Human existence is fundamentally spatial. Space is not a neutral backdrop, but something produced and shaped through everyday interactions, practices, and processes (Lefebvre 1991; Massey 2005). For historians, this matters because spatial relations can themselves be treated as historical evidence. Rather than serving only as illustration, they provide an entry point for reconstructing past structures from fragmented and uneven traces (Ethington 2007).

Yet within much historical scholarship, spatial analysis remains unevenly adopted and is often used illustratively rather than as a systematic means of connecting heterogeneous evidence. This tendency did not automatically change with the rise of geospatial “big data”, which has frequently encouraged data-led description of large-scale patterns, while the more demanding work of layering small, multidisciplinary traces has remained comparatively rare (Cooper-Green 2016; McCoy 2017). Rather than privileging only the “best” sources or defaulting to those that are most abundant, we treat space as a join key: a way to link fragments across different evidential regimes and to compare them under explicit assumptions. In line with theory-driven approaches to archaeological data and spatial modelling, the aim is not to force a single reconstruction, but to make visible which spatial correlations hold across evidence types and which depend on particular modelling choices (Gillings et al. 2019; Vargha 2022).

Methodologically, the paper develops a stepwise spatial linking workflow that assembles and compares heterogeneous traces, archaeological, onomastic, and documentary, by treating space as a shared analytical scaffold rather than as an outcome to be illustrated. The workflow is designed to make linking assumptions explicit and testable, so that correlations can be evaluated against alternative baselines rather than fixed in advance.

We show how this form of spatial linking can connect onomastic, historical, and archaeological traces in order to propose candidate institutional nuclei that are not explicitly recorded in the historical record, and to clarify which linked traces support them (Ferenczi-Vargha 2024).

A related methodological question, therefore, deserves renewed attention: whether later, better-documented evidence can be used responsibly to address earlier periods that are preserved more indirectly. The wager here is not that later data can stand in for earlier realities, but that some spatial structures display sufficient persistence to function as proxies, particularly when multiple, independently produced traces are linked and cross-checked rather than treated in isolation (Ferenczi-Vargha, 2024; Vargha, 2022).

We develop and test this approach in a comparative historical setting where these issues are especially acute: the Christianisation of rural landscapes and the emergence of local church networks in the newly Christianised kingdoms of Bohemia and Hungary. Written evidence is uneven and strongly shaped by elite perspectives, and the few fragments that address rural foundations and initiative are heavily debated. The central question is therefore how institutional nuclei for this process can be identified, understood as early cores of administrative and ritual gravity from which later ecclesiastical space appears to have developed. Our answer begins with explicit modelling, assembling and spatially linking all relevant traces, including evidence that is often dismissed as too fragmented to be useful when considered on its own (Vargha 2022; Vargha et al. 2024).

We begin from the spatial puzzle itself. Early rural churches appear across landscapes where neither secular power centres nor monasteries provide a stable and ubiquitous background, yet a local church network nevertheless takes shape. The first question is therefore not where “centres” were located, but what forms of social and institutional organisation could have supported the establishment of rural churches in places that do not lie in the shadow of obvious central nodes. Our approach treats this as a problem of spatial explanation and tests competing forms of support through layered evidence, rather than attempting to fill what the record does not say.

As a first modelling step, we construct a baseline representation of early ecclesiastical space from the best available church evidence and model its local areas of influence using Thiessen or Voronoi catchments. This baseline is explicitly constructed and incomplete, but it becomes substantially more informative once archival archaeological observations are combined with standing monuments, published excavation results, and the small and contested written record (Volkmann 2018; Vargha 2022). The catchments are not presented as historical parishes. Instead, they function as a comparative scaffold that enables asking where churches appear “early” relative to one another and where the network's edges and gaps are more likely to reflect documentation than absence.

Only then do we introduce proxy geographies for the otherwise poorly visible institutional substrate. In this case, service-village toponymy and related onomastic labels are treated as a spatial imprint of service and administrative organisation that may point to early estate and curial infrastructures. The aim is not to establish the existence of a curia system, but to test whether this proxy layer shows spatial correlations with the constructed ecclesiastical baseline, and whether those correlations persist under alternative baselines and linking assumptions (Ferenczi-Vargha 2024). Where correlations hold, they provide candidate institutional nuclei that help explain how rural church space could expand beyond the reach of well-attested central places. Where they fail, the workflow makes that failure visible and analytically productive.

Mortuary landscapes provide a further check. Field cemeteries represent local burial organisation prior to the consolidation of funerary space into central churchyards, and therefore offer a material baseline for the pre-nucleation phase (Vargha 2022; Vargha et al. 2024). When considered alongside the constructed church network, they help distinguish between church-centred consolidation and earlier community burial regimes, and they render later churchyard nucleation legible as a spatial reorganisation rather than as a simple chronological marker.

The comparison between Bohemia and Hungary functions as a test of transferability. The two kingdoms differ in political configuration, evidential profiles, and trajectories of Christianisation, with Hungary characterised by a more centralised institutional context. Applying the same linking protocol to both cases makes it possible to assess whether comparable spatial regularities emerge when onomastic proxies, constructed ecclesiastical baselines, and pre-nucleation mortuary landscapes are layered together, and where they diverge in historically meaningful ways (Vargha et al. 2024).

Taken together, this approach demonstrates a reusable protocol for spatially linking small, heterogeneous historical datasets in ways that foreground modelling choices, uncertainty, and interpretive consequences.

Within DH frameworks, the article contributes not a spatial analytical perspective but equally an infrastructural and data-integration approach. The process is carried out in OpenAtlas, serving as a spatial foundation that represents diverse evidence, as a web of entities and connections aligned with CIDOC CRM categories and attributes. This facilitates the connection of archaeological and documentary data without erasing their distinct epistemological characteristics (Vargha-Eichert 2025). Modelling decisions remain transparent because categories are managed through type hierarchies that can be expanded, linked to external vocabularies, and explored across different fields. Temporal and spatial uncertainty are regarded as elements of the record represented as time intervals and levels of spatial accuracy instead of being substituted with misleading certainty (Doerr 2003; Eichert-Filzwieser 2020; Eichert 2021). Interpretation subsequently occurs through interface-guided investigation: entities possess identifiers, connections remain apparent, outcomes can be reviewed on interactive maps and dashboards, and data are accessible in standard formats, such as CSV and GeoJSON.

Although the research question and historical context inevitably shape the specific sets of evidence gathered here, the process itself is not limited to a single case or dataset. Its component stages, such as creation, spatial association, proxy validation and comparison that accounts for uncertainty, can be adapted to other types of evidence. This adaptability is facilitated by the data management and modelling system provided by the OpenAtlas framework, which enables diverse data to be compiled, integrated, and thoroughly examined in a uniform and reusable manner.

Crucially, the workflow is designed so that both recognisable spatial regularities and the absence of expected correlations constitute meaningful historical results. Recurring patterns provide qualified support for specific institutional hypotheses, whereas negative or unstable results delineate the conditions under which such explanations fail, yielding historically informative “no” answers alongside provisional and context-dependent “yes” answers.