The friendships and play partners of lonely children - SAGE Journals

139 downloads 28 Views 128KB Size Report
Pamela Qualter. University of Central Lancashire. Penny Munn. University of Strathclyde. ABSTRACT. The research examines the play partners and friendship.
05 qualter (ds)

18/5/05

3:22 pm

Page 379

The friendships and play partners of lonely children Pamela Qualter University of Central Lancashire

Penny Munn University of Strathclyde

ABSTRACT

The research examines the play partners and friendship patterns of a sample of children (N = 409) from four separate clusters (lonely, rejected, lonely and rejected, and nonlonely/nonrejected controls). Controls were more likely than rejected, lonely and lonely/rejected children to have a best friend. Control and rejected children differ from lonely and lonely/rejected children by reporting significantly more support from a close friend. Observations of play showed that lonely, lonely/rejected and rejected children are not isolated and dyadic interactions between children in these three clusters were likely to be positive in nature, involving prosocial acts, and acts that were positively received by others. We note that there are discrepancies between children’s reports of friendships and observations of play partner choice and we consider why this might be the case. KEY WORDS:

children • friendships • loneliness • observational analysis • peer relationships • play partners • sociometry

There are two major theories that have gained prominence in the research on adult loneliness: the cognitive process approach (Peplau, Miceli, & Morasch, 1982; Peplau & Perlman, 1979) and the social needs theory (Weiss, 1973). There are conceptual differences between the two and it would seem that much of the research effort in the area of childhood loneliness has taken the cognitive process perspective. This presumes that loneliness is an emotional response experienced because of social deficits (Maragoni & Ickes, 1989). The assumption of this theory is that social dissatisfaction invariably leads to feelings of loneliness, but critiques of this

All correspondence concerning this article should be addressed to Pamela Qualter, Department of Psychology, University of Central Lancashire, Preston, PR1 2HE, UK [e-mail: [email protected]]. Stanley O. Gaines was the Action Editor on this article. Journal of Social and Personal Relationships Copyright © 2005 SAGE Publications (www.sagepublications.com), Vol. 22(3): 379–397. DOI: 10.1177/0265407505052442

05 qualter (ds)

380

18/5/05

3:22 pm

Page 380

Journal of Social and Personal Relationships 22(3)

model argue that this is not always the case (Rook, 1984, 1988; Shute & Howitt, 1990; Wood, 1986; Young, 1982). So, current theories of childhood loneliness have been derived from research that has taken the cognitive process approach to the study of childhood loneliness, and in doing so, have proceeded on the unexamined assumption that social loneliness (the physical absence of others or peer rejection) and emotional loneliness (the psychological absence of others) are the same thing. Weiss’ model may be more relevant for conceptualizing childhood loneliness. In reports of loneliness (e.g., Hayden, Taruili, & Hymel, 1988), children make reference to lack of companionship and sense of belonging and to ‘being left out of things’, but they also talk about a lack of emotional support and affection and of having ‘no-one to share your private thoughts with’. These ideas are reminiscent of Weiss’ (1973) theoretical distinction between social and emotional loneliness. Weiss’ writings imply that ‘social loneliness’ refers to ‘aloneness’, meaning the physical absence of other people. In contrast, ‘emotional loneliness’ stems from the absence of a close attachment. ‘Social loneliness’, then, does not refer to a symbolic, psychological, or emotional construct, and writers have looked at how the ‘loneliness of social isolation’ refers to notions of spending time alone, and of not spending time with others (DiTommaso & Spinner, 1997). ‘Social loneliness’ is about a lack of social networks, whereas ‘emotional loneliness’ is about a lack of attachment and this can be either ‘felt’ or real. Weiss (1973, 1989) and others (see DiTommaso & Spinner, 1997) argue that ‘emotional loneliness’ can only be alleviated by a satisfying attachment relationship. This distinction between social and emotional loneliness is supported by research on loneliness among adults (e.g., Russell, Cutrona, Rose, & Yorko, 1984). Until recently, however, formal assessment of loneliness among children has not examined this distinction (Hoza, Bukowski, & Berry, 2000; Qualter & Munn, 2002; for an exception, see Hayden, 1989). Hoza et al. (2000) developed a questionnaire that assessed emotional (feelings) versus social (aloneness/rejection) loneliness. As predicted, it was found that social loneliness (questions relating to experiences of social rejection) related to lack of group acceptance, whereas emotional loneliness was more highly related to a lack of a best friendship (or low-quality friendship). Weiss’ distinction is a useful one: in our earlier research (Qualter & Munn, 2002), we did not use his terms, but his ideas. For the purposes of our work, we used ‘rejection’ to mean one version of social loneliness and ‘loneliness’ to mean emotional loneliness. We investigated whether children were ‘rejected’, ‘lonely’ or both, and we investigated the particular behavioural functioning and thinking associated with each of these groups. In the Qualter and Munn (2002) study, differences were found between the groups on direct observation measures of solitariness, sociability, and aggression; peer reports of shyness, aggression, prosocial behaviour, disruptive behaviour and inability to take teasing; self-reports of self-worth and competence, self-reports of supportive relationships; and measures of language use. The results indicate that it is loneliness (emotional loneliness) and not rejection (a version of social loneliness) that co-occurs with low

05 qualter (ds)

18/5/05

3:22 pm

Page 381

Qualter & Munn: Lonely children

381

self-worth, nonself-serving attributional style, and inaccurate perceptions of self, solitariness and lack of sociability. Criticisms can be raised regarding our earlier study: we had information about how children felt about their relationships, but we only had observational data on their actual interactions. So, we did not have information about number of actual friends, the quality of these friendships, and play partner choice on the playground. Thus, we did not have a complete objective picture of what lonely and rejected children’s actual interactions with peers and friends were like. In order to extend the theory of social and emotional loneliness in childhood, it is important to investigate the friendships and social exchanges of socially and emotionally lonely children and, in this study, we attempt to do this. Investigating the friendships and play partners of socially lonely and emotionally lonely children Research shows that having a mutual relationship with a peer and being popular (being liked and accepted by member’s of one’s peer group) have independent effects on a child’s feelings of self-worth (Bukowski & Hoza, 1989). Indeed, feelings of self-worth were significantly predicted by the experience of having at least one mutual friend within the class, even after the effects of overall popularity were statistically removed (Bukowski & Newcomb, 1987), and it is possible that having at least one mutual friend may alleviate feelings of emotional loneliness (originally suggested by Cohn, Lohrmann, & Patterson, 1985). Given also that lonely children report a lack of emotional support provided within the family (Qualter & Munn, 2002), peer friendship may be a good, if not the only, source of emotional security for lonely children. Having someone with whom a child can either play or confide in may be crucial in determining how a lonely child copes with a particular stressor or with his or her loneliness in general. Thus, the study of lonely children’s friendships is important, particularly in the light of possible intervention programmes. Assessing friendships is important for a study on lonely children. A single best friend can serve as a buffer from the many negative outcomes associated with peer rejection and emotional loneliness (Howes, 1988; Stocker, 1994). Parker and Asher (1993) reported that approximately 54% of the low-accepted children in their study did have a best friend, and that those with best friends did seem to be buffered from loneliness. In addition, lowaccepted children tended to be friends with other low-accepted children. Furthermore, not only the existence of a friendship, but also the quality of the friendship, predicted a child’s loneliness. Indeed, not all friendships are alike, and some relationships will be characterized by conflict, and a lack of intimacy and support. Thus, for some functions, such as the provision of emotional support, the qualities of an individual’s friendships may be more important than whether that individual has a friend with whom they play at playtime. Hartup (1996) argued that the identity (characteristics) of one’s friends is important when addressing the functions of children’s friendships. Therefore,

05 qualter (ds)

382

18/5/05

3:22 pm

Page 382

Journal of Social and Personal Relationships 22(3)

we wanted to see who socially lonely (to use our term, rejected) and emotionally lonely (lonely) children tended to play with. So, in this study, we were interested in whether friendship varies according to whether one is rejected or lonely, but we also asked the question: Who are the rejected and lonely children’s friends and play partners? This question refers to the categorization of children’s friends as rejected or lonely and not to personality characteristics. This concept of ‘identity’ then, is bound up in children’s conceptions of themselves. With regard to the first question, there is no research looking at who are the friends and play partners of emotionally lonely children. The second of these two questions has been asked before in relation to rejected children, but the data collected were based on a different way of identifying rejected children. In previous research, children were grouped into the sociometric groups of popular, average, rejected, neglected, and controversial children using computerized raw scores from sociometric testing. In this research, aggressive–rejected children were often seen to play with each other (Ladd, 1983) and we wondered if similar findings would result when a different way of identifying rejected children (i.e., through a cluster analytic solution) was used. Measures of friendship and play partner choice Almost any research paradigm for investigating children’s friendships and interactions within them depends upon the selection of children who are friends, and there are different ways in which to do this. Some researchers operationalize friendship as mutual choice from the sociometric interview (for example, Ashley, 1992; Foot, Chapman, & Smith, 1977; Margarita, Tur-Kaspa, & Most, 1999), whereas others (Marshall & McCandless, 1957; Ramsey, 1995) observe children during play to see with whom they played. This raises the immediate problem that children who identify each other verbally as friends may not necessarily interact with each other any more frequently than with unselected children. Given that the evidence is confusing concerning whether children’s sociometric choices generally correspond with preferences manifested in actual play associations (Berndt, 1996; Biehler, 1954; Chapman, Smith, Foot, & Pritchard, 1979; Gottman, 1977; Ramsey, 1995), can we be sure that both methods are assessing the same thing? The earlier reports of low concordance suggest that these two approaches may tap different aspects of friendship and underscore the need to utilize different methods to provide a more complete picture of friendship. In this study, then, both self-report (sociometry) and observational measures of friendship were utilized to avoid the problems of a partial assessment of children’s friendship. Goals The questions we address in this article are: 1. Do lonely (emotionally lonely) and rejected (socially lonely) children have friends at school?

05 qualter (ds)

18/5/05

3:22 pm

Page 383

Qualter & Munn: Lonely children

383

2. Is it the case that having at least one mutual friend alleviates feelings of emotional loneliness? 3. Who are the friends and play partners of rejected and lonely children? Method Participants A total of 409 children (205 females and 204 males) aged between 5 and 8 years participated in the research. There were 47 children from reception classes (ages 4–5); 140 children from Year 1 (5–6 years old) and 222 from Year 2 (6–8 years old). They made up the full population of their age levels in seven different schools (a mix of rural and urban and local authority and private) in the north west of England (with the exception of two children whose parents did not give consent for them to take part in the study). The mean age of the sample was 76 months (6 years and 4 months) with a range from 50 to 96 months. The classification of the original sample The children who participated in this study included a sample of the children from the clusters (N = 846) identified in earlier research (Qualter & Munn, 2002). The children from the original sample that took part during the current study were from schools that agreed to take part in ongoing research, and so represent an opportunity sample. Sociometric measure. We used sociometry to derive a popularity index for each child. Each child was asked to pick children from their class photograph that they like and do not like to play with. They were allowed to pick as many children from the class list as they like. This allowed for the fact that some classes may be more sociable than others. To permit comparison of number of nominations across classrooms that varied in size, a proportion score was computed for each child (i.e., the number of nominations divided by the number of classmates contributing to the sociometric data), and the proportion score was standardized for each classroom. We used peer acceptance as composed of two continuous variables (standardized liked and disliked nominations) rather than as a categorical variable. Interview measure of loneliness. We used the Parent and Peer Related Loneliness Questionnaire (Marcoen & Brumagne, 1985) as it focuses on school and parent related loneliness, as well as assessing boredom, and children’s desire to be alone. The measure was revised here to allow statements to be asked in a more suitable format for young children. Research by Marcoen et al. (1987) has demonstrated the psychometric integrity of this questionnaire with its high internal consistency and construct validity. We found satisfactory test–retest reliability (N = 117) after 3 months (r = 0.78, p < .01). Only the peer-loneliness scores were placed into the cluster analysis. Teacher rating Items from section one of the Classroom Adjustment Rating Scale (formally the Connors Teacher Rating Scale) (CARS; Connors, 1969) were completed by the teacher to assess the child’s behavioural problems. Section one of the CARS consists of 41 behavioural-oriented items describing school adjustment

05 qualter (ds)

384

18/5/05

3:22 pm

Page 384

Journal of Social and Personal Relationships 22(3)

problems. The other sections from the CARS (which relate to family background and the teacher’s feelings about the child) were not used here. This measure is among the few recommended teacher assessments of behaviour because it produces factors similar to those found for peer behavioural assessment measures (Rubin, LeMare, & Lollis, 1992). The cluster analysis As we (Qualter & Munn, 2002) did not have a particular number of clusters in mind, we conducted hierarchical cluster analysis on two random subsets of cases (n = 300 per sub-sample). We then used these results (e.g., dendrograms) to specify the number of clusters that we entered into the K-means cluster analysis. We found that the two replicable sub-samples showed a high degree of similarity (see Qualter & Munn, 2002) and that four clusters made theoretical sense. All children in the sample were then assigned to one of the clusters on the basis of having a smaller distance to one cluster centre than to the centres of any other cluster (K-Means Cluster Analysis). The clusters were internally validated and externally validated by Qualter and Munn (2002). Control: Cluster A. Not lonely; well accepted by peers; well adjusted. This group consisted of 374 children (173 males and 201 females; 58.5% of the sample; M age = 76 months, SD = 9.4). These children saw themselves as low on school-related loneliness (M = 8.9) and were well liked by their peers (liked score: M = 3.3; disliked score: M = 2.0). They had low teacher-rated externalizing (M = 7.5) and internalizing maladjustment (M = 8.7). Rejected: Cluster B. Not lonely; not accepted by peers; high on externalizing and low on internalizing behaviour. This second group consistent of 60 children (53 males and 7 females; 9.4% of sample; M age = 75 months, SD = 10.3). These children report low levels of school-related loneliness (M = 7.9) and were not well liked by their peers (M = 2.7); in fact, they were the most disliked group of children in their class (M = 7.9). They were reported by teachers as showing a high level of externalizing difficulties (M = 17.8 ) and low levels of internalizing difficulties (M = 10.1). Lonely/rejected: Cluster C. Lonely; not accepted by peers; high on externalizing and high on internalizing behaviour. This group contained 61 children (35 males and 26 females; 9.5% of sample; ages: M = 76 months, SD = 9.6). These children perceived themselves to be lonely at school (M = 18.9), and their peers disliked them (disliked score: M = 5.6; and liked score: M = 1.6). Their teachers rated them as demonstrating extremely maladjusted externalizing (M = 20.4) and internalizing behaviour (M = 13.5). Lonely: Cluster D. Lonely; well accepted by peers; high on internalizing and low on externalizing behaviour. The final group contained 145 children (58 males and 87 females; 22.6% of sample; ages: M = 73 months, SD = 9.6). These children felt most lonely at school (M = 19.4). They were well liked by their peers (liked scores: M = 2.5; and disliked score: M = 2.5). They were rated by teachers as demonstrating a high level of internalizing maladjustment (M = 13.9). In addition, they also obtained low scores on teacher reports of externalizing behaviour (M = 8.1).

05 qualter (ds)

18/5/05

3:22 pm

Page 385

Qualter & Munn: Lonely children

385

The clusters used in the current study In the current sub-sample, there were 248 children from the control group, 41 children from the rejected cluster, 41 children from the lonely/rejected cluster and 79 children from the lonely cluster. Comparisons of the sub-sample and whole sample gender proportions among the four clusters using a chi-squared analysis revealed no significant differences (␹2 (3, N = 409) = 3.1, p > .05). The data indicate that there are more males (89%) and fewer females (12%) in the rejected cluster and more females (60%) than males (40%) in the lonely group, which is consistent with the gender proportions found in the total population. A comparison of the clusters regarding children’s grade for the sub-sample and whole sample was carried out using a Chi-squared analysis. The results revealed no significant differences (␹2 (6, N = 409) = 9.7, p > .13.) demonstrating that the grade proportions of the sub-sample are consistent with those of the total population. Measures used to study friendship, friendship quality and play partners Mutual friendship choice. Sociometric testing was completed again for the subsample of children involved in the current research, using a similar procedure to that noted earlier: the difference is simply that rather than pick children they like and dislike from the school photograph, children are asked to pick children they consider to be their friends. Mutual friendships were identified whenever two children nominate each other as their ‘friend’. Quality of friendship. The Social Support Scale/’People in My Life’ Questionnaire (Harter, 1985) looks at four possible sources of social support: (a) parents, (b) teachers, (c) classmates, and (d) close friends. Only section (d) was used in this study. The child is first asked to decide which of two children described is most like them. The child is then asked whether this is only sort of true/a little bit true or really true of them. (e.g., ‘Some kids have a close friend who they can tell problems to BUT other kids don’t have a close friend who they can tell problems to’.); Harter (1985) reports internal consistency reliabilties ranging from .72 to .88 across subscales for children in grades 3 to 8. Internal consistency for section (d) close friends for our sample was .82. Observations of play associations. Observations were used to examine the play partners and associations of children from the separate clusters. Instances where children were allowed to mix with whom they liked were observed rather than having the choice made for them by teacher-directed seating plans. So, children were observed during break or lunchtime. Within 2 weeks of completion of sociometric interviews, unobtrusive observations (taken from a classroom with a clear view of the playground) were recorded for the children. Each child was observed for a 5-minute period on four separate occasions by the researchers. An auditory cue signalled the observer(s) to code the child’s behaviour at 15-second intervals, using instantaneous time. After the 5 minutes of observation was completed for a particular child, the rater observed the next child according to a pre-arranged schedule (children were not observed in the same order on each of these four occasions). A total of 20 minutes were recorded for each child. Thus, a total of 80 15second episodes were collected for each of the 409 children. The child’s play partners were recorded. The pictures used during the sociometric testing were

05 qualter (ds)

386

18/5/05

3:22 pm

Page 386

Journal of Social and Personal Relationships 22(3)

used by observers to identify children’s playmates (by code number) during the observations. It was assumed that the child had the same play partner for a whole 15-second segment, and children involved in interactions with more than one cluster were discounted from the analyses. The total number of observed episodes not considered for the analyses represented 3% of the total number of episodes observed. To provide reliability estimates for the coding procedure of play partner, an undergraduate student who was unfamiliar with the hypotheses of the study was trained as an additional independent coder for 15% of the selected subsample. Inter-rater reliability was calculated using the Kappa coefficient with the reliability rating for this category being .87. Coding the ‘quality’ of play partner interactions. Interactions and behaviours were coded into one of ten categories. The observer(s) placed the child into one category at the sample point, and this would be the category that was most representative of the child’s behaviour/interaction. Interactions were coded as follows: (1) alone – the child is simply standing by him/herself and is uninvolved in play or interaction or is standing watching others play; (2) playing alone – the child is playing by him/herself away from peers; (3) initiates interaction, positively received by others – the child makes an overture for social contact that is positively received by others; (4) initiates interaction, negatively received by others – the child makes an overture for social contact that is negatively received by others; (5) other initiates interaction, positively received by focal child – a peer makes an overture for social contact that is positively received by the focal child; (6) other initiates interaction, negatively received by focal child – a peer makes an overture for social contact that is negatively received by the focal child; (7) in process of interacting – the child engages in prosocial interactions with his/her peers; (8) in process of interacting, but alone – the child is engaged in interactive play but is not actually involved in direct social contact (e.g., playing football but standing on the sidelines waiting for the ball); (9) negative/aggressive interaction – the child engages in interaction that is either verbally or physically hostile (e.g., name calling or nonplayful hitting or kicking); and (10) uncodable – behaviour does not fall into one of the nine categories outlined above. Inter-coder reliability (Kappa coefficient), on 15% of observations ranged from .79 (for category 8 – in process of interacting but alone) to .98 (category 1 – alone). Data from the different codes were later summed together to provide categories of positive, negative and neutral interaction. Such categories represent the quality of interactions that would later be analysed. Positive interactions were calculated as the total number of episodes during which the child engaged in the process of interacting (7), initiated interaction that was then positively received (3), and other child initiated interaction that was positively received (5). Negative interactions were calculated as the total number of episodes spent in negative aggressive interactions (9), initiated interaction that was negatively received (4), and other initiated interaction that was negatively received (6). Neutral interactions were calculated as the total number of episodes engaged in the process of interacting, but alone (8). Procedure Several months after the initial study, which was used to cluster children, we went back into schools to collect data for the current study. The children

05 qualter (ds)

18/5/05

3:22 pm

Page 387

Qualter & Munn: Lonely children

387

completed sociometric testing again, to derive information about friendships, and completed the friendship quality questionnaire. We also collected observational data over the course of several months.

Results Mutual friendship nominations The average number of mutual friends ranged from 1.36 for the control cluster to 0.6 for the lonely/rejected cluster members. The rejected cluster had an average of one mutual friendship, as did those from the lonely cluster. The ANOVA indicated that mutual friendship nominations differed significantly across clusters, F(3,408) = 3.5 p < .01. Follow-up Tukey’s HSD post-hoc paired comparisons indicated that control cluster members were significantly more likely to be chosen as a mutual friend than children from the lonely/rejected cluster. No other post-hoc comparisons reached significance. Quality of friendships An ANOVA was conducted with cluster membership and number of mutual friendships as independent variables and the score on the close friend subscale of the Social Support Scale/ People in My Life Questionnaire as the dependent variable. The ANOVA revealed a significant main effect of cluster (F(3,196) = 14.3, p < .001), but no effect of number of mutual friendships (F(3,196) = 1.6, p = .39). Follow-up Tukey’s HSD post-hoc paired comparison for cluster indicated that the control and rejected clusters reported significantly higher levels of support from a close friend than lonely/rejected and lonely children. The ANOVA revealed no significant interaction between number of mutual friends and cluster group on the reported level of support from a close friend (F(8,195) = .4, p = .94). Thus, it is not the case that those in the lonely cluster who have no mutual friend report less support than lonely children who have one or more mutual friends, but it is the case that lonely and lonely/rejected children (whether they have a best friend or not) report less support from a close friend than children from the control or rejected clusters (Table 1). Using sociomatrices (mutual choice) to determine who children’s friends are We were able to identify the target child’s nominated friends’ cluster membership. From this information, we could investigate who lonely children nominate as their friends. In addition, we asked whether rejected children nominated other rejected children as play partners. Comparisons of play partner choice among the four clusters using a chi-squared analysis revealed a significant difference (␹2(1, N = 409) = 44.7, p < .001). Table 2 illustrates the number of nominations for each of the four clusters alongside the target child’s group membership. Using EXACON (Bergman & El-Khouri, 1987), after Bonferroni correction, the significant cells are: control–lonely; rejected–rejected; rejected–lonely; and rejected/lonely–lonely. Hence, it occurs that control children nominate lonely children as their friends more often than expected by chance and that rejected children nominate other rejected children. However, it occurs less often than chance that rejected and the rejected/lonely children nominate lonely children.

05 qualter (ds)

18/5/05

3:22 pm

Page 388

388

Journal of Social and Personal Relationships 22(3)

TABLE 1 Amount of perceived support from a close friend for each of the four clusters Cluster

M SD

Control

Rejected

Lonely/Rejected

Lonely

22.0 3.8

17.1 6.9

15.7 6.4

14.5 5.8

TABLE 2 Number of friendship nominations (and expected frequency) for each of the four clusters Target child’s cluster membership

Control Rejected Rejected/Lonely Lonely

Control

Rejected

Lonely/Rejected

Lonely

427 (445) 49 (46) 35 (31.5) 100 (87.4)

65 (73) 16 (7.7) 5 (5.2) 15 (14.4)

64 (64) 10 (7.4) 11 (5.1) 13 (14)

250 (215) 9 (22.4) 6 (15.2) 30 (42.3)

␹2 (1, N = 409) = 44.7; p < .001.

Do lonely and rejected children spend more time alone? There was a significant main effect for cluster on spending time alone (F(3,314) = 11.38, p < .001), but not on playing alone (F(3,314) = 1.98, p = .17). Follow-up Tukey’s HSD post-hoc pair comparisons indicated that children from the control cluster spent less time alone than did children from all other clusters. Children from the other three clusters were more often engaged in onlooker behaviour, standing alone, but they were not engaged in uninvolved play more than controls. Who do lonely and rejected children play with and what is the quality of these interactions? Even though lonely and rejected children spent less time interacting than controls and more time alone, they still spent the majority of their time in interaction. Who are they interacting with? The proportion of a child’s interacting time they spent with members of each cluster group was calculated for the entire sample. Table 3 illustrates the mean number of observed episodes spent in interaction with the members of the four clusters (of 80 observed episodes for each child). In order to determine the extent to which play partners differed across clusters, a one-way MANOVA was performed, using target child’s cluster membership as the independent variable and the proportion of observed

05 qualter (ds)

18/5/05

3:22 pm

Page 389

Qualter & Munn: Lonely children

389

episodes spent in interaction with children from each of the separate clusters as the dependent variable. The MANOVA indicated a significant main effect of target child’s cluster membership, F(3,315) = 6.0, p < .001. Follow-up ANOVAs revealed significant main effects of target child’s cluster membership with differences in the proportion of interaction time spent playing with members of cluster groups. There was a significant difference between clusters in the proportional number of episodes engaged in interaction with children from the control cluster (F(3,315) = 17.3, p < .001). Clusters also differed in the proportional number of episodes engaged in interaction with children from the rejected cluster (F(3,315) = 14.9, p < .001), children from the lonely/rejected cluster (F(3,315) = 10.8, p < .001), and with the lonely cluster children, F(3,315) = 14.0, p < .001). Follow-up Tukey’s HSD post-hoc paired comparisons revealed that children from the control and rejected clusters were significantly more likely than children from the lonely/rejected and lonely clusters to play with other control and rejected children. In contrast, children from lonely/rejected and lonely clusters were significantly more likely to engage in interaction with other lonely children than their nonlonely peers. In order to investigate whether the clusters differed significantly from one another in terms of the number of episodes they spent engaged in the different types of interaction (positive, negative, and neutral) with members of each cluster group, three one-way MANOVAs were conducted. Thus, separate MANOVAs were conducted for the positive, negative and neutral interactions,

TABLE 3 Mean number of episodes (each 15-seconds in duration) spent in interaction with members of the four clusters Target child’s group membership Play partner membership Control Rejected Rejected/Lonely Lonely

M SD M SD M SD M SD

Control

Rejected

Rejected/lonely

Lonely

34.00 (30.6) 25.20 (27.2) 1.82 (4.3) 3.53 (12.7) 64.50

25.70 (26.6) 14.02 (15.5) 2.50 (4.8) 3.02 (10.2) 45.30

8.60 (13.8) 11.80 (19.2) 10.70 (19.8) 11.93 (19.0) 43.00

14.10 (21.5) 7.53 (13.0) 9.60 (14.2) 19.40 (26.2) 50.62

21.60 (21.7)

25.50 (20.6)

21.05 (22.5)

7.93 (14.3)

6.42 (12.4)

6.38 (13.6)

Total no/episodes engaged in interaction Episodes spent alone M 7.99 (standing alone, uninvolved SD (15.4) in play, watching others) Episodes spent playing M 3.24 alone SD (7.7) Total number of episodes observed = 80

05 qualter (ds)

18/5/05

3:22 pm

390

Page 390

Journal of Social and Personal Relationships 22(3)

with the number of episodes of interaction with the different cluster groups as the dependent variables, and the target child’s cluster membership as the independent variable. Table 4 illustrates the proportion of interactions with each cluster that was of negative, positive, and neutral quality. Play partners during positive interactions. The MANOVA conducted for positive interaction episodes yielded a significant effect of target child’s cluster membership, F(3,315) = 9.8, p < .001. Follow-up univariate analyses revealed that the number of positive interactions with clusters differed as a function of the target child’s cluster membership. Clusters differed in terms of the proportional number of episodes they spent in positive interactions with control children (F(3,315) = 21.4, p < .0001), rejected children (F(3,315) = 23.1, p < .001), lonely/rejected children (F(3,315) = 19.3, p < .001), and lonely children (F(3,315) = 27.1, p < .001). Follow-up Tukey’s HSD post-hoc paired comparisons indicated that both the control and rejected children had significantly more positive interactions with other members of these two nonlonely groups than did children from the lonely/rejected and lonely clusters. Lonely children (lonely/rejected and lonely clusters), in contrast, were significantly more likely to have positive interactions with other lonely children from both lonely clusters than were children from control and rejected clusters. Play partners during negative interactions. The one-way MANOVA conducted on the target children’s negative interactions with members of the cluster groups yielded a significant effect of target child’s cluster membership, F(3,315) = 2.3, p < .01. Follow-up univariate analyses revealed that clusters

TABLE 4 Mean proportion of interactions with each cluster that is of negative, positive and neutral quality Target child’s group membership Play partner’s membership and quality of interaction Control

Rejected

Lonely/Rejected

Lonely

positive negative neutral positive negative neutral positive negative neutral positive negative neutral

Control

Rejected

Lonely/Rejected

Lonely

.88 .03 .09 .88 .05 .06 .46 .39 .14 .82 .11 .06

.71 .14 .15 .67 .25 .11 .06 .92 .03 .31 .50 .20

.48 .30 .21 .33 .46 .20 .56 .19 .25 .89 .02 .09

.49 .18 .33 .42 .39 .32 .78 .03 .18 .90 .003 .08

05 qualter (ds)

18/5/05

3:22 pm

Page 391

Qualter & Munn: Lonely children

391

differed in terms of the proportional number of negative interaction episodes they engaged in with control children (F = 6.4, p < .001), rejected children (F(3,315) = 3.9, p < .01) and lonely/rejected children (F(3,315) = 3.2, p < .05). However, no two clusters differed from one another by being more likely to engage in negative interactions with children from the lonely cluster (F(3,315) = 2.1, p > .05). Follow-up Tukey’s HSD post-hoc paired comparisons indicated that both the lonely clusters members (lonely/rejected and lonely clusters) differed significantly from control children by being more likely to engage in negative interactions with control and rejected children. Comparisons also indicated that children from the rejected cluster differed significantly from children from the lonely cluster by being more likely to interact negatively with lonely/rejected cluster children. Play partners during neutral interactions. A MANOVA was conducted on the mean number of neutral interaction episodes target children from the different clusters were observed with the members of the other cluster groups. The MANOVA yielded no significant main effects of cluster.

Discussion Friendship quality protects against loneliness Based on the work of Cohn et al. (1985), Howes (1988), Hoza et al. (2000), Parker and Asher (1993) and Stocker (1994), we expected rejected and lonely children to have fewer friends than nonlonely, nonrejected children. The results confirmed this, showing that lonely, rejected and rejected/lonely children had fewer friends than control cluster children when friendship is measured by sociometric nominations. However, the quality of the friendship distinguishes lonely and lonely/rejected from rejected and control children. As Parker and Asher (1993) suggested, the quality of the friendship may be more predictive of a child’s loneliness level and in this study lonely and lonely/rejected children both reported low levels of perceived support from a close friend. Lonely children report a lack of emotional support from their family (Qualter & Munn, 2002) and also report low levels of support from close friends. We may ask whether lonely children’s perceptions of support are divorced from objective reality: lonely children may have plenty of support available from close friends and parents, but they may be less likely to ask for help. It may be reluctance to ask for help that is characteristic of lonely children rather than an objective lack of support. Research that assesses mutual perceptions in child–friend and parent–child dyads will help us to understand the specific nature of the problem lonely children have in friendship interaction. Who’s friends with whom? Sociometric interviewing compared with observations of children’s play showed different pictures of social contacts. Observations of play partners

05 qualter (ds)

392

18/5/05

3:22 pm

Page 392

Journal of Social and Personal Relationships 22(3)

showed that lonely children from both the lonely and lonely/rejected clusters tended to play together at playtimes. This contradicted the sociometry data on lonely/rejected children that indicated that they were not friends with other lonely children. The behaviour of children from the ‘control’ group also contradicted sociometry data indicating that they were friends with lonely children. In fact, children from the ‘control’ group seldom actually played with lonely children; when they did, the interactions were more often negative than positive. Rejected children’s behaviour – they tended to play with other rejected children – was consistent with what they said in the sociometric interview, and their play interactions were most often positive in nature. Our observations of rejected children’s friendships are also consistent with previous research on rejected children’s play behaviour (see Asher & Coie, 1992 for reviews). So, our findings replicate earlier findings on rejected children, even though we used a different way of creating groups. However, the cluster of children we refer to as rejected are actually a certain type of rejected child (aggressive rather than withdrawn) and some of our observations ran contrary to what we expected. One would expect rejected children to be engaged in many negative exchanges with a playmate, because we defined them by teacher reports of externalizing behaviours, but we did not find this, except for those few interactions they had with lonely/rejected children. This raises the possibility that teachers’ reports of externalizing behaviour do not relate to their observations of children’s encounters, but to children’s reports of their peers’ behaviour. Why do lonely and lonely/rejected children play with each other more than they play with children from either the rejected or control clusters? And why are these interactions among lonely and lonely/rejected children often positive in nature? A simple explanation may be that their interactions represent merely ‘opportunistic’ meetings. Lonely children may wander around the playground looking for someone to play with, and end up playing together because they are both alone. It is also possible that the interactions between lonely–lonely and lonely–lonely/rejected pairs are the result of being drawn to one another through recognition of similarity, as would be predicted by the similarity–attraction hypothesis. A simple explanation of the positivity might be that if these interactions are merely opportunistic meetings, the children have no real reason to fight with one another. Researchers (Hartup, 1989; Rizzo, 1989) have reported that arguments are the preserve of those within established relationships. Children engage more in conflict within friendships than outside them (Doll, 1996). So, opportunistic meetings between lonely and lonely–rejected children may explain both the positive quality of their interactions with one another and the lack of emotional support they report: their friendships are neither aggressive nor supportive because they do not represent close or established relationships.

05 qualter (ds)

18/5/05

3:22 pm

Page 393

Qualter & Munn: Lonely children

393

Strengths, limitations and directions for future research Our observational measures of lonely children raise some interesting issues for theories of children’s friendships. In some ways, the observational data support report data, and in some ways they do not. One explanation for the contradictions in the data might be that rejected and lonely children had few mutual friends on sociometric measures, so all their interactions with any child on the playground would be inconsistent with their sociometric data. A more comprehensive explanation is that some friendless children nominated a variety of classmates as friends, but the nomination was unreciprocated. In this case, sociometry would classify only bilateral friendships as ‘friendships’, whereas observational method would code both unilateral and bilateral friendships as ‘friendships’. The observational method allows us to observe companionship, which is an important feature of friendship, as opposed to dimensions such as intimacy, loyalty, and selfvalidation, to which children may be referring when they make their decision of ‘friend’ during sociometric interviewing. From this viewpoint, sociometric interviews tap quite different aspects of children’s social experience than do observations. Sociometry indicates friendship-in-thehead as measured by reciprocal choice, whereas observation indicates peer acceptance as measured by objective companionship. Can there be such a one-dimensional opposition between friendship and peer acceptance? If we take a multidimensional, discourse analytic, view of the sociometric process, we would pay rather more attention to the power relations that might be shaping the children’s friendship nominations. For example, some authors (Denscombe, Szulc, Patrick, & Wood, 1986; Hallinan, 1981; James, Jenks, & Prout, 1998) have argued that we should pay more attention to the fact that children are positioned in particular sets of power relations in the classroom. Children’s accounts of their social experiences will be directly related to their own understanding of these power relations. In particular, their responses to sociometric interviews will be influenced by the social order constructed by teachers and parents. The interviewer is an adult, so these measures of the social meanings that children hold for friendship will be quite different from more direct measures, because they draw on adult versions of social order. During sociometric interviews, children are asked to state their preferences, and this is what they do. Researchers then interpret these answers in terms of popularity, acceptance, and social status as they themselves understand these concepts, not as the children experience them. The outcome of this is that the sociometric interview provides the researcher with detailed information about the social structure in the classroom (that is, who is ‘popular’ and who is not) according to adult definitions of this structure. Because the interviewer asks nothing about the children’s experiences of friendships on the playground, there is little information about the children’s objective view of the group’s social structure. For this reason, sociometric measures may give misleading indicators of friendship status. What is interesting in our data is not so much that there is a discrepancy between what the children report to adults and what the researcher

05 qualter (ds)

394

18/5/05

3:22 pm

Page 394

Journal of Social and Personal Relationships 22(3)

observes, but the size of that difference. By definition, one would expect rejected children to have fewer friends and play with others less than nonrejected children. However, the rejected children from our sample are not socially isolated, they do have friends, and they play positively with both control and rejected cluster members. This raises questions about the growing evidence that social rejection is a predictor of later maladjustment (see Parker & Asher, 1987 for review). If peer rejection is not directly related to the experience of social isolation (as is indicated here), or to the number of friends one has, then what aspect of the children’s social and subjective experience is actually affecting their development? A possible explanation is that it is the discrepancy between what the child experiences, as measured by observation, and what she/he wants, as measured in terms of social preference using sociometry, that causes problems in development and later adjustment. Maladjustment problems later in life may be the result of having a smaller or less satisfying social network than the one the person desires. The absolute size of the network will not matter – what will affect development is the size of the network relative to the child’s aspirations. A significantly smaller network than desired might lower self-esteem, and affect development negatively. It seems unlikely that the rejected children’s friendship quality contributes to later maladjustment, as rejected children did not report lower levels of social support from their friends. A more plausible explanation is that having the reputation in adult eyes of being rejected is the social event with deleterious consequences. The predictors of later maladjustment may be bound up with children’s representations of their social structure rather than with the interactive realities of peer popularity per se. Conclusions Our data show that sociometric interviews give a quite different picture of children’s social reality than do direct observations. This is not a new or surprising finding, but what we do find new and surprising is that ‘rejected’ children have far better social experiences with their peers than sociometric studies have led us to believe. Because the literature reports that peer rejection is related to later maladjustment, we need to examine our assumptions about the mechanisms that mediate this link. We have drawn the outline of an approach to children’s friendships that interposes an adult-defined social hierarchy between children’s playground experiences and their representations of their friendships with other children. We have also suggested that the mechanism for the link between peer rejection and later maladjustment might exist in children’s social representations rather than in their observed social experiences. Specifically, the common factor might be a large discrepancy between social experiences and social aspirations. Because adults’ representations of children’s social worlds have been so little commented upon, our main conclusion is that a future research agenda should map out the ways in which children perceive and react to adults’ representations of their social world. A related conclusion is that researchers should not rely on sociometric interviews to yield

05 qualter (ds)

18/5/05

3:22 pm

Page 395

Qualter & Munn: Lonely children

395

empirical data that are closely linked to children’s experienced social reality. This proposed research agenda would be particularly important for interventions in rejection that identify subjects by sociometry and use the same method to illustrate success. Where psychological theory is used to intervene in social events, we need to be particularly clear about whether we are manipulating children’s social reality or their social representations.

REFERENCES Asher, S. R., & Coie, J. D. (1992). Peer rejection in childhood. London: Cambridge University Press. Ashley, M. (1992). The validity of sociometric status. Educational Research, 34, 149–154. Berndt, T. J. (1996). Exploring the effects of friendship quality on social development. In W. M. Bukowski, A. F. Newcomb, & W. W. Hartup (Eds.), The company they keep: Friendship in childhood and adolescence (pp. 326–365). Cambridge: Cambridge University Press. Biehler, R. F. (1954). Companion choice behaviour in the kindergarten. Child Development, 25, 45–50. Bukowski, W. M., & Hoza, B. (1989). Popularity and friendship: Issues in theory, measurement and outcome. In T. J. Berndt & G. W. Ladd (Eds.), Peer relations in child development (pp. 15–45). New York: Wiley. Bukowski, W. M., & Newcombe, A. F. (1987, April). Friendship quality and the ‘self’ during early adolescence. Paper presented at the biennial meeting of the Society for Research in Child Development, Baltimore, MD. Chapman, A. J., Smith, J. R., Foot, H. C., & Pritchard, E. (1979). Behavioural and sociometric indices of friendship in children. In M. Cook & G. D. Wilson (Eds.), Love and attraction. Oxford: Pergamon Press. Cohn, D. A., Lohrmann, B. C., & Patterson, C. (1985). Social networks and loneliness in children. Paper presented at the biennial meeting of the Society for Research in Child Development, Toronto, Ontario. Connors, C. (1969). A teacher rating scale for use in drug studies with children. American Journal of Psychiatry, 126, 884–888. Denscombe, M., Szulc, H., Patrick, C., & Wood, A. (1986). Ethnicity and friendship: The contrast between sociometric research and fieldwork observation in primary school classrooms. British Educational Research Journal, 12, 221–235. DiTommaso, E., & Spinner, B. (1997). Social and emotional loneliness: A re-examination of Weiss’ typology of loneliness. Personality and Individual Differences, 22, 417–427. Doll, B. (1996). Children without friends: Implications for practice and policy. School Psychology Review, 25, 165–183. Foot, H. C., Chapman, A. J., & Smith, J. R. (1977). Friendship and social responsiveness in boys and girls. Journal of Personality and Social Psychology, 35, 401–411. Gottman, J. M. (1977). Toward a definition of social isolation in children. Child Development, 48, 513–517. Hallinan, M. (1981). Recent advances in sociometry. In S. R. Asher & J. M. Gottman (Eds.), The development of children’s friendships (pp. 91–115). Cambridge: Cambridge University Press. Harter, S. (1985). Manual for the social support scale for children. Denver, CO: University of Denver. Hartup, W. W. (1989). Social relationships and their developmental significance. American Psychologist, 44, 120–126. Hartup, W. W. (1996). The company they keep: Friendships and their developmental significance. Child Development, 67, 1–13.

05 qualter (ds)

396

18/5/05

3:22 pm

Page 396

Journal of Social and Personal Relationships 22(3)

Hayden, L. (1989). Children’s loneliness. Unpublished doctoral dissertation, University of Waterloo, Waterloo, Ontario, Canada. Hayden, L., Taruili, D., & Hymel, S. (1988) Children talk about loneliness. Paper presented at the biennial University of Waterloo Conference on Child Development, Waterloo, Ontario, Canada. Howes, C. (1988). Peer interaction of young children. Monographs of the Society for Research in Child Development, 53. Hoza, B., Bukowski, W. M., & Berry, S. (2000). Assessing peer network and dyadic loneliness. Journal of Clinical Child Psychology, 29, 119–128. James, A., Jenks, C., & Prout, A. (1998). Theorizing childhood. Cambridge: Polity Press. Ladd, G. W. (1983). Social networks of popular, average, and rejected children in school settings. Merrill-Palmer Quarterly, 29, 283–307. Maragoni, C., & Ickes, W. (1989). Loneliness: A theoretical review with implications for measurement. Journal of Social and Personal Relationships, 6, 93–128. Marcoen, A., & Brumagne, M. (1985). Loneliness among children and young adolescents. Developmental Psychology, 21, 1025–1031. Marcoen, A., Goosens, L., & Ceas, P. (1987). Loneliness in pre-through-late adolescence: Exploring the contributions of a multidimensional approach. Journal of Youth and Adolescence, 16, 561–577. Margarita, M., Tur-Kaspa, H., & Most, T. (1999). Reciprocal nominations, reciprocal rejections and loneliness among students with learning disorders. Educational Psychology, 19, 79–90. Marshall, H., & McCandless, B. (1957). A study of prediction of social behaviour of preschool children. Child Development, 28, 149–159. Parker, J. G., & Asher, S. R. (1987). Peer relations and later personal adjustment: Are low accepted children ‘at risk’? Psychological Bulletin, 102, 357–389. Parker, J. G., & Asher, S. R. (1993). Friendship and friendship quality in middle childhood: Links with peer group acceptance and feelings of loneliness and social dissatisfaction. Developmental Psychology, 29, 611–621. Peplau, L. A., Miceli, M., & Morash, B. (1982). Loneliness and self-evaluation. In L. A. Peplau & D. Perlman (Eds.), Loneliness: A sourcebook of current theory, research, and therapy (pp. 135–151). New York: Wiley-Interscience Peplau, L. A., & Perlman, D. (1979). Blueprint for the social psychological theory of loneliness. In M. Cook & G. Wilson (Eds.), Love and attraction: An interpersonal conference (pp. 101–110). New York: Pergamon Press. Qualter, P., & Munn, P. (2002). The separateness of social and emotional loneliness. Journal of Child Psychology and Psychiatry 43, 233–244. Ramsey, P. G. (1995). Changing social dynamics in early childhood classrooms. Child Development, 66, 764–773. Rizzo, T. A. (1989). Friendship development among children in school. Norwood, NJ: Ablex Press. Rook, K. S. (1984). Promoting social bonding: Strategies for helping the lonely and socially isolated. American Psychologist, 39, 1389–1407. Rook, K. S. (1988). Toward a more differentiated view of loneliness. In S. W. Duck (Ed.), Handbook of personal relationships (pp. 571–589). New York: Wiley. Rubin, K. H., LeMare, L. J., & Lollis, S. (1992). Social withdrawal in childhood: Developmental pathways to peer rejection. In S. R. Asher & J. D. Coie (Eds.), Peer rejection in childhood (pp. 219–249). New York: Cambridge University Press. Russell, D., Cutrona, C., Rose, J., & Yorko, K. (1984). Social and emotional loneliness: An examination of Weiss’s typology of loneliness. Journal of Personality and Social Psychology, 46, 1313–1321. Shute, R., & Howitt, D. (1990). Unravelling the paradoxes of loneliness: Research and elements of a social theory of loneliness. Social Behaviour, 5, 169–184. Stocker, C. M. (1994). Children’s perceptions of relationships with siblings, friends, and

05 qualter (ds)

18/5/05

3:22 pm

Page 397

Qualter & Munn: Lonely children

397

mothers: Compensatory processes and links with adjustment. Journal of Child Psychiatry, 35, 1447–1459. Weiss, R. S. (1973). Loneliness: The experience of emotional and social isolation. Cambridge, MA: MIT Press. Weiss, R. S. (1989). Reflections on the present state of loneliness research. In M. Hojat & R. Crandall (Eds.), Loneliness: Theory, research and applications (pp. 51–56). London: Sage. Wood, L. A. (1986). Loneliness. In R. Harré (Ed.), The social construction of emotions. Oxford: Blackwell. Young, J. E. (1982). Loneliness, depression and cognitive therapy: Theory and application. In L. A. Peplau & D. Perlman (Eds.), Loneliness: A sourcebook of current theory, research and therapy (pp. 379–406). New York: Wiley.