The region of present-day Moravia and western Slovakia, along with their immediate surroundings, constituted the periphery of the Carolingian Empire. Notably, this frontier zone witnessed some of the most significant events in European history of the period; these include the collapse of the Avar Khaganate (AD 788–803), the Christianisation of the first Slavic tribes (AD 831), and the conflicts with the Hungarians (AD 900–907) — which brought lasting changes to the cultural mosaic of Europe^{1, 2, 3}.

The emergence and development of the Moravian dukedom around AD 830—commonly known as Great Moravia, the first Christian proto-state of the Slavs⁴—is closely linked to the Carolingian concept of the limes and the imperial strategy of managing “peripheral institutions”. This approach, based on the notion similar to modern “near abroad concept”, aimed to establish a chain of politically dependent buffer polities that were militarily and economically controlled; as well as ideologically, through Christianisation and ecclesiastical administration. These territories were governed by tribal dukes who were accepted by the Carolingian rulers^{5, 6, 7}. Such buffer zones extended from the North Sea and the Baltic, along the Elbe and Danube rivers, to the borders of Byzantium^{8, 9}.

Although the establishment of these polities often required military intervention—amounting to a form of shock therapy¹⁰—and frequently met with resistance, there is little doubt that the outcome for these regions was ultimately transformative. The process brought about remarkable opportunities for social development and civilisational advancement^{11, 12}. Regardless of the Carolingians’ motivations for imperial expansion, the elites of these newly formed buffer entities gained access to the full spectrum of Western culture, knowledge, and technology—particularly in the fields of engineering and military organisation. This was materially manifested in the flourishing of craftsmanship, elite culture, and the development of defensive and ecclesiastical architecture¹³. Indeed, the history and material culture of Great Moravia remain among the clearest examples of these processes¹⁵.

The early medieval hillfort agglomeration at Bojná (western Slovakia, Inovec Mountains), which forms the focus of this article, represents one of the most important archaeological discoveries of the past two decades in Central Europe¹⁶. The identification of a complex in the forests of western Slovakia—including massive hillforts, smaller fortified sites, burial mounds, as well as unique military artefacts, gold liturgical items, and iron hoards—marked a major breakthrough in archaeological scholarship¹⁷. These finds have fundamentally reshaped our understanding of the development of Slavic settlement structures and defensive architecture in the Middle Danube region, particularly within its mountainous zones.

Earlier archaeological research indicated that the most intensive phase of investment in defensive construction across the Bojná agglomeration occurred in the late 9th century, with the erection of the Bojná I stronghold and its formidable fortifications^{18, 16, 19, 20}. The chronology and character of the associated material culture reveal close ties to the Carolingian cultural sphere, and to the period when Moravia constituted a border region of the Carolingian Empire¹⁹. However, more detailed investigations have suggested that this activity is likely to have built upon an earlier tradition associated with the site, indicating that the choice of location was not solely dictated by topography. Crucially, structures predating the Bojná I fortress have also been identified; although their chronology remained unclear in previous research, and lacked regional analogies²¹.

In recent years, there has been growing interest in the reassessing of the chronology of Carolingian hillforts, particularly in light of emerging large-scale radiocarbon datasets from eastern regions of former Carolingian states such as Austria and Saxony^{22, 23}. By contrast, the publication of radiocarbon data from Moravia, Slovakia, and more broadly from Central Europe, remains limited. Much of the existing research has focused on presenting general results, while often omitting detailed analyses, chronological modelling, and transparent interpretations of site sequences. Therefore, data presented in this article are intended to change that situation and stimulate a broader scholarly discussion concerning the chronology of Carolingian-period hillfort construction in Central Europe, with a particular emphasis on the Middle Danube Basin.

The traditional view of early medieval hillfort construction in western Slovakia held that a developed network of fortified strongholds had emerged already in the late 8th or early 9th century. This would imply that the local population rapidly achieved a scarcely credible rate of socio-economic and technological advancement—especially when contrasted with conditions in the 8th century—and did so amid a regional context of political instability and warfare. Nevertheless, this view has long remained largely unchallenged. It is only in light of recent research²⁴ that a new perspective has emerged, indicating that the process of intensive fortification construction in western Slovakia most likely began towards the end of the 9th century^{25, 26}.

In this context, it is necessary to reassess the Carolingian Empire’s potential influence on—or at least its inspiration for—the emergence of these structures, both in technological and administrative terms. At the same time, it remains important not to neglect the underlying local foundations that may have created the conditions for such development.

In this study, we analysed 79 radiocarbon samples from five sites that form the core of the Bojná agglomeration (Fig. 1): Bojná I (Valy Hill), Bojná II (Hradisko Hill), Bojná III (Žihľavník Hill), Bojná IV (Iron Valley), and Bojná VI (the likely prehistoric hillfort Bojná V has not yet been excavated). A key element in understanding the dynamics of the agglomeration is the chronology of its individual components. Through radiocarbon analysis and a comparison with dendrochronological data, this article seeks to address the following questions: (1) When were the individual components of the agglomeration established, and during what periods were they in use? (2) What were the relationships between these sites? (3) How do these findings contribute to a revised understanding of the chronology of early medieval defensive architecture in the Middle Danube region?

Chronological Structure of the Agglomeration. To construct a general framework of the chronological structure of the agglomeration and the temporal relationships between individual features, we used data derived from 51 out of 79 radiocarbon samples (Supplementary Table S1). These data were processed using OxCal software^{27, 28}. The analysis encompassed simple and combined calibrations, non-parametric frequentist methods (kernel density estimation), as well as Bayesian modelling based on excavation data and dendrochronological assessments.

The earliest features of the agglomeration. Based on the analysis, the hillfort at Bojná III appears to be both the earliest and the longest-functioning site within the agglomeration. Radiocarbon and archaeological data jointly corroborate the view that its initial establishment dates to around the mid-7th century, with a terminus post quem (TPQ) of 1470 ± 30 years before the present (BP); human activity at the site ceased during the 9th century AD. However, the decline in activity was not simultaneous with the physical abandonment of the site; the hillfort had most likely ceased to fulfil its primary function already at the turn of the 8th and 9th centuries.

Stratigraphic observations made in conjunction with the established radiocarbon dates allow the refinement and extension of this interpretation. As no finds predating early medieval ceramics were recorded within the rampart, nor was any imposed cultural layer identified that might indicate earlier permanent occupation of the area, it can be reasonably assumed that the rampart construction coincided with the onset of early medieval activity at the Bojná III site. Consequently, the rampart construction’s TPQ may also be regarded as that of the start of intensive early medieval fortified settlement activity within the agglomeration.

In order to verify the chronology of the earliest settlement horizon, it was necessary to consult the archaeological material. A group of artefacts, including ornaments and weapons, which are indicative of an elite presence, suggests that such elites may have been present at the Bojná III site from as early as the turn of the 6th and 7th centuries, and certainly from the mid-7th through to the early 9th century^{29, 21}. The chronology of the earliest finds aligns almost precisely with the established TPQ, thus placing the foundation of the Bojná III hillfort in the mid-7th century. The radiocarbon date Poz-186694 (1,470 ± 30 BP) is particularly significant in this context, as it corresponds closely with the general dating of numerous artefacts not only from the hillfort itself, but also from its immediate and broader surroundings. Furthermore, the sample that produced this date originated from a concentration of ceramics, which strengthens its archaeological context. As such, this date may be considered broadly indicative of the beginning of early medieval settlement activity in the Bojná region.

As in the case of the ramparts, an extended interpretation of the burial mounds’ dating was only possible through integration with stratigraphic and archaeological data. The chronological sequence of the mounds in relation to early medieval settlement is confirmed by the presence of an occupation layer, which was used as fill material in constructing the mounds. It appears, however, that the burial mounds’ construction coincided with the cessation of settlement and the abandonment of the hillfort. This interpretation is supported by the archaeological assemblage, whose youngest datable components correspond to the turn of the 8th and 9th centuries; as well as by the absence of any contemporaneous or later cultural layer in the mounds’ vicinity.

The “temporal gap” observed between the two radiocarbon clusters (Fig. 3) corresponds precisely to the occupational period of the hillfort as previously inferred by archaeologists—i.e. broadly spanning the 7th to 8th centuries. To verify this projected chronology, future investigations should focus on identifying settlement features and structural remains that can be directly associated with the hillfort’s period of use. Based on simulations of the settlement phase, their radiocarbon dates are expected to cluster around calibrated values of approximately between 1,450 and 1,240 BP (± 30).

This also raises the question of whether the animal bones recovered from the base of the mounds—which were excluded from the integrated analysis at the preliminary stage, due to concerns over sample reliability—might in fact originate from redeposited settlement layers beneath the mounds, rather than representing traces of funerary practices as initially assumed. The simulation further reinforces the proposition that the date Poz-186694 (1,470 ± 30 BP) may indeed reflect the approximate founding of the hillfort, around the mid-7th century.

The earliest radiocarbon cluster from the Bojná III site aligns closely with the only available date obtained from the rampart at Bojná IV: Poz-16305/20/13 (1,525 ± 30 BP). Interpretation of this date is meaningful only when considered in conjunction with the analytical model constructed for Bojná III, as both share an identical stratigraphic context and sample character. Given that the rampart at Bojná IV follows the same line as the fortifications at Bojná III, it is reasonable to assume that they form parts of a single defensive complex constructed within a similar timeframe. This suggests that we may be dealing with one of the earliest complexes of early medieval Slavic hillforts in Central Europe; certainly within the Middle Danube basin.

Successors. The sites that succeeded the hillfort at Bojná III were the fortifications at Bojná I and Bojná II. Based on radiocarbon evidence correlated with archaeological data—and, in the case of Bojná I, also supported by dendrochronological results—it is possible to place the establishment of both sites broadly within the 9th century, with their decline occurring during the 10th century. However, their chronologies are not perfectly synchronised. The absolute duration of use at Bojná II appears to have been slightly longer than that of Bojná I.

Bojná II Hillfort. The interpretation of models based on radiocarbon data and archaeological material has allowed the formulation of two alternative hypotheses regarding this site’s chronology. Unfortunately, at the current stage of investigation, neither clearly prevails. What is certain is that construction activity at the site took place either at the end of the 9th century or during the 10th century. Unlike the radiocarbon samples from Bojná III, the earliest dates from Bojná II fall within a relatively narrow range, forming a distinct phase dated approximately to AD 790–865. Importantly, this chronological phase aligns well with the dating of archaeological finds (a spur and an earring) from the Bojná II hillfort. Therefore, the charcoal fragments recovered from the rampart may represent evidence of early medieval human activity at the site, which directly preceded construction of the fortifications. These charcoal inclusions were most likely incorporated into the rampart material during its construction. However, at this stage, it is not possible to determine the temporal depth of the TPQ. Therefore, we cannot yet establish whether: [Hypothesis A] the fortifications had already been built during an earlier phase, and the younger radiocarbon cluster reflects subsequent repairs, reconstruction, or expansion of the ramparts; or [Hypothesis B] the fortifications were constructed in a single phase shortly after the TPQ, and the charcoal reflects settlement activity on an unfortified site prior to this construction. From a methodological perspective, the latter hypothesis is more cautious and conservative interpretation (see Fig. 4).

The Bojná I hillfort is the most thoroughly investigated site within the Bojná agglomeration. Its fortifications have been dated dendrochronologically to the period after AD 866 to after AD 908; this allows the construction to be placed around AD 890¹⁶, and indicates the possibility of later repairs. The radiocarbon dates of samples taken from the fortifications’ structural elements therefore serve primarily as supplementary information, and do not significantly influence the general chronology of the fortifications. Nonetheless, they have been instrumental in verifying the existing chronological framework, and in revealing previously obscured nuances of both the chronology and technical aspects of the construction.

Detailed analysis of the radiocarbon data from the fortifications revealed the following chronological patterns: (1) the earliest dates derive from the massive load-bearing elements of the rampart structure, while the latest originate from its external walls; and (2) samples taken from the eastern gate form a distinct, younger cluster within the eastern rampart. Both of these observations explain the relatively wide chronological spread of individual samples, and suggest repeated repairs to the outer walls and gate—elements that are more prone to decay than the internal structural framework. These findings align with observations made through dendrochronological analyses.

The radiocarbon dataset for Bojná I has demonstrated a general consistency with the dendrochronological results and the chronological framework established through archaeological investigations. The models support the hypotheses derived from the archaeological record: they confirm initial occupation of the site from the early 9th century, an intensification of settlement activity around the mid-9th century, the construction of the large-scale fortifications around AD 890, and their continued use into the first half of the 10th century (Fig. 5).

Bayesian modelling and the interpretation of data obtained from internal features indicate several plausible scenarios for the phasing of the site. Each scenario, however, necessitates the acknowledgement of at least one earlier settlement phase predating the preserved rampart’s construction. At present, it is not yet possible to precisely define the lower chronological boundary of this earlier phase.

Bojná VI. The sequence of radiocarbon dates for the agglomeration is further supplemented by two determinations obtained from investigations at the Bojná VI site. These derive from two burial contexts, but do not permit the formulation of detailed conclusions regarding their chronology. Only the archaeological assemblage recovered from the graves, alongside the site’s spatial context, permits the tentative hypothesis that the Bojná VI cemetery was associated with the Bojná III hillfort and its period of activity between the 7th and 9th centuries. This interpretation is not contradicted by the radiocarbon determination of 1,225 ± 30 BP from Barrow 3; nor by a nearly identical result of 1,230 ± 30 BP, obtained from charcoal that probably originated from a destroyed cremation grave.

These dates place the burials at Bojná VI within the same chronological horizon as the decline of the Bojná III hillfort; they also align with the earliest dates obtained from the associated burial mounds, and with the oldest phases of activity documented at Bojná I and Bojná II (circa AD 800). However, in the absence of a larger dataset of coherent and stratigraphically secure radiocarbon dates, the precise chronology of the Bojná VI burial ground cannot be reliably determined at present; this must remain a task for future research. The site appears to serve as a chronological and spatial intermediary between the earlier phase of the agglomeration (represented by Bojná III and IV sites) and its later phase (Bojná I and II).

Milestones in the Development of the Agglomeration. Analysis of the corpus of radiocarbon dates from the five investigated sites identified a pattern of overlapping events; this allows the establishment of key chronological thresholds, and the definition of chronological milestones in the development of the Bojná agglomeration.

The cluster of the earliest radiocarbon determinations (1,650 ± 40–1,525 ± 30 BP), derived from the embankments of the Bojná III and IV hillforts, marks a TPQ for the onset of development in the Bojná region. This formative phase most likely began during the 7th century AD; a view corroborated by the earliest radiocarbon dates associated directly with archaeological features and by artefacts. These radiocarbon signatures fall within the range of 1,470 ± 30–1,370 ± 90 BP; however, these values should not be interpreted as dating specific constructions or features, but rather as chronological markers indicative of intensified human activity.

A subsequent and strongly pronounced cluster of chronological markers is observable at the turn of the 8th and 9th centuries. This period corresponds to the decline of the earliest hillfort, Bojná III, and the emergence of occupation at the sites Bojná I and Bojná II. During the same interval, the Bojná VI cemetery was also in use. Nonetheless, it remains unclear whether the initial settlement at Bojná I and II was contemporaneous with the construction of fortifications there. Evidence suggests that the main phase of construction and extensive fortification activity occurred only at around AD 900, which aligns closely with the dendrochronological data. It therefore appears that the fortifications postdate the initial occupation.

The AD 900 threshold thus represents not only a peak in construction activity, but also serves as a TPQ for the decline of the early medieval agglomeration. The evidence suggests that the agglomeration ceased to function during the first half of the 10th century. Both dendrochronological data and Bayesian radiocarbon models indicate that the final maintenance or repair work on defensive structures occurred within the period of AD 920–950.

The chronological framework established for the principal sites within the Bojná agglomeration was also compared with the radiocarbon chronology of St George’s Rotunda in Nitrianska Blatnica, a site within the immediate settlement hinterland of the Bojná complex. The original investigations focused on dating the construction of the rotunda itself³⁰. However, a reassessment of the data^{20, 31} has revealed that the set of radiocarbon determinations from samples taken from the church’s walls also contains a cluster that reflects the chronology of the surrounding settlement context. The chronological range of this cluster (1,440 ± 30–1,140 ± 25 BP) corresponds precisely with the temporal span of the Bojná agglomeration. It replicates key chronological markers—the onset of settlement activity in the 7th century AD, and the construction boom around AD 900—and thus establishes a TPQ of AD 880 for the rotunda’s construction. The radiocarbon chronology of the structure itself, based on dates obtained from the earliest mortar layers, was estimated at 1,115 ± 13 BP, or 1,152 ± 12 BP, or alternatively calAD 881–1002 (95.4% confidence), depending on the modelling approach employed. This suggests that the rotunda was constructed either at the turn of the 9th and 10th centuries, or around the mid-10th century. Both possibilities are equally plausible, and correspond chronologically with the period of construction or abandonment of the Bojná I hillfort. In light of the analyses presented here, the convergence of chronological signals between the Bojná agglomeration and the Nitrianska Blatnica site does not appear coincidental.

The results of the radiocarbon analyses presented in this study offer a new perspective on the cultural transformations associated with the emergence of early Slavic culture in the Carpathian region and, more broadly, in Central Europe. They also shed new light on the dynamics of the development of the so-called Great Moravia—the first advanced political entity of the Slavs, which acquired the characteristics of a proto-state.

Overall, the radiocarbon data from the Bojná agglomeration (particularly from sites Bojná III and IV, with supplementary data from Nitrianska Blatnica) support the hypothesis of Slavic settlement expansion into upland areas of the Middle Danube basin at the turn of the 6th and 7th centuries AD. The first hillforts are likely to have emerged around the mid-7th century, as a result of the stabilisation of settlement patterns and social structures. This phenomenon, previously unrecorded in the Middle Danube region, has thus far been observed only in areas of present-day Poland and Ukraine³². A key to interpreting such early fortifications is to understand them not as strictly military structures, but as symbolic constructions associated with a ruling group’s authority over a given territory.

The identification of such features in Slovakia suggests that they may represent a broader cultural model characteristic of early Slavic culture across its range—although appearing in different regions at different times. Generally, the further the sites are located from west from the presumed Slavic homeland in the Middle Dnieper basin³³, the later these features appear. It is noteworthy that, in the case of Moravia and Slovakia, the emergence of such features coincides chronologically with the appearance of the so-called “Samo’s state”. This was a federation of local Slavic tribes, recorded in historical sources as existing ca. AD 631–654, which were formed under the leader called Samo to represent their collective political interests³³. While no direct causal link is proposed between this political entity and the construction of the Bojná III hillfort, the chronological coincidence is significant. It may reflect the achievement of a cultural threshold that allowed the social consolidation of populations that were previously characterised by mobility, as is reflected in both historical and archaeological sources.

More technically advanced, military-type fortifications may have begun to appear only in the second half of the 8th century or early 9th century. This chronological marker aligns well with the decline and eventual collapse of the Avar Khaganate’s political and military dominance in Central Europe. This collapse triggered both the acceleration of tribal socio-political development among Slavic populations in the Danube region, and a broader phase of political fragmentation and instability. A genuine building and engineering boom, focused on advanced military infrastructure, is evident only from the late 9th century onwards, within the proto-state structures of Great Moravia as a peripheral component of the Carolingian Empire. This does not, of course, exclude the possibility of earlier administrative-defensive constructions, particularly during the second half of the 9th century. The disuse of these structures is associated with the political downfall of Great Moravia during the first half of the 10th century, and the gradual assumption (rather than destruction) of regional administration by the emerging Hungarian state.

Importantly, the traditionally accepted date of AD 907 as the point of Moravian political collapse is not corroborated by archaeological evidence as a moment of clear cultural discontinuity. This is confirmed by radiocarbon dates from the hillfort at Bojná I, and by comparative material from the nearby rotunda at Nitrianska Blatnica.

Furthermore, the integrated chronological model of the Bojná agglomeration presented here may serve as a valuable point of reference not only for site-specific chronological frameworks, but also for broader developmental schemes of the Early Middle Ages in Central Europe. Most significantly, it provides important arguments concerning the chronology of Slavic settlement and hillfort construction. More detailed hypotheses, however, can only be resolved through the acquisition of new, large-scale radiocarbon datasets that would enable the testing of more complex and refined models.

The results of the analyses indicate a prolonged, successive, and uninterrupted development of the Bojná agglomeration throughout the entire early medieval period, beginning in the 7th century (or possibly as early as the turn of the 6th and 7th centuries) and continuing until at least the mid-10th century. The most significant chronological markers and the observed developmental pattern are closely aligned with the broader chronology of the Early Middle Ages in Central Europe, and support chronological hypotheses that were previously based on archaeological data. These findings underscore the importance of complementing archaeological and historical research with detailed analysis of scientific (natural) data, in order to assess the intensity of settlement processes and, consequently, cultural transformations.

The modelling of the radiocarbon data has also revealed an important issue: calibrated dates for features originating from the first half of the 9th century tend to extend chronologically back into the 8th century; a finding that is not supported by the archaeological record. The identification of this tendency highlights a serious challenge for archaeologists, as it may result in mistakenly antiquated features and distorted chronological interpretations. In our view, this problem may stem from the properties of the calibration curve itself—as is particularly evident in the simulation of dates for AD 800 from Bojná III, which lack broad uncertainty ranges—or it may be related to specific characteristics of samples from this region. If, however, the issue proves to be systematic, similar distortions should be expected at other early medieval sites. This would suggest a broader problem that has significant implications for the entire field of early medieval archaeology, particularly for the chronology of the Carolingian period. Identifying this issue here provides a foundation for future efforts to address and resolve it. It should also be emphasised that the results of mathematical modelling must be interpreted with caution: never in isolation from the archaeological context, and, where possible, in comparison with historical sources.

Radiocarbon samples. The analysis is based primarily on radiocarbon dating of charcoal samples collected from archaeological features. These include both previously published samples and newly acquired, unpublished data that are presented and evaluated here for the first time (see Supplementary Table S1). Most of the earlier data had never been subjected to detailed analysis, with the exception of the earliest series of dates (2007–2010) obtained using AMS at the now-defunct Erlangen laboratory in Germany²⁵. Subsequent research was carried out at laboratories in Kraków, Poland (code: MKL), using the LSC technique (years 2010–2016), and in Poznań, Poland (code: POZ), using AMS (since 2017). These methods are considered mutually consistent, and their results statistically comparable^{34, 35, 36}.

A total of 79 radiocarbon determinations were made available for analysis. Five of these—obtained from animal bone samples found at Bojná III—were not included in the integrated analysis and are discussed separately. The integrated analysis encompasses 74 radiocarbon dates, including samples of charcoal (68), slag (2), wood (2), human bone (1), and an animal tooth (1), all originating from the five sites that constitute the core of the Bojná agglomeration.

The dataset was thoroughly evaluated with regard to the origin and method of sample collection, using excavation field documentation provided by the Institute of Archaeology of the Slovak Academy of Sciences (AÚ SAV) in Nitra. This was done to exclude samples that were unsuitable for analysing the chronology of the early medieval agglomeration, or those clearly affected by post-depositional disturbance.

A total of 20 dates were classified as clearly inconsistent with the chronological framework relevant to this study, or demonstrably incompatible with the stratigraphic context of the sampled features. In two cases, the anomalous dating was attributed to contamination by roots or mould; in one instance, to prolonged atmospheric exposure of the sample, as established through excavation records; and in another, an implausibly early date derived from an iron-smelting furnace, which is likely to have been affected by alterations in carbon content due to high-temperature processing. The remaining 16 excluded dates correspond to prehistoric or post-medieval periods, well outside the broadest expected range of the early medieval occupation. Although these fall beyond the scope of the current article, it is important to note that they are likely to be accurate, and to reflect earlier or later phases of human activity within the agglomeration area; this is supported by the presence of associated finds.

During detailed analysis, three further dates from the Bojná VI site were deemed unreliable: two slag dates were rejected following failed agreement tests, and one was excluded based on the dating laboratory’s own recommendation, due to insufficient nitrogen and carbon content.

In total, 51 radiocarbon dates were ultimately included in the final analysis. It is important to emphasise that none of the excluded dates originated from construction elements of early medieval features; rather, they derived exclusively from small charcoal fragments recovered from fill layers, or from mound and rampart deposits.

In the integrated analysis, five radiocarbon dates from animal bones recovered from the fills of two burial mounds at the Bojná III site were excluded, due to concerns regarding their stratigraphic context—and, by extension, their origin and pre-depositional treatment—as well as broader doubts surrounding the reliability of radiocarbon dates from bone material without proper pretreatment³⁷. These concerns arose following the receipt of results that contradicted the expected chronology of the contexts from which the samples were reportedly taken. The decision to exclude these five animal bone samples from the analysis was further reinforced by a similarly problematic radiocarbon result from a human bone (Poz-20138/22/4) from the Bojná VI site, which also proved inconsistent with the archaeological chronology. In both cases, the discrepancies were relatively minor, and the resulting dates still fell within the broader chronological framework of the Bojná agglomeration.

A detailed investigation of the Bojná VI sample revealed insufficient nitrogen and carbon content, thus rendering the date unreliable^{38, 39}. In the case of the Bojná III animal bone samples, however, measurements of carbon and nitrogen content, atomic carbon–nitrogen ratios, and collagen quality provided values that were consistent with reliable radiocarbon dating. This opened up alternative interpretive possibilities, suggesting that the dates themselves may be accurate, and that it was the archaeological context that was misidentified—possibly the bone disposition was unrelated to funerary activity. Given that radiocarbon dating of bone is inherently complex and requires a methodological approach grounded in sound archaeological understanding⁴⁰, additional analyses using a larger number of samples from Bojná III are planned for the near future.

Analysis and modelling. The calibration, analysis, and modelling of the data were conducted using OxCal version 4.4.4.175²⁸. Bayesian chronological modelling functions provided by the software were applied in interpreting the results. Bayesian statistics offer an explicit, probabilistic framework for integrating different types of evidence in order to estimate the timing of past events and quantify the uncertainty associated with these estimates. This approach—now a standard technique in the construction of chronological models based on radiocarbon datasets^{41, 42}—enables, in short, the refinement and improved precision of dating by combining statistical outputs with archaeological knowledge.

In assessing the evidentiary weight of radiocarbon data for constructing chronological models, a fundamental principle was adopted: dates obtained from charred structural elements are of greater value, as they are likely to reflect—at least approximately—the structure’s period of use. In contrast, isolated charcoal (bone, wood, etc.) fragments retrieved from the fills of archaeological features or stratified layers must be treated with greater caution. As a rule, such samples can only provide a TPQ for the formation of specific features or layers, and more broadly may indicate the general chronology of a site, or highlight phases of intensified human activity⁴⁴.

In all such cases, there remains the possibility that the charcoal sample was not directly associated with the feature in question, having entered it through post-depositional processes. In the case of hearths, for example, it is possible that fuelwood of considerable age and unknown origin was used. The problem of so-called “old wood” (inbuilt age) can also arise when dating structural wooden elements⁴³. However, in these cases, the uncertainty can be mitigated through appropriate modelling, and outliers can be identified and excluded. This principle has already been successfully applied in the development of a chronological model for the St. George’s Rotunda in Nitrianska Blatnica, part of the broader Bojná agglomeration^{20, 31}. Specific issues related to individual models are discussed in detail below, and in the Supplementary Materials.

Markov chain Monte Carlo (MCMC) analysis was used for almost all multi-parameter Bayesian analyses performed in OxCal. The only exception is the use of the Combine() function, which reduces the problem to a single independent parameter so that it can be solved analytically. In OxCal, a model is considered plausible when its Amodel value exceeds 60. The Amodel parameter represents a measure of agreement between the model’s prior structure and the observational data (likelihood)²⁸.

In our modelling, we adopted the principle of striving for models with high internal consistency, aiming for Amodel values close to or exceeding 100. A value in this range indicates a statistically well-fitting model that is likely to contain no major outliers, and whose structure—such as phasing and boundaries—accurately reflects the underlying data. Amodel values exceeding 150 must, however, be interpreted with caution, as they may indicate overfitting—that is, excessive precision due to overly restrictive assumptions, underestimation of uncertainties, or artificially tight constraints. Such values do not automatically indicate an error, but should be carefully assessed in the broader context of the archaeological and statistical evidence.

Aggregate radiocarbon data for individual sites were also analysed using a kernel density estimation (KDE) model. This method allows a smoothed representation of the temporal distribution of dates while incorporating the full structure of the Bayesian model, including phases, boundaries, and outliers. By applying KDE, it becomes possible to estimate periods of peak activity within a given phase, and to compare these activity patterns across different sets of dates⁴⁵.

Bojná IV. The only radiocarbon date obtained from the rampart at the Bojná IV site—Poz-16305/20/13 (1,525 ± 30 BP)—is insufficient to construct a standalone model. Its interpretation is therefore only possible in conjunction with the analytical model developed for the Bojná III site (see Supplementary Fig. S5), as described in the Results and Discussion section.

Bojná III. For the Bojná III site, the radiocarbon dataset was grouped into two clusters already at the initial stage of analysis. The first cluster consists of dates obtained from scattered charcoal samples collected from the fortification structures (ramparts) and from the cultural layer located beneath one of the barrows. These dates fall within the range of 1,650 ± 40 to 1,470 ± 30 BP. The second cluster includes dates derived from the construction elements of the barrows, ranging from 1,210 ± 30 to 1,190 ± 30 BP (or 1,208 ± 18 combined BP, to 1,190 ± 30 BP), as well as a single date from the fill of the ditch (1,195 ± 30 BP).

The dates from the first cluster fall within a relatively broad range of 1,650 ± 40 to 1,470 ± 30 BP (Supplementary Fig. S5). These should not be interpreted as indicators of the fortifications’ construction period, as the samples from which they were obtained do not originate from structural elements of the rampart. Most likely, these charcoal fragments entered the rampart fill along with surrounding soil used in its construction. At best, these dates can serve as TPQs for the erection of the fortifications (Supplementary Fig. S1), while nonetheless providing clear confirmation that the defensive structures of Bojná III are of early medieval origin. This latter point is particularly significant, as it effectively rules out the possibility that the rampart represents reused prehistoric fortifications.

The second data cluster consists of four radiocarbon dates obtained from two burial mounds. These derive from samples collected from the cremation pyre and structural elements of the mounds, and thus provide an approximate indication of the time of their construction. The dates are tightly grouped within a narrow range of 1,210 ± 30 to 1,190 ± 30 BP (or 1,208 ± 18 combined BP, to 1,190 ± 30 BP). Although this cluster demonstrates remarkable consistency and dating precision, its utility for establishing an exact absolute chronology is limited. The dates fall within the so-called “early medieval plateau” on the calibration curve, which constrains resolution and permits only a broad estimate for the construction of these features—somewhere between the late 8th and late 9th centuries.

A third, separate date—Poz-186685 (1,195 ± 30 BP)—was obtained from the fill of the hillfort’s inner ditch. This date is nearly identical to those from the younger of the two barrows. While it does not directly contribute to narrowing the chronology, it may tentatively be interpreted as confirmation of some form of human activity at the site during the period of barrow construction; this possibility is also supported by the associated archaeological material.

In the basic chronological model (Supplementary Fig. S2), developed on the basis of many years of archaeological research, it was assumed that the fortifications of the hillfort are contemporaneous with the initial phase of settlement, while the barrows were constructed in the final stage of site use—either during or after the decline of occupation. Since no traces of earlier cultural layers or associated artefacts were recorded beneath or within the rampart body, it is reasonable to presume that the rampart was among the earliest structures at the site, erected on terrain that was previously unoccupied or only lightly utilised. Over time, continued use of the hillfort led to the formation of a cultural layer. In this context, dates obtained from unstratified charcoal of uncertain origin found within the archaeologically sterile rampart fill must be treated as TPQs for the fortifications’ construction. The most reliable and informative of these is sample Poz-186694 (1,470 ± 30 BP), taken from the outer face of the rampart and found in direct association with ceramic material. This date may therefore serve as a boundary marker for the onset of occupation at the site. Similarly, sample Poz-186596 (1,545 ± 30 BP)—derived from charcoal found in the cultural layer beneath Barrow 2—should also be considered a TPQ for the burial mound’s construction. The stratigraphic sequence is supported by the presence of mixed cultural material from the occupation layer within the barrow fill, and by its absence in the surrounding area. The lack of settlement-related deposits in the immediate vicinity of the mounds indirectly suggests that intensive human activity at the site had ceased by the time the barrows were erected.

The model made it possible to establish a TPQ for the fortifications’ construction and the onset of early medieval settlement at the Bojná III site (Supplementary Fig. S2), which corresponds approximately to the point defined by date Poz-186694 (1,470 ± 30 BP). Due to the lack of further data, the duration of the settlement cannot be precisely determined; however, the approximate end of site use is marked by the barrows’ construction dates.

Between these chronological boundaries lies a “temporal gap” in the radiocarbon record, which most likely corresponds to the actual period during which the Bojná III hillfort was in use. On this basis, the maximum span of settlement activity may be estimated as extending from the mid-7th to the early 9th centuries.

The absence of radiocarbon dates for the period between the mid-7th and early 9th centuries does not reflect a true archaeological hiatus. This is clearly contradicted by the excavated material, which is consistently dated within that range²¹. The archaeological assemblage indicates uninterrupted activity from the mid-7th century to the early 9th century—roughly matching the period for which radiocarbon data are lacking. The proposed model was subsequently tested through simulation (Supplementary Fig. S3), which confirmed its validity.

Test. The preliminary test models simulate radiocarbon dates for the period between the early 7th and early 9th centuries. Several simulations were conducted using calendar years from AD 600 to 800, at 25-year intervals and with uncertainties of 50, 40, and 30 years—thus reflecting the precision typical of modern radiocarbon laboratories. One such example is presented in Supplementary Fig. S3.

In the second test, for which two models were constructed, it was assumed that the charcoal fragments found within the rampart represent traces of early medieval human activity at the site—activity that directly preceded the fortifications’ construction and the establishment of permanent settlement. The modified simulation of the settlement phase was incorporated into the broader chronological sequence. The test models (Supplementary Fig. S4) demonstrate that simulated dates corresponding to the period before AD 625 (± 30–40) should be excluded, while calibrated values approximating AD 625 appear to represent a threshold. These models support the chronological observation derived from the simple model: that the settlement phase is likely to fall between the mid-7th century and the turn of the 8th and 9th centuries (ca. AD 650–800).

Bojná II. The radiocarbon dates that define the chronological framework of the Bojná II hillfort form two clusters (Supplementary Fig. S10). The first cluster comprises charcoal samples collected from the fortifications and ditch in the north-eastern part of the site. As these were not taken from structural elements, the dates—ranging from 1,250 ± 30 to 1,190 ± 30 BP—should be treated with caution; and, following the adopted methodology, they should be interpreted primarily as TPQs for construction of the fortifications. Notably, no older material was identified within this context, thereby supporting the interpretation that the ramparts were constructed during the early medieval period.

The second cluster of radiocarbon dates consists of samples obtained from burnt structural elements of the fortifications, recovered from the bottom of the ditch in the north-western part of the site (Supplementary Fig. S9). These dates indicate that this section of the defences is likely to have been constructed between the late 9th and late 10th centuries. Importantly, they build upon the TPQ established through the analysis of the first cluster, and thus reinforce the overall chronological framework for the Bojná II hillfort’s development.

In the chronological model (Fig. 4), the charcoal samples from the north-eastern rampart (which were not structural elements) were treated as TPQs, as they are likely to have entered the rampart fill from the surrounding area at the time of construction—analogous to the situation at Bojná III. Within this model, the TPQ for the fortifications’ construction falls at the end of the 9th century. This aligns perfectly with the earliest dates obtained from the burnt structural elements found at the bottom of the north-western ditch, and hence reinforces the interpretation of a construction phase in the late 9th to early 10th centuries.

This model is both safe to interpret and methodologically sound, as it rests on the reasonable assumption that the charcoal within the rampart originated from a phase of human activity predating its construction. Archaeological finds indicate the possibility of occupation at the site during the first half of the 9th century—a scenario supported by the model. However, these finds cannot confirm the existence of fortifications during that early phase.

Test. In the test, two models were developed (Supplementary Fig. S11; Supplementary Fig. S12) based on the assumption that the Bojná II site was occupied prior to the rampart’s construction. Since the radiocarbon data from the NE and NW sections of the rampart clearly cluster into two distinct chronological horizons, the test models applied a combination of data from these sections as indicators of two hypothetical phases—an earlier and a later one.

In the models, it was assumed that the radiocarbon data from the NE rampart reflect the chronology of a hypothetical event—either the construction of two phases of the rampart (earlier and later), or of two distinct sections (older and newer). The modelling results in both cases closely resemble those of the preliminary model, but they do not clearly confirm the existence of a discrete “event”. Instead, they require us to assume the presence of an earlier settlement phase, which the models constrain to the period between the late 8th and late 9th centuries. This interpretation is consistent with the archaeological evidence. However, it remains unclear whether the rampart was already constructed during this earlier phase—with the later phase reflecting a repair or modification—or whether it was built only at the end of the 9th century. From a methodological perspective, the latter hypothesis is more cautious and conservative interpretation.

Bojná I. The radiocarbon dates used to construct the chronological models for the Bojná I hillfort were grouped into three clusters. The first cluster comprises data obtained from the fortifications of the main enclosure, while the other two clusters correspond to archaeological features identified within the hillfort and the adjacent suburbium. For analytical purposes, the first cluster was further subdivided into two groups corresponding to samples from the eastern and western sections of the rampart. All samples in this first cluster were taken from structural elements of the fortifications and, in line with the adopted methodology, can be treated not merely as TPQs but as approximate indicators of the construction or use of the defences. Samples were collected from three locations along the rampart (the eastern section, the adjacent eastern gate, and the western section), and are interpreted as belonging to a single construction phase, with the exception of sample MKL-1420 (1,290 ± 60 BP). According to the field documentation, this sample may derive from an earlier (stratigraphically distinct) structure; and indeed, modelling consistently identifies it as an outlier across all scenarios. Although individual radiocarbon determinations from the remaining samples may suggest a relatively broad chronological spread—ranging from 1,290 ± 80 to 1,020 ± 35 BP—this variation falls within the 2σ confidence intervals for each sample, and is not inconsistent with the dendrochronological dating framework (see Supplementary Fig. S13)

Since dendrochronology has provided precise dating for structural elements associated with the final phase of the hillfort’s fortifications (after AD 866 – after AD 908), and given that the broader chronological framework for the Bojná I site (not necessarily limited to the hillfort) has been outlined through archaeological investigation, these findings were incorporated into several test models as a posteriori modifiers. Based on this combined evidence, the following assumptions can be made: (a) the rampart was constructed around AD 890; (b) the last repairs to the rampart were carried out after AD 908; and (c) the overall chronology of the Bojná I site, based on the archaeological material, spans approximately AD 800–950, and is likely to include more than one phase of use. The purpose of the test models was to evaluate whether, and to what extent, the radiocarbon data support these assumptions, and whether Bayesian modelling could reveal additional insights into the site’s chronological development.

The basic single-phase chronological model did not provide answers to detailed questions concerning the chronology of the site. Its main outcome was the clear indication that the radiocarbon data needed to be subdivided into distinct phases. Moreover, it was not possible to apply the TPQ-based approach used at the other sites, since no loose charcoal samples were collected from the Bojná I fortifications; only structural elements were dated. It was only through the construction and evaluation of multiple test models that a more refined and reliable chronological model for the development of the Bojná I hillfort could be established (Fig. S5).

Test. Repeated modelling of the dataset revealed a high degree of statistical consistency among the radiocarbon dates. In the first model, the samples were arranged in a depositional sequence reflecting the stratigraphic positions of structural elements: load-bearing components → inner wall elements → outer wall elements → outer basketwork. In the second model, all elements were assumed to belong to a single construction phase. The third model combined the assumptions of the previous two. These models (Supplementary Fig. S15) enabled the identification of outliers and elements that weakened the model’s robustness. A subsequent test model assumed that the rampart was in use between AD 800 and 1000 (Supplementary Fig. S16).

The models are weakened only by samples with chronologically extreme dates; however, these do not invalidate the models, and it is known with certainty that they originated from structural elements of the rampart. Moreover, the 1σ ranges of the modelled chronological phase in all test models align—either fully or partially—with the range of absolute dendrochronological dates, thus supporting the argument that the fortifications were constructed around AD 900. It is likely that older elements or long-lived timber (e.g., sample MKL-1422) were incorporated into the construction. It is also possible that the rampart replaced an earlier structure that was dismantled. During its use, the outer part of the structure appears to have undergone maintenance and repairs, as evidenced by the youngest radiocarbon dates.

The radiocarbon dates obtained from residential and economic features discovered within the hillfort and its suburbium exhibit greater chronological variation than those from the fortifications. These dates suggest the possibility of an earlier settlement phase predating the preserved fortifications (Supplementary Fig. S13). This, in turn, raises the possibility that the ramparts themselves may have had an earlier construction phase—a hypothesis already proposed on the basis of archaeological observations^{18, 19}. Radiocarbon dating of samples from these features, combined with targeted chronological modelling, appears to support this interpretation (Supplementary Fig. S17; Supplementary Fig. S18; Supplementary Fig. S19; Supplementary Fig. S20).

At present, two distinct clusters of radiocarbon dates can be observed within the area of the hillfort: the first ranging from 1,290 ± 40 to 1,260 ± 30 BP preceded by two markedly earlier dates: 1400 ± 30 and 1370 ± 90 BP; and the second from 1,225 ± 30 to 1,090 ± 35 BP (or 1,197 ± 19 BP combined, to 1,090 ± 35 BP). Interestingly, neither this older cluster nor the very early individual dates have yet been identified within the suburbium (Supplementary Fig. S13 B; Supplementary Fig. S14), where only a small number of features have been investigated. The samples from those contexts show striking consistency, clustering tightly within the range of 1,185 ± 30 to 1,165 ± 30 BP (or 1,178 ± 18 BP combined, to 1,165 ± 30 BP); this suggests a later and relatively short-lived phase of occupation in the suburbium—an interpretation that is supported by archaeological observations.

The only earlier date from the suburbium—Poz-16305/20/6: 1,245 ± 30 BP—was obtained from a large posthole pit, possibly associated with fortification construction. Notably, this date is surprisingly close to the outlier from the rampart, MKL-1420: 1,290 ± 60 BP, which perhaps indicates the existence of an earlier phase of fortification. However, these results cannot yet be considered definitive for determining the absolute chronology of Bojná I, or for clearly distinguishing individual chronological phases. The data are derived from charcoal samples retrieved from feature fills; they must therefore be treated primarily as indicators of periods of human activity, and as TPQs. These dates cannot be confidently correlated with specific phases of fortification construction. That said, their overall range is strikingly similar to that of the radiocarbon chronology for the ramparts, which suggests a degree of temporal alignment between settlement activity and fortification use.

In the final test models (Supplementary Fig. S21; Supplementary Fig. S22; Supplementary Fig. S23), the phasing of the Bojná I site was simulated based on archaeological data and the results of previous radiocarbon analyses of the fortifications. The modelling and interpretation of samples from archaeological features indicated several equally plausible ways to group the data. However, all options require acknowledging the existence of at least one settlement phase predating the construction of the preserved rampart. This rampart was built at the end of the 9th century and underwent repairs during the first decade of the 10th century. This chronological transition is clearly visible in every radiocarbon model.

The earlier settlement phase most likely dates generally to the 9th century, although it is currently not possible to definitively determine its chronological boundaries. Based on the adopted methodology, it is possible to hypothesise that the duration of this earlier phase may be defined by the TPQs of the respective radiocarbon clusters. According to models for features 1–2 (Supplementary Fig. S17; Supplementary Fig. S18) and the overall test model 2 (Supplementary Fig. S22), the settlement phase preceding the rampart’s construction can be framed within ca. AD 840–900. In contrast, models 3 and 4 for individual features (Fig. S19; Fig. S20) and overall test model 1 (Supplementary Fig. S21) extend this phase much further into the 8th century. While not an implausible hypothesis, this extension does not correlate with the archaeological evidence gathered during excavations at the site. This discrepancy remains unexplained, and it is suspected that the radiocarbon calibration curve may be partly responsible. The modelling process helps to highlight this issue. The two oldest dates from Bojná I (1,400 ± 30 and 1,370 ± 90 BP) are so far removed from the mean values of other samples that they can only serve as very general TPQs, roughly corresponding to the initial stages of regional settlement (see Supplementary Fig. S5). Among all the tested scenarios, test model 3 (Supplementary Fig. S23), which assumes that intensive settlement activity ended before AD 1000, appears to be the most probable.

The test chronological models for the complete dataset (Supplementary Fig. S24) indicate that the period of intensive use of the site is likely to fall between ca. AD 840 and 940. This finding allows the terminus ante quem (TAQ) for the radiocarbon dates to be further narrowed in the final chronological model (Fig. 6), thus potentially setting it at AD 950.

All presented models exhibit a high degree of statistical plausibility; however, the most robust and informative are those that assume the existence of more than one occupational phase at the Bojná I site. These results clearly indicate that future archaeological investigations must seriously consider the possibility of a multi-phased settlement development. The final chronological model (Fig. 6) may serve as a testable hypothesis, capable of guiding subsequent archaeological investigations. Particular attention should be given to the earliest phase of occupation, which extends significantly earlier than the chronology inferred from the artefactual assemblage. Additionally, the refinement of the terminal date for settlement activity would further enhance the precision of the modelling process.

Bojná VI. The dataset obtained from the Bojná VI site demonstrates a high degree of incoherence—most likely due to the limited number of samples—which hinders the identification and potential rejection of outliers. This issue is compounded by the selection of materials that are inherently problematic for radiocarbon dating (e.g., slag and human bone). As a result, the two dates obtained from slag samples taken from a furnace (1,165 ± 30 and 1,015 ± 30 BP) fail the χ² agreement test when combined (T = 12.498, critical value for 5% = 3.8), despite appearing individually plausible. The situation is further complicated by an entirely unreliable radiocarbon date of 10,150 ± 60 BP derived from charcoal found in the same furnace, which strongly reinforces doubts regarding the slag dates⁴⁶. Another questionable result is the date of 1,120 ± 30 BP obtained from a human bone burial beneath Mound 3. While this would imply an interment in the 10th century AD—which is not inherently inconsistent with the overall chronology of the Bojná agglomeration—it does not align with the archaeological evidence recovered from Bojná VI. Chemical analysis of sample Poz-20138/22/4 revealed 0% nitrogen and only 0.8% carbon, which confirms its unsuitability for radiocarbon dating³⁹, and the result must therefore be deemed unreliable. Consequently, the dating of Bojná VI must rely exclusively on two samples—POZ-4 and Poz-172285—whose calibrated dates are consistent with the archaeological record (see Supplementary Fig. S28). These two dates were used in the KDE modelling.

General model. For all sites, a chronological modelling test was also conducted using the KDE method (Supplementary Fig. S26; S27). However, in this case, models constructed for individual sites reflect relatively short chronological spans, and do not yield any conclusive or transformative results beyond indicating peaks of activity (example on Supplementary Fig. S25). Potential occupational phases are not revealed within the models⁴⁵.

Bayesian modelling was demonstrably more effective, as confirmed by comparative analysis. KDE modelling of the entire Bojná agglomeration dataset did, however, confirm the general chronological succession of settlement activity between sites (Fig. 2), though it failed to disclose the finer detail evident in the Bayesian models. This limitation arises from the relatively narrow time window under investigation, and the uneven sample representation, which is heavily skewed by data from a single site.