Professional relationships among school staff are vital for professional learning and school improvement. Social network analysis research confirms that the structure of staff networks — who turns to whom, how reciprocally, and how centrally — shapes outcomes ranging from inclusive teaching practices to professional wellbeing. Yet, most school network studies examine only a single relational type, typically advice or general collaboration, leaving a gap regarding whether the factors that shape tie formation in one relational domain operate similarly in others. This gap is particularly consequential when expressive ties such as emotional support and inspiration are considered alongside instrumental ones such as advice.

This study addresses this gap by examining three relational types — advice, emotional support, and inspiration — across seven primary and secondary schools in Finland, Scotland, and Sweden. Drawing on the subject-centered sociocultural approach to professional agency (Eteläpelto et al., 2013) and social capital theory (Lin, 2001), we investigate how these three network types relate to one another, how their structures differ, and whether staff members’ professional agency and seniority predict tie formation differentially across these relational layers. By applying exponential random graph models (ERGMs) with meta-analytic pooling across schools in three national policy contexts, this study provides one of the first cross-national, multiplex analyses of staff network tie formation in schools.

The theoretical framework centres on Eteläpelto et al.’s (2013) subject-centered sociocultural approach, which conceptualises professional agency as analytically separable from yet mutually constitutive with the social conditions of work. Building on this framework, Vähäsantanen et al. (2022) operationalised professional agency along three dimensions: Influencing at work (the capacity to participate in decision-making and shape one’s work community), Participation at work (active collaboration and engagement with colleagues), and Negotiating professional identity (the ability to realise professional goals and develop one’s professional self). While this line of work has advanced understanding of what professional agency entails at the individual level, its implications for how staff members interact with one another — that is, for the formation of professional network ties — remain largely unexplored.

The concept of relational agency (Edwards, 2005) provides a theoretical bridge between individual professional agency and network tie formation. Edwards (2005) described relational agency as a professional’s capacity to work with others by recognising what they bring to a joint enterprise. Pantić (2015) extended this to teacher agency, arguing that agency is not only an individual enactment but also a relational accomplishment constituted through purposeful interaction with colleagues. Taken together, these accounts suggest that professional agency and network tie formation are mutually implicated: agentic staff may actively seek professional interaction (a sender mechanism) and simultaneously become visible to colleagues as valuable partners, attracting incoming ties (a receiver mechanism). This dual theoretical possibility motivates testing both sender and receiver effects of professional agency dimensions in our network models.

The following three research questions guide the study:

RQ1: To what extent are ties in staff members’ networks for advice, emotional support and inspiration related to one another in each school?

RQ2: How is the structure of staff members’ networks (dis)similar across advice, emotional support and inspiration?

RQ3: Do staff members’ professional agency and seniority explain their social tie formation in the networks for advice, emotional support and inspiration respectively?

By situating professional agency within a network-analytic framework across three countries, this study extends the agency theory toward more explicit treatment of how agency is perceived and recognised by others in school staff communities.

This study surveyed 439 staff members across seven schools in Finland (two schools), Scotland (three schools), and Sweden (two schools), encompassing both primary and secondary settings. All staff — teachers, teaching assistants, and other educational professionals — were eligible. Participation was voluntary and anonymous, with response rates ranging from 73% to 80%. The study protocol received ethical approval from the participating universities and relevant education authorities. Data were collected via an online survey comprising sociometric network questions and individual-level measures. Three directed network relations were captured using complete within-school staff rosters: advice on supporting students’ learning or wellbeing, emotional support during difficulties, and inspiration in any aspect of work. Non-respondents were retained as nodes to preserve the full network boundary. Professional agency was measured using the short form of the Professional Agency Measure (PAM, Vähäsantanen et al., 2022), comprising nine items across three subscales: Influencing at work, Participation at work, and Negotiating professional identity. Seniority was measured as years of experience in education across five categorical levels. The analysis proceeded in three stages corresponding to the research questions. For RQ1, pairwise Quadratic Assignment Procedure (QAP) correlations between all three network adjacency matrices were computed for each school. For RQ2, descriptive network statistics — density, degree centralisation, reciprocity, and transitivity — were calculated for each of the 21 networks (three relation types × seven schools). For RQ3, exponential random graph models (ERGMs; Lusher et al., 2013) were fitted to each school–relation combination. Because all individual-level variables contained missing values, 100 imputed datasets were generated using multiple imputation prior to ERGM estimation. The ERGM specification included structural terms (edges, mutual) and attribute-based terms: sender, receiver, and homophily effects for each professional agency dimension, and seniority homophily. Within each school, estimates from the 100 imputed datasets were combined using Rubin’s (1987) rules, and the resulting school-level estimates were then synthesised across the seven schools using random-effects meta-analysis with the Knapp–Hartung adjustment for relatively small samples (i.e., 7 schools). To guard against scaling bias, average marginal effects (Duxbury & Wertsching, 2023) were computed for each fitted ERGM, providing tie-probability estimates comparable across networks of different sizes and densities.

The three network types were moderately correlated yet structurally distinct across all seven schools. Advice networks were consistently denser, more centralised, and more reciprocal than emotional support and inspiration networks. Inspiration and emotional support exhibited the strongest pairwise correlation in six of seven schools, suggesting a shared expressive foundation distinct from instrumental advice ties. Results revealed that seniority homophily was a robust predictor of tie formation, significant across all three networks with zero or near-zero between-school heterogeneity. Staff at similar career stages consistently formed ties with one another, independent of national context. This carries direct implications for professional learning policy: structured cross-seniority arrangements may be necessary to bridge experience-based clustering and ensure early-career teachers access resources concentrated among senior colleagues. Professional agency operated exclusively through a receiver-side mechanism. Staff who reported greater capacity to influence workplace decisions were more likely to be sought out for advice, emotional support, and inspiration, yet no corresponding sender effects were found. Being recognised as agentic matters more for attracting professional ties than enacting agency does for seeking them. This extends the subject-centered sociocultural agency framework (Eteläpelto et al., 2013) by demonstrating that agency operates in networks through others’ recognition rather than self-initiated outreach — aligning with Edwards’ (2005) relational agency concept, but adding quantitative specification: it is the attribution of agentic capacity by colleagues that reliably predicts tie formation across relational layers. Participation at work contributed a receiver effect only in emotional support networks, suggesting that collaborative visibility signals approachability for affective rather than instrumental support. Negotiating professional identity showed no significant effects, consistent with its internally oriented nature. These findings suggest that school leaders should identify staff who serve as relational hubs across network layers and design collaborative structures that counteract homophilic clustering to strengthen professional learning.

Duxbury, S. W., & Wertsching, J. (2023). Scaling bias in pooled exponential random graph models. Social Networks, 74, 19-30. https://doi.org/10.1016/j.socnet.2023.02.003 Edwards, A. (2005). Relational agency: Learning to be a resourceful practitioner. International Journal of Educational Research, 43(3), 168-182. https://doi.org/10.1016/j.ijer.2006.06.010 Eteläpelto, A., Vähäsantanen, K. Hökkä, P., & Paloniemi, S. (2013). What is agency? Conceptualizing professional agency at work. Educational Research Review, 10, 45–65. http://dx.doi.org/10.1016/j.edurev.2013.05.001 Lin, N. (2001). Social capital: A theory of social structure and action. Cambridge, UK: Cambridge University Press. Lusher, D., Koskinen, J., & Robins, G. (Eds.). (2013). Exponential random graph models for social networks: Theory, methods, and applications. Cambridge University Press. Pantić, N. (2015). A model for study of teacher agency for social justice. Teachers and Teaching, 21(6), 759-778. https://doi.org/10.1080/13540602.2015.1044332 Vähäsantanen, K. (2015). Professional agency in the stream of change: Understanding educational change and teachers' professional identities. Teaching and Teacher Education, 47, 1-12. https://doi.org/10.1016/j.tate.2014.11.006