Relational infrastructure and distributed leadership: A longitudinal study of school system development

Distributed leadership in school systems: From structural models to relational conditions

Distributed leadership reconceptualizes leadership as practice rather than position, recognizing that leadership activities are constituted through the interactions of multiple actors across organizational contexts rather than residing in individual roles (Spillane, 2005; Gronn, 2002). Research supports its contribution to school improvement, organizational learning, and teacher effectiveness when enacted under the right conditions (Harris, 2004; Hallinger and Heck, 2009; Tian et al., 2016). Yet as Harris (2004) noted, that relationship depends critically on how leadership is distributed and on the organizational conditions that support or constrain collaborative practice.

The literature has progressively identified what those conditions look like. Trust emerges as the most consistently identified enabling factor across cultural contexts, enabling genuine professional empowerment rather than mere task delegation (Tian et al., 2016). Organizational routines and infrastructure shape what collaborative practice is possible (Diamond and Spillane, 2016; Larsson and Löwstedt, 2020), and Liljenberg's (2015) longitudinal study in Swedish schools showed that sustainable distribution of leadership depends on the progressive construction of shared frameworks, expanded collaboration and professional culture. How these conditions develop over time—through what processes and over what timescales—remains less fully addressed than the question of what they are (Harris and DeFlaminis, 2016; Timperley, 2005).

A recurring tension in this literature concerns the relationship between structural and relational dimensions of distributed leadership development. Efforts to strengthen distributed leadership have tended to prioritize structural redistribution—redesigning roles, formalizing responsibilities, and restructuring meetings—even when the relational and cultural conditions the literature identifies as necessary are less amenable to structural design (Denis et al., 2012; Harris and DeFlaminis, 2016; Svensson and von Knorring, 2025). This reflects in part the normative adoption of distributed leadership as a model to implement rather than an analytical lens (Svensson and von Knorring, 2025), and in part a deeper assumption: when leadership is conceived as a configuration of roles, structural refinement is the natural response to developmental challenges. Swedish empirical evidence illustrates the cost of this asymmetry: Liljenberg and Blossing (2021), in a 3-year collaborative study of Swedish schools and preschool units, found that well-developed organizational structures—grouping systems, distributed roles, and improvement processes—failed to produce changed practice when school leaders prioritized teachers’ personal and relational needs over the organizational goals those structures were designed to serve, and vice versa; organizational building that neglected teachers’ relational engagement generated compliance without commitment. The relational and cultural dimensions were not supplementary to structural design but constitutive of whether it worked. What structural responses cannot fully address is the relational and cultural fabric through which distributed leadership actually operates.

This is not to say that relational dimensions are absent from the distributed leadership literature—trust, culture, dialogue, and shared meaning recur across studies as enabling conditions (Liljenberg, 2015; Tian et al., 2016; Bolden, 2011). But as Lumby (2013) argue: the field has persistently narrated redistribution while leaving power structures intact—an “inclusivity lite” that does not genuinely engage with how leadership opportunities and decision-making power are actually distributed. What is missing is not recognition that relationships matter, but a conceptual account of the relational conditions themselves: what they consist of, how they develop, and what organizational processes sustain them over time.

Toward relational infrastructure

A relational perspective implies understanding leadership as an emergent process constituted through the quality of interactions, dialogues, and meaning-making between organizational members over time—something that happens between people rather than residing in them (Uhl-Bien, 2006; Cunliffe and Eriksen, 2011). A genuinely relational ontology requires understanding leadership as a way of being-in-relation-to-others: constituted in the lived quality of interactions, not in the structural arrangements within which they occur. Helstad and Møller (2013) ground this empirically: in school improvement contexts, authority, and trust are continuously constituted and renegotiated in concrete relational encounters rather than fixed by formal structure.

From this perspective, distributed leadership is not something formal leaders distribute—it is an ongoing accomplishment of collective practice (Raelin, 2016; Svensson and von Knorring, 2025). Lindgren and Packendorff (2011) describe this accomplishment in terms of three interrelated processes: the construction of direction (shared goals and paths), co-orientation (developing mutual understanding across diverging interpretations), and action-spacing (the relational construction of what is possible, legitimate and meaningful in local organizational contexts). At the same time, Endres and Weibler (2016) insist that not all relational processes constitute leadership: influence must actually manifest in changed agency and decision making, not merely in improved relationships or efficient cooperation. The conditions for distributed leadership therefore require people to be in sustained relation with one another—building the trust, dialogic capacity and shared meaning that direction, co-orientation, and action-spacing require—in ways that demonstrably shift where decisions are made and who holds responsibility for outcomes. These conditions cannot be installed through structural design; they must be cultivated. This is what makes the organizational context not merely a backdrop to leadership practice but an active constituent of how it unfolds—what Spillane et al. (2004) term the situational dimension of distributed leadership.

Larsson and Löwstedt (2020) develop this situational dimension into a concept of educational infrastructure: the meeting structures, organizational routines, cultural-cognitive beliefs, and interaction spaces that shape how leaders make sense of their work together. Effective infrastructure requires conscious design and purposeful use to enable collaborative practice rather than merely constrain it (Diamond and Spillane, 2016). Yet existing frameworks describe what infrastructure contains—what meetings exist, what beliefs circulate, and what routines are in place—without theorizing how the relational conditions within it are cultivated. Miesner et al. (2022), in a study of eight elementary schools, found that while formal infrastructure shapes the collaborative opportunities available to teachers, relational trust is the necessary condition for those opportunities to become genuine practice: teachers with high relational trust collaborated regardless of formal structures; those without it avoided collaboration even when time and space were formally provided. Structure creates opportunity—but relationship creates practice. What is missing from existing frameworks is a theorization of how those relational conditions are built and sustained over time.

We propose relational infrastructure as a sensitizing concept in the sense developed by Blumer (1954): not a predefined analytical framework whose content is determined in advance, but a theoretical orientation that directs attention toward the relational and cultural conditions that distributed leadership requires. What relational infrastructure consists of in practice, how its dimensions develop across time, and how it interacts with the organizational conditions surrounding it are empirical questions. To address them, we draw on 3 years of longitudinal data from a Swedish municipal school system, asking: how do the relational conditions for distributed leadership develop over time, and what does that developmental process look like?

Research context and participants

The municipality of a medium-sized Swedish university town initiated a 3-year school development program in collaboration with a nongovernmental R&D institute specializing in practice-based school development. The explicit goal was to develop distributed leadership across the system and thereby strengthen the overall improvement capacity. The school system serves approximately 25,000 students across 80 schools at four educational levels: preschool, primary school, upper secondary, and adult education. The development program worked at leadership level, engaging approximately 130 principals and managers across the system. The institute's program director approached the research team to join as academic partners, and we were embedded in the program throughout its 3-year duration as co-designers of development activities and active participants in its seminars and network meetings.

No individual ethical review was required, as participation took place within the participants’ professional roles and the program was a formal organizational initiative. All participants are anonymized in this publication.

The program's organizational structure created multiple levels of participation that enabled us to study distributed leadership development across different organizational scales. At the system level, a steering group consisting of senior administrators provided strategic oversight and resources. At the operational level, 16 experienced principals served as Process Learning Leaders (PLLs), each facilitating learning groups of six to eight principals within their respective school forms. This structure engaged approximately 130 school principals in systematic development work over the 3-year period.

Research approach and analytical strategy

This study emerged from a dual engagement that shaped both its design and its findings. The study thus constitutes what Gustavsen (1992) terms developmental research—in which understanding and contributing to organizational development are pursued simultaneously—and what Reason and Bradbury (2003) describe as participatory inquiry: knowledge co-constructed through sustained interaction between researchers and practitioners rather than extracted from an external position (Coghlan and Brannick, 2010).

Relational processes unfold incrementally, are often tacit, and can only be understood through sustained proximity to the organizational life in which they are embedded. Retrospective accounts alone would have been insufficient to capture how trust developed, how dialogue deepened, how shared meaning was constructed—and how these processes were related to the developmental work we were simultaneously designing and enacting. Longitudinal access from within the program made this visible in a way that external observation could not. The concept of relational infrastructure, and the dimensions through which it operates, emerged from 3 years of iterative engagement with empirical material rather than being imposed upon it (Bowen, 2006).

In terms of positionality, both authors were present throughout as co-designers and program leaders—not as external evaluators arriving with a preformed research instrument. The academic analytical work—developing the concept of relational infrastructure, constructing the phase model—ran alongside the practical work of program development rather than following it. This means our interpretations are shaped by sustained involvement with the phenomenon we are studying, which we address through triangulation across multiple data sources and the member-checking process described below. We acknowledge in the limitations section that our findings describe a researcher-supported developmental process rather than naturally occurring organizational change.

Data collection

Data collection occurred through multiple complementary methods over the 3-year period (see Table 1). This was beneficial as distributed leadership development involves both explicit activities (meetings, seminars, and formal decisions) and subtle relational processes (trust-building, identity formation, and cultural change) that require different forms of documentation and analysis.

OPEN IN VIEWER

Table 1.

Dataset overview (2018–2021)

Data source	Period	Participants	Purpose
In-depth interviews	2018–2020	16 Process Learning Leaders (PLLs); 3 rounds (2018, 2019, 2020); 42 interviews total	Longitudinal tracking of distributed leadership development: expectations, evolving practice, sustainability, and cultural change
Development seminars	2018–2021	∼130 principals across all school levels; PLLs; program leader; research team	Structured distributed leadership development across the system; field notes and audio recordings; member checking of phase model at final seminar (May 2021)
PLL network meetings	2018–2021	16 PLLs; program leader; research team	Between-seminar facilitation practice, collegial reflection, and problem-solving
Steering group observations	2018–2021	6–8 senior administrators, program leader	Strategic oversight; program monitoring; crisis response and system-level evaluation
Document analysis	2018–2021	Various sources	Contextual and institutional evidence: program documents, school development plans, participant reflections

Interviews were semistructured and lasted 60 to 90 min. All seminars were documented through field notes and audio recordings. Steering group observations were supplemented by informal conversations with senior administrators throughout the program period.

Analytical approach

Analysis was oriented by relational infrastructure as a sensitizing concept (Bowen, 2006; Blumer, 1954): rather than a predetermined coding framework, it provided theoretical direction—attending to how relational and cultural conditions for distributed leadership developed, were experienced and changed across time. In-depth interviews with PLLs at three time points (2018, 2019, and 2020) constituted the primary analytical source; seminar field notes, network meeting records, and steering group observations served corroborative and contextualizing functions. Coding proceeded iteratively across the three data rounds: first-stage open coding generated categories clustering around the relational conditions the sensitizing concept directed attention toward—including action space, safe-to-fail conditions, trust-as-precondition, and structural-versus-relational change. Second-stage analysis identified developmental patterns across time points, producing the temporal structure underpinning the four-phase model. We fed findings back to participants throughout the research period, and the final development seminar in May 2021 provided systematic member-checking: participants validated and extended the phase model through collective reflection on their own developmental trajectory.

Findings: A four-phase process model

Our analysis revealed a discernible developmental pattern in how distributed leadership took hold across the school system. Rather than following a predetermined program logic, the pattern emerged from iterative cycles of collective reflection, structural experimentation, and cultural consolidation. We describe this pattern in terms of four phases, each characterized by a distinct configuration of leadership practice, organizational challenge, and relational dynamic. As Table 2 indicates, these phases are not strictly sequential: organizations moved back and forth between them, and different schools within the same system were often in different phases simultaneously. What oriented our analysis across all four was the persistent question of what relational conditions were being built, tested, or consolidated at each moment.

OPEN IN VIEWER

Table 2.

Four-phase process model of distributed leadership development

Dimension	Phase 1: Identifying action space	Phase 2: Experimenting with new approaches	Phase 3: Embedding in processes	Phase 4: DL as organisational culture
Core focus	Expanding individual and collective understandings of what is possible and legitimate	Exploring new forms of collaborative practice in bounded, safe-to-fail contexts	Integrating distributed approaches into ongoing organizational routines and processes	Distributed leadership becomes taken-for-granted—“the way we do things here"
Relational infrastructure	Trust-building begins through structured vulnerability; initial dialogue practices; individual sensemaking of distributed leadership concept	Trust networks expand through shared experimentation; dialogue deepens as dilemmas are shared; collective sensemaking of what distributed leadership means locally	Trust networks institutionalized across school forms; dialogue practices embedded in routines; shared meaning-making drives development planning	Trust “built into the walls”; dialogue is default mode; shared meaning sustains culture through leadership transitions and external crises
Key tensions	Purpose uncertainty; role ambiguity; hierarchical constraints; resistance to re-imagining action space	Tension between distributed ideals and hierarchical structures; cultural vs. structural change; maintaining safe-to-fail conditions	Sustaining momentum; avoiding reversion to projectification; balancing formalization with organic development	Leadership transition and turnover risk; external pressures (e.g. COVID-19); preventing cultural regression
Developmental work	Structured dialogue about current practice; appreciative inquiry; articulating shared values; confronting hierarchical assumptions; creating conditions for professional uncertainty to become expressible; disrupting existing relational configurations	Structural redesign experiments; pilot initiatives; testing organizational arrangements; supporting peer-based experimentation and cross-boundary learning through collegial exchange; building trust sufficiently for vulnerability and failure to become shareable	Embedding distributed leadership in quality work cycles; formalizing new roles; consolidating shared language and frameworks; integrating experimental practice into ongoing organizational routines as developmental work progressively shifts from externally supported to internally generated	DL operates without constant reinforcement; new leaders socialized into culture; organizational members co-produce the developmental narrative; formal architecture sustained while relational infrastructure operates self-sufficiently
Transition to next phase	Awareness of own leadership practice; collective recognition that DL requires more than structure	Recognition that cultural transformation is required, not just structural adjustment	Distributed practice becomes desired and sought after rather than imposed; persists despite disruption	(Ongoing attention required; institutionalization is not permanent)

Relational processes across the phases

Looking across all four phases, what stands out analytically is the consistency with which relational processes shaped the pace, direction and depth of distributed leadership development—and the degree to which these processes were irreducible to the structural arrangements that surrounded them. Three relational dynamics recurred across the data in ways that cut across the phase boundaries.

The first was the gradual development of trust networks that enabled genuine professional vulnerability. This was not a precondition for the program but emerged through it—unevenly, at different speeds in different groups, and always contingent on continued relational investment. The absence of prestige—the willingness to be uncertain and publicly so—was itself a relational achievement, not a personality trait. Caroline (19-12, PLL, compulsory school) captured the quality of what this created:

We have come very close to each other, very tight. So today we had dilemma discussions, and it just becomes so clear, because you open up completely. You have like no… There's no prestige left.

The second was the evolution of dialogue from information exchange to collective inquiry, something that cannot be installed through meeting structures alone. It describes a conversational quality that had to be cultivated over time. Asta (19-18, PLL, preschool) described what this felt like in practice:

Teachers are very questioning and that's good. Which makes it a little dancing, a little tango back and forth, all the time. ‘How do you see that …? So how do you yourselves see your role as leaders?’

The third was the progressive construction of shared meaning—a common language and set of reference points through which participants could make sense of what distributed leadership meant in their specific context. The durability of these concepts—persisting as orienting reference points across 3 years—illustrates how shared meaning-making, when genuinely collective, creates a kind of organizational memory that structural documents cannot replicate. Sharon (20-05, PLL, compulsory school) reflected on how this accumulated:

The first year, all principals sat together in the municipality, and tried to find… We picked out concepts that would be some cornerstones for us. And I know it's about courage, trust and participation… were the words that came out. And they have actually influenced me all the time.

Relational infrastructure

We entered this study with relational infrastructure as a sensitizing concept—a theoretical orientation that directed our attention without predetermining what we would find (Bowen, 2006). We identified three empirically distinct dimensions, each necessary and none sufficient on its own, proposing them as the constitutive elements of relational infrastructure in the context of distributed leadership development. These three dimensions did not emerge as separate threads from distinct parts of the data. Rather, they were the analytical convergence of categories that appeared repeatedly across all four phases and all three interview rounds—participants returned to trust, dialogue, and shared meaning regardless of which phase they were in. What differed across phases was not which dimensions were present but how developed each was, and what each made organizationally possible.

The relational infrastructure dimensions we identify are not politically neutral forms of relationship-building. As Lumby (2013) documents in her analysis of distributed leadership as a political phenomenon, relational processes can mask unchanged power distributions, creating the appearance of shared leadership without genuine redistribution of agency and decision making. In our data, the critical difference between earlier and later phases is not merely the depth of trust or sophistication of dialogue, but the degree to which these relational conditions actually enabled and sustained redistributed decision-making power. This aligns with Lindgren and Packendorf'’s (2011) argument that genuine distributed leadership is subversive to existing hierarchies—it genuinely relocates the locus of legitimate action from position to relationship. The relational infrastructure in phases 3 and 4 therefore carries structural consequence: it is the condition through which power actually moves in organizations, not merely the condition through which communication improves.

The first dimension is trust networks: the web of interpersonal relationships across organizational levels and school forms through which leaders extend professional vulnerability, share uncertainty, and hold one another accountable without recourse to hierarchy (Miesner et al., 2022). Our data extend this reasoning in an important direction: the trust that developed across 3 years was not dyadic but organizational in character, constituted through patterns of relationship that gave individuals a sense of systemic support. This organizational quality is what Uhl-Bien (2006) anticipates in relational leadership theory—leadership capacity emerging from the quality of the relational context as a whole rather than from specific pairings. It is also what explains why distributed leadership in our case could persist through leadership transitions and external disruptions: the trust was embedded in the organization rather than vested in individuals. Building this organizational trust required the sustained cross-school contact that the formal architecture provided—the repeated collegial encounters, across organizational levels and school forms, through which professional vulnerability could be gradually extended and relational accountability built.

The second dimension is dialogue practices: the conversational capabilities through which leaders move from information exchange to collective inquiry, from reporting to sensemaking, from managing disagreement to learning from it. Gustavsen (1992) established democratic dialogue as a developmental practice that enables organizations to generate new understanding through structured communicative interaction—a process that requires deliberate cultivation rather than simply time and goodwill. Our data confirm and extend this: participants in 2018 described meetings characterized by confusion and passive disengagement; by 2020 they were describing a genuinely reciprocal conversational quality. This shift required structured dialogue formats, sustained practice within process learning leader groups, and the gradual internalization of inquiry as an organizational norm. Dialogue practices, in our terms, are the organizational routines through which that communicative capacity is built and sustained. Cultivating these practices was the explicit developmental work of the program: structured inquiry formats, the gradual internalization of democratic dialogue (Gustavsen, 1992) as an organizational norm, and the deliberate shift from externally scaffolded to self-sustaining conversational practice across the four phases.

The third dimension is shared meaning-making processes: the ongoing organizational work through which participants constructed a common language, a collective understanding of what distributed leadership meant in their specific context, and a shared narrative of development. Seashore Louis et al. (2013) identify collective sensemaking as the mechanism through which distributed leadership produces organizational coherence—enabling common action to emerge from individual initiative. Our data show how this collective sensemaking was itself an achievement of the development process rather than its precondition: the concepts that crystallized collectively in phase 1 continued to orient participants 3 years later, constituting what Weick (1995) describes as a shared interpretive frame that shapes what individuals notice, how they read situations and what responses they reach for. Liljenberg (2015), in her longitudinal study of distributed leadership development in Swedish schools, found that sustainable distribution of leadership was achieved not through structural role redistribution but through the gradual construction of shared frameworks for understanding what leadership meant collectively—a process that required persistent collegial dialogue and took considerably longer than the structural changes that typically precede it. Her finding that this conceptual and relational work was the condition most directly enabling lasting collaborative practice anticipates our own: shared meaning-making is the dimension of relational infrastructure most responsible for the cultural depth that phase 4 requires, because it is the dimension that most directly transforms individual relational experiences into organizational culture. The shared conceptual foundations constructed in the phase 1 seminars—the common language around trust, courage and participation that participants described carrying throughout their development—were the seeds of this dimension, deliberately cultivated through co-constructed collective concept-building before experimentation began.

Together, these three dimensions constitute infrastructure in the specific sense developed by Larsson and Löwstedt (2020) and Spillane et al. (2004): they are the organizational conditions that actively shape what is possible, expected and meaningful in leadership practice. But they extend the educational infrastructure perspective in a crucial direction. Existing frameworks describe structural and cultural-cognitive components—what meetings exist, what beliefs circulate, what routines are in place. Relational infrastructure names the dynamic, developmental process through which the relational and cultural conditions within that infrastructure are cultivated. It is infrastructure not as a set of components but as an ongoing organizational accomplishment—and one that, as our data show, requires sustained time and investment to build.

Analytically, the notion of relational infrastructure concept speaks to the tension between descriptive and normative uses of distributed leadership that Harris (2007), Denis et al. (2012), and Svensson and von Knorring (2025) have identified as a persistent difficulty in the field. That tension arises from a conflation of two different questions: the analytical question of how leadership is distributed in practice, and the normative question of how it should be distributed to achieve desired outcomes. Treating these as the same question—or using the same concept to address both—has produced the conceptual confusion and implementation failures that the literature has documented.

The notion of relational infrastructure does not resolve this tension by collapsing the distinction but by redirecting inquiry altogether: away from the structural and toward the relational. It provides a developmental account of the conditions under which genuinely distributed leadership becomes organizationally possible—conditions that are not describable in a single observation, nor achievable through structural prescription, but cultivatable through sustained relational investment over time. This relational turn reframes the central question from “How is leadership distributed, and what are its main characteristics?” to “What organizational conditions enable collaborative leadership to emerge and endure, and how are those conditions cultivated?” This question can be investigated empirically in any organizational context, and the findings can inform development efforts without generating the structural bias that normative prescription produces. In Svensson and von Knorrin'’s (2025) terms, it keeps the analytical and developmental registers in productive dialogue rather than collapsing one into the other—while grounding both in the relational processes that give distributed leadership its practical substance.

The four-phase model gives this argument empirical grounding. Each phase transition in our data was enabled not by structural intervention but by a relational threshold being crossed: the shift from phases 1 to 2 required sufficient trust for experimentation to feel safe; the shift from phases 2 to 3 required sufficient shared meaning for distributed practice to be desired rather than imposed; the shift from phases 3 to 4 required sufficient depth of relational infrastructure for it to sustain itself without deliberate maintenance. These thresholds are irreducibly temporal—they cannot be accelerated by increasing project resources or tightening timelines, because relational threshold crossings require the kind of sustained investment over time that structural interventions and bounded project logics cannot by themselves provide.

Distributed leadership development should not be understood as moving toward some sort of stable equilibrium. School systems face persistent forces that favor centralization and unitary command: crisis response, accountability pressures, leadership transitions, and budget constraints. The literature has emphasized how to build distributed practice but given less attention to what works against it. In our case, the COVID-19 pandemic functioned as a natural stress test: it suspended formal structures while leaving relational infrastructure intact. Relational infrastructure, when sufficiently immersed in the organization, can resist institutional pressure toward centralization—but without active maintenance, organizations will tend to revert. Investment in relational infrastructure is therefore not a one-time developmental effort but an ongoing organizational commitment. Absent these deeper conditions, adaptive collaborative responses risk remaining temporary accommodations to crisis rather than sustainable shifts in organizational culture.

Implications for practice

For school leaders and system administrators, the practical implication of this study is a reframing of what distributed leadership development actually requires. The four-phase model suggests that the question is not “How do we implement distributed leadership?” but “What relational conditions does our organization currently have, and what is the next developmental investment that would strengthen them?” Answering this honestly requires attending to trust, dialogue and shared meaning as organizational phenomena that can be cultivated, not assumed (Sahlin et al., 2025).

The phase model also offers guidance about sequencing. Attempts to establish distributed leadership in processes (phase 3) before the experimental trust-building of phase 2 has been completed tend to produce compliance without commitment—distributed structures that are formally enacted but relationally hollow. Equally, investing in structural redesign before phase 1 has created sufficient relational security tends to generate the anxiety our 2018 data documented. The implication is not that development must proceed rigidly through each phase, but that skipping earlier relational work typically displaces rather than resolves the challenges it addresses. Reaching phase 4 is not the end of the developmental work but the beginning of a sustainability question: relational infrastructure can be eroded by leadership transitions, budget pressures or competing institutional demands, and school systems need to attend continuously to what sustains it.

There is also an argument against projectification (Jałocha, 2024). If relational infrastructure cannot be built within the temporal and institutional constraints of project logic, development funding organized as successive discrete initiatives will systematically undermine the conditions it intends to create. The alternative is sustained, open-ended investment measured not by the outputs of a bounded initiative but by the depth of the relational infrastructure it leaves behind.

Limitations and future research

Several limitations should be acknowledged. The study is grounded in a single Swedish municipal school system, and the Swedish educational context—with its institutional emphasis on collegial development, relatively flat hierarchies and democratic organizational norms—may have created conditions more favorable to relational infrastructure development than other national contexts. The transferability of the phase model and the concept require empirical investigation in different settings.

The formal governance structures of the school system's top management—built around line management and accountability reporting—reflected general municipal governance principles rather than the distributed and relational logic being cultivated at principal and middle leadership level. Relational infrastructure as theorized here describes conditions at the level of professional leadership culture, not the full governance architecture of a school system. How the two logics interact remains an important question for future research.

Finally, longer-term longitudinal research—tracking organizations 5 to 10 years after a sustained development effort—would illuminate questions our data cannot address: how durable relational infrastructure is across leadership transitions, and whether phase 4 cultural embedding is a stable achievement or requires ongoing maintenance. Our partial evidence suggests a productive asymmetry: earlier-phase infrastructure is more vulnerable to disruption than later-phase infrastructure, and erosion may be faster than construction.