Salt Production in Central Italy and Social Network Analysis Centrality Measures: An Exploratory Approach

2.1 Salt Production Sites in Central Italy Within the European Context

Prior to the invention and widespread adoption of refrigeration (Rees, 2014), which occurred at the turn of the nineteenth and twentieth centuries, there were only a few methods available for food preservation: drying, smoking, and salting. Among these methods, salting was the most effective and efficient. In this context, salt played a fundamental role in the socio-economic dynamics of ancient societies (Harding, 2021; Weller & Brigand, 2015). It is believed to have been a cornerstone of the early power structures in Rome (Giovannini, 1985) and certainly played a pivotal role in ancient Venice (Hocquet, 1973, 1979). However, due to the high solubility of sodium chloride (salt) in water, it has been quite challenging to trace its presence in ancient times, despite its vital importance. Nonetheless, in the latter part of the twentieth century, the significant integration of scientific approaches into archaeological research provided new tools for detecting salt production, both in terms of its manufacturing and consumption (Alessandri et al., 2019; Flad et al., 2005; Sandu, Weller, Stumbea, & Alexianu, 2012; Sordoillet, Weller, Rouge, Buatier, & Sizun, 2018; Vasilache, Kavruk, & Tencariu, 2020). In Europe, this development has recently sparked a renewed interest in the study of salt and its significance in ancient societies (Alessandri & Attema, 2022).

In general, there are three recognized phases that distinguish ancient salt production: concentration, crystallization, and conditioning. The initial phase, which leads to the formation of brine, involved the natural evaporation of saltwater, possibly enriched with sodium chloride (NaCl). Subsequent crystallization could occur in large artificial basins or by artificially heating the brine inside ceramic vessels placed in specially designed kilns. The former method is still in use in modern salterns. The latter, known as briquetage, based on current knowledge is the older method: the earliest evidence dates back to the Copper Age in Romania (Weller & Dumitroaia, 2005), but the technique quickly spread throughout Europe (Harding, 2013, 2021).

In the briquetage technique, once the salt had crystallized, it was necessary to break the container to extract the salt cake. Sites where briquetage was employed are typically characterized by extensive deposits of ashes and charcoal, the presence of kilns and pits, and, most importantly, large quantities of ceramic fragments attributed exclusively to so called ollae (in which the brine was placed for evaporation), usually in reddish hues. These features, found throughout Europe, were first highlighted along the Tyrrhenian coast of Italy by Pacciarelli at the end of the last century and linked to salt production (Pacciarelli, 1991, 2001). Following that, further investigations have led to the identification of similar sites, all located along the shoreline and often near ancient lagoons (Alessandri et al., 2019, with a list of Italian briquetage sites) (Figure 1).

Figure 1

Sites along the Italian Tyrrhenian coast where briquetage salt-production has been hypothesized and a simplified chronological scheme for Central Italy Bronze Age (from Alessandri et al., 2019 and Belardelli, 2011).

The most recently identified site with these features, and the oldest in Tyrrhenian Italy (Middle Bronze Age), is Caprolace in the Pontine Plain (Alessandri et al., 2019). However, many aspects remain unresolved in the Italian context, and two, in particular, have been the subject of extensive research in recent years. The first concerns the ceramic forms associated with briquetage: while those found in the European context are mainly cylindrical or truncated-conical pots, the Italian ones exhibit a much broader variability, including closed forms (barrel or ovoid shapes) that are not well-suited for the evaporation process. This has led some authors to question the salt hypothesis and instead propose the processing of other edible products, such as fish (Belardelli, 2011, 2013; Di Fraia, 2023). The second issue pertains to the chronology and motivations underlying the transition from the briquetage production method to proper saltworks. In this study, we take a novel additional approach which is the combination of network indices with spatial data to provide novel insights and produce new hypotheses on the second issue: on the important moment of change from briquetage technique to proper saltern mode of production.

2.2 Regional Sites and Routes

For this work, settlements from Latium vetus and Southern Etruria from the Bronze Age to the end of the Archaic Period have been considered. These sites are very well known and documented thanks to a long tradition of studies that goes back to the first topographic studies conducted within the tradition of the aristocratic grand tours of Rome and the Roman countryside during the eighteenth century. British and German aristocrats, fascinated by the possibility of interacting and getting closer to ancient authors through the contemplation and study, were the first to produce catalogues and descriptions of the monuments and environment of the so-called Campagna Romana including both the immediate surroundings of Rome and the southern Etruscan region, respectively, to the south and north of the Tiber river (Dennis, 1848; Gell, 1934; Nibby, 1837). Subsequently, this early activity of survey and documentation was continued by the Antiquarian tradition of the late nineteenth to early twentieth century (Lanciani, 1909; Nissen, 1883–1902; Solari, 1915–20) and the more recent landscape and topographic traditions before (Carta Archaeologica del Territorio di Roma, prepared in the 1920–1930 but only published in the 1960, Lugli, (1962)) and after World War II both by Italian (e.g.: Giuliani, 1966, 1970; Giuliani & Quilici, 1964) and other subsequent works of the Institute of Topography of the University of Rome La Sapienza and the volumes of the Forma Italiae published by them; or Quilici (1974) and other volumes of the Latium vetus series (see below) and international scholars.^[1]

Finally, in the last few decades of the twentieth century many important research projects have been conducted by Italian and/or international teams with modern standards and up-to-date methodologies that have greatly improved the knowledge of the region not only at key excavated sites but also capillary in the territory around them. Among these works, the volumes of the Forma Italiae (Roman School of Topography), the volumes of the Latium vetus series (by Lorenzo and Stefania Quilici Gigli), and the work of John B. Ward-Perkins (1961, 1962) and Tim Potter (1979, 1985) in southern Etruria ought to be mentioned. In addition, an important survey of pre-and proto-historical sites in the territory of Rome was led by Anna Maria Bietti Sestieri (Bietti Sestieri & Sebastiani, 1986; Bietti Sestieri, 1989), while various survey projects and synthesis works have been conducted both in Latium and Etruria, by scholars of the Roman school of prehistory and protohistory, founded by Renato Peroni (e.g. di Gennaro, 1986; di Gennaro, 1988; Guidi, 1984, 1986a,b; di Gennaro & Peroni, 1986; and more recently Alessandri, 2007, 2013; Barbaro, 2010; Pacciarelli, 2001; Schiappelli, 2008). At the same time, the coastal area around the mouth of the Tiber was investigated by the Malafede survey Project (Arnoldus Huyzendveld, Gioia, Mineo, & Pascucci, 1995), while the Pontine Plain and the Nettuno area were intensively surveyed by teams led by Peter Attema from the University of Groningen.^[2]

While a project directed by Helen Patterson and Christopher Smith with the collaboration of Helga di Giuseppe and Rob Witcher focused on an enhancement project of the original Tiber Valley Project of the British School of Rome,^[3] the Sabine region, partially included in this work, has also been investigated by the Galantina project (Guidi, Gabrielli, & Santoro, 2003; Guidi, Rioda, & Santoro, 2008). Finally, the Suburbium Project led by Andrea Caradini and Paolo Carafa has conducted a systematic survey and documentation of both Rome (Carandini, 2012) and its territory, resulting in the Geographical Information System (GIS) available on-line for scholars and the public (http://www.archeositarproject.it/). This work is now continued and enhanced by the New Latium vetus project, directed by Paolo Carafa to produce a similar tool at the regional level in collaboration with Regione Lazio (Figure 2).

Figure 2

Survey Project conducted in central Italy with modern fieldwork methodologies and recording standards (from Fulminante, 2023).

At the same time, in recent years, some synthetic works have been published, which have been points of reference for the present study. First, a project co-ordinated by Regione Lazio has revised all previous studies and produced the Repertorio dei Siti Preistorici e Protostorici del Lazio, a very special and useful tool to approach a great quantity of data with a synthetic but also a very detailed approach. For Latium vetus, we have used data already collected for the work The Urbanization of Rome and Latium vetus from the Bronze Age to the Archaic Era (Fulminante, 2014), also compared with the important works on the same region by Luca Alessandri (2007, 2013). For Etruria, the most important synthetic works for the Final Bronze Age are the works by Barbaro (2010), di Gennaro (1986, 1988), and di Gennaro and Peroni (1986); for the Mid Tiber valley, the work by Schiappelli (2008) is very useful; while Iaia and Mandolesi (2010) have produced a synthesis of the later Early Iron Age sites on the whole Etruria. Finally, for the later Etruscan phases, the work by Marco Rendeli on the territorial organization of southern Etruria in the Orientalizing and Archaic Period (Rendeli, 1993) has been important. All these works have been up-dated based on the review of Studi Etruschi and major Italian and international journals, proceedings of specialistic conferences (e.g. the Proceedings of the biannual conference Preistoria and Protostoria in Etruria or International Conference on Italian Archaeology) and exhibition catalogues (e.g. Della Fina & Pellegrini, 2013).

These settlements are primarily known from either excavation or by survey. However, geophysical prospection has been used in recent years (teams led by the British School at Rome, the University of Siena and the Groningen Institute of Archaeology) (e.g. for salt production sites, see Sevink et al., 2021), which provide new data especially on the built environment and the organization of the space within the nucleated settlements and opens new perspective for multi-scalar analyses. To build the database of sites with their extensions and geographical location, some assumptions have been made following a long tradition of study present in central Italy. According to these assumptions, settlements are hypothesized in the case of coherent oro-graphical units even if only few sherds are available from the survey and/or excavation. This normally applies mainly to Bronze Age and Early Iron Age sites, for which the evidence is scarcer. In addition, some other sites identified because of literary sources but for which archaeological evidence was not available have also been included. This second assumption mainly applies to the Latin region rather than the Etruscan one, for which literary sources are less abundant.

We have considered the maximum period in which the settlements co-existed without major changes and obtained six time slices:

• Final Bronze Age 3 (FBA3): 1050/1025–950/925 BC

• Early Iron Age 1 Early (EIA1E): 950/925–900 BC

• Early Iron Age 1 Late (EIA1L): 900–850/825 BC

• Early Iron Age 2 (EIA2): 850/825–730/720 BC

• Orientalizing Age (OA): 730/720–580 BC

• Archaic Period (AP): 580–500 BC

Both terrestrial and fluvial communications have been considered in this study. To construct the terrestrial and the fluvial route networks, a bidirectional link has been established between two settlements directly adjacent on a terrestrial or a fluvial route without any settlement in between. The fluvial routes have been based on digital data of modern rivers provided by Regione Lazio and published on the website of Ministero dell’Ambiente (http://www.pcn.minambiente.it/viewer/). While some studies are available on the changes in the Tiber River route through time,^[4] to our knowledge, there are no studies available at the regional level. To eliminate recent channels and irrigation works and to obtain the network most likely to have been present in antiquity, the modern rivers have been selected with a query performed in GIS about the superimposition of modern rivers to alluvial deposit, because these are the most likely channel that were probably present also in antiquity.

The terrestrial communication and transportation routes have been reconstructed from hypotheses advanced by various scholars. For Latium vetus, the reconstruction by Lorenzo and Stefania Quilici Gigli (in Colonna, 1976) elaborated at the regional level for the Archaic Period has been used. For the Etruscan region, instead, a comprehensive study is still lacking, although important work on Orientalizing and Archaic road cuts has been done by Tuppi (2014). In order to hypothesize the terrestrial links reconstructive proposal of various authors have therefore been considered: Bonghi Jovino (2008), Brocato (2000), Enei (2001), Potter (1979, 1985), Schiappelli (2008), Tartara (1999), and Zifferero (1995).

The different interpretations have also been tested by considering their alignment with settlements discovered more recently after the publication of those works. According to the topographical principle, the existence and/or use of older tracks has been assumed, if older settlements coherently align with later archaeologically attested roads (Roman and later) and or with natural morphological routes (e.g. River valleys, ridges, etc.) and/or significant archaeological landmark such as funerary tumuli and/or bridges, forts etc. This principle has not only been traditionally and commonly applied in topographical studies both in Italy (see e.g. volumes Forma Italiae) and in Germany (Müller, 1904) but also in the archaeology of the New World (Trombold, 2011[1991]).

Figure 3 shows a summary of the terrestrial network of path in Iron Age southern Etruria and Latium vetus with indication of the sources of the interpretations. To the work cited as used in this study, also the work by Wetter (1962) has to be added that unfortunately has been overlooked at the stage of the reconstruction of the networks and analyses. “Position S” indicates those paths hypothesized on settlement alignments. Unfortunately, it has not been possible to find enough information on paths and routes in southern Etruria during the beginning of the Final Bronze Age (FBA 1–2) and therefore this time slice for terrestrial routes in Etruria has been omitted from the analyses. Both settlements and communication routes have been considered constant within each time slice. In this sense, the analysis concerned static networks rather than an evolving system.

Figure 3

Reconstruction of Early Iron Age Terrestrial routes in Southern Etruria and Latium vetus according to various scholars (from Fulminante, 2023).

However, this does not mean that the system is constant in the six periods. Some sites are abandoned, and some others are founded; therefore, routes are not the only thing that changes but also the settlements. Finally, it ought to be mentioned here that routes and paths have not only been reconstructed topographically at a detailed small scale but also schematically at a large scale for the purpose of building the networks. In future work, we are planning to reconstruct paths and routes within a GIS platform that will make available the detailed topographical data for specialists and for the public.

3.1 Betweenness Centrality

Betweenness centrality indicates the degree to which an actor controls or mediates the “relations between other pairs or dyads of actors that are not directly connected. Actor betweenness centrality measures the extent to which other actors lie on the geodesic path (or the shortest distance), between pairs of actors in the network” (Knoke & Yang, 2008, p. 68). At this point, it ought to be noted that distance in a network is not a geographical distance but the number of links which connects two nodes. In other words, the betweenness centrality measures the extent to which a node or an actor lies on the shortest route connecting each pair of other nodes/actors in the network (Knoke & Yang, 2008, pp. 68–69; but see also the seminal papers by Freeman, 1977; White & Borgatti, 1994). As originally proposed by Freeman, betweenness centrality is given by the formula:

where g jk is the number of geodesic paths between the two nodes j and k (dyad) and g jk ( N i ) is the number of geodesics between j and k that contain i .

Then, dividing g jk ( N i ) by g jk measures the proportion of geodesic path connecting j and k in which node i is involved. Summing across all the dyads not including node i measure the extent to which i sits on the geodesic path of the other network members. Wasserman and Faust suggest normalizing the actor betweenness centrality by dividing it by the maximum theoretical value of ( g − 1 ) ( g − 2 ) 2 assuming each pair has only one geodesic according to the formula:

The standardized actor betweenness centrality is 0.0, when the original betweenness centrality is 0, and it is 1.0 when node i falls on the geodesic path of every dyad of the remaining g − 1 nodes. Therefore, the closer the standardized betweenness centrality is to 1.0, the more the actor controls or mediates relations in the network (Knoke & Yang, 2008, p. 68).

3.2 Closeness Centrality

The closeness centrality of an actor/node, developed by Sabidussi (1966), measures the extent to which a node/actor is close to all other actors/nodes in the network. It is based on the total distance between the node/actor and all other nodes/actors, where larger distances imply lower closeness centrality values. Closeness and distance refer to how quickly an actor/node can interact with others, e.g. by communicating directly or via few intermediaries; again, the geodesic or the shortest path is a key concept and distance means the number of links which connect two nodes and not the geographical distance.

In fact, the closeness centrality ( C C ) of an actor/node ( N i ) is based on the geodesic to all other nodes “and is computed as the inverse of the sum of the geodesic distances between actor i and the g – 1 other actors” (Knoke & Yang, 2008, p. 65):

Again, closeness centrality might vary with network size. Therefore, Beauchamp suggested that the index of actor closeness centrality should be standardized by multiplying it by the maximum of nodes in the network minus 1 (Beauchamp, 1965 quoted in Knoke & Yang, 2008, p. 66):

3.3 Normalized Degree Centrality

The degree centrality measures “the extent to which a node connects to all other nodes in a social network” and indicates how easily information can reach a node. It is based on the assumption that more links and neighbours a node has, the higher the probability of that node to receive information and is given by the following equation (Knoke & Yang, 2008, p. 63):

This means that in a simple and undirected network (where g is the total number of nodes or actors), the degree centrality ( C D ) of an actor or node i ( N i ) is given by the sum of the number of its direct links to the g − 1 other j nodes of the networks that is, in simple term, the number of its neighbours (Knoke & Yang, 2008, p. 63). Actor degree centrality, however, may vary with the size of the network (g or the number of nodes/actors). In fact, the larger the network, or the number of its nodes/actors, the higher the potential of each single node/actor to be directly linked to other nodes/actors. For example, an absolute degree centrality of 3 (which means direct links to three other actors) might represent a high value in a network of 5 nodes/actors but would be a very low value in a network of 50 or more nodes/actors.

Therefore, Wasserman and Faust, to eliminate the effect of variation in degree centrality caused by the size of the network (g), suggest it to be normalized according to the following formula (Wasserman & Faust, 2007, p. 179 quoted in Knoke & Yang, 2008, p. 63):

Then, the normalized degree centrality ( C D ′ ) of a node i ( N i ) is given by its degree centrality ( C D ( N i ) ) divided by the maximum number of possible connections with other nodes/actors that is the total number of nodes (g) minus one, the node itself ( g − 1 ) . In this way, it is possible to yield the “proportion of the network members with direct ties to actor i. Proportion varies between 0.0 indicating no connections with any actor (i.e. isolate), to 1.0 reflecting direct ties to everyone. Normalised degree centrality measures the extent to which an actor is involved in numerous relationships. Actors with high scores are the most visible participants in a network” (Knoke & Yang, 2008, pp. 63–64).