andstarting fights. They were also frequentlynominated for "seeks help." This latter itemwas highly correlated with an item in theoriginal set describing children who wereslow to learn. On the other hand, they wereperceived, as were the popular children, asbejng leaders in the peer group. They werenot perceived as being cooperative in the waythat popular children were viewed by peers,but neither were they viewed as lacking se-riously in this behavior, as were the rejectedchildren. Rather, they were perceived as sim-ilar to average children in this regard. Fi-nally, the fact that the controversial childrenwere the only group significantly below themean on the shyness item supports the viewof them as visible, active, and assertive chil-dren. Sometimes this activity takes the formof leadership and sometimes it puts them indemand as leaders. This group we refer to

566 J. COIE, K. DODGE, AND H, COPPOTELLI

Table 4Mean Standardized Scores for Each Behavioral Description by Status Group

Status group

Behavioral description

CooperatesDisrupts

groupActsshy Fights

Seekshelp

Leadspeers

Group means

PopularControversialRejectedNeglectedAverage

1.23.14

-.64-.32

.05

-.321.18-1.02

-.46-.13

-.02-.33-.07

.17

.12

-.291.12 -.94

-.37-.17

-.25.43.77

-.37-.08

.87

.91-.29-.43-.06

F ratios

Main effect"Planned contrasts5

Popular vs.controversial

Controversial vs.rejected

67.01**

41.53**

25.38**

71.05**

77.11**

2.97*

17.8**

55.06**

60.33**

27.01**

14.50**

37.96**

57.03**

"F(4, 427). bf(l , 456).*p<.Q5. **p<.001.

as controversial children has been identifiedpreviously by Roff et al. (1972), althoughtheir analyses were not systematic. Theiranecdotal follow-up data indicated thatamong populations of lower socioeconomicstatus, these children are likely to becomejuvenile delinquents. They suggested thattheir leadership qualities establish them asdominant figures in delinquent peer groups.The pattern of results found in this study isnot incongruent with the picture provided byRoff et al.

Status group interaction effects. Threesignificant multivariate interaction effectswere found, all involving the status groupvariable. First, a significant multivariate in-teraction between status and gender wasfound, F(24, 1682) = 3.09, p < .001. Threesignificant univariate interactions involvingthe descriptive items were also found. Theseincluded "cooperates," F(4,427) = 2.60,p <.04; "fights," F(4, 427) = 11.65, p < .001;and "seeks help," F(4, 427) = 2.49, p < .05.None of these effects altered the main effectsof status and gender described earlier; thatis, the differences among status groups wereconstant for each gender, and the differencesbetween genders were constant for each sta-tus group. The interactions only reflected

differences in the magnitude of effects. Forthe "cooperates" item, female scores had agreater range (i.e., the difference betweenthe highest and lowest group mean) amongthe status group means (2.16) than did themale scores (1.53). Thus there was greaterdifferentiation among female status groupsfor this item than for male subgroups. Theopposite pattern held true for two items onwhich males had higher scores—"fights"and "seeks help." The ranges in male scoreswere 2.37 and 1.36, respectively for theselast two items, whereas for the female groupsthe ranges were .74 and .81.

A significant multivariate interaction ef-fect was obtained between status and race,F(24, 1682) = 2.48, p < .001. There werethree significant univariate interactions forstatus and race. One was for "disrupts," F(4,427) = 6.25, p < .001, and another was for"fights," F(4, 427) = 6.27, p<.001. Forboth of these effects the pattern of meansdid not alter the main effects for race orstatus group, but both effects were due todifferences in range. Because the black con-troversial group had more extremely highscores for these two items than the whitecontroversials, the overall range for theseitems was greater for the black status group.


r

The third univariate effect was for the"leader" item, F(4, 427) = 4.74, p < .001.The pattern of means reflects the fact thatpopular white children were viewed as lead-ers (M = 1.05), but popular black childrenwere not (M = .08); they were viewed asaverage in this regard. Only controversialblack children were seen as leaders (M =.85), on a par with controversial (M = .97)and popular white children.

The third significant multivariate inter-action effect was that of Status X GradeLevel, F(48, 2223) = 1.69, p < .003. (Sincethe behavioral items were standardizedwithin grade level, obviously no main effectsfor grade level could result.) Although noneof the univariate interactions of Status XGrade were significant, five of the six effectsapproached significance. Analyses of the sta-tus group means for each behavioral descrip-tion at each grade level revealed that noneof the main effects of status reported earlierwas altered by these interactions. Indeed, theprofiles described earlier were identical acrossgrade levels. As was the case for the Status XGender interactions, these interactions onlyreflected differences in the magnitudes ofeffects. It is the decrease in the magnitudeof status group differences from grades 3 and5 to grade 8 that explains the interactioneffects.

^~^ General Discussion

The data from the two studies underlinethe importance of using both positive andnegative sociometric choice questions to ob-tain a more differentiated picture of both thedimensions of social status among children'speer groups and the types of status that canbe found among these groups. AlthoughAsher and Hymel (1981) noted some of thehazards of using negative choice questions,and we agree that they must be used cau-tiously and carefully, there is much that can-not be learned about snqa| relations among

^children without introducing the negativechoice issue.

nre~preceding data provide a child's-eyeview of the basis for social status. Since itis the peer group with whom the child largelymust relate, the peer group's perception ofthe behaviors associated with different kinds

of social status is informative even thoughthat data may involve a confounding of thecausal relationships between actual behaviorand social choice. Certainly peers have ac-cess to a wider sample of social interactionsand a wider range of interaction settingsthan do adult observers.

The kinds of behavior associated with thepositive end of the social preference dimen-sion are similar to those described in otherstudies of children of these ages, studies thatemployed the "Guess Who?" or "ClassPlay" technique (e.g., Tuddenham, 1951,with elementary school children and Elkins,1958, with eighth graders.) There were noreal age differences found for these patternsThere were, however, some age-relatedchanges in .the behaviors associated withnegative social preference. In Experiment 1we found the younger children had a simpleand undifferentiated picture of unpopularchildren compared to the eighth-grade chil-dren. For the younger children there wasessentially one reason for rejecting anotherchild; for older children there could be sev-eral different reasons. At all ages, indirectaggression was associated with low prefer-ence, but there was less emphasis on activemisconduct among the junior high schoolgroup. Whether this is because the processof social 'perception is more global and sim-plistic among the younger .children or be-cause there are more varied patterns of be-havior tied to unpopularity among the oldergroup cannot be determined from the pres-ent data. Similar age shifts in the nature ofpeer descriptions of low social status childrenhave been found (Coie & Pennington, 1976),although the fact that fewer predicters ac-counted for more variance in the criterionvariable for both the social preference andsocial impact regressions for the youngergroups does support the hypothesis that so-cial perception is more global among theyounger children.

Experiment 2 was undertaken to obtaina better picture of the controversial statusgroup as well as to resolve questions aboutthe basis for the mixed collection of behav-iors connected with high social impact scores.As it turns out, the mixed description gen-erally applies to the controversial group withthe exception of the one prosocial behavior—


cooperation—which was highly correlatedwith the "supportive" item that emerged inthe regression analyses. Thus, it is likely thatthe high social impact scores of very popularwhite children accounted for the high load-ing of this prosocial item on the social impactdimension because controversial childrenwere not perceived to be above average onexhibiting this behavior. The status groupanalyses of Experiment 2 also shed morelight on the neglectedgroup (but only a littlelight) than did Experiment 1, since no be-havioral items had significant negative load^ihgs on social impact, except for the mar-^gmaTreiationship of shyness in the eighth-grade results. It is clear from Table 4 thatthe neglected group is truly the polar op-posite of the controversial group, since on allbut the "cooperates" item they were at op-posite ends of the distributions for the groupmeans. They are a low visibility group. Forthis reason it may have been difficult evenfor their peers to reach a consensus aboutnaming them as shy.

The controversial children are an inter-esting group to consider and one with whomother children seem to have to reckon. Theydo not show up in large numbers and for thisreason may have been overlooked, except forthe very large-scale sociometric study ofRoff et al. (1972). The peer descriptions con-nected with the controversial group are com-patible with the anecdotal picture providedby Roff et al., particularly for the lower SEScontroversial children of our sample. Thepeer description of the controversial childrenis a blend of the description provided for thepopular and rejected children. Controversialchildren appear to engage in the activelyantisocial behavior associated with ex-tremely rejected children, and they tend toseek help with their school work. (Although

/ t h e item, "seeks help." does not specify that/ it is school help they seek, the item is highlyI correlated with the item "slow to learn inL^ school-" 1 On the other hand, controversial

children are viewed as leaders in the peergroup, and in this way they are like the pop-ular group (or at least the white populargroup). They are not viewed as shy. Theyare viewed as neither highly cooperative noras uncooperative. One might speculate that

controversial children possess more positivesocial skills than they are described as hav-ing, simply because it must be hard for peersto describe them as good to have in a groupwhen they also tend to see them as disruptiveand aggressive. That they give rise to con-siderable affective ambivalence may not bejust a matter of mixed opinion among peergroup members but of mixed feelings withinindividual group members as well.

Roff et al. compiled case studies on chil-dren who received large ^numbers of bothpositive and negative sociometric mentionsin an extensive survey of sociometric statusamong preadolescents in Texas and Min-nesota. They described males from lowerSES backgrounds who were not coldly hos-tile and antagonistic toward peers but whodid appear to be at risk for juvenile delin-quency. Our speculations regarding the peerdescriptions of these children suggest the im-portance of more objective observer assess-ments of the'social behavior of controversial'children, as well as the compiling of longi-tudinal evidence on both the continuity ofsocial status for these children and their po-tential involvement in serious antisocial ac-tivity. The rejected group with whom theyapparently shared many antisocial tenden-cies has already been identified as a riskgroup for long-term problems. It would beinteresting to know whether the leadershipcapacities wiih which tfte controversial are Mcredited is a mitigating factor with respect 11to their risk potential.

The Race X Status Group interaction forthe leader item gives rise to some specula-tions about what it means to be a popularor a controversial child. In the group studiedhere, the black children found themselves inthe racial minority. (In this community ithad only been a matter of some half-dozenyears that this school had been racially in-tegrated.) One possible consequence of thisminority situation is that black children whoare assertive and who otherwise possess theattributes of popular children may generateresentment among majority children whenthey attempt to dominate situations and thusget into conflict. In an all-black peer groupthey might simply become popular children,but in the mixed situation they become con-

~v^


troversial and take on some of the perceivedcharacteristics of unpopular children. A sec-ond possibility is that children who normallybecome popular with peers do so by first siz-ing up social situations and then taking rolesin those situations that are nonthreateningto the larger group. As they acquire popularstatus, they also acquire the freedom to bemore assertive and less uniformly prosocial.Hollander (1958) referred to such a processas the acquisition of idiosyncracy credit.Minority children who are popular in mixedracial groups either may not acquire suchcredit as quickly as majority children or maybe more hesitant to test such a possibilityand thus take less active leadership rolesthan they otherwise might. Or it may be thatthey do behave as leaders in some ways butare not perceived as such by the larger group.

The method of defining independent di-mensions of social preference and social im-pact and selecting types of social statusgroups on the basis of these dimensions of-fers a means of identifying groups, such asthe controversial and neglected children,who might otherwise be confused with chil-dren whose behavior patterns and social re-lations are quite different. Our data seem tosuggest distinctive patterns of behavior forthe four extreme status groups we identi-fied—patterns that were quite similar at allthree age groups investigated. Given the re-newed interest in using social status as apredictor variable for long-term problems ofadjustment and as a basis for identifyingchildren for preventive intervention (Fur-man, Rahe, & Hartup, 1979; Gottman,Gonso, & Schuler, 1976; Ladd, 1981; Oden& Asher, 1977), these distinctions in typesof social adjustment take on a special im-portance.

References

Asher, S. R., & Hymel, S. Children's social competencein peer relations: Sociometric and behavioral assess-ment. In J. D. Wine & M. D. Syme (Eds.), Socialcompetence. New York: Guilford, 1981.

Bower, E. M. Early identification of emotionally handi-capped children in school. Springfield, 111.: CharlesC Thomas, 1960.

Coie, J. D., & Pennington, B. F. Children's perceptionsof deviance and disorder. Child Development, 1976,47, 407-413.

Cowen, E. L., Pederson, A., Babigian, M., Izzo, L. D.,& Trost, M. A. Long-term follow-up of early detectedvulnerable children. Journal of Consulting and Clin-ical Psychology, 1973, 41, 438-446.

Dion, K., & Berschied, E. Physical attractiveness andpeer perception among children. Sociometry, 1974,37, 1-12.

Dunnington, M. H. Behavioral differences in sociomet-ric status groups in a nursery school. Child Devel-opment, 1957,25, 103-111.

Elkins, D. Some factors related to the choice status ofninety eighth-grade children in a school society. Ge-netic Psychology Monographs, 1958, 55, 207-272.

Furman, W., Rahe, D. F., & Hartup, W. W. Rehabil-itation of socially withdrawn preschool childrenthrough mixed-age and same-age socialization. ChildDevelopment, 1979, 50, 915-922.

Gottman, J. M. Toward a definition of social isolationin children. Child Development, 1977, 48, 513-517.

Gottman, J. M., Gonso, J., & Schuler, P. Teaching so-cial skills to isolated children. Journal of AbnormalChild Psychology, 1976, 4, 179-197.

Green, K. D., Forehand, R., Beck, S. J., & Vosk, B. Anassessment of the relationship among measures ofchildren's social competence and children's academicachievement. Child Development, 1980, 5/, 1149-1156.

Gronlund, N. E. Sociometry in the classroom. NewYork: Harper, 1959.

Gronlund, N. E., & Anderson, L. Personality charac-teristics of socially accepted, socially neglected, andsocially rejected junior high school pupils. Educa-tional Administration and Supervision, 1957, 43,329-338.

Hartup, W. W. The origins of friendship. In M. Lewis& L. A. Rosenblum (Eds.), Friendship and peer re-lations. New York: Wiley, 1975.

Hartup, W. W. Children and their friends. In H.McGurk (Ed.), Issues in childhood social develop-ment. London: Metheun, 1979.

Hollander, E. P. Conformity, status and idiosyncracycredit. Psychological Review, 1958, 65, 117-127.

Hymel, S., & Asher, S. R. Assessment and training ofisolated children's social skills. Paper presented atthe biennial meeting of the Society for Research inChild Development, New Orleans, April 1977. (ERICDocument Reproduction Service No. ED 136980)

Johnson, S. O. Hierarchical clustering schemes: An al-gorithmic method. Psychometrica, 1967, 32, 241-254.

Kleck, R., Richardson, S., & Ronald, L. Physical ap-pearance cues and interpersonal attraction in chil-dren. Child Development, 1974, 45, 305-310.

Kohlberg, L., Lacrosse, J., & Ricks, D. The predict-ability of adult mental health from childhood behav-ior. In B. B. Wolman (Ed.), Manual of child psy-chopathology. New York: McGraw-Hill, 1972.

Ladd, G. W. Effectiveness of a social learning methodfor enhancing children's social interaction and peeracceptance. Child Development, 1981,52, 171-178.

Langlois, J. H., & Stephen, C. The effects of physicalattractiveness and ethnicity on children's behavioral


attributions and peer preferences. Child Develop-ment, 1977, 48, 1694-1698.

Moreno, J. L, Who shall survive? A new approach tothe problems of human interrelations. Washington,D.C.: Nervous and Mental Disease Publishing, 1943.

Northway, M. L. Outsiders: A study of the personalitypatterns of children least acceptable to their agemates. Sociometry, 1944, 7, 10-25.

Oden, S., & Asher, S. R. Coaching children in socialskills for friendship making. Child Development,1977, 48, 495-506.

Peery, J. C. Popular, amiable, isolated, rejected: A re-conceptualization of sociometric status in preschoolchildren. Child Development, 1979, 50, 1231-1234.

Roff, M., Sells, S. B., & Golden, M. M. Social ad-justment and personality development in children.Minneapolis: University of Minnesota Press, 1972.

Rutter, M. Helping troubled children. London: PlenumPress, 1976.

Shantz, C. U. The development of social cognition. InE. M. Hetherington (Ed.), Review of child devel-opment research (Vol. 5). Chicago: University ofChicago Press, 1975.

Singleton, L. C., & Asher, S. R. Racial integration andchildren's peer preferences: An investigation of de-velopmental and cohort differences. Child Develop-ment, 1979, SO, 736-941.

Speed, F. M., Hocking, R. R., & Hackney, O. P. Meth-ods of analysis of linear models with unbalanced data.Journal of the American Statistical Association,1978,73, 105-112.

Tuddenham, R. D. Studies in reputation: III. Correlatesof popularity among elementary school children.Journal of Educational Psychology, 1951, 42, 257-276.

Received October 9, 1981

dimensions and types of social status: a cross-age perspective

Documents