Legacy data from previously published and unpublished research have a critical role to play in future archaeological research. Traditionally, it has been difficult or even impossible to discover, access, and use data from both previously published research and gray literature (McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017:239). Consequently, archaeologists have made concerted efforts to expand the accessibility of data by investing in their long-term curation. These efforts include the expansion of digital repositories, such as the Digital Archaeological Record (McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017), Open Context (Kansa Reference Kansa2010), the Digital Index of North American Archaeology (Wells et al. Reference Wells, Kansa, Kansa, Yerka, Anderson, Bissett, Myers and Carl DeMuth2014), the Comparative Archaeology Database (Drennan et al. Reference Drennan, Hanks, Maguiña-Ugarte and Martín2019), and the Canadian Archaeological Radiocarbon Database (Martindale et al. Reference Martindale, Morlan, Betts, Blake, Gajewski, Chaput, Mason and Vermeersch2016). Archaeologists can promote ethical stewardship of archaeological resources by integrating this increasingly accessible legacy data into new research (Altschul Reference Altschul2016; Cook et al. Reference Cook, Çakirlar, Goddard, DeMuth and Wells2018; Kintigh Reference Kintigh2006; McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017).
To meet this goal, archaeologists must pursue strategies that transform legacy data into new research questions and hypotheses. As we develop new theoretical approaches to understanding past lifeways, existing datasets may provide a critical launch point. A dataset that archaeologists compiled to answer one research question can help archaeologists answer another. Indeed, there has been a recent push for archaeologists seeking new insights on major anthropological and archaeological topics to start their inquiry with existing data (Altschul Reference Altschul2016; Cook et al. Reference Cook, Çakirlar, Goddard, DeMuth and Wells2018; Kintigh Reference Kintigh2006; Kintigh et al. Reference Kintigh, Altschul, Kinzig, Fredrick Limp, Michener, Sabloff, Hackett, Kohler, Ludäscher and Lynch2015, Reference Kintigh, Spielmann, Brin, Selçuk Candan, Clark and Peeples2018; McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017; Witcher Reference Witcher2008). These recent calls have emphasized the need to develop new analytical techniques to transform legacy data into testable hypotheses.
There are well-documented challenges and potentials of working with legacy data (see Bauer-Clapp and Kirakosian Reference Bauer-Clapp and Kirakosian2017; Kansa and Kansa Reference Kansa and Kansa2018; Kintigh Reference Kintigh2006; McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017). In many parts of the world, archaeologists have recorded site locations as well as other site characteristics (e.g., site size, socioeconomic activities, temporal and cultural affiliations) via large-scale systematic survey. Researchers have developed new techniques to use systematically collected legacy databases to address questions about regional organization and social dynamics in the past (see Casarotto et al. Reference Casarotto, Pelgrom and Stek2016, Reference Casarotto, Pelgrom and Stek2019; Spencer and Bevan Reference Spencer and Bevan2018; Ullah Reference Ullah2015).
In many parts of the world, however, archaeologists have not conducted large-scale systematic surveys. In these areas, the discovery of archaeological sites was often driven by opportunity. Professional archaeologists often learn about these sites from amateur archaeologists, farmers who turn up artifacts while plowing their fields, or local informants who have knowledge of cultural landscapes. Our knowledge of these sites is often incomplete. Find spots may only be dots on a map, with no other information about site characteristics. Without this information, it is difficult to discuss aspects of social, economic, political, and ideological organization that are of interest to anthropological archaeologists. These partial datasets, however, are an important resource available to archaeologists. Legacy settlement datasets that lack detail or regular coverage will remain an underused resource without appropriate methodologies for making sense of them. Geospatial analytical techniques may provide the tools necessary to make use of these legacy data.
In this paper, we present a combined GIS and network analysis methodology that uses unsystematically collected site location data to generate testable hypotheses about site characteristics and the organization of regional settlement systems. While the individual components of this methodology are common in landscape archaeology, this methodology represents a novel combination of these techniques. We argue that legacy data can be a critical resource to help predict characteristics of sites and socioeconomic systems (Witcher Reference Witcher2008). We demonstrate the utility of this approach with a case study: Bronze Age (2700–1100 BC) settlement patterns in the mining region of Hunedoara in southwest Transylvania, Romania.
SETTLEMENT SYSTEMS AND EMERGENT HIERARCHY IN BRONZE AGE MINING LANDSCAPES
For decades, archaeologists have analyzed settlement systems to understand past socioeconomic organization (e.g., Blanton et al. Reference Blanton, Appel, Finsten, Kowalewski, Feinman and Fisch1979; Flannery Reference Flannery1976; Wilkinson Reference Wilkinson2000). The ways that people organize themselves in space, connect with each other, and interact with their environment have distinct material consequences (see Barrier and Horsley Reference Barrier and Horsley2014; Birch Reference Birch2012, Reference Birch2014; Duffy Reference Duffy2015; Gyucha Reference Gyucha2019; Wright Reference Wright2014; Wright and Henry Reference Wright and Henry2013). With new technologies and conceptual models, archaeologists continue to improve their understanding of the relationship between people and place. These new approaches include innovative satellite imagery and lidar analyses (e.g., Ebert et al. Reference Ebert, Hoggarth and Awe2016), incorporation of drones into data collection (e.g., Olson and Rouse Reference Olson and Rouse2018), and least-cost path and social network analysis of spatial data (e.g., Hill et al. Reference Hill, Peeples, Huntley and Jane Carmack2015; White and Surface-Evans Reference White and Surface-Evans2012).
For archaeologists working on Bronze Age Europe, settlement systems are an important line of evidence in the study of socioeconomic organization and transformation (Duffy Reference Duffy2014, Reference Duffy2015; Earle and Kristiansen Reference Earle and Kristiansen2010; Galaty Reference Galaty2005; Quinn and Ciugudean Reference Quinn and Ciugudean2018). During the Bronze Age, socioeconomic organization became increasingly centralized and hierarchical (Earle Reference Earle2002). Ultimately, these social transformations resulted in the transcendence of village autonomy and the emergence of complex regional polities with institutionalized inequality (Earle and Kristiansen Reference Earle and Kristiansen2010). Population aggregation and growth led to the development of settlement hierarchies with contemporaneously occupied large towns, small villages, and more isolated hamlets and farmsteads. Monitoring the changing ways in which people positioned themselves relative to other communities and features in the environment can help archaeologists elucidate the socioeconomic processes that led to the emergence of complex regional polities.
There remains significant debate about how hierarchical polities emerged in Bronze Age Europe. Political economic approaches, which are the most common explanatory framework, focus on elite control (see Earle et al. Reference Earle, Ling, Uhnér, Stos-Gale and Melheim2015). Emerging elites may have gained advantages by turning differential access into control of the flows of people and material (Earle Reference Earle2002; Earle and Kristiansen Reference Earle and Kristiansen2010; Earle et al. Reference Earle, Ling, Uhnér, Stos-Gale and Melheim2015). Even within political economic models, however, there are numerous potential bottlenecks in material flows that emerging elites could control. In Bronze Age Europe, this might have included controlling (1) the extraction of raw materials such as copper, tin, gold, and salt; (2) access to nonlocal resources, such as metal, salt, amber, obsidian, and faience; (3) the labor needed to manufacture goods such as metal objects, ceramics, and boats; (4) the modes and paths of movement of materials including rivers, roads, boats, wagons, oxen, horses, and carts; and (5) the human power needed to defend communities and resources, conduct raids or more organized military excursions, and construct fortifications. Each of these political economic mechanisms, which are not necessarily mutually exclusive, has its own set of material consequences. For some of these factors, we can use spatial analyses of settlement systems to reconstruct the socioeconomic processes that underpinned the creation and persistence of social hierarchies.
As a case study, we focus our analysis on the Bronze Age settlement systems in Hunedoara County in southwest Transylvania, Romania (Figure 1). Hunedoara is an ideal context in which to study how socioeconomic processes fueled the emergence of complex hierarchical polities. Southwest Transylvania is home to the largest gold deposits in Europe along with major deposits of copper, salt, silver, and tin (Quinn et al. Reference Quinn, Ciugudean, Bălan and Hodgins2019). Resources from this region were traded widely during the Bronze Age (Stos-Gale Reference Stos-Gale, Meller, Risch and Pernicka2014). Unlike many areas where the control of access to metal has been argued to have been a major factor in the emergence of hierarchical polities (e.g., Earle and Kristiansen Reference Earle and Kristiansen2010), metal is abundant and locally available. Hunedoara is also a key crossroads between the Carpathian Basin to the west and the Transylvanian Plateau to the east. The Mureș River, a major river and corridor for exchange during the Bronze Age (see O'Shea Reference O'Shea2011), passes through the center of the county. In the southern and central parts of the county, an important overland trade route connects the Carpathian Basin to Transylvania through the Hațeg region. By investigating how communities in this critical resource procurement zone positioned their settlements relative to metal resources and trade routes, we can begin to test models about political economic mechanisms that led to the development of complex polities in the region.
Map of Hunedoara County with major topographical features and rivers.
Despite Hunedoara's critical geographic position and mineral resources, archaeologists investigating emergent social complexity in the European Bronze Age have largely overlooked the region. This omission is primarily due to the lack of systematic archaeological surveys in the county. The counties around Hunedoara—Alba to the east and Arad to the west—have been the focus of recent research into the development of social complexity in the Bronze Age (see Nicodemus Reference Nicodemus2014; O'Shea Reference O'Shea1996, Reference O'Shea2011; O'Shea and Nicodemus Reference O'Shea, Nicodemus and Gyucha2019; O'Shea et al. Reference O'Shea, Parditka, Nicodemus, Kristiansen, Sjögren, Paja, Pálfi and Milašinović2019; Quinn Reference Quinn2017). Hunedoara also encompasses the headwaters of the Körös River, known as the Criș in Romania, which has seen significant regional archaeological work (Duffy Reference Duffy2014, Reference Duffy2015; Duffy, Paja et al. Reference Duffy, Paja, Parditka and Giblin2019; Duffy, Parditka et al. Reference Duffy, Parditka, Giblin and Paja2019; Duffy et al. Reference Duffy, Parkinson, Gyucha, Yerkes and Birch2013). Therefore, Hunedoara is an important next piece of the regional puzzle.
The first step in designing systematic archaeological research in Hunedoara is to develop testable models of the socioeconomic organization of Bronze Age communities. This is where extant legacy datasets can be most impactful. In the next section, we present an integrated least-cost path and network analysis approach to developing testable hypotheses about the structure and evolution of community organization in Hunedoara. We use legacy data derived from the unsystematically generated archaeological site gazetteer compiled by Luca (Reference Luca2008). Luca (Reference Luca2008) describes the chronological affiliation of sites as assigned by the original archaeologists or museum specialists who accessioned finds. We digitized settlements (excluding cemeteries and other special purpose sites) affiliated with each of the three chronological subphases of the Bronze Age (Table 1): Early (EBA; 2700–2000 BC), Middle (MBA; 2000–1500 BC), and Late (LBA; 1500–1100 BC). The diachronic study of survey data can highlight fluctuations in settlement pattern centralization and use of different areas of the landscape (Spencer and Bevan Reference Spencer and Bevan2018:71).
Bronze Age Chronology and Archaeological Cultural Affiliations for Southwest Transylvania.
LEAST-COST PATH AND NETWORK MODEL IN TRANSYLVANIAN LANDSCAPES
The development of testable hypotheses about settlement and socioeconomic systems from unsystematically collected survey data involves two broad analytical steps: (1) conducting least-cost path (LCP) analysis in a GIS to build a settlement network, and (2) conducting network analysis to examine the characteristics of the network and individual settlements therein. While LCP and network analyses are not new to archaeologists, the method here represents a novel combination designed to address social questions. The model we present here is simple enough to be replicable but is also amenable to layering on additional complexities where appropriate (e.g., Monte Carlo simulation, multi-cost surfaces). In this section, we outline the process used in the creation of hypotheses about socioeconomic organization in Hunedoara during the Bronze Age (Figure 2).
Conceptual map of the least-cost path and network analysis method.
Building the Network
LCP analysis provides a way to better approximate the movement of people in Hunedoara than simply using the geographic (“as the crow flies”) distance between settlements. LCP analyses have become a staple of landscape archaeological approaches (see Howey Reference Howey2007, Reference Howey2011; White and Surface-Evans Reference White and Surface-Evans2012). LCP analyses create a path between two known features that minimizes the costs for the traveler. The most likely paths people took between two sites was not a straight line. People consider the characteristics of the landscape, often taking a more circuitous route if it will be significantly easier. In mountainous landscapes, such as Hunedoara, LCP analysis can provide data that better match theoretical models of movement and interaction. We focus on the time it would take to walk between sites. We believe that travel time would have been a relevant variable for Bronze Age communities, as it would have been more easily recognized than other ways in which costs can be measured (e.g., calorie expenditure). We next outline the steps and justifications for the model, while the ArcGIS model itself can be found in Supplemental Material 1.
The first step in building the settlement network was creating a cost surface map for LCP analysis. A cost surface is a raster map in which each pixel is given a value for the cost it takes to traverse that pixel. For the mountainous landscape in Hunedoara, we modeled the cost surface using slope. We used an Aster Digital Elevation Model (DEM) with 30 m pixels. The slope for each pixel was calculated in ArcGIS, which is derived from the changes in elevation between pixels in the DEM. A cost surface map was then generated by assigning each pixel a cost (in this case, the time it would take someone to walk across a 30 m pixel with a given slope). The time it takes to walk a set distance at a particular slope can be calculated using Tobler's Hiking Function (Seubers Reference Seubers, Attema, Seubers and Willemsen2016; Tobler Reference Tobler1993). We grouped slopes together into categories, (e.g., 0 to 3 degrees of slope; 3 to 6 degrees of slope, etc.) and assigned them a time based on Tobler's Hiking Function (Table 2).
Cost Surface Values for Walking across the Landscape.
The second step was to digitize the site locations from the legacy dataset for Hunedoara. Site locations (UTM; Zone 34 N) were derived from maps published in the Hunedoara County site gazetteer (Luca Reference Luca2008). The information available about these Bronze Age settlements is highly variable. Some sites have been systematically excavated, whereas others are find spots reported by local community members that have not been investigated further. We used ArcGIS to create shapefiles for each Bronze Age subphase: EBA, MBA, and LBA. The differences between the ceramics of these subphases in Transylvania are robust and are likely reliable indicators of this chronological assessment (see Table 1 for cultural affiliations associated with each subphase). However, finer-grained chronological categorization within each subphase is not possible without excavation and radiocarbon dating (see Quinn et al. Reference Quinn, Ciugudean, Bălan and Hodgins2019).
Third, we calculated cost distance maps for each site in ArcGIS and used the resulting raster to calculate the time it would take to move along the least costly path to each contemporary Bronze Age settlement. Cost distance maps compile the time it would take to move from the original settlement to any other part of the DEM. The amount of time it would take to travel between all sites, calculated using the site shapefiles and cost distance rasters for each site, was compiled for all pairs of sites.
The fourth and final step in creating the settlement network was to make an “edge” between settlements that are within a four-hour walk of each other. Anthropologists and archaeologists have established that complex regional polities are usually limited in spatial extent by a need to control their territory while lacking internally specialized administrative units (of more complex state-level societies) that can allow for delegation of authority to lower-tier settlements (Livingood Reference Livingood, White and Surface-Evans2012). In particular, the upper-level distance from the center of a regional polity that political control could extend is limited to within a half-day's journey (Livingood Reference Livingood, White and Surface-Evans2012:174–175; Spencer Reference Spencer1990:6–8). This distance would allow a chief to interact with all communities in the polity without having to impose on their hospitality, and to mobilize a force to defend more distant settlements and return within a day (Livingood Reference Livingood, White and Surface-Evans2012:175). For nonelites, being located within a half-day's journey of the center would also allow them to access social and economic opportunities in the regional center (Livingood Reference Livingood, White and Surface-Evans2012:175). This last point is critical for middle-range societies lacking centralized political authority, as even without chiefs, interaction would be much more common between communities within a half-day's journey than with communities at greater distances. We use four hours as a cutoff for a half-day's journey. All other potential links between sites that were longer than four hours were discarded. This is not to say that people would have never interacted with communities farther than four hours away. Rather, it assumes that the interaction between communities within a half-day's walk would have occurred much more often than with settlements that would have required overnight stays. With this last modification, we are left with a settlement network for each Bronze Age subphase made up of nodes (settlements) and edges (links between sites within a four-hour walk).
Analyzing Networks and Generating Hypotheses
Network analyses are ideally suited for exploratory analysis of legacy site location data. The applications of archaeological network analysis have grown significantly in their depth and diversity over the past decade (see Brughmans Reference Brughmans2010, Reference Brughmans2013; Knappett Reference Knappett2011, Reference Knappett, Schoep, Tomkins and Driessen2012; Knappett, ed. Reference Knappett2013; Mills et al. Reference Mills, Peeples, Randall Haas, Borck, Clark and Roberts2015; Peeples et al. Reference Peeples, Mills, Randall Haas, Clark, Roberts, Brughmans, Collar and Coward2016). In the Carpathian Basin, network analyses have played a growing role in characterizing the organization and evolution of settlement systems (see Duffy et al. Reference Duffy, Parkinson, Gyucha, Yerkes and Birch2013; Quinn Reference Quinn, Németh, Rezi and Berecki2018). Network analyses are relatively easy and cheap to conduct, can help generate quantitative data for comparison across datasets at different temporal and spatial scales, and are designed to explore central topics in archaeological research, such as interaction and integration. From patchy or incomplete datasets, archaeologists can develop multiple testable hypotheses for how settlement systems were organized. Using the settlement network from Hunedoara, based exclusively upon site locations and local topography, we illustrate how testable hypotheses can be developed using network analysis techniques.
The first step of network analysis was to convert our database of linked sites into a network database in UCINET, a free software platform for network analysis. The settlement network from Hunedoara does not assume directionality on the relationship between sites (e.g., for tribute or redistribution). Instead, we modeled the edges as undirected; that is, where people, goods, and ideas could have flowed equally between both settlements.
In the second and final step, we used UCINET to calculate centrality measurements for the networks for EBA, MBA, and LBA settlement systems in Hunedoara. We focus on two centrality measures in particular: degree centrality and betweenness centrality. While considerable conceptual overlap exists between different measures of network centrality, they are distinct from each other and measure different characteristics of the network (Valente et al. Reference Valente, Coronges, Lakon and Costenbader2008). Degree measures the number of other nodes to which a node is connected. In our model, degree is the number of sites one could walk to within a four-hour radius. Betweenness measures the extent to which a node lies between other actors in a network. Betweenness is how often a particular site falls along a path between other sites. Sites with high betweenness are ones that are difficult to avoid when traveling across a network. Even within the same network, these alternative network centrality measures may identify different critical nodes within the network (Figure 3).
Network showing how different nodes within the same network will have higher centrality scores based on the centrality measure used. The nodes in red have high degree (d) centrality (connected to the most other nodes), while the node in blue has high betweenness (b) centrality (most critical to the flow of materials, people, or information across the network).
The use of multiple centrality measures can help archaeologists test alternative models of socioeconomic organization. Settlements that have high degree centrality can directly influence the highest number of other communities, suggesting that they may be able to mobilize the most people for communal labor projects, warfare, and defense. Settlements that have high betweenness centrality will be able to control the flow of information and nonlocal goods and resources through a system, creating bottlenecks that emergent elites may be able to manipulate. Archaeologists can use the result of these network analyses to identify key sites to prioritize for future systematic investigation. If site characteristics (e.g., site size, quantity of prestige goods) match one of the predicted networks, we would have evidence of the differential importance of the mobilization of labor (if degree centrality predicted the key sites) or the control of trade and exchange (if betweenness centrality predicted the key sites).
RESULTS: MODELED NETWORKS FOR BRONZE AGE HUNEDOARA
The GIS and network analyses reveal significant transformations in settlement systems throughout the Bronze Age.
Early Bronze Age Networks
Thirty-five settlements are attributed to the EBA (Supplemental Material 2). All but five of these sites are integrated into one settlement interaction network, with the other five sites positioned high in the Apuseni Mountains (Figure 4). The settlement network is densest in the Mureș River valley, with connections to the Criș (Hungarian: Körös) river system to the northwest. Four of the six sites with the highest degree centrality are in the modern town of Deva positioned along the Mureș River, on its southern terrace (Table 3). While this may be partially the result of oversampling due to the presence of the modern town, it is important to note that this area is not overrepresented in other time periods. Deva is strategically located at the intersection of the Strei and Mureș Valleys, with views up toward the metal-rich southern Apuseni Mountains. The other sites with high degree centrality, Boholt-Ciuta and Cărpiniș-Comoara, are located as close to the Mureș as possible, while still being within easy access of nearby mineral deposits. The site with the highest betweenness centrality is Dealu Mare-Ruști, positioned near the region's rich metal resources and the land between the Mureș Valley and the headwaters of the Criș. The next two sites with the highest betweenness centrality (Brănișca-La Tău 2, Brad-Dealul Ștefanului) are also located in the metal-rich Apuseni Mountains in the northwest quadrant of the county.
EBA settlement network with nodes reflecting degree centrality and betweenness centrality. Key sites, labeled by their Site IDs, are (3) Boholt-Ciuta, (6) Brad-Dealul Ștefanului, (7) Brănișca-La Tău 2, (13) Cărpiniș-Comoara, (17) Dealu Mare-Ruști, (18) Deva-Cartierul Viile Noi, (19) Deva-Cimitirul Reformat, (20) Deva-Dâmbul Popii, and (21) Deva-Magna Curia.
Sites Predicted to Be Prominent Based on Legacy Data in Luca (Reference Luca2008).
Middle Bronze Age Networks
Significantly more sites in Hunedoara, 79 in total, have been assigned to the MBA than the EBA (Supplemental Material 3). All but three of the 79 sites are integrated into a single settlement network (Figure 5). Unlike the EBA, the MBA settlement system centered on the Mureș and Strei Valleys. The Strei Valley connects to the Timiș River system southwest of Hunedoara through the Hațeg region (also known as the “Transylvanian Iron Gates”). This would have been a path for the movement of people and goods overland, as this pathway does not fall along a highly navigable river like the Mureș. The Mureș Valley continues to be critical during the MBA, while the Apuseni Mountains are less of a focus for settlement than during the EBA. The four sites with the highest degree centrality are all positioned along trade routes: one at the confluence of the Mureș and Strei Valleys (Deva-Cimitirul Ceangăilor) and three further south along the Strei (Bretea Streiului-Grumedea, Peștișu Mare-La Țărmure Sud, Silvașu de Jos-Între Ogăși; Table 3). Similarly, the four sites with the highest betweenness values also focus on the Mureș (Rapoltu Mare-Șeghi) and Strei Valleys (Bercu-Vârcolin, Silvașu de Jos-Între Ogăși, Strei-Canton CFR). None of the sites with the highest degree or betweenness centrality values are in areas where communities would have had direct access to the most abundant metal deposits.
MBA settlement network with nodes reflecting degree centrality and betweenness centrality. Key sites, labeled by their Site IDs, are (11) Bercu-Vârcolin, (14) Bretea Streiului-Grumedea, (28) Deva-Cimitirul Ceangăilor, (60) Peștișu Mare-La Țărmure Sud, (61) Rapoltu Mare-Șeghi, (69) Silvașu de Jos-Între Ogăși, and (71) Strei-Canton CFR.
Late Bronze Age Networks
Only 13 sites are attributed to the Late Bronze Age (Supplemental Material 4). The sites are separated into two clusters, with two sites alone on the southern and northern edges of the county (Figure 6). The larger cluster, connecting eight sites, centers on the western side of the Strei Valley. Most sites in this cluster are set back in the uplands above the valley, though two are located to the north along the Mureș River. There are only three sites in the smaller network, along the southern terrace of the Mureș River in the eastern half of the county. Valea Nandrului is predicted to be a critical site by both centrality measures (Table 3). This is in contrast to the EBA and MBA, when no single site has the highest value across both centrality measures. This is the type of pattern we would expect for the presence of institutionalized regional hierarchies where elites have control across multiple dimensions of socioeconomic organization. The site has a strategic position within the network that would allow elites to both marshal labor from nearby communities as well as control the flow of goods, resources, and people within the local area. However, Valea Nandrului is set back from the trade corridors along the Mureș River and the floor of the Strei Valley that had been key to the MBA settlement network.
LBA settlement network with nodes reflecting degree centrality and betweenness centrality. Key sites, labeled by their Site IDs, are (3) Cerișor-Peștera Mare, (5) Deva-Cartierul Viile Noi, (7) Peștișu Mare-La Tarmure Sud, (8) Peștișu Mic-(no name), and (12) Valea Nandrului-(no name).
DISCUSSION: BRONZE AGE HUNEDOARA SOCIOECONOMIC ORGANIZATION AND TESTABLE HYPOTHESES FROM LEGACY DATA
The results of the GIS and network analyses allow us to construct a model for the development of settlement systems in Hunedoara. Over the course of the Bronze Age, there was an increased focus on monitoring interregional trade routes in the lowlands at the expense of controlling ore-rich mountain landscapes. During the EBA, communities placed their sites and constructed interaction networks that connected people living in the Mureș Valley with communities in the southern Apuseni Mountains. More so than in other periods, EBA communities centered their interactions within metal-rich landscapes. Beyond settlement patterns, evidence in burial practices supports the conclusion that there was significant interaction among communities in the mountains. EBA communities in southwest Transylvania primarily buried their dead under stone- or earthen-covered tombs along mountain ridges (Ciugudean Reference Ciugudean, Berecki, Németh and Rezi2011). By placing tombs in highly visible positions, EBA communities could mark territory and secure access to local resources through their ancestors (see Goldstein Reference Goldstein, Chapman, Kinnes and Randsborg1981). The investment in these monuments makes sense if people were routinely traversing these highland landscapes as they moved between settlements.
By the MBA, settlement networks had changed significantly. MBA communities placed their sites along the most important interregional trade routes along the Mureș River and the Strei Valley at a greater rate than before. These valleys are significant interregional trade routes, especially along the Mureș River and the overland route through the Strei Valley and Hațeg region. Because of the significant increase in the number of sites attributed to the MBA, the settlement network was much denser, and more connected, than other subphases of the Bronze Age. Some of this may be the product of assuming contemporaneity, as not all sites were occupied for the entire 500-year sequence. However, Duffy and coauthors (Reference Duffy, Parkinson, Gyucha, Yerkes and Birch2013) have also documented an increase in network connectivity due to an increase in the number of sites in the MBA in the Körös region.
There was another shift in settlement networks in the LBA. During this period, the drop in the number of sites could have been the product of population aggregation, as seen in neighboring regions at this time (e.g., Szentmiklosi et al. Reference Szentmiklosi, Heeb, Heeb, Harding, Krause and Becker2011). Communities continued to construct settlement networks that centered on the major interregional trade routes. Some LBA communities, however, placed their sites at a distance from the Mureș River and Strei lowlands in more prominent topographic positions. From these positions, communities would have been able to monitor exchange networks but would also have been insulated from the risks (e.g., raiding) and benefits (less costly access to trade) of living along these key pathways.
The overall pattern in Hunedoara settlement networks that we reconstruct here is similar to the patterns seen in neighboring regions during the Bronze Age. Using a different GIS analytical method, Quinn and Ciugudean (Reference Quinn and Ciugudean2018) demonstrated the shift away from metal procurement landscapes toward interregional trade routes from the EBA to the MBA. The shift in settlement networks from the MBA to the LBA is also consistent with disruptions in the organization of settlement systems seen after 1500 BC throughout the Carpathian Basin and Transylvania (see Ciugudean and Quinn Reference Ciugudean, Quinn, Németh, Rezi and Berecki2015; Duffy, Parditka et al. Reference Duffy, Parditka, Giblin and Paja2019; O'Shea Reference O'Shea2011; O'Shea et al. Reference O'Shea, Parditka, Nicodemus, Kristiansen, Sjögren, Paja, Pálfi and Milašinović2019). The emergence of large fortified sites during the LBA is consistent with the presumed population aggregation and selection of more-defensible site locations seen in this model (e.g., Gogâltan and Sava Reference Gogâltan and Sava2010; Szentmiklosi et al. Reference Szentmiklosi, Heeb, Heeb, Harding, Krause and Becker2011; Uhnér et al. Reference Uhnér, Ciugudean, Bălan, Burlacu-Timofte, Hansen, Rustoiu, Hansen and Krause2018).
We are now ready to ask the question: What were the socioeconomic processes that underlay this model of changing community organization within Hunedoara throughout the Bronze Age? To answer this question, we must conduct systematic fieldwork in Hunedoara. We can use the results of the GIS and network analyses to help design future research.
We now have two alternative hypotheses (see Table 3). If Hunedoara communities were organized around controlling labor, then we predict that the sites with the highest degree centrality would be the most prominent sites within the settlement systems. If differential positioning within the network to control the flow of materials locally and through long-distance exchange were more critical to socioeconomic organization at the regional level, then we predict that the sites with the highest betweenness centrality would be the most prominent. Prominence, in this case, may be measured in site sizes (demographic prominence) or evidence for differential access to material wealth (e.g., exotics, fineware ceramics, ornamental metalwork). Alternatively, sites that have low centrality (see Supplemental Materials 2, 3, and 4) would be expected to have lower regional prominence, depending on whether control of labor or exchange networks were more important factors in the regional settlement system.
To evaluate these two hypotheses, we must design and conduct systematic archaeological investigations based on assessing the predicted site characteristics. We recommend a research program that combines mapping the horizontal extent of settlement, assessing the depth of deposits, and collecting datable material associated with diagnostic material culture. Documenting site sizes and economic activities through additional survey and excavations will provide the data needed to support one of these two hypotheses. If neither degree centrality nor betweenness centrality accurately predicts the key sites within the region, then it is likely that other factors that are not part of these models would have been the most important factors in organizing people in space during the Bronze Age in Hunedoara. We can use the models derived from legacy data to inform our choice of which sites to prioritize for future research. Archaeologists should target sites that the model predicts are important based on the centrality measures (see Table 3 for specific sites) as well as sites that are predicted to be less important to the overall settlement network. These models also generate pilot data, which are necessary to secure funding and permits within Hunedoara.
Legacy data are notoriously unreliable, even in regions that archaeologists have systematically surveyed (e.g., Ullah Reference Ullah2015). In Hunedoara, where there has been no systematic survey, there are likely many unrecorded sites. This may affect the models we have developed in different ways. For example, there is a gap between two settlement clusters during the LBA. If there were sites in this region that were not included in the gazetteer, they would instantly become critical nodes (with high betweenness) in the settlement network model. This model is more sensitive to missing data when there are few known sites (e.g., LBA) than when there is a more robust sample (e.g., EBA and MBA). There are additional analytical concerns, including whether all sites were contemporaneously occupied (e.g., during the MBA), or whether certain types of sites are overrepresented or underrepresented. During the MBA, for example, ceramics belonging to the Wietenberg Culture were decorated with diagnostic motifs and techniques (Bălan et al. Reference Bălan, Quinn and Hodgins2016). The visibility of these motifs, and comparably fewer diagnostic designs for the EBA and LBA, might have resulted in an overrepresentation of MBA sites when compared with the other subphases of the Bronze Age. Recently revised chronological models also suggest that the Wietenberg persisted into the LBA in southwest Transylvania (Quinn et al. Reference Quinn, Ciugudean, Bălan and Hodgins2019). The potential problems with the legacy dataset cannot be resolved without new systematic survey, excavation, and radiocarbon dating programs.
As we gain more insights into the nature of the archaeological record in Hunedoara, we can start to add more complexity to our model. With stochastic approaches such as Monte Carlo simulations, we can test how the duration of occupation affects the organization of the settlement network. To construct these simulations, we must better understand duration of occupation of sites in Hunedoara—something that requires more fieldwork and radiocarbon dating. We could also vary the travel costs within the landscape. We could include the cost of crossing waterways, time diverted from traveling for social obligations when passing a village, and efficiency of different modes of travel such as by boat or horse. We have chosen to wait to add these other criteria to the cost model presented here, as each has built-in assumptions (e.g., available technology, how prehistoric land use varied from modern land use, etc.). While we argue that the topography of the mountainous Hunedoara would have been the most significant cost, we hope that future modeling can involve running multiple scenarios with multi-criteria cost surfaces. These more complex approaches will have greater value if the simple models presented here do not accurately predict site characteristics.
More broadly, archaeologists can use the method described here to model the socioeconomic processes that underpinned settlement networks in any region where site location information is available. As shown here, the method works with piecemeal legacy datasets. However, the method also has significant potential to add to our understanding of settlement systems in places with systematically collected survey data. Many of the classic archaeological surveys, in locations such as the Viru Valley (Willey Reference Willey1953) and the Valley of Oaxaca (Blanton et al. Reference Blanton, Kowalewski, Feinman and Appel1982), remain critical datasets that archaeologists continue to use to conduct research. Archaeologists can apply the LCP and network analysis method to test alternative hypotheses about how these settlement systems were organized. The data to evaluate hypotheses in these contexts, including site sizes, presence of exotics, and evidence of wealth inequality, may already be contained in these survey reports. This is especially important in some of the older systematic survey datasets. In many of these places, human-led development and climate change have destroyed the sites in the years and decades since archaeologists recorded them. In these cases, legacy data from large-scale surveys will remain the only archaeological data about settlement patterns after the destruction of these nonrenewable resources.
CONCLUSION
In this article, we demonstrate how archaeologists can use legacy data to develop testable hypotheses about socioeconomic organization in the past. We present a new integrated GIS and network analysis model for predicting site characteristics using site location data. Using a case study from southwest Transylvania during the Bronze Age, we show that this approach is useful for characterizing settlement networks. In Hunedoara, settlement systems increasingly centered on interregional trade routes over metal-rich mountain landscapes throughout the Bronze Age. Using the results of GIS and network analyses, archaeologists can more efficiently design a fieldwork-based research program to explore when, and how, regional communities became hierarchically organized. Researchers can employ this method in any archaeological context where site location and terrain information are available. Across North America, there are many regions where legacy site location data could be explored using this geospatial technique. The method is particularly useful in contexts where there have been no previous systematic archaeological surveys, and therefore inconsistent or incomplete information about each site within the region. This method can also be applied to systematically collected settlement pattern data, with the potential to contribute new insights into previously surveyed regions. Network analyses and LCP analyses continue to provide new perspectives into past settlement systems as they become an increasingly important part of landscape archaeologists’ tool kit. As site location data becomes increasingly accessible through digital repositories, state archaeology site files, dissertations, theses, and technical reports, archaeologists can use geospatial techniques such as the method presented here to develop novel insights into settlement systems and socioeconomic organization in almost any archaeological context. As we have shown, these methods can reinvigorate legacy data as part of future archaeological research design.
Supplemental Material
For supplemental material accompanying this article, visit https://doi.org/10.1017/aap.2019.37.
Supplemental Material 1: Least-cost path model used in this study.
Supplemental Material 2: Complete list of Early Bronze Age settlement locations and results of network analyses.
Supplemental Material 3: Complete list of Middle Bronze Age settlement locations and results of network analyses.
Supplemental Material 4: Complete list of Late Bronze Age settlement locations and results of network analyses.
Acknowledgments
We would like to thank Lisa Young for guidance in this study. This study has also been significantly improved by insightful comments and suggestions made by Lacey Carpenter, Bree Doering, Christina Sampson, Horia Ciugudean, Sarah Herr, and three anonymous reviewers. We also thank Sofía Pacheco-Fores for help preparing the Spanish abstract.
Data Availability Statement
Data employed in this study are included as supplemental material.
