21
Journal of Abnormal Child Psychology, Vol. 10, No. 3, 1982, pp. 389-410 Behavior Patterns of Socially Rejected and Neglected Preadolescents: The Roles of Social Approach and Aggression 1 Kenneth A. Dodge 2 Indiana University at Bloomington John D. Coie Duke University N. Paul Brakke Indiana University at Bloomington Sociometric nominations were used to select groups of popular, average, rejected, and neglected third- and fifth-grade children. In two studies, the peer interactive behaviors of these children were naturalistically observed in their classrooms and on the playground. In contrast to popular children, rejected children displayed fewer task-appropriate behaviors and more task-inappropriate and aggessive behaviors. Whereas rejected children prosocially approached peers as frequently as did popular children, peer responses to the approaches of rejected children were more likely to be negative. Neglected children, on the other hand, displayed relatively few task-inappropriate and aggressive behaviors, and socially approached peers infrequently. Their approaches also met with frequent rebuff by Manuscript received in final form January 25, 1982. 1This research was supported by BRSG Grant 507 RR0703, National Institutes of Health, to the first author and by a Duke University Research Council Grant to the second author. The authors wish to acknowledge the support of the administrators and teachers of the Durham County School System and the Monroe County Community School Corporation. Betty Bisno, Janice Bremer, John DeHority, Jean Ann Matter, Theresa Rader, and Polly Wheeler were instrumental in the conduct of this research. A report of study 1 was presented at the Fifth Biennial Southeastern Conference on Human Development, Atlanta, 1978. 2Address all correspondence to Kenneth A. Dodge, Department of Psychology, Indiana University, Bloomington, Indiana 47405. 389 0091-0627/82/0900-0389503.00/0 1982 Plenum Publishing Corporation

Behavior patterns of socially rejected and neglected preadolescents: The roles of social approach and aggression

Embed Size (px)

Citation preview

Journal o f Abnormal Child Psychology, Vol. 10, No. 3, 1982, pp. 389-410

Behavior Patterns of Socially Rejected and Neglected Preadolescents: The Roles of Social Approach and Aggression 1

Kenneth A. Dodge 2 Indiana University at Bloomington

John D. Coie Duke University

N. Paul Brakke Indiana University at Bloomington

Sociometric nominations were used to select groups o f popular, average, rejected, and neglected third- and fifth-grade children. In two studies, the peer interactive behaviors o f these children were naturalistically observed in their classrooms and on the playground. In contrast to popular children, rejected children displayed fewer task-appropriate behaviors and more task-inappropriate and aggessive behaviors. Whereas rejected children prosocially approached peers as frequently as did popular children, peer responses to the approaches o f rejected children were more likely to be negative. Neglected children, on the other hand, displayed relatively f ew task-inappropriate and aggressive behaviors, and socially approached peers infrequently. Their approaches also met with frequent rebuff by

Manuscript received in final form January 25, 1982. 1This research was supported by BRSG Grant 507 RR0703, National Institutes of Health, to the first author and by a Duke University Research Council Grant to the second author. The authors wish to acknowledge the support of the administrators and teachers of the Durham County School System and the Monroe County Community School Corporation. Betty Bisno, Janice Bremer, John DeHority, Jean Ann Matter, Theresa Rader, and Polly Wheeler were instrumental in the conduct of this research. A report of study 1 was presented at the Fifth Biennial Southeastern Conference on Human Development, Atlanta, 1978.

2Address all correspondence to Kenneth A. Dodge, Department of Psychology, Indiana University, Bloomington, Indiana 47405.

389

0091-0627/82/0900-0389503.00/0 �9 1982 Plenum Publishing Corporation

390 Dodge, Coie, and Brakke

peers. The findings were discussed in terms o f the behavioral bases o f sociometric status. Suggestions were made for clinical researchers interested in behavioral change with rejected and neglected children.

The study of sociometric status in children's peer groups has increased markedly in recent years. This movement has been related to a growing awareness of the role of peer relations in social development (Hartup, 1979), to the fact that many children are being referred to mental health clinics for difficulties in peer relations (Woody, 1969), and to the discovery that negative peer status in childhood is predictive of a variety of maladap- tire outcomes in adolescence and adulthood, including deliquency (Roff, Sells, & Golden, 1972), psychopathy (Robins, 1966), and psychiatric impairment (Cowen, Pederson, Babigian, Izzo, & Trost, 1973). Measures of sociometric status are now being used to select children for preventive intervention programs, and as change measures to evaluate the succes of those programs (e.g., Gottman, Gonso, & Schuler, 1976; Ladd, 1981; Oden & Asher, 1977). Those intervention programs are predicated on the existence of social skills deficits in certain children or behavior differences among various status groups. The goal of the present research was to determine specific behavioral differences among children in several social status groups, for the purpose of coming to a better understanding of the role of behavioral processes in the development and maintenance of peer status, and for the purpose of focusing behavior change efforts by clinical researchers.

The most commonly used index of peer status has been the sociometric nomination scores one obtains by using Moreno's (1934) peer nomination technique. Frequently, researchers have used low scores on positive status items (such as "Whom do you like?") as the criterion for selection to a low status group. Recently, a number of researchers (Goldman, Corsini, & de Urioste, 1980; Gottman, 1977; Hymel & Asher, 1977) have pointed out, however, that use of this single criterion confounds two groups of low status children: those who are not liked and are highly disliked and those who are not liked but are not disliked either. Coie, Dodge, and Coppotelli (1982) use the terms actively rejected for the former group and socially neglected for the latter group. Through interviews with the p~ers of these children, they found that rejected children are perceived as more aggressive and less cooperative than popular children, whereas neglected children are perceived as neither aggressive nor cooperative but as shy an fion- involved.

Whereas most of the descriptive findings on the behavioral correlates of rejected and neglected peer status are based on peer or teacher percep- tions, a number of studies have employed direct observational methods (for reviews, see Asher & Hymel, 1981; Asher, Oden, & Gottman, 1977).

Rejected and Neglected Preadolescents 391

In general, the observational data match the peer perception data quite well. That is, rejected boys have been found to display more hostile aggression toward peers (McGuire, 1973) and to dispense fewer prosocial reinforcers than do popular children (Hartup, Glazer, & Charlesworth, 1967). Un- fortunately, most of the observational studies of peer status have been conducted with preschool children. Very few observational studies of older, preadolescent children have been conducted, presumably because of dif- ficulties in effectively observing older children and because of the increased complexity of behaviors that must be observed among older children. Since most of the empirical bases for thinking that childhood social maladjust- ment is a good prediction of future adjustment problems are founded on social status data for older children (e.g., Roff et al., 1972), the need for observational studies of status groups of older children is all the more apparent.

The few observational studies of peer status in preadolescent groups that do exist provide a mixed pattern of findings. Lippitt and Gold (1959) used the observational logs of camp counselors to conclude that the rejected children in a camp group exhibited more inappropriate aggression and destructiveness than their nonrejected peers. On the other hand, Campbell and Yarrow (1981) used observational specimen records to conclude that both popular and rejected children display more aggression than neglected children and do not differ from each other in aggression. However, rejected children were the targets of peers' aggression more frequently than were other children. Also, popular children initiated more prosocial interactions than the other children.

Both of the preceding studies involved the use of a type of specimen record method of observation. The measures of reliability reported in those studies simply described the reliability with which behavioral category judgments were made from a written narrative, and not the reliability with which the narratives themselves were taken. Given that the narrators (camp counselors) knew the observed children and could have been biased in their logging of behavior, it is possible that the results of those studies are confounded by the approach to data collection.

A third study, by Gottman, Gonso, and Rasmussen (1975), did involve the use of observers who were unaware of the social status of the third- and fourth-graders they observed. The data were recorded in terms of behavior-event categories rather than o f specimen records. Social status was defined by the median split on the positive sociometric item (number of "best friend" nominations received). High friendship children more frequently received positive reinforcement from other children and were less often engaged in solitary off-task behavior than low friendship children. A marginal effect (p < .09) for dispensing positive reinforcement also tended to corroborate the Hartup et al. (1967) findings. The category most

392 Dodge, Coie, and Brakke

closely approximating an aggression measure (distributing negative reinforcers) did not provide any significant social status differences, but as noted above, the use of only a positive nomination status measure to determine social status results in a comination of children having very different social characteristics. Thus, a more definitive observational study would require that these two groups-rejected and neglected children- be separated as quite distinct criterion groups.

The purpose of the first study reported here was to obtain behavioral data on the social behavior of actively rejected preadolescents in contrast to their popular and average status peers and to collect such data across a range of settings in which the school peer group interacts. In addition to assessing behavior pattern differences among children of varying status groups, we were interested in the ways children of varying social status differentiated their behavior according to environmental context. We were interested both in the ways that these children act toward others and in the ways in which they are treated by other children. The categories of observa- tion used in this study included prosocial approaches by one child toward a peer, antisocial aggressive acts directed at a peer, and the child's ap- propriate and inappropriate solitary behavior. In the case of prosocial approaches toward a peer, the valenced response by the peer was also noted.

In the first study, rejected, average, and popular fifth-grade boys and girls in a semirural North Carolina public school were observed by trained coders who sat nearby. In the second study, the social behaviors of rejected, neglected, average, and popular third- and fifth-grade boys in a Mid- western school were recorded by a remote-control zoom lens video camera and later coded by trained personnel. The differences in subject populations and observational methodologies of these two studies allow for some evaluation of the generalizability of the findings. Also, the teachers of the children were asked to assess the academic performance and social adjustment of the children under study.

STUDY 1

Method

Subject Selection. Subjects were selected from the fifth grade of a North Carolina public elementary school serving a lower-middle-SES population, on the basis of peer sociometric nominations. Following the receipt of parental permission, 101 fifth-graders were administered a socio- metric interview in group sessions. Children were asked to select from

Rejected and Neglected Preadolesccnts 393

a roster the names of three peers whom they liked most and then three peers whom the liked least. Liked most (LM) and liked least (LL) scores for each child were computed as the sum of all nominations received from peers, and these scores were combined to form a measure of social preference of popularity (cf. Coie et al., 1982). The eight children who received the lowest social preference scores (calculated as LM-LL) were selected as the actively rejected status group (LM mean = 1.3, range = 0 to 2; LL mean = 8.8, range = 6 to 16). The eight children who received the highest social preference scores were selected as the popular status group ( L M m e a n -- 8.8, range = 6 t o 12; L L m e a n = 1.6, range -- 0 t o 4 ) . An average status group of eight children was selected from among those children receiving social preference scores that were close to the median score for the population (LM = 2.5, range = 1 to 4; LL mean = 1.8, range = 1 to 3). Since the selection criteria were based solely on socio- metric scores, slight differences in the racial and gender composition of the groups emerged . .Of the eight children in each group, the rejected group included three black males, three white males, and two black females. The popular group consisted of two black males, two white males, one black female, and three white females. The average group consisted of one black male, four white males, two black females, and one white female.

Observations. A six-category behavioral event coding system was developed for this investigation. A specified child was targeted for observa- tion and his or her behaviors were coded for 5-minute periods. Two of the categories described solitary activity. Solitary-task-appropriate behavior was defined as the activity displayed by the child when he or she was not in the physical proximity of others, or was physically proximate to others but was not interacting with them, and that was deemed by the observer to be appropriate to, and in accordance with, the tasks placed before the child. This category included seat work, attending to the teacher, and standing in line at appropriate times. This category was differentiated from solitary-task-inappropriate behavior, which was defined as that solitary activity displayed by the child that was deemed by the observer to be inappropriate to the tasks placed before the child and was therefore a violation of classroom norms. This category included clowning, daydreaming, and aimlessly walking around the room. It is essential to point out that these two categories were defined in reference to the observer's understanding of the norms and tasks of the classroom group. The same behavior (standing) could be either appropriate or inappropriate, depending on the task at hand.

Four categories described social interactive behaviors among peers. Two of these were aggressive acts, which were defined as verbal or physical hostile or destructive behaviors that are assaultive, taunting, obstructive, or threatening in nature and are directed at a specific peer. These categories

394 Dodge, Coie, and Brakke

included physical hits, shoving, pushing, name calling, teasing, and berating a peer. A distinction was made as to whether the target child initiated aggression or was the object of aggression initiated by a peer.

The remaining two categories described prosocial approaches by one child toward a peer or group of peers for the purpose of initiating play or conversation. These approaches have been called "group entry at- tempts" by Putallaz and Gottman (1981) and "initiations" by Vandell and George (1981). A prosocial approach usually involved moving into the physical proximity of a peer and displaying a verbal behavior designed to initiate an ongoing social activity. A distinction was made between those prosocial approaches initiated by the subject toward a peer and those prosocial approaches initiated by a peer toward the subject. For th.e approaches initiated by the subject, the valenced response by the peer was also noted. This response was coded as "accept" if the peer responded positively to the subject and the two children continued to interact in cooperative ways, and "reject" if the peer responded negatively to the subject by ignoring, rebuffing, or rejecting the approach. Valenced responses to prosocial approaches were analyzed separately from the rest of the observational data.

Four female graduate student observers (naive to the status of the children) were trained to code the occurrence and duration of each of the six categories. They employed a 15-second interval recording method. Observers sat at the back of the classroom or on the fringes of the play- ground and recorded behaviors of a target child on a clipboard equipped with a timing device. Any time one of the six behavior categories occurred within a particular 15-second interval, it was recorded in a box corre- sponding to that interval. If a behavior continued beyond the interval, the duration was noted in terms of the number of intervals it spanned. Distinction was made between a behavior that lasted more than one interval and two same-category behaviors that occurred in adjacent intervals. Observers coded the behavior of one child for a 5-minute period and then located a second child and coded his or her behavior for the next 5- minute period, and so on, according to a randomized schedule. After 2- weeks of training,-measures of interobserver agreement between each possible pair of observers were collected by having two observers simul- taneously code the behaviors of the same child. Twenty-four 5-minute periods (480 15-second intervals) constituted the test of interobserver agreement. Agreement was calculated as the number of intervals in which both observers agreed that a particular behavioral category had occurred divided by the number of intervals in which they had agreed plus the number of intervals in which they had disagreed. The overall weighted percent agreement for this measure was 68%. The percent agreements for

Rejected and Neglected Preadolescents 395

"Fable i. Measures of Interobserver Agreement for Each Behavioral Category

Code

Percent intercoder agreement u

Study 1 a Study 2"

Solitary-task-appropriate 86 Solitary-task-inappropriate 52 Prosocial approach by subject 69 Accept versus reject response by peer 77 Aggressive acts by subject 43 Prosocial approach toward subject 70 Aggressive acts toward subject 50 Teacher-subject interaction

94 83 67

100 72 72 68 75

"Agreement was defined as the number of agreements divided by the number of agreements plus disagreements. For the peer's response to a prosocial approach, only those instances in which both observers agreed that an approach had occurred were considered.

bFigures are based on 120 minutes (480 15-second intervals) of coding.

'Figures are based on 480 minutes 12,680 10-second intervals) of coding.

the individual observation categories ranged from 43~ to 86070. These figures are displayed in Table I.

After training, each subject was observed for six 5-minute periods by each of two observers. Half of these periods took place in the classroom setting when children were supposed to be in their seats doing independent work, and half took place on the playground when children were at recess. The environmental context (classroom vs. playground) constituted an independent variable in later analyses. Approximately 12 weeks after the first set of observations were completed, in the spring of the school year, each subject was observed for 12 more 5-minute periods? This second set of observations constituted a replication of the first set. Time of observa- tion (fall vs. spring) was also treated as an independent variable in sub- sequent analyses.

The intercoder agreement data suggest that caution be taken in using these data-to make comparisons of individuals on some of the categories of behavior. The impact of questionable reliability is somewhat different where between-groups differences are at issue, as is the case for this study.

3The subjects in this study were the control group for a pilot intervention program that yielded no significant outcome effects. For this reason, observations were made at two different times.

396 Dodge, Coie, and Brakke

As both Johnson and Bolstad (1973, p. 17) and Yarrow and Waxler (1979, p. 56) have argued, "Other things being equal, if a positive finding appears even with considerable error in the measures, a relatively strong relation can be assumed. On the other hand, observer unreliability leaves one less certain as to what can be concluded from negative findings." Thus, an important issue is the generalizability of significant between-groups differences found in the present study. This generalizability is examined within the structure of the design by measurment across time (fall vs. spring sessions) and setting (classroom vs. playground).

Teacher Assessments. The teacher of each subject rated the academic performance and social adjustment of each of her students on two 9-point scales at the time of the first set of observations. These ratings were repeated during the second set of observations. Test-retest reliabilities for the two measures were .84 and .82, respectively. The initial ratings were used as the dependent measures in subsequent analyses.

Results

Observations. Frequency and duration (coded in terms of 15-second intervals) scores were computed for each of the six behavior categories. Since duration scores correlated highly (r > .90 for all categories) with frequency scores, only the analyses of frequency scores are reported. The six frequency scores served as dependent variables in a multivariate analysis of variance, in which subject status (popular, average, or rejected) was a between-subjects factor and environmental context (classroom vs. playground) and time of observation (fall vs. spring) were within-subjects factors. Because of slight gender differences in cell frequencies, all analyses were repeated with a males-only sample. The findings of these analyses were similar to those reported for the whole sample: When a significant multivariate effect was found, significant univariate effects were explored. Status effects were evaluated by means of a Newman-Keuls test of cell dif- ferences.

The mean frequencies of each of the six behavior categories for each of the status groups in the classroom and on the playground are found in Table II. The means reported are based on 60 minutes of observation, A significant multivariate main effect of status was found, F(12, 32) = 2.41, p < .03. A univariate main effect for the category solitary-task-ap- propriate, F(2, 21) = 11.49, p < .001, and subsequent Newman-Keuls tests of status cell means revealed that the rejected group was significantly less frequently engaged in task-appropriate solitary activity than either the average or the popular group (each p < .05). The latter two groups did not differ in this category. A main effect for the category aggressive

Rejected and Neglected Preadolescents 397

Table !|. Mean Frequencies of Behaviors According to Status Group and Environmental Context: Study 1 ~'

Status group

Popular Average Rejected

Classroom Play Classroom Play Classroom Play

Solitary-appropriate 23.3 17.6 22.9 21.4 18.6 11.6 Solitary-inappropriate 2.2 2.6 2.0 1.1 4.5 1.5 Social approaches 16.5 36.0 18.0 35.6 28.3 28.9

Percent rejected by peer 6.7 4.2 5.6 5.3 17.0 6.6 Aggressive acts 2.4 3.4 1.5 1.9 4.3 7.5 Social approaches toward 21.7 24.4 16.1 19.5 17.6 20.6

subject Aggressive acts toward 2.4 4.1 3.0 1.9 2.6 3.4

subject

'+Scores reported represent the mean frequency of each behavior pet hour of observation, except percent rejected by peer.

acts by the subject, F(2, 21) = 4.90, p < .02, and subsequent Newman- Keuls tests revealed that the rejected group engaged in significantly more aggressive acts toward peers than the average and popular groups (each p < .05), whJ.ch did not differ. Since it might be argued that the status group differences in the frequency of aggression were merely a function of differences in interpersonal activity level, the proportion of aggressive acts to total interpersonal behaviors was calculated for each child and assessed by an additional univariate ANOVA. This analysis also revealed a significant main effect of status, F(2, 21) = 6.51, p < .01. Whereas 16.8% of the interpersonal behaviors of rejected children were aggressive acts, only 6.5% of the behaviors by average children and only 7.5% of the behaviors by popular children were aggressive. The differences between the rejected group and the other two groups were each significant (Newman-Keuls test, p < .05). Because none of these univariate main effects were qualified by a significant interaction with other factors, and because no multivariate interaction effects involving status were found to be significant, the preceding findings may be deemed generalizable across the time of observation and the environmental contexts in which the observations were made. Even though the multivariate interaction between status and context was not significant, F(12, 32) = 1.34, n.s., one significant univariate interaction will be reported, albeit with caution. A significant interaction was found for the category social approach by the subject, F(2, 21) = 4.41, p < .03. Newman-Keuls tests revealed that the rejected group made significantly more prosocial approaches than the average and popular groups did in the classroom, whereas the reverse was true on the playground (eachp < .05).

398 Dodge, Coie, and Brakke

In order to investigate the responses of peers to the prosocial ap- proaches by subjects, the proportions of approaches that were accepted and that were rejected by peers were computed for each status group and compared by means of Z-score tests (Bakeman, 1978). These proportions are listed in Table II. These analyses revealed that, summed across both contexts, the proportion of prosocial approaches that were rejected by peers was significantly greater for rejected children (11.4%) than for popular children (5.0%0), Z = 2.83, p < .01, and for average children (5.2%) Z = 2.61, p < .01. In other words, when a prosocial approach was attempted, rejected children were significantly more likely to be rebuffed in that attempt than were popular or average children.

The multivariate main effect of environmental context was found to be significant, F(6, 16) = 7.23, p < .001. Significant univariate main effects revealed that task-appropriate solitary activity was more frequent in the classroom than on the playground, F(1, 21) --- 24.44, p < .001, whereas social approaches were more frequent on the playground than in the class- room, F(1, 21) = 19.16,p < .001.

A significant multivariate main effect of time of observation, F(7, 15) = 4.16, p < .01, and significant univariate effects revealed more inter- personal activity during the spring observations than during the fall ob- servations. More social approaches F(I , 21) --- 17.15, p < .001, more social approaches by peers to the target child, F(1, 21) = 10.15, p < .001, and more aggressive acts by the subject, F(1, 21) = 4.68, p < .05, were found in the spring than in the fall. Time of observation did not, however, significantly interact with any other factor.

Teacher Assessments. Teacher assessments were evaluated by two one-way analyses of variance in status groups with subsequent Newman- Keuls tests. The group means for these analyses are found in Table III.

Table !I1. Status Group Means for Teacher Assessments: Study 1 and Study 2"

Status group

Assessment I'opular Average Rejected Neglected

Study 1 Academic performance 7.3,, 5.1 I, 2.3,. - Social adjustment 7.% 5.6j, 3 .5 , . -

'-;t udy 2 Favorability of peer relations 7.6,, 5.4~, 2.6c 4.5~, Frequency of aggression 2.91, 2.%, 6.5,, 3.11, Frequency of social approaches 6.9,, 5 . 8 , , 6.2, 4.1 i,

"On 9-point scales, higher scores represent greater academic performance, better social adjustment, more favorable peer relations, greater frequency of ag,~ression, and greater frequency of social approaches. Scores designated by the subscript a differ from those designated by b, which differ from those designated by c, by Newman-Keuls tests, p < .05.

Rejected and Neglected Preadolescents 399

Teachers assessed the academic performances of rejected children as sig- nificantly lower than those of average children, which, in turn, were sig- nificantly lower than those of popular children, F(2, 21) = 4.09, p < .03, Newman-Keuls tests, each p < .05. Also, teachers found the rejected children to be significantly more well adjusted socially than the average children, who were judged as significantly less well adjusted socially than the popular children, F(2, 21) -- 5.54, p < .02, Newman-Keuls tests, eachp < .05.

Discussion

The data from this study provide a picture, in relatively molar inter- actional terms, of contrasts in the social behavior of popular, average, and rejected status groups of fifth-grade children. Several findings emerged that, taken together, may describe some of the bases for peer rejection and may help clinical researchers focus their behavior change efforts with rejected children. First, in the classroom, rejected children spent less time in appropriate solitary activity than average and popular children. Second, they approached peers in the classroom more frequently than did other children, an activity that may be considered inappropriate for that context (work time). Apparently, the rejected children's peers thought that these approaches were inappropriate, for they rejected these approaches at a higher rate than they did the approaches of other children. In an earlier study (Coie et al., 1982), fifth-grade peers described the rejected children in their grade level as more disruptive and more likely to start fights than other children. It may be the case that some of the social approach behavior observed in this study is what was referred to as dis- ruptive behavior by the peers. Finally, and possibly most important, the data of this study corroborate the peer descriptions of rejected children as starting fights. Rejected children displayed over twice as many peer- directed aggessive acts as popular and average children. These findings suggest that behavior change efforts with rejected status children should focus on task-appropriate activity, social approach patterns, and aggression. We will defer further discussion of these findings until Study 2 has been reported.

STUDY 2

Even though the data of the first study strongly suggested that both social approach patterns and aggression are correlates of sociometric status in fifth-graders, several questions remained for a second observational study. First, because of the difficulties in observing children

400 Dodge, Coie, and Brakke

of this age, interobserver agreement ratios in the first study were not as high as would have been preferred. In an attempt to improve the agree- ment among observers, the second study utilized video recordings of children's naturalistic behaviors so that observers could view and review the behaviors until greater agreement was reached. Also, slight changes in the coding system were made. Second, in the first study, matching of children in status groups according to race and gender was not perfect, because only sociometric criteria were used to select children. In the second study, only boys were observed, and children were matched according to race. Third, the first study did not include an interesting sociometric status group, neglected children, defined as those children who are neither liked nor actively disliked by their peers. Sociometrically, this group is quite distinct from actively rejected children. Many intervention programs in- clude both kinds of children in treatment but do not distinguish rejected from neglected children. It is possible that these two groups display quite different behavioral patterns. The neglected group was therefore included in the observations of the second study. Fourth, to explore possible age differences in status group behavioral patterns, children from grade 3 and grade 5 were observed in the second study. Finally, the second study could be viewed as an attempt to replicate the findings of the first study with children from a different geographical and cultural setting.

Method

Subject Selection. Following the receipt of parental permission, each of 101 third-grade children and 118 fifth-grade children from a Midwestern public elementary school serving a middle-class population was ad- ministered the same sociometric interview used in Study 1. At each grade level, the 8 boys who received the lowest social preference scores (number of liked most (LM) nominations less number of liked least (LL) nomina- tions) were selected as the rejected status group (LM mean = 1.08, range = 0 to 4; LL mean -- 11.2, range = 6 to 17). Matching for race (there were 2 minority boys selected for each status group) and classroom, the 8 boys at each grade level who received the highest social preference scores were chosen as the popular group (LM mean = 5.3, range = 4 to 8; LL mean = .9, range = 0 to 2). The neglected group was selected according to a procedure described by Coie et al. (1982), and Goldman et al. (1980). Also matching for race and classroom, the 8 boys who received the lowest social impact scores (LM + LL) were selected as the neglected status group (LM = 1.0, range = 0 to 2, LL mean = 1.0, range = 0 to 2). The average .status group (also matched to the rejected group) was selected from those boys receiving LM and LL scores that were each close to the modal scores

Rejected and Neglecled Preadolescents 401

for the population (LM mean = 3.0, range = 2 to 4; LL mean = 2.6, range = 2 to 3). After selection of these 64 subjects, 9 boys were not able to be observed because of lack of permission or absenteeism, so 55 boys were observed. Of these, 15 were popular, 11 were average, 16 were neglected, and 13 were rejected. The reported means for the sociometric measures are based only on the observed children.

Observations. Children's behaviors in the classroom and on the play- ground were recorded and coded with the same coding categories used in Study 1. One additional behavior was coded, called teacher-child interac- tions. This activity occurred in the classroom only and was defined as any verbal interaction between the teacher and the child. Since most of the time the children were supposed to be engaged in independent seat work, most of these interactions involved directions and reprimands by the teacher or requests for help by the child. A second change in coding was integrated into Study 2. Rather than code the occurrence of each behavior separately, as in Study 1, observers coded the number of intervals in which each behavior occurred. That is, in Study 2, if the same behavioral act spanned two intervals, it was counted twice, whereas in Study 1 it would have been counted only once. This change was included because it made the coding task easier for observers. For example, if a child displayed aggressive behavior in two adjacent intervals, the observers in Study 2 coded the presence of aggression twice. For this same example, the observers in Study 1 would have been faced with the difficult task of determining how many different acts of aggression occurred in those two intervals (one, two, or more). As a result, the intercoder agreement in Study 2 was greater than that in Study 1.

When a specified child was targeted for observation in the classroom, his behavior was recorded for a 6-minute period by a remote-control zoom lens video camera that had been placed in a corner of the ceiling in the classroom and was operated by a technician in a control room elsewhere in the school. Five microphones were hidden in the ceiling of each class- room for audio recording. This elaborate procedure was possible because the school classrooms had been equipped with video cameras by the school system, and video recording in these classrooms was not uncommon. In each classroom, the technician followed a predetermined schedule of locating a specified target child, recording his behavior a 6-minute period, and then moving on to a second child, and so on. The technician had previously consulted with the teacher to help identify children. Children were aware that they were being filmed, but no child could determine when he was being targeted for observation. Teachers reported that the procedure was less obtrusive than having observers sit in the classroom. For the playground observations, the technician stood on the fringes of the area and used a portable zoom lens video camera with a single microphone

402 Dodge, Coie, and Brakke

to record the behaviors of each child for 6-minute periods on a similar rotating basis.

Each child was scheduled to be observed for four 6-minute periods in the classroom and four 6-minute periods on the playground. Absenteeism prevented the completion of this schedule for six children; however, all children were observed for at least five periods.

Two undergraduate research assistants were trained to code the be- haviors of each child from the video record in 10-second intervals. A time generator recorded time in seconds on the video record to aid the synchronization of coding by observers. Posttraining assessment of inter- observer agreement (using 30 minutes of behavior) revealed an overall agreement ratio of 87~ calculated as the number of agreements that a particular behavior has occurred divided by the number of agreements plus disagreements. Individual category agreement ratios ranged from 75 ~ to 100070, although no aggression occurred during these 30 minutes. Inter- observer agreement was repeatedly assessed during actual coding as well. To facilitate efficiency of coding, each observer coded one-half of all of the periods, plus one-fourth of the periods coded by the other observer (or 62.5~ of all periods). Observers did not know which of their codes were being checked. The interobserver agreement ratio for the actual coding, based on 8 hours of coding, averaged 86~ over all categories, and ranged from 6707o to 100~ These figures are displayed in Table I.

Teacher Assessments. The teachers of each child were asked to make three assessments on 9-point scales. First, they rated the favorability of each child's peer relations. Second, each teacher assessed the relative frequency of each child's peer-directed aggression. Finally, each teacher assessed the relative frequency of social approaches to peers (defined as attempts to initiate interaction with peers) by each child.

Results and Discussion

Observations. The frequencies of occurrence for each of the seven behavior categories were calculated as the number of 10-second intervals in which the behavior occurred per 6-minute period. The mean frequencies of each behavior for each status group in the classroom and on the play- ground are displayed in Table IV. The first six dependent measures were subject to a multivariate analysis of variance in which subject status (popular, average, rejected, or neglected) and grade level (3 vs 5) were between-subjects factors and environmental context (classroom vs. playground) was a within-subjects factor. Since the seventh measure, teacher-child interactions, was not observed on the playground, a separate univariate ANOVA was conducted for this measure. When the multivariate

'Fab

le I

V.

Mea

n O

ccu

rren

ces

of

Beh

avio

r A

cco

rdin

g t

o S

tatu

s G

rou

p a

nd

En

vir

on

men

tal

Co

nte

xt:

Stu

dy

2"

m.

e,,

Sta

tus

gro

up

Po

pu

lar

Av

erag

e R

eiec

ted

Neg

lect

ed

Beh

avio

r C

lass

P

lay

('la

ss

Pla

y C

lass

P

lay

Cla

ss

Pla

y

So

lita

ry-a

pp

rop

riat

e 23

0.6

11

8.1

21

6.8

119.

3 20

4.0

108.

4 31

0.0

187.

1 S

o[i

tary

dn

app

rop

riat

e 32

.1

3.2

36.6

3.

0 10

0.6

16.2

52

.2

4,5

Soc

ial

app

roac

hes

12

.4

44.1

16

.8

z6.3

15

.0

40.3

7.

9 15

,9

Per

cent

rej

ecte

d by

pee

r 1.

6 .0

.6

.7

2.

1 3.

8 .0

6.

6 A

ggre

ssiv

e ac

ts

2.1

8.5

3.0

4.4

3.3

19.4

2.

1 9.

0 So

cial

ap

pro

ach

es t

ow

ard

su

bje

ct

11.9

11

.3

5.0

12.8

5.

2 14

.2

12.5

13

.9

Agg

ress

ive

acts

to

war

d s

ub

ject

3.

9 6.

3 2.

2 4.

1 4.

9 6.

3 2.

0 5.

4 T

each

er-c

hild

in

tera

ctio

ns

7.1

- 8.

9 --

16

.5

- 3.

7 --

"Sco

res

rcp

ort

ed

repr

esen

t th

e m

ean

n

um

ber

of

10-s

econ

d in

terv

als

per

h

ou

r o

f o

bse

rvat

ion

, ex

cept

fo

r th

e ca

teg

ory

pe

rcen

t ~ e

ject

ed b

y pe

er.

404 Dodge, Coie, and Brakke

effect was significant, univariate effects were assessed by unweighted analyses of variance. Status group differences were assessed by Newman- Keuls tests.

A significant multivariate main effect of status was found, F(18, 130) = 2.08, p < .01, as was a significant multivariate interaction between status and environmental context, F(18, 130) = 1.74, p < .04. Analyses of the univariate main effects and interactions yielded findings that were generally consistent with those of Study 1, and additional findings con- cerning the neglected group. A significant main effect for the category of task-appropriate solitary activity was found, F(3, 51) = 3.46, p < .03. Subsequent Newman-Keuls tests revealed that the neglected group engaged in this activity more than each other group (each p < .05). Whereas the order of group means for the popular and rejected groups was the same as in Study 1, the Newman-Keuls test of this difference was not significant. A significant main effect for task-inappropriate solitary activity was also found, F(3, 51) = 3.86, p < .01. The rejected group was found to display more task-inappropriate solitary activity than each other group (by Newman-Keuls test, each p < .05). A significant interaction for this measure, F(3, 51) = 3.15, p < .03, indicated that this effect was present primarily in the work context. A significant main effect for the category of social approaches, F(3, 51) = 4.71, p < .01, and subsequent Newman- Keuls tests revealed that the neglected group made fewer prosocial approaches than did each of the other three groups (each p < .05). This finding is consistent with the perception by peers that neglected children are shy in their peer interactions (Coie et al., 1982). As in Study 1, the popular, average, and rejected groups did not differ overall in the frequency of prosocial approaches. A marginally significant interac- tion between status and environmental context for this category, F(3, 51) = 2.24, p < .10, is reported here because the same interaction effect was significant in Study 1. The pattern of means as shown in Table III is con- sistent with that of Study 1 in that the rejected group made more prosocial approaches than did the popular group in the classroom, whereas on the playground, the reverse was found. Newman-Keuls tests of this pattern of differences were not significant, however. Still, the consistency of this pattern across both studies suggests that it may be a reliable finding.

As in Study 1, a main effect for aggressive acts was found, F(3, 51) = 2.73, p = .05. Newman-Keuls tests demonstrated that the rejected group engaged in more aggression than did any other group (each p < .05). Popular, average, and neglected children did not differ in aggression. A significant interaction between status and context for aggression, F(3, 51)

= 3.03, p < .04, indicated that the differences between the rejected group and the other groups were more pronounced on the playground than

Rejected and Neglected Preadolescents 405

in the classroom. Status groups did not differ in the frequencies of social approaches and aggressive acts directed toward them. Finally, the class- room observations revealed a significant main effect of status for the category teacher-child interactions, F(3, 51) -- 2.91, p < .05. The rejected group interacted with the teacher more than each other group did (by Newman-Keuls test, each p < .05). Given that most interactions with the teacher involved help-seeking or reprimands, this finding supports the hypothesis that rejected children are overly involved in disruptive activities in the classroom. No other main effects or interactions involving status as a factor were found to be significant.

The multivariate main effect of environmental context was also significant, F(6, 46) -- 66.84, p < .001. More solitary activity, both task- appropriate, F(1, 51) = 36.45, p < .001, and task-inappropriate, F(1, 51) = 58.84, p < .001, took place in the classroom than on the playground. Interactive behaviors, including prosocial approaches by the subject, F(1, 51) = 32.54, p < .001, prosocial approaches to the subject, F(1, 51) = 6.87, p < .02, aggressive acts by the subject, F(1, 51) = 19.64, p < .001, and aggressive acts by peers to the subject, F(1, 51) = 3.72, p < .06, were more frequent on the playground than in the classroom. All other multivariate effects, including those for grade level, were found to be nonsignificant.

The valenced responses by peers to the prosocial approaches by subjects were subject to similar Z-score analyses as in Study 1. As Table IV shows, the proportion of prosocial approaches that were rejected by peers in Study 2 was much lower than that in Study 1. The Z-score tests revealed that the proportion of prosocial approaches that were rejected by peers was greater for rejected children (4.0%) than for popular children (.3%), Z -- 3.02, p < .01, and average children (.9%), Z -- 1.98, p < .05. This finding replicates that of Study 1. Also, the proportion of prosocial ap- proaches that were rejected by peers was greater for neglected children (4.6%) than for popular children, Z = 3.47, p < .01, and average children, Z = 2.37, p < .05. Because these proportions are so low, however, caution must be expressed concerning the reliability of these statistics. Still, these data suggest that rejected and neglected children are more likely to be rebuffed in their prosocial approach attempts than are popular and average children.

Teacher Assessments. The three teacher assessment scores were subject to a status by grade level multivariate analysis of variance. Status group means for these analyses are shown in Table III. A significant multivariate main effect of status group was found, F(9, 110) = 10.16, p < .001. Univariate analyses and Newman-Keuls tests revealed that teachers assessed the peer relations of the rejected group as significantly

406 Dodge, Coie, and Brakke

less favorable than those of neglected and average children, which, in turn, were assessed as less favorable than those of popular children, F(3, 51) = 23.64, p < .001, and, by Newman-Keuls tests, each p < .05. Rejected children were viewed by teachers as being involved in more aggression than each other group, F(3, 51) = 11.48, p < .001, and, by Newman-Keuls tests, each p < .05. The popular, average, and neglected groups did not differ on this measure. Finally, teachers judged that neglected children made fewer social approaches than did each other group, F(3, 51) -- 7.33, p < .001, and, by Newman-Keuls tests, each p < .05. Popular, average, and rejected children did not differ in the frequency of social approaches attempted by them, according to teachers. No other multi- variate main effects or interactions were found for teacher assessments.

GENERAL DISCUSSION

In spite of highly divergent subject populations and procedures, the observational data of these two studies were quite similar and are con- gruent with the profiles of these status groups provided in previous studies by peers and in this study by teachers. Although significance levels for status group differences in the two studies varied, the patterns in group means were similar.

In general, the vast majority of interpersonal activity among the children could be termed prosocial, and the majority of social initiations were met with a positive response. The rate of aggressive interactions among the children was quite low. Each of these statements holds true for the rejected sample as well as the other groups. However, the significant difference in aggressive behavior initiated by the rejected children compared to the other two groups was dramatic in terms of the relative percentages (more than double) involved, and this difference is probably a major factor in these children being disliked by their peers.

In comparison to the popular status group, the rejected group spent more time engaged in task-inappropriate solitary activities such as clown- ing, daydreaming, and walking aimlessly around the room. Teachers had to intervene and interact with them more than with any other group. In the classroom, rejected children also made more social approaches toward peers than did popular children. Given that the times selected for the class- room observation were seat work periods, these social approaches can be considered inappropriate for that context. Indeed, peers rejected the social approaches of rejected children in the classroom at a comparatively high rate.

Several explanations for this pattern are plausible. It may be that rejected children have not learned to discriminate the work context from

Rejected and Neglected Preadolescents 407

the play context. Indeed, the behavior of rejected children varied less as a function of the context than did the behavior of popular children. Thus, rejected children may come to be rejected by peers because of the inap- propriateness of the timing of their social overtures to peers. This inap- propriateness in the timing of social initiations may interact with a lack of skillfulness in the nature of the social approaches themselves and thus account for the negative responses that rejected children more frequently receive from peers than do other children. Putallaz and Gottman (1981) have reported findings that support such a conclusion from their studies of social entry behavior among second-graders. Gottman et al. (1975) and Richard and Dodge (1982) have found that popular children are able to think of more and better ways of initiating peer interactions than other children. The present study suggests that these cognitive skill differences are matched by behavioral pattern differences in initiations.

The picture of rejected children spending less time on their assigned academic tasks (solitary-appropriate behavior) and more time attempting to engage others fits with teacher assessments of their relatively poor academic progress. This pattern also is congruent with the findings of Lahey, Green, and Forehand (1980) that unpopular children often are observed by both teachers and trained observers to fit a pattern of hyperactivity and off-task behavior.

Rejected children may perform less well academically because they spend less time at academic tasks and more time in social activity, and do so for one or more of the reasons cited above. Conversely, they may be less engaged with the academic tasks because they have fallen behind academically at earlier grade levels and now lack some of the skills essential to making a reasonable effort with their academic work. Since these academic tasks have thus become neither meaningful nor stimulating to them, they then seek out other forms of stimulation-social stimulation. The paradoxical consequence of this socially oriented classroom activity seems to be a greater degree of social rejection than most other children experience.

In addition to deviant social-approach behavior, rejected children displayed over twice as many peer-directed aggressive acts as popular and average children. These data are consistent with teacher assessments and with the perceptions of peers. Clearly, then, improving the timing and quality of social approaches and reducing the frequency of aggression must be considered essential features of an intervention program designed to improve the peer relations of rejected children. However, an equally plausible argument can be made for an intensive program of academic remediation designed to bring their classroom performance up to a level that enables them to become engaged with the classroom tasks and less likely to become a source of distraction and irritation to their classmates and teachers. Of course, academic remediation could have only an indirect

408 Dodge, Coie, and Brakke

effect on the playground behavior of rejected children, which was also deviant.

The behavior patterns of neglected children differed from those of rejected children but were also different from those of socially well-adjusted children. Neglected boys did not display inappropriate or aggressive class- room behaviors, nor did they socially approach peers at inappropriate times. However, during recess, when social approach is the most efficient way of integrating successfully into the peer group, the neglected children failed to approach peers as frequently as did the average and popular children. This fact matches the impressions of teachers who assessed their social approaches as infrequent. Like those of rejected children, the social approaches of neglected children frequently met with rebuff by peers. It seems clear, however, that the social skills training that is most ap- propriate for rejected children is not the program that is best suited to this neglected group. The data of Study 2 suggest that the latter group could use training and coaching in the initiation of social interactions, whereas the former group needs help in controlling some of its social impulses. In- terestingly enough, some of the recent efforts to train groups of low-status children-groups that very likely included both rejected and neglected children- were based on skill-training programs that involved both types of remedial foci (e.g., participation, cooperation, and perspective-taking) (Gottman et al., 1976; Ladd, 1981; Oden & Asher, 1977). These programs may be somewhat lacking in the kinds of training necessary to reduce the aggressive behavior of rejected children, however.

One must keep in mind that the data base for these speculations has been correlational. We observed children who had already acquired status with their peers as rejected or neglected. Longitudinal studies of the de- velopment or rejected and neglected status over time would establish more firmly the causal role of these behaviors in determining and maintaining deviant peer status. Still, the data reported in these studies do provide a picture of the middle school-aged child that complements one already established for preschoolers (e.g., Charlesworth & Hartup, 1967) and it provides a more adequate empirical basis for the behaviorally oriented social skill training programs appropriate for this older group of children.

REFERENCES

Asher, S., & Hymel, S. Children's social competence in peer relations: Sociometric and behavioral assessment. In J. Wine & M. Smye (Eds.), Social competence. New York: Guilford Press, 1981.

Asher, S. R., Oden, S. L., & Gottman, J. M. Children's friendships in school settings. In L. G. Katz (Ed.), Current topics in early childhood education (Vol. 1) Hillsdale, New Jersey: Erlbaum, 1977.

Bakeman, R. Untangling streams of behavior: Sequential analyses of observation data. In G. P. Sackett (Ed.), Observing behavior (Vol. 2) Baltimore: University Park press, 1978.

Rejected and Neglected Preadolescents 409

Campbell, J. D., & Yarrow, M. R. Perceptual and behavioral correlates of social effective- ness. Sociometry, 1961,24, 1-20.

Charlesworth, R., & Hartup, W. W. Positive social reinforcement in the nursery school peer group. Child Development, 1967, 38, 993-1002.

Coie, J. D., Dodge, K. A., & Coppotelli, H. Dimensions and types of social status: A cross-age perspective. DevelopmentalPsychology, 1982, 18, 557-570.

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

Goldman, J. A., Corsini, D. A., & de Urioste, R. Implications of positive and negative sociometric status for assessing the social competence of young children. Journal of Applied Developmental Psychology, 1980, 1, 209-220.

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

Gottman, J. M., Gonso, J., & Rasmussen, B. Social interaction, social competence, and friendship in children. Child Developmen t, 1975, 46, 709-718.

Gottman, J. M., Gonso, J., & Schuler, P. Teaching social skills to isolated children. Journal ofAbnormal ChildPsychology, 1976, 4, 179-197.

Hartup, W. W. Children and their friends. In H. McGurk (Ed.), Issues in childhood and development. London: Methuen, 1979.

Hartup, W. W., Glazer, J. A., & Charlesworth, R. Peer reinforcement and sociometric status. Child Development, 1967, 38, 1017-1024.

Hymel, S., & Asher, S. R. Assessment and training of isolated children's social skills. Paper presented at the Biennial Meeting of the Society for Research in Child Develop- ment, New Orleans, 1977. (ERIC Document Service Reproduction Service No. ED146 930)

Johnson, S. M., & Bolstad, O. D. Methodological issues in naturalistic observation: Some problems and solutions for field research. In L. A. Hamerlynck & E. J. Mash (Eds.), Behavior change: Methodology, concepts, and practice. Champaign, Illinois: Research Press, 1973.

Ladd, G. W. Effectiveness of a social learning method for enhancing children's social interaction and peer acceptance. Child Development, 1981,52, 171-178.

Lahey, B. B., Green, K. D., & Forehand, R. On the independence of ratings of hyperac- tivity, conduct problems, and attention deficits in children: A multiple regression analysis. Journal of Consulting and Clinical Psychology, 1980, 48, 566-574.

Lippitt, R., & Gold, M. Classroom social structure as a mental health problem. Journal of Social Issues, 1959, 15, 40-49.

McGuire, J. M. Aggression and sociometric status with preschool children. Sociometry, 1973, 36, 542-549.

Moreno, J. L. Who shall survive? Washington, D.C.: Nervous and Mental Disease Publishing, 1934.

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

Putallaz, M., & Gottman, J. M. Social skills and group acceptance. In S. R. Asher & J. M. Gottman (Eds.), The development o f friendship: Description and intervention. New York: Cambridge University Press, 1981.

Richard, B. A., & Dodge, K. A. Social maladjustment and problem-solving among school- aged children. Journal of Consulting and Clinical Psychology, 1982, 50, 226-233.

Robins, L. N. Deviant children grown up. Baltimore: Williams and Wilkins, 1966. Roff, M., Sells, S. B., & Golden, M. M. Social adjustment and persona#ty development

in children. Minneapolis: University of Minnesota Press, 1972. Vandell, D. L., & George, C. B. Social interaction in hearing and deaf preschoolers: Successes

and failures in initiations. Child Development, 1981, 52, 627-635. Woody, R. H. Behavioral problem children in the schools: Recognition, diagnosis, and

behavioral modification. New York: Appleton-Century-Crofts, 1969. Yarrow, M. R., & Waxler, C. Z. Observing interaction: A confrontation with methodology.

In R. B. Cairns (Ed.), The analysis of social interactions. Hillsdale, New Jersey: Erlbaum, 1979.