The social behavior of peer-identified Aggressive, Withdrawn, and Aggressive/Withdrawn children

Journal o f Abnormal ChiM Psychology, Vol. 16, No. 5, 1988, pp. 539-552

The Social Behavior of Peer-Identified Aggressive, Withdrawn, and Aggressive/Withdrawn Children

Judith Lyons , l.2 Lisa A . Serbin,13 and Keith Marchessault I

The behavioral patterns associated with peer ratings o f aggression and withdrawal were explored. First, a discriminant function analysis (N = 74) using seven observational variables was found to significantly identify groups o f Aggressive, Withdrawn, and Contrast fourth- and fifth-grade girls and boys. Aggressive/Withdrawn children were not distinguishable f rom Con- trasts. In subsequent analyses comparing the behaviors o f children in the four groups at two schools (total N = 117), children in the Aggressive and With- drawn groups each showed distinctive patterns o f social behavior, which were consistent across the two schools. The behavior o f the Aggressive/Withdrawn children was not significantly different from that o f Contrast children. However, results from one school suggested that Aggressive~Withdrawn children may receive a disproportionate amount o f aggression from peers. Fi- nally, the behavior patterns displayed by the deviant groups were similar for girls and boys, allowing for sex differences in base rates o f playground behavior. These results confirm the observability o f peer-identified patterns o f aggression and withdrawal, and provide a detailed description o f the behavior o f such children in a free-play situation.

Manuscript received in final form February 22, 1988. This research was supported by grants from the National Health Research and Development Program, Canada, and the F.C.A.R. of the Ministry of Education of Quebec. We are extremely grateful for the cooperation of the Commission des t~coles Catholiques de Montr6al. The authors also gratefully acknowledge the helpful comment on an earlier version of this manuscript made by Drs. Anna-Beth Doyle and Debbie Moskowitz. tDepartment of Psychology, Concordia University, Montreal, Quebec, Canada H3G 1M8. 2Judith A. Lyons is now at the V.A. Medical Center and the University of Mississippi Medical Center, Jackson, Mississippi 39216.

3Address all correspondence to Lisa A. Serbin, Centre for Research in Human Development and Department of Psychology, Concordia University, 1455 de Maisonneuve Boulevard West, Montreal, Quebec, Canada H3G 1M8.

539

00914~627/88/1000-0539506.00/0 �9 1988 Plenum Publishing Corporation

540 Lyons, Serbin, and Marchessault

Internalizing/withdrawal and externalizing/aggression have been factor- analytically identified as the two broad dimensions under which nearly all childhood psychological problems can be classified (Achenbach & Edelbrock, 1978). The research strategy of identifying "high-risk" children on the basis of extreme scores on these dimensions is becoming increasingly popular. Based on evidence that peers are often better informants than adults in identifying classmates who are likely to have continuing difficulties (Bower, 1969; Co- wen, Peterson, Babigian, Izzo, & Trost, 1973; Zax, Cowen, Izzo, & Trost, 1964), the strategy of using peer nominations to identify Aggressive children, Withdrawn children, and children who score high on both dimensions (Ag- gressive/Withdrawn) has been adopted by several research teams.

Several independent studies have used the Pupil Evaluation Inven- tory (PEI; Pekarik, Prinz, Liebert, Weintraub, & Neale, 1976), a group- administered peer-nomination instrument that can be used to screen large populations, to identify groups of Aggressive, Withdrawn, Aggres- sive/Withdrawn, and Contrast (statistically nondeviant) children. Such studies have confirmed that these peer-identified categories represent distinct groups at risk for particular academic, social, and psychological problems (Milich & Landau, 1984; Schwartzman, Ledingham, & Serbin, 1985).

One limitation of this precedure, however, is that peer nominations do not provide a detailed picture of the social behavior patterns actually exhibited by each group. If children in these groups do represent specific at-risk populations, early or preventive intervention strategies would require detailed knowledge of the social behavior patterns of each group, in comparison with the social behavior of nondeviant children of the same age. The behavioral distinctiveness of two peer-identified groups, Aggressive and Aggres- sive/Withdrawn, was first demonstrated by Milich and Landau (1984). However, their study was limited to kindergarten boys in these two groups, and it focused on confirming the distinctiveness of these classification groups, rather than on a detailed description of their behavior.

In a subsequent study, Serbin, Lyons, Marchessault, Schwartzman, and Ledingham (1987) confirmed that children in the peer-identified groups (Ag- gressive, Withdrawn, Aggressive/Withdrawn, and Contrast) were behaviorally distinct. In a discriminant function analysis using four behaviors as predictor variables, 65% of a sample of 43 girls and boys in grades 5 and 6 were accurately identified as belonging to the correct peer-identified group, on the basis of their behavior on the playground. A major limitation of this procedure was that it "capitalized on chance" in selecting the four strongest predictors from a larger pool of available observational measures. In order to confirm the behavioral discriminability of the four peer-identified groups, an independent replication would be required, in which similar observational measures emerged as significant predictors of group membership.

Peer-Identified Children 541

A second important limitation of the initial study by Serbin and her associates was that too few children were included to conduct a detailed examination of behavioral patterns within the four groups, or to examine sex differences in the behavioral patterns found. Further, the study was restrict- ed to children from a single school, where baseline levels of specific types of social behavior may have influenced the patterns identified and thus limited the generality of the findings. To date, a comprehensive examination of the social behavior of school age girls and boys identified by peers as aggressive, withdrawn, or aggressive/withdrawn has not been carried out.

This study was designed to further investigate the behavioral correlates of peer-identified aggression and withdrawal. First, we attempted a replication of the earlier results of Serbin et al. (1987) with a larger sample (N -- 74), to confirm that observational variables can identify members of peer- identified Aggressive, Withdrawn, Aggressive/Withdrawn, and Contrast groups in a new, independent sample. A second goal was to examine in de- tail the patterns of social behavior shown by Aggressive, Withdrawn, and Aggressive/Withdrawn children in a free-play situation. These analyses included both the current sample and the original sample observed by Serbin et al., so that behavioral patterns shown by the four groups could be compared between two schools, and the generality of the findings examined directly. Third, the study examined sex differences in patterns of social behavior within and across specific peer-identified groups. There is little information about aggressive or aggressive/withdrawn girls, in particular, in the litera- ture, and there seem to be no observational studies of such children in a naturalistic setting. Finally, the behavior of peers toward children in the four peer-identified groups was examined. Since peers had initially identified the children in this study as atypical, we expected that they might also display distinct behavior patterns when interacting with children they per- ceived as socially deviant.

M E T H O D

Subjects

Subjects were selected from two French-language elementary schools located in different working-class neighborhoods in Montreal. Subjects were identified on the basis of the administration of a French translation of the Pupil Evaluation Inventory (PEI). The PEI was developed by Pekarik et al. (1976) and measures Aggression/disruption (A) and Withdrawal (W) on the basis of frequency of nominations by classmates. There are 20 items on the


A scale (e.g., "Those who start a fight over nothing") and 9 items on the W scale (e.g., "Those who are too shy to make friends easily"). Each child in the participating grades was asked to nominate, from class lists, up to four classmates who best fit the description given in each item of the PEI. The inventory was administered twice, once for nominations of same-sex classmates and once for nominations of other-sex classmates, so children were selecting from a single-sex list during each administration. Each child's total scores on the Aggression and Withdrawal factors of the PEI were converted to standard scores relative to those of same-sex classmates.

The selection procedures and resulting group scores for the 43 fifth- and sixth-grade children observed in the original study are described in Ser- bin et al. (1987). Similar selection procedures used for the new sample involved identification of Aggressive children in grades 4 and 5 at a second school who scored above the 90th percentile on Aggression and below the 75th percentile on Withdrawal. This yielded 10 boys and 7 girls, with mean percentile scores of 97 on Aggression and 40 on Withdrawal. For the Withdrawn group, the reverse criteria were applied, yielding 9 boys and 7 girls, with mean percentile scores of 97 on Withdrawal and 26 on Aggression. For the Ag- gressive/Withdrawn group, children scoring above the 75th percentile on both dimensions were selected, yielding a sample of 10 boys and 10 girls, with means of 95 on each dimension. Finally, Contrast children were randomly selected from those receiving scores between the 27th and 67th percentile on each dimension. Ten boys and 11 girls were chosen for this group, with mean scores of 38 on Aggression and 44 on Withdrawal. This new sample included a total of 74 children, selected from the total enrollment of 282 children in grades 4 and 5.

Videotaping Procedure

Three graduate students and five undergraduates, paid as research as- sistants, were involved in videotaping the target children on the playground. They were uninformed of the classification group status of the subjects. The film crews positioned themselves in second-story windows overlooking the playground. There, they could view the entire playground yet were minimal- ly intrusive. Children had outdoor recess for 12 to 15 minutes each morning and 10 to 15 minutes each afternoon, weather permitting. Subjects were videotaped in random order, for 2 to 4 minutes at a time, over the school year. The 43 subjects at School 1 were videotaped an average of 3.2 times (SD = 0.6). Approximately 9 minutes of data were collected per subject. At the second school, an increase in funding allowed us to collect an average of 13.7 segments (SD = 0.9) for each of the 74 target children, yielding an average total of approximately 28 minutes of data per subject.


Observational Coding of the Videotapes

Based on pilot work, a quantitative code was devised that focused both on the target child's behavior and on peers' behavior that was directed toward the target. Coders scoring the tapes viewed each video segment four times. On the first pass through the tape, the coder focused on identifying the target child and became familiar with the context of the segment. During the second pass, the duration of three categories was recorded: (1) time spent with a group, (2) time spent in a dyad, (3) time spend alone. Next, on the third pass, the frequencies of nonaggressive touching and of physical aggression (e.g., punch, slap, kick, push, "wash face in snow") were scored. Ag- gression by the target was coded either as "initiated" or "retaliated." Similarly, peer touching of target and peer aggression toward the target were recorded, with aggression coded as "initiated" or "retaliated." Finally, on the fourth pass, coders were asked to rate the target's (1) motor activity, (2) involvement in the ongoing group activity, and (3) solicitation of peer attention, and also to rate (4) peers' degree of involvement with the target. Each of these ratings was made using a 3-point scale (high-medium-low). Each category scored during the fourth pass was rated over a 7-second interval on a rotating basis, so that each of these four categories was rated every 28 seconds.

From this initial set of measures, 12 variables (some based on ratios between two measures) were used in subsequent analyses. These are listed in Table I.

Six coders, ages 22 to 26, took part in the study. Three were male, three female. All were university students. Coders were uninformed of the classification group status of the subjects. MORE data-microprocessing units (Observational Systems, Seattle, Washington) were used in recording coded observations.

Reliability

Each segment of videotape was coded by two randomly paired observers. Interobserver agreement on the coding of each segment was measured using Pearson product-moment correlation coefficients. This measure of agreement was selected because the nature of the variables (including duration measures and 3-point rating scales) and other features of the automated recording procedure did not permit the use of an occurrence-by-occurrence agreement on frequency measure, such as Cohen's Kappa (Johnson & Bol- stad, 1973; Hartmann, 1977; Kent & Foster, 1977). Across categories, the average level of interobserver agreement was .77. For nonaggressive touching, the rate of interobserver agreement was .59 for "target touches" and .63

544 Lyons, Serbin, and Marehessault

Table I. Means (and Standard Deviations) of Observational Variables for the Four Clas- sification Groups

Aggressive/ Aggressive Withdrawn withdrawn Contrast (N = 29) (N = 27) (N = 28) (N = 33)

Duration variables (percentage of total observed) Total time spent 76~ 64~ 73~ 73%

with a group (14) (19) (11) (14) Total time spent with 13~ 17~ 16~ 15%

a single peer (dyad) (10) (11) (8) (10) Total time spent 10070 17~ 11 ~ 11 ~

alone (5) (11 ) (6) (7)

Physical contact variables (mean frequency per minute)

Nonaggressive 1.54 0.90 touching by target (0.58) (0.47)

Aggression incited 0.86 0.43 by target (0.57) (0.41)

Aggression incited 0.45 0.48 by peers (0.22) (0.50)

Percentage of peer-incited aggression that was 41o70 27o70 retaliated by target (25) (23)

Ratio of peer-incited to 0.67 1.41 target-incited aggression (0.55) (1.53)

Rated variables (based on a scale of 0 to 1, rated over 7-second

Target's level of 0.88 0.73 involvement with peers (0.08) (0.19)

Target's level of solicitation of peer 0.70 0.52 attention (0.07) (0.15)

Peers' level of involvement 0.38 0.25 with target (0.11) (0.14)

Ratio of target's level of solicitation to peers' level 1.95 3.64 of involvement with target (0.56) (4.14)

Target's level of 0.48 0.41 motor activity (0.11) (0.11)

1.17 1.15 (0.40) (0.45) 0.56 0.49

(0.44) (0.54) 0.61 0.42

(0.48) (0.38)

28~ 26% (20) (22)

1.87 1.35 (2.56) (1.80)

intervals)

0.85 0.84) (0.07) (0.11)

0.66 0.64 (0.09) (0.10) 0.36 0.35

(0.14) (0.13)

2.14 1.99 (0.94) (0.59) 0.46 0.44 (0.9) (0.08)

Significant school by category interaction:Ratio ofpeer-Incited to target-incited aggression"

School 1 0.66 2.17 3.92 2.26 (0.74) (2.40) (4.43) (2.95)

(N= 11) ( N = 9) ( N = 7) ( N = 10)

School 2 0.67 0.98 1.15 0.91 (0.41) (0.36) (0.81) (0.59)

(N = 17) (N = 16) (N = 20) (N = 21)

~N = I 11 for this analysis (omitting 6 children with zero demoninators in the ratio).


for "touches by peers." Agreement on all other variables ranged from .71 to .87. Because some of these rates were low, implying that more error variance was being added than was desirable, scores for each subject were aver- aged across the ratings made by the two observers, rather than using the raw data from each coder (Moskowitz & Schwarz, 1982).

RESULTS OF DISCRIMINANT FUNCTION

The first goal was to replicate the results of the discriminant function analysis by Serbin et al. (1987), showing that peer nominations on Aggres- sion and Withdrawal scales could be used to identify behaviorally distinct groups. With the new sample of 74 children from School 2, the 7 best predictors (approximately 1 per 10 degrees of freedom) were identified from the 12 observational variables on the basis of changes in Rao's V, a generalized distance function reflecting the degree of separation between groups (Tabach- nick & Fidell, 1983).

The discriminant analysis generated three functions. The first, X 2 = 42.04, df = 21, p < .004, accounted for 57% of the between-group varia- bility and maximally separated the Aggressive and Withdrawn groups. The second function was marginally significant, X 2 = 18.77, df = 12, p < . 10, accounted for 34~ of the variance between groups, and discriminated the Aggressive and Contrast groups. The third function was not significant.

In the present sample, group membership of 54~ of the 74 children was identified correctly (chance rate for a four-way classification system is, of course, 25%). Aggressive subjects had the highest identification rate of 71o70, followed by the Withdrawn and Contrast subjects, at 56~ and 57% correct, respectively. The Aggressive/Withdrawn group could not be discriminated from the Contrast group. Thirty-five percent of the Aggres- sive/Withdrawn group was correctly identified, while another 35~ was incorrectly classified as belonging to the Contrast group.

The variables selected by the equation as best predictors included (1) level of solicitation of peer attention by the target, (2) nonaggressive touching by the target, (3) peer aggression toward the target, (4) ratio between peer- and target-initiated aggression, (5) ratio of target retaliation to peer aggression, (6) time spent alone, and (7) level of peers' involvement with the target. Three of the four variables selected as best predictors at the first school (solicitation of peer attention by the target, ratio between peer and target aggression, and peers' involvement with the target) were once again identified as predictors by the equation at the second school. Only one variable selected at School 1, amount of group play by the target, was not repeated as a significant predictor.


R E S U L T S OF U N I V A R I A T E A N A L Y S E S

Following this multivariate analysis, an examination of differences between the groups on specific behaviors using univariate analyses was conduct- ed. Both samples, School 1 and School 2, were included in these analyses (with school included as a factor) to identify directly the behaviors that dis- tinguish the four groups across both schools. Thus, while there were many main effects of school in these analyses, it was interactions between school and classification group that were noted since these would limit the poten- tial generality of between-group differences. Sex was also included as a factor in these analyses since establishing the generality of behavioral patterns in deviant groups across the two sexes was an important goal of the study. Results are organized to present a description of the behavior of each of the four peer-identified groups. Sex differences are presented separately at the end of the section. Data were analyzed using three-way analyses of variance, with three between-subject factors, Classification group (4), Sex (2), and School (2). Post hoc analyses were done using Tukey tests with a .05 proba- bility as the criterion for significance. Means and standard deviations for each classification group on each variable are presented in Table I.

Aggressive Group

The Aggressive children were generally highly active and very involved in social interaction. The main effect of Classification on amount of time spent with a group (rather than alone or in a dyad) was significant, F(3, 101) = 4.00, p < .01. The four groups fell along a continuum, with the With- drawn subjects lowest and the Aggressive group highest. Post hoc Tukey tests revealed that the two extreme groups (i.e., Aggressive and Withdrawn) differed significantly.

Physical contact was also high for this group. A significant main effect of Classification was found for nonaggressive touching, F(3, 101) -- 9.56, p < .001, with the Aggressive group higher than all other groups. Ag- gressive children also initiated more aggression toward peers. The main effect of Classification on frequency of aggression was significant, F(3, 101) = 5.52, p < .01, with post hoc tests showing that the Aggressive group was significantly more aggressive than the Contrast or the Withdrawn subjects. Interestingly, the Aggressive children were also more apt to retaliate when peers initiated aggression against them (as defined by the percentage of aggression directed toward them to which the Aggressive children retaliated), F(3,100) = 3.34, p < .05, with post hoc tests showing the Aggressive group higher than all other groups.


Finally, the Aggressive group was the most physically active group, with the main effect of Classification on level of motor activity significant, F(3, 101) -- 3.57, p < .05, and post hoc tests showing the two extreme groups (Aggressive and Withdrawn) differing significantly.

No interactions of Sex or School with Classification were found in these analyses. That is, Aggressive children at each school behaved in this way relative to the other groups, and both Aggressive boys and girls showed this pattern in contrast to same-sex children in the other groups.

Withdrawn Children

Children in the Withdrawn group spent more time alone than the others. The main effect of Classification on this variable was significant, F(3, 101) = 5.07, p < .01, as were post hoc comparisons of the Withdrawn group versus all others. The Withdrawn children also stood out on the variables "level of target involvement with peers" and "level of target attempts to solicit peer attention." Main effects of Classification group were significant on both these factors, F(3,101) = 10.04, p < .001 and 19.91,p < .001, respectively, and post hoc analyses showed the Withdrawn group to be significantly different from the three other groups on both variables. Not surprisingly, peers were rated as less involved with these children, F(3, 101) = 5.29, p < .01, with post hoc tests showing the Withdrawn children receiving less attention from peers than all other groups. There were no sex or school interactions in these findings.

Aggressive/Withdrawn Children

As we expected from the results of the discriminant function analyses, the behavior of the Aggressive/Withdrawn group did not stand out on any of the observational measures. They appeared to be intermediate on most variables, and their behavior did not differ significantly from Contrast subjects on any measure. In contrast, the Aggressive/Withdrawn group appeared to stand out in terms of peer aggression toward the target children. A significant effect of Classification was found for the rate of peer aggression initiated toward target subjects, F(3, 101) = 2.99, p < .05. However, although the Aggressive/Withdrawn group was highest on this dimension (with a mean rate of aggression by peers of .61 per minute) and the Contrast group lowest (mean rate of .42 times per minute), the post hoc comparisons were not significant.


When rates of peer-incited and target-incited aggression were compared by means of a ratio, a significant classification effect again indicated the presence of between group differences, F(3, 95) = 4.85, p < .01. The Ag- gressive/Withdrawn group received almost twice as many aggressive initiations as they gave, whereas the Aggressive group received only about half as much aggression initiated by peers toward them as they initiated toward others. Post hoc comparisons showed this difference to be significant. This pattern was consistent across both sexes. However, a significant interaction between school and category was found, F(3, 95) = 2.77, p < .05 (see bot- tom of Table I). Post hoc tests confirmed that the pattern was stronger at School 1, where peers attacked Aggressive/Withdrawn targets almost four times as often as these children attacked peers (mean ratio = 3.92). At School 2, the Aggressive/Withdrawn group's ratio of peer to target aggression was not significantly different from those of other groups. In this setting, rates of peer aggression toward targets were consistently low and less variable than at School 1. Descriptively, at School 2 there was less aggression because teachers required that all children participate in organized games within their classroom groups (e.g., dodgeball, jump rope). These groups were closely supervised, leaving fewer opportunities for physical conflict. Even within this restrictive setting, Aggressive children initiated almost twice as much aggression as peers initiated toward them (mean ratio = .67), while Aggres- sive/Withdrawn children attacked peers slightly less often than peers attacked them (mean ratio = 1.15).

Sex Differences

There were many significant main effects of sex. In particular, amount of nonaggressive touching, amount of aggression incited by the target, amount of aggression incited by peers toward the target, and level of motor activity all showed significant sex effects, with boys' scores higher than girls (see Ta- ble II). There were no significant interactions with group on these variables, but it is noteworthy that the magnitude of the sex difference in aggression was actually higher within the Contrast group, where boys engaged in three times more aggression than girls, than it was in the three deviant groups. Thus, the sex difference in aggression was not due primarily to the contribu- tion of a subgroup of highly aggressive boys.

The only significant interaction between Sex and Classification Group was in the rate of peer retaliations to target initiations of aggression, F(3, 95) = 4.16, p < .01. This percentage was significantly lower for Contrast girls than for girls in the three deviant groups. Contrast girls received retaliations for only 14% of their initiations of aggression, while rates of peer retaliation were between 30 and 39~ for the girls in the deviant groups. For

Tab

le 1

1. S

ex a

nd C

lass

ific

atio

n G

roup

Mea

ns f

or V

aria

bles

sho

win

g S

igni

fica

nt M

ain

Eff

ects

or

Inte

ract

ions

wit

h Se

x

Agg

ress

ive/

A

ggre

ssiv

e W

ithd

raw

n w

ithd

raw

n C

ontr

ast

Tot

al

Boy

s G

irls

B

oys

Gir

ls

Boy

s G

irls

B

oys

Gir

ls

Boy

s G

irls

(N

=

16)

(N =

13

) (N

=

15)

(N =

12

) (N

=

14)

(N =

14

) (N

=

16)

(N

= 17

) (N

=

61)

(N

= 56

)

Mai

n ef

fect

s o

f se

x

Non

aggr

essi

ve

1.63

1.

43

1.06

0.

70

1.25

1.

09

1.14

1.

16

1.27

1.

10

touc

hing

by

targ

et

(0.6

7)

(0.4

5)

(0.4

6)

(0.4

1)

(0.4

0)

(0.4

0)

(0.4

5)

(0,4

5)

(0.5

5)

(0.4

9)

(rat

e pe

r m

inut

e)

Agg

ress

ion

inci

ted

1.06

0.

61

0.60

0.

22

0.75

0.

37

0.77

0.

24

0.80

0.

35

by t

arge

t (0

.64)

(0

.34)

(0

.47)

(0

.19)

(0

.46)

(0

.33)

(0

.67)

(0

.16)

(0

.58)

(0

.29)

(r

ate

per

min

ute)

A

ggre

ssio

n in

cite

d 0.

51

0.38

0.

60

0.34

0.

85

0.36

0.

58

0.26

0.

63

0.33

by

pee

rs

(0.2

4)

(0.1

9)

(0.4

8)

(0.5

0)

(0.5

2)

0.28

) (0

.45)

(0

.22)

(0

.44)

(0

.30)

(r

ate

per

min

ute)

L

evel

of

mot

or

0.49

0.

46

0.45

0.

35

0.51

0.

42

0.46

0.

42

0.48

0.

41

acti

vity

(0

.07)

(0

.14)

(0

.10)

(0

.11)

(0

.08)

(0

.09)

(0

.09)

(0

.07)

(0

.08)

(0

.11)

(r

ated

fro

m 0

to

1)

Per

cent

age

of t

arge

t-

inci

ted

aggr

essi

on

that

was

ret

alia

ted

by p

eers

Inte

ract

ion:

Sex

•

clas

sifi

cati

on g

roup

20%

30

%

29%

31

%

33%

39

~ 35

%

14%

29

%

28%

(0

8)

(18)

(2

3)

(29)

(2

2)

(32)

(2

4)

(14)

(2

0)

(25)


the boys~ the lowest rate of peer retaliation was 20~ for the Aggressive group, while the other three groups ranged from 29 to 35~ In sum, Contrast girls seem to be relatively ignored when they aggress against peers, while deviant girls, and boys in general, receive an aggressive retaliation on about one- third of the occasions when they aggress. There were no significant interactions of School X Sex, or of School X Sex X Classification Group.

DISCUSSION

A primary goal of this study was to examine descriptively the behavior patterns displayed in a "natural" setting by children in specific deviant groups. Most previous studies of this type (including both studies based on peer nominations and those using other criteria for identifying aggressive and withdrawn children) have focused on preschoolers or the early primary grades. It was especially urgent to observe older children because the stability of peer nominations in children below third grade has been shown to be relatively weak, especially for the dimension of withdrawal (Moskowitz, Schwartzman, & Ledingham, 1985). For this reason, it was important to compare and contrast the behavior patterns displayed by children in the four classification groups at an age when peer evaluation of these dimensions has stabilized. The descriptions of each group reported here provide the first actual description of peer-identified children in these four groups, at an age when peer perceptions of the dimensions of aggression and withdrawal are well defined and stable. Distinct behavioral profiles emerged for the Aggressive and With- drawn groups, which were consistent across the two schools and across both sexes.

In contrast, the results for the Aggressive/Withdrawn children failed to reveal any clear behavioral pattern. This is especially puzzling, consider- ing that previous studies have suggested that this group may be the most deviant on a variety of indices, and may be at extremely high risk for psychopathology and other difficulties (Ledingham & Schwartzman, 1984; Masten, Morison, & Pellegrini, 1985; Michael, Morris, & Soroker, 1957; Milich & Landau, 1984; Robins, 1979; Schwartzman et al., 1985). The behavior of peers toward these children, especially at School 1 where there was less adult-imposed structure and supervision, does suggest that there may be some aspect of their behavior that is provocative. They may initiate con- tacts in a manner that is inappropriate or immature (e.g., interrupting the ongoing activity, avoiding eye contact, or being verbally provocative). Previ- ous reports have indicated that Aggressive/Withdrawn children are less liked, and have fewer mutual friends, than other children (Feltham, Doyle, Schwartzman, Serbin, & Ledingham, 1985). The high-risk status of this group


does suggest that they are lacking social skills, and that the peer hostility observed toward these children may play a role in the development and maintenance of problems in this group.

The sex differences observed confirm widely reported findings that boys are generally more physically active and aggressive in free-play situations. A new and potentially important finding, however, is that the behavior patterns of highly aggressive or withdrawn girls are very similar to those of their male counterparts, allowing for the sex differences in base rates of these behaviors. Interestingly, the rates of aggression for girls in all groups, while lower than for boys, were still quite high, averaging once every 3 to 4 minutes. Clear- ly, the playground is a good setting for studying girls' aggression, which occurs relatively rarely in classroom or laboratory settings.

In sum, the present findings support the validity of the Pupil Evalua- tion Inventory as an instrument for identifying Aggressive and Withdrawn children, and provide a detailed picture of the behavior of girls and boys in these groups. They also confirm the behavioral similarity between the sexes in the ways in which the groups are distinct behaviorally. Peers seem to report and interact somewhat differently with children in each group, which may contribute to the establishment and maintenance of specific problems. Fur- ther research will be necessary to confirm the risk status of these groups for specific forms of later psychopathology, and to establish more clearly the conditions under which Aggressive/Withdrawn children may receive relatively more aggression f rom peers. Meanwhile, it is clear that it is possible to identify specific patterns of social behavior associated with peer-identified deviant groups. This information should eventually be useful in designing preventive and early intervention strategies.

R E F E R E N C E S

Achenbach, T. M., & Edelbrock, C. S. (1978). The classification of child psychopathology: A review and analysis of empirical efforts. Psychological Bulletin, 85, 1275-1301.

Bower, E. M. (1969). Early identification of emotionally handicapped children in school (2nd ed.). Springfield, IL: Charles C Thomas.

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

Feltham, R. F., Doyle, A. B., Schwartzman, A. E., Serbin, L. A., & Ledingham, J. E., (1985). Friendship in normal and socially deviant children. Journal of Early Adolescence, 5, 371-382.

Hartmann, D. P. (1977). Considerations in the choice of interobserver reliability estimated. Journal of Applied Behaviour Analysis, 12, 103-116.

Johnson, S. M., & Bolstad, O. D. (1973). Methodological issues in naturalistic observation: Some problems and solutions for field research. In L. A. Hamerlynck, L. C. Handy, & E. J. Mash (Eds.), Behavior change: Methodology, concepts, and practice (pp. 7-68). Champaign, IL: Research Press.


Kent, R. N., & Foster, S. L. (1977). Direct observational procedures: Methodological issues in naturalistic settings. In A. R. Ciminero, K. S. Calhoun, & H. E. Adams (Eds.), Hand- book of behavioral assessment (pp. 279-328). New York: Wiley.

Ledingham, J. E., & Schwartzman, A. E. (1984). A 3-year follow-up of aggressive and withdrawn behavior in childhood: Preliminary findings. Journal of Abnormal Child Psychology, 12, 157-168.

Masten, A. S., Morison, P., & Pellegrini, D. S. (1985). A revised class play method of peer assessment. Developmental Psychology, 21, 523-533.

Michael, C. M., Morris, D. P., & Soroker, E. (1957). Follow-up studies of shy, withdrawn children II: Relative incidence of schizophrenia. American Journal of Orthopsychiatry, 27, 331-337.

Milich, R., & Landau, S. (1984). A comparison of the social status and social behavior of aggressive and aggressive/withdrawn boys. Journal of Abnormal Child Psychology, 12, 277-288.

Moskowitz, D., Schwartzman, A. E., & Ledingham, J. (1985). Stability and change in aggression and withdrawal in middle childhood and adolescence. Journal of Abnormal Psy- chology, 94, 30-41.

Moskowitz, D. S., & Schwarz, J. C. (1982). Validity comparison of behavior counts and ratings by knowledgeable informants. Journal of Personality and Social Psychology, 42, 518-528.

Pekarik, E. G., Prinz, R. J., Liebert, D. E., Weintraub, S., & Neale, J. M. (1976). The Pupil Evaluation Inventory: A sociometric technique for assessing children's social behavior. Journal of Abnormal Child Psychology, 4, 83-97.

Robins, L. N. (1979). Follow-up studies. In H. C. Quay & J. S. Werry (Eds.), Psychopatho- logical disorders of childhood (2nd ed., pp. 483-513). New York: Wiley.

Schwartzman, A. E., Ledingham, J., & Serbin, L. A. (1985). Identification of children at risk for adult schizophrenia: A longitudinal study. International Review of Applied Psychology, 34, 363-380.

Serbin, L. A., Lyons, J. A., Marchessault, K., Schwartzman, A. E., & Ledingham, J. E. (1987). Observational validation of a peer nomination technique for identifying aggressive, withdrawn, and aggressive/withdrawn children. Journal of Consulting and Clinical Psychol- ogy, 55, 109-110.

Tabachnick, B. G., & Fidell, L. S. (1983). Using multivariate statistics. New York: Harper & Row. Zax, M., Cowen, E. L., Izzo, L. D., & Trost, M. A. (1964). Identifying emotional disturbance

in the school setting. American Journal of Orthopsychiatry, 34. 447-454.

Documents

The social behavior of peer-identified Aggressive, Withdrawn, and Aggressive/Withdrawn children