Our findings using a nationally-representative sample of adolescents with an ASD indicate that half experience no or very limited social activities with friends and only one-third participate in social activities in the community with peers. These participation rates were significantly lower than those observed in three other disability groups: adolescents with speech/language impairments, learning disabilities, and mental retardation.
Rates of social activities with groups were lower than reported personal interactions with friends. Whereas about half of the adolescents with an ASD got together with friends, received phone calls, and were invited to activities by friends, only one-third participated in group social activities. They were most likely to volunteer or provide community service (35.1%) or take lessons or classes outside of school (30.6%). In terms of belonging to community groups, about one-quarter belonged to a religious group and the same number belonged to a disability specific group. Fewer belonged to sports teams or performing groups.
Overall, these findings show that the majority of adolescents with an ASD experience major obstacles to social participation. It appears that experiences with peers are more likely to occur one on one, and perhaps at home rather than in the community. One mechanism for promoting social relationships is by fostering participation with peers in group activities. With only one-third of adolescents with an ASD accessing such opportunities, there is an obvious need for greater supports and services to promote community inclusion for this population.
Social participation with friends was the factor that most differentiated adolescents with an ASD from those in the other three disability groups included in the study. Those with an ASD had fewer experiences with their friends outside of school and were three to five times more likely never to get together with friends compared with peers from all three other disability groups. These findings concur with those of Solish et al., who found that half of their sample of children and adolescents with an ASD had no friends, compared with less than one-quarter of adolescents with an intellectual disability (mental retardation) 
The one notable exception to the lower rates of community participation by adolescents with an ASD was their higher rate of participation in disability-related groups. These findings suggest that the extracurricular activities of adolescents with an ASD may frequently take place in non-inclusive settings. Our data cannot answer whether this disproportionate rate of participation is driven by the motivations and choices of adolescents or parents or some other factor. Understanding the dynamics underlying this disproportionate rate of participation could be deepened by mixed methods studies involving a qualitative component.
Not surprisingly, conversational impairment and low social communication skills were associated with a lower likelihood of social participation. Having impaired conversational ability was associated with an elevated risk for friends never calling, never being invited to activities, and having no involvement in extracurricular activities. Having a higher social communication score (based on questions about joining groups, making friends, social confidence, and conversation initiation) was associated with lower odds of never seeing friends, friends never calling, and never being invited to activities. This is consistent with prior work which found a lower likelihood of friendships among those with an ASD if they scored poorly on an ADI-R item about social impairment 
. Our findings highlight the centrality of social communication abilities and suggests these adolescents continue to need the kind of supports typically provided by speech/language pathologists. Research into effective ways of promoting social communication abilities should be prioritized if we want to increase the social success of adolescents with an ASD.
After controlling for other variables, we found several notable correlates of four measures of limited social participation among adolescents with an ASD. Two of the measures were about friendships (never sees friends, friends never call) and two were about activity participation (never invited to social activities, having no extracurricular activities). Adolescents from families with lower income had an elevated risk for no involvement in activities, but not an elevated risk for limited contact with friends. This is consistent with prior research using NLTS2 data on all high school students enrolled in special education (i.e. not broken out by disability category) that youth involvement in extracurricular activities is significantly more likely among wealthier families 
. In contrast, this prior research found that students from higher-income families had a significantly lower likelihood of never seeing friends. The correlation of income and participation in our findings clearly suggests that future research should include better measures of access related to social activity.
Higher functional cognitive skills were associated with a lower risk for limited social participation across all four measures. This is consistent with the finding of Mazurek and Kanne that children and adolescents with an ASD with IQ<85 were less likely to have friendships as measured by a single ADI-R item 
. Again, our work extends prior research by pointing to the pervasive association between individual abilities and a wide range of social participation indicators.
Notably, having neither comorbid ADHD nor high externalizing behaviors was significantly associated with any indicators of social participation. Difficulties with friendships and social participation are well documented among children with ADHD and those with externalizing behaviors 
. Our lack of confirmation could be due to using a measure of externalizing behaviors that is weak on reliability and validity.
Age, sex, race, ethnicity, and school context factors were not significantly related to social participation outcomes in the multivariate models. In results not reported, we examined a variety of indicators of inclusion in general education settings beyond the one we entered in our final regression model (whether the student spends any part of their day in a special education classroom). None were significantly associated with social participation indicators in multivariate models that controlled for other factors. This is consistent with our prior work, which did not find an association between regular education inclusion and friendships for adolescents with an ASD 
Several recent reviews have summarized findings about intervening to improve social skills in children with an ASD 
. As in most areas of research on ASDs, much less is known about how to intervene with adolescents. Our findings simultaneously underscore the fact that impairments in social communication are a core challenge for adolescents with an ASD and that these core challenges are strongly associated with a wide range of social participation outcomes. Some research suggests that social communication skills can be improved through targeted intervention. However, successful generalization of skills remains a substantial challenge. Improving social communication skills may not automatically result in increased social participation. Our findings suggest it is also important to look at the family socioeconomic context and how that may affect access to social opportunities.
Inclusion in the NLTS2 sample was based on schools' determination of meeting eligibility criteria for special education services under the autism category. Strictly speaking, our findings generalize to adolescents with an ASD who are enrolled in the special education autism category rather than to all adolescents with an ASD. How representative of all adolescents with an ASD are our results? We cannot answer this definitively. The male:female ratio of 6.5:1 in this ASD special education population is slightly higher than the mean of 4.5:1 across recent epidemiological surveillance site estimates but still within the range of prior research 
.The distribution by race and ethnicity in this study was similar to population-based surveillance findings 
. The higher rate of conversational impairment relative to adolescents from other eligibility groups is consistent with the fact communicative impairment is a hallmark diagnostic feature of ASDs. Household income among adolescents with an ASD in this study tended to be higher relative to those from other special education categories. This is consistent with other research indicating that autism is identified at a higher rate among more affluent families, suggesting that poorer children with autism are commonly under-identified 
. Lower household income was associated with limited social participation in our study. If our sample under-identifies the true prevalence of ASD among lower income households then the income association may be biased.
Overall, our findings suggest that a mix of individual and socioeconomic factors can significantly influence a range of social participation outcomes. What are the implications for intervention and policy? At the broadest level of interpretation, there is a need for researchers and practitioners to focus on expanding opportunities for social participation and
on improving individual social competence. This conclusion is consistent with the conceptual model of disability put forth in the World Health Organization's International Classification of Functioning, Disability, and Health 
. This landmark document challenges a purely clinical focus on treating individual impairments by depicting the outcomes of disability as a function of both individual and contextual factors.
Our study has some limitations. The survey-based measures of social participation were not as fine-grained as in some other studies and were reported by parents. Furthermore, the available survey measures were focused on the quantity of social participation rather than the quality. It is possible that the quality of social connections is more important than the number. We did not have information about the size of each youth's social network. Our study also lacked a comparison group of typically developing peers.
Our study also has several important strengths. The external validity of our findings was greatly enhanced by the representativeness of the sampling strategy and the size and diversity of the sample. We were able to test hypotheses about race, ethnicity, and socioeconomic status which are often neglected in ASD research. Finally, the availability of comparison groups from other special education categories enhanced our ability to contextualize findings.
In summary, compared with prior research, our study significantly expands inquiry in this area by broadening the range of social participation indicators examined, by increasing the external validity of findings, by focusing on the under-studied developmental stage of adolescence, and by taking an ecological approach that included many potential correlates of social participation. Future research needs to extend this inquiry into adulthood while also looking at individual and contextual characteristics, such as access, which may influence social participation. Finally, research in this area needs to catch up with the rapidly changing landscape of normative social participation by developing measures of new forms of electronically mediated social interaction, by broadening intervention goals to include helping adolescents with an ASD participate in these new forms of interaction, and by considering how to take advantage of a prevalent proclivity for computer use to see how electronic media might become a platform for delivering effective social competence interventions.