Analyses below are based on 120 children, 60 children with ASD and a paired sample of 60 typically developing peers (children matched on age and gender from the same classroom). There were no differences on outcome measures between children diagnosed with Asperger syndrome (n = 16) and autism (n = 44); therefore, the following analyses included all 60 children with ASD. We first report descriptive data for the groups on measures of peer nomination and friendship ratings: social network centrality, friendship nominations (indegrees, outdegrees, connections, rejects, reciprocity of best and top 3 friends), and friendship quality ratings. In each case, we tested for group and grade-related differences. Next we report the descriptive data for the playground observations of children with ASD, and finally report individual differences in measures for the children with ASD.
Descriptive Data: Self-and Other Perceptions of Social Connections for Children With and Without ASD
Social Network Centrality. For the group of children with ASD, 8 children were isolated, 25 had peripheral status, 22 had secondary status, and five had nuclear social status. In contrast, none of the typically developing children were isolated, six had peripheral status, 35 had secondary status and 19 had nuclear status. See Fig. 2.
Consistent with our earlier studies, an ANOVA indicated that there was a significant group difference for social network centrality, F(1, 116) = 38.57, p < .0001. Children with ASD had significantly poorer social network centrality (1.38 ± .09) compared to typically developing matched peers (2.20 ± .09). There was a significant main effect of grade level, F(1, 116) = 4.87, p = .03 where children in the older grades (3rd–5th) had lower social network centrality (1.65 ± .10) than children in the younger grades (1st–2nd; 1.93 ± .09). There was no significant interaction between grade and diagnostic group.
Friendships, Connections, and Rejections. A MANOVA with grade and diagnostic group as independent variables was used to compare the number of children’s friendships, connections, and rejections within his/her classroom between typically developing children and children with ASD. Of the 120 children, 114 children were used in this analysis because three children with ASD did not complete the rejections portion of the friendship survey, as it was added after the study began; thus, these students (and their matched peers) were excluded from this analysis.
The multivariate result was significant for group, Wilks Lambda = .79, F(1, 106) = 7.13, p ≤ .001, indicating a difference in friendships between children with ASD and their typically developing peers. The univariate F tests showed a significant difference between children with ASD and their matched peers for the number of friends they nominated within the classroom (outdegrees; F(1, 106) = 8.57, p = .004), the number of received friendship nominations by other children (indegrees; F(1, 106) = 18.84, p ≤ .001), and the number of classroom connections (connects; F(1, 106) = 14.61, p ≤ .001). Children with ASD nominated fewer peers as friends (3.76 ± .34), were nominated fewer times as a friend by peers (1.48 ± .24) and had fewer overall classroom connections (i.e., smaller social networks; 2.76 ± .33) than their typically developing matched peers (5.17 ± .34; 2.92 ± .24; 4.54 ± .33, respectively). Children with ASD did not differ from typically developing children in their percentage of connections to peers by gender. Boys were more often connected to boys and girls to girls for both groups. Lastly, when examining children’s number of rejections, there was no significant difference in the number of rejection nominations received by children with ASD relative to their matched peers. There was no main effect of grade or a group by grade interaction for these outcomes (see Fig. 3).
Reciprocal Friendships. Both reciprocated top 3 friends and best friendships were examined using an ANOVA. For their top 3 friends, the overall model was significant, F(3, 99) = 13.12, p < .0001 with a significant main effect for group, F(1, 99) = 39.22, p < .0001.The percentage of children’s reciprocal friendships with their nominated top three best friends was significantly lower for children with ASD (17.91% ± 5.32) in comparison to their typically developing matched peers (63.91% ± 5.07). There was no main effect of grade or a group by grade interaction.
The overall model was significant for reciprocal best friends, F(3,61) = 3.94, p = .0123, with a significant main effect of group, F(1, 61) = 10.82, p = .0017. Children with ASD had fewer reciprocal best friends than did typical children, (11.33% ± 8.4 compared to 44.97% ± 7.08), respectively. There was no main effect of grade or a group by grade interaction. See Fig. 4.
In addition, there was no difference between children with ASD and typically developing children in whether they selected a same-sex best friend (χ2(1, N = 116) = 3.42, p = .09). The majority of both groups chose same sex best friends, 56 out of 60 typically developing children and 46 out of 56 children with ASD. Four children with ASD did not list any peer as a friend.
Friendship Quality Scale. A MANOVA with grade and diagnostic group as independent variables was used to compare the five domains of child-rated friendship quality (i.e. companionship, help, security, conflict, and closeness) between typically developing children and children with ASD. Of the 120 children in the matched sample, 116 children were used in this analysis. Four children with ASD did not list a best friend and therefore did not complete the FQS; therefore, they were excluded from the analysis.
The multivariate result was significant for group, Wilk’s Lambda = .84, F(1, 108) = 4.13, p = .002, indicating a difference in friendship quality between children with ASD and their typically developing peers. The univariate F tests showed there was a significant difference between children with ASD and their matched peers for closeness, F(1, 108) = 17.87, p ≤ .001, security, F(1, 108) = 4.45, p = .04, helpfulness F(1, 108) = 15.00, p ≤ .001, and companionship, F(1, 108) = 8.60, p = .004, in that children with ASD reported poorer friendship quality in all four domains (see Table 2). Children’s perceptions of conflict with respect to their best friendships were not significantly different between the two groups. There was no main effect of grade or a group by grade interaction for any domain of friendship quality (see Fig. 5).
Table 2 Estimated mean differences and standard errors in friendship quality between children with autism and their matched controls
Within the Autism Group Analyses: Playground Observations of Children with ASD
On the playground, children with ASD were engaged with their peers for just over a third of the observed intervals on the playground (38.6% of the total intervals). Children engaged in structured games with rules for approximately 20% of the observed intervals, and 18.6% of the observed intervals in joint engaged activities, such as having a conversation. For the remaining percentage of observed intervals, children were either solitary/unengaged (33.4%), or in lower levels of engagement: parallel play (6%); in proximity to other children (8%); parallel aware (engaging in similar activities with mutual social awareness; 7%); and onlooking (watching another group of children engaged in a game or activity; 7%). The rate of initiations to peers was, on average, once every 3 intervals (mean of 5.13 initiations during 15.79 observed playground intervals). Peers responded to the child with ASD in approximately 66% of the opportunities observed. Peers also initiated to the child with ASD an average of once every four intervals and the child with ASD responded to the peer in 75% of the opportunities observed.
Children with ASD who had a higher percentage of intervals observed in joint engagement and games on the playground also initiated to other children more often on the playground, r = .45, p ≤ .001, and responded more to peers’ initiations, r = .57, p ≤ .001. In addition, children with ASD who were more engaged on the playground were significantly less likely to have a 1:1 aide, r = −.27, p = .04.
Qualitative notes from observations: In order to determine if children with an aide were more likely to be interacting with their aide rather than with other children, we examined the qualitative comments made by the independent coders for each observation. During the playground observations the observer noted if the child was interacting with their aide or with peers. Two raters further coded the qualitative comments about what the child was doing during recess. These raters agreed 100% on the statements about the child’s behavior and the role of the aide. Categories included child unengaged but peers nearby, completely unengaged with peers or adults, wandering or unfocused, engaged with the aide or engaged with peers. Half of the children assigned an aide were observed as unengaged on the playground (18/36 children or 50%), 12 were wandering or unfocused and 6 were unengaged with peers but other children were nearby (e.g., eating nearby, or digging in the sand nearby). Another 14% (5 children) were observed interacting with their aide only. The rest of the sample (13/36 or 36%) was observed interacting with peers or engaged in games. Children with an aide were most often unengaged on the playground, neither interacting with peers or with the aide.
Connections Between Playground Observations, Peer, Self and Teacher Reports for Children with ASD
Correlations were run to determine if playground variables (engagement/games, unengaged/solitary, initiations, responses) were associated with peer nominations (indegrees, outdegrees, rejects, connects), social network centrality (SNC), and friendship quality for the children with ASD. None of the correlations reached significance.
We then explored whether children with ASD who were more engaged on the playground differed by teacher report of social skills. Using chi square statistics, we found that teachers rated children who were more engaged on the playground as having higher social skills, although the association was only marginally significant, χ2(1, N = 56) = 3.78, p = 06.
Next we tested whether the children who had a reciprocal friendship were more engaged on the playground, and whether peers rated these children differentially. Twenty percent of the children with ASD (N = 12) had at least one reciprocal friendship. These children had significantly higher social network centrality scores (2.17 ± .20) as compared to children with ASD who did not have a reciprocal friendship (1.24 ± .11; χ2(1, N = 49) = 11.59, p = .001). Having a reciprocal friendship, however, was not associated with being more engaged on the playground (χ2(1, N = 49) = .67, p = 1.00).