Friendship ties among American Indian adolescents were driven by structure, not substance use
Friendship formation in American Indian reservation schools followed a pattern that diverges from much of the adolescent network literature. In a three-school cohort of 9th- and 10th-grade students on a Northern Plains reservation, the strongest forces shaping peer ties were reciprocity, transitivity, and demographic similarity, not alcohol, tobacco, or marijuana use. Depression and AI identity also failed to show a substantive association with who nominated whom as a friend. The result is a clear reminder that peer networks in AI-majority schools cannot be assumed to behave like networks drawn from racially heterogeneous settings.
The study used social network data from 247 students and modeled friendship formation with exponential random graph models, a framework suited to teasing apart the pull of network structure from individual traits. That choice matters because adolescent friendship is never just a simple reflection of shared behavior. It is also a product of who is already connected, who sits at the center of the network, and which social similarities make a tie feel natural before a single behavior is considered.
A single pattern stood out sharply: substance use did not consistently predict sending or receiving friendship ties.
Reciprocity and triadic closure organized the peer network
The network was built on the familiar architecture of adolescent friendship. Reciprocity, the tendency for a friendship nomination to be returned, remained a central mechanism, and transitivity also helped hold the network together. In plain terms, students were more likely to connect when a tie was already moving in both directions or when they shared a mutual friend. Those processes are not unusual in adolescent groups, but they take on special importance in reservation schools where students share overlapping family, school, and community lives.
That social overlap creates a dense relational environment in which mutual recognition and friend-of-a-friend connections can accumulate quickly. Transitivity is especially relevant in settings like these because it captures the closure of indirect ties into cohesive clusters. The study’s framing suggests that the network may be less about status competition and more about the reinforcement of already existing social pathways.
Degree-related effects were also evaluated, reflecting whether some students disproportionately attracted ties or whether popularity was constrained. The analysis did not present friendship as a pure popularity marketplace. Instead, the most reliable signals were structural and relational, indicating that the network’s shape emerged from how students were embedded in their school community rather than from any one high-risk behavior.
Demographic similarity mattered more than risky behavior
Homophily by gender, grade, and sexual orientation helped explain friendship formation. Those findings fit basic social network theory: adolescents often gravitate toward peers who resemble them in visible or socially salient ways. In this cohort, those similarities were more predictive than substance use or depression, which frequently receive attention in prevention work aimed at peer contagion.
The absence of a strong substance-use signal is clinically important because much of the prevention literature is built on the idea that risky behaviors cluster through selection and influence. In many school settings, adolescents who drink, smoke, or use marijuana are more likely to be connected to one another, either because they seek each other out or because their behavior converges once the friendship begins. Here, that pattern did not emerge in a meaningful or consistent way. The practical consequence is straightforward: intervention models imported from heterogeneous school populations may misread the social drivers of risk in AI-majority contexts.
Depression was also not substantively associated with friendship formation. Neither a vulnerable internal state nor AI identity itself appeared to structure the direction of peer nominations in a strong way. For clinicians, educators, and public health teams, that finding argues against assuming that peer networks in these schools are organized around adversity in the same way as networks studied elsewhere.
Cultural context changes what network theory predicts
The authors situated the findings in a cultural context shaped by kinship, collectivism, and community interdependence. Those social principles likely help explain why reciprocity and triadic closure were so prominent. In a setting where relational obligations are reinforced by community norms, friendships may be less contingent on risky behavior and more grounded in mutual recognition, shared setting, and demographic commonality.
That framing also challenges a common assumption in adolescent prevention science: that substance use is one of the primary magnets drawing youth into friendship groups. The present analysis suggests a different relational ecology. Within AI-majority reservation schools, peer ties appear to rest on structural and demographic foundations that are not reducible to alcohol, tobacco, marijuana, or depressive symptoms. Prevention strategies built around the assumption that risk behaviors themselves are the main organizing force will likely miss part of the social terrain.
This does not make substance use irrelevant. It makes the social mechanism around it more specific. If risky behavior is not the main driver of tie formation, then prevention efforts that rely on identifying socially central users may have less reach than expected. Social influence can still operate after friendships form, but the entry point into the network may not be shaped by behavior in the way many programs presume.
Reservation schools may require a different prevention logic
The study’s implications extend beyond network modeling. It asks public health programs to confront the possibility that adolescent peer ecology differs meaningfully across settings, especially in AI-majority schools that have not been the primary focus of previous network studies. The authors built on prior work showing that AI adolescents in heterogeneous school contexts may have fewer reciprocated friendships than White peers, but this analysis moved deeper into the mechanisms generating ties. Here, reciprocity was present, yet substance use was not a primary organizing principle.
For school-based prevention, that means peer-focused interventions should be designed with local relational structure in mind. If friendship is driven mainly by shared grade, gender, and orientation, then the social channels through which health messaging spreads may not map neatly onto models borrowed from majority-population settings. Intervention teams will need to account for the fact that cohesion can coexist with risk without risk serving as the main social glue.
The study also reinforces the value of first-hand network data collected in AI communities rather than relying on broader datasets that fold AI youth into racially mixed samples. Reservation schools are not simply smaller versions of other schools. Their social organization may reflect different relational norms, different community histories, and different pathways through which support and risk circulate.
The data also come with an important restriction: the underlying dataset is not publicly available because of tribal ownership and research review requirements. That limitation does not weaken the finding, but it does narrow what outside analysts can independently test. Future work will need to build on locally governed partnerships and comparable network data from other AI communities before these patterns can be treated as general across reservation settings.
Citable evidence for a distinct adolescent network pattern
Zhang and colleagues published the study in PLoS One as a cross-sectional network analysis of AI adolescents in three reservation schools. The central conclusion is precise: friendship ties were organized primarily by endogenous network structure and demographic homophily, while substance use, depression, and AI identity did not meaningfully shape who connected with whom. For adolescent medicine, school health, and public health planning, that is not a small correction. It is a different map of the peer environment altogether.
The study leaves one pressing question open for future research: how stable are these friendship mechanisms over time, and do they change as students move through adolescence? Longitudinal network data will be needed to see whether the same structural drivers persist as cohorts age, or whether behavior, context, and identity begin to matter differently across later school years.
- • PLoS One article DOI 10.1371/journal.pone.0362373: https://doi.org/10.1371/journal.pone.0362373