Marble trade in the Roman Mediterranean: a quantitative and diachronic study

Abstract

Marble provenance studies in archaeology have become increasingly popular in recent decades. This has resulted in a large quantity of analytical data becoming available for archaeological marbles. This article presents the results of a quantitative study of the distribution of white marble in the Mediterranean based on an analysis of the available provenance data for the Roman period. The study shows increased distribution of white marble between the late 1st c. BCE and the end of the 2nd c. CE. A decline in distribution from the 3rd c. CE was less abrupt than traditionally believed and shows object-, material-, and region-specific trajectories. The marble distribution data is finally evaluated within a wider socio-economic frame, considering factors such as the marble trade system and broader Roman economy, changes in cultural practices related to statue erection, importance of reuse and recycling, growing ruralization, and reduced interest of the elite in urban capital investment in the later Roman periods.

Recent decades have witnessed a growing importance of archaeological data in studies of how past economies functioned and evolved through time. Much of this progress results from advances in scientific (provenancing) techniques, the application of (geo-)statistical methods, and increased interest among archaeologists and historians in what the material record can tell us about past socio-economic systems.¹ Commonly used archaeological proxies for the economy of the Roman period include amphorae, various kinds of fineware pottery, shipwrecks, wine and olive presses, and fish-salting infrastructure.² While studies of high-quality raw materials, such as marbles, are not uncommon in archaeology, the full potential of this material category as proxy data for the Roman economy has not yet been fully explored in the field of economic archaeology/history.

Roman society was highly stratified, and the upper classes were constantly striving to showcase, maintain, and increase their power and prestige. Monumental architecture and sculpture, in both public and (semi-)private contexts, were among the most powerful means of doing this. The grandeur of monumental architecture was expressed not only by the sheer investment of energy, manpower, or funds involved in the building of its structural components, but also by its elaborate decor; in essence, the scale and elaboration of monumental buildings largely exceeded their functional requirements.³ Ancient cities were lavishly adorned with marble statuary and marble(-clad) architecture, mainly through benefaction by members of the elite.⁴ Architectural and sculptural decoration were considered integral parts of any monumental building, contributing to the message the building conveyed as a whole. As such, Roman architectural and sculptural decoration was highly functional. It served as an organizing and structuring element within a building,⁵ providing the physical setting where people lived, met, and interacted. This function is evident for public buildings, but elite housing also served a public purpose as a location of contact between the elite owners and their clientele. More fundamentally, decor had a distinct social function. It was intended to be seen by visitors. Lavish decor displayed wealth and prestige, and was associated with the Roman elite, who had substantial financial means to acquire luxury materials. It acted as a social marker with a clear communicative effect on outsiders. It was used to express and reinforce membership of the elite, as well as to display, perpetuate, and increase the political, social, and economic power and status of the resident/benefactor.⁶ Monumental marble statuary and marble(-clad) architecture were therefore an essential part of the Roman urban environment, and their study can contribute to our understanding of the level of urban investment by the upper classes. Roman cities were adorned to celebrate and reinforce the power of benefactors. As such, marble decoration can, in addition to building histories, function as an indicator for the level of urbanization. In addition, the act of monumental building shows the presence of a societal class that had the necessary means to mobilize and coordinate the manpower required for construction. Monumental architecture thus serves also as an indicator of social complexity and hierarchy.⁷

The Mediterranean basin has abundant attractive white marbles, many of which were exploited in the Roman period. The distribution of these marble resources is, however, very uneven, with the major exploitation districts clustered mainly in the Italian Peninsula, mainland Greece, the Aegean Islands, and Asia Minor. Smaller production centra can be found, for example, in the Iberian Peninsula (Fig. 1). This uneven distribution of resources means that trade in marble was unavoidable. Marble objects were traded in enormous quantities and over long distances in the Roman period, much like other objects (wine, olive oil, pottery, etc.), and as such reflect wider economic practices.⁸ The importance of marble for Roman society, combined with its durability, provenancing potential, and chronological resolution – marble objects (e.g., building elements, sculpture, sarcophagi) are generally well dated – make it a very promising subject of research for archaeologists and historians interested in the economy of antiquity.

Fig. 1. Geographical distribution of the marble quarry districts and Roman sites with white marble provenance determinations of objects dated between 200 BCE and 500 CE. (D. Taelman.)

Since the mid-1980s, when scientific marble provenancing in archaeology started, research has concentrated mainly on developing techniques and methodologies for material characterization and provenance determination.⁹ Advances in petrographic and archaeometric methodologies, in combination with extensive quarry sampling, have in recent years allowed marble artefacts to be provenanced with high reliability. Typically, archaeologists have explored the Roman marble economy through case studies, that is, individual monuments, collections, or archaeological sites. While this case study-based approach is undeniably valuable, there are certain questions that can only be answered by studying the marble provenance data in its entirety. Such synthetic, regional, or supra-regional research is still scarce.¹⁰ In particular, quantitative approaches to the Roman marble economy are still lacking, despite the general consensus on the value of more and better quantification and the need to reach firmer, falsifiable hypotheses in the understanding of ancient economies.¹¹

In this article, I focus on published white marble provenance data of quarry districts that exported on a supra-regional scale in the Roman Mediterranean (200 BCE–500 CE). Thanks to the rising popularity of white marble provenance studies in recent years, there are now sufficient reliable provenance determinations for white marble, with a large enough regional and chronological spread, to allow for quantification on a large scale. As a result, provenance determinations for white marble can now serve as a meaningful proxy indicator for studying long-distance trade, changes in urbanization, and elite urban investment in the Roman world, as well as for examining the structure, behavior, and performance of the wider Roman economy. Analysis of the data focuses on revealing and visualizing general trends, patterns and associations in the rise and fall of market shares and the distribution of individual marble resources across the Mediterranean. Without ignoring the effect of taphonomic processes, changes in consumption and distribution patterns are charted here with the aim of identifying relative changes in the volume of marble traded throughout the Roman Mediterranean. Distribution patterns of white marbles are compared diachronically through standard descriptive and exploratory quantitative methods, looking specifically at similarities and dissimilarities between regions and object categories. Finally, the quantitative data are discussed with the aim of identifying the mechanisms that may have produced the observed patterns and trends. Changes in production systems of quarries, changes in the exchange system (level of market integration, regionalization of the markets, etc.), and changes in the consumption patterns of marble (consumer taste, urbanization rates, etc.). are considered.

Materials and quantitative approach

White marble provenance dataset

Data from 417 publications with provenance determinations for archaeological white marble have been collected, resulting in a corpus of 8,722 records of individual white marble artefacts from 434 archaeological sites, mainly around the Mediterranean Sea. Marble objects were only added to the database if their provenance determination was considered reliable, that is, based on a generally accepted multimethod approach combining at least mineralogic-petrographic and archaeometric analysis.¹² The topic of this article is the Roman Mediterranean between 200 BCE and 500 CE. All marble objects that fall within these chronological and geographical limits were extracted from the dataset for further analysis. This constituted a subset of 7,017 objects from 350 Roman sites, provenanced to different 65 quarry districts (Fig. 1). The dataset was then further refined to focus on the quarry districts that exported mainly on a supra-regional scale in Roman times. Quarry districts included in the analysis are Aphrodisias, Carrara, Dokimeion, Ephesos, Göktepe, Naxos, Paros (Paros-1, Paros-2(3)), Pentelikon, Prokonnesos and Thasos (Thasos-1(2), Thasos-3).¹³

Each database entry represents an archaeological marble artefact, and the record includes findspot, storage location, object type, chronology, analytical data, material provenance, laboratory, and literature reference, as well as the immediate and wider contexts of the object (e.g., historical background of the site, building or monument). Chronology was assessed on the basis of the information presented in the literature. Objects were typologically classified into six main classes: architectural elements (ashlar, veneer, column elements, etc.), sculptural objects (statuary, relief sculpture, etc.), epigraphic documents, utilitarian objects (vases, plates, chandeliers, cosmetic plates, etc.), sarcophagi, or undefined (Fig. 2A). The two most frequent functional classes, that is, architectural and sculptural white marble objects, were a further focus.

Fig. 2. White marble object categories included in the study: (A) count per object category; (B) chronological resolution per object category. Object type abbreviations: ADC = architectural element, SCU = sculptural object, ISC = inscription, UTI = utilitarian object, OTH = undefined type, SRC = sarcophagus. (D. Taelman.)

While the strengths of marble as a proxy indicator for the Roman economy have already been highlighted (popularity, durability, provenancing potential, quantity of available data, and chronological and geographical resolution), the material is obviously not free of biases that typically affect the amount and type of archaeological information at hand. These are primarily related to depositional and post-depositional processes, as well as research and publication history. Understudied (or underpublished) regions are certainly the Eastern Adriatic and the Balkans, North Africa, the inner Iberian Peninsula, northwestern Europe, and inner Asia Minor. Nonetheless, the database for this article represents the most complete overview of white marble in the Roman Mediterranean currently available.

Biases that relate specifically to marble essentially involve reuse/recycling/redistribution, as well as differential sampling strategies and intensities. Differential sampling strategies and intensities relate primarily to the geographical and chronological spread of case studies and to the choice and number of objects analyzed. The most common approach to sampling in marble provenance studies has been to maximize the lithological variation of marble types at each findspot. Quantification of absolute counts of provenanced marble objects is, therefore, not always straightforward, representative, and/or comparable between findspots. While less problematic from a production perspective, this is a significant issue when focusing on distribution and consumption patterns. It must also be noted that differences in the quantity of provenance determinations available for certain findspots are related not only to differential sampling intensities, but also indirectly to differences in the quantities of marble material imported, and therefore to historical realities. This is particularly the case for Rome, where without doubt significantly more marble was imported than at most other sites in the Mediterranean. The site of Rome, for example, accounts for almost 1,300 provenanced marble objects in the database. Other sites with high levels of marble use in the Roman period, such as Constantinople, Leptis Magna, etc., are, however, less well represented in the database due to different research focuses and/or different sampling strategies and intensities. To counter these effects of research focus and sampling strategies and intensities, this article concentrates on the presence/absence of marble varieties at findspots in a particular period instead of looking at the total sample count for findspots.¹⁴

Because of the physical properties and intrinsic value of the material, life histories of marble objects can be complex, sometimes with several cycles of reuse, recycling, and redistribution. Much of the material, especially in the Late Empire and Late Antiquity, was salvaged and reused or recycled for re-application, recarving, or use in lime production.¹⁵ Where the context of the objects indicates signs of reuse/recycling, this was accounted for upon entry in the marble database by focusing on the earliest identifiable date for the object.

While these biases surely affect the dataset, I believe that its size, the focus on presence/absence counts, and having accounted for the immediate and wider context of the sampled objects reduces the effect of these weaknesses to a large extent.

Chronology and temporal uncertainty

The chronology of the marble objects under study is derived either from their stratigraphic context or, where the objects originated from open contexts or secondary depositions, from stylistic characteristics. Regardless of depositional context, exact dates of production and/or consumption are generally not available. As in most archaeological material studies, chronology is normally given using date ranges. Apart from the intrinsic temporal uncertainty of the collected archaeological material, we are confronted with objects that have different levels of temporal uncertainty. While some marble objects can be dated very narrowly (e.g., Imperial portraits, inscriptions), others can only be assigned to cultural periods (e.g., Early Empire, Late Roman, Hadrianic) or centuries (e.g., 1st c. CE). Figure 2B, shows, however, that with the exception of the undefined object types, it safe to say that all object categories show a relatively narrow chronological resolution, with mean values generally well below 100 years. Figure 3 further displays the relation between the temporal resolution of the marble objects included in the study and their distribution over time, using an alternative visual method known as the MR-diagram (midpoint-radius diagram), developed by Zenon Kulpa.¹⁶ Here, every marble object is represented as a point in time on the basis of its date range (x-axis) and temporal resolution (y-axis). Using this method, it is possible to represent visually how many objects are dated to a particular time period and also what objects have relatively higher or lower temporal reliability. Figure 3A, for example, shows that most architectural marble objects date from 50 BCE to 250 CE, with most having a temporal resolution below 100 years. For the period after the 3rd c. CE, dating of the objects becomes less precise, as illustrated by the higher values on the y-axis. Figure 3B illustrates that sculptural marble objects included in the study from the 2nd/1st c. BCE have varying temporal resolutions, and that most of the sculptural objects date from the late 1st c. BCE to the middle of the 3rd c. CE. In this period, temporal resolutions for most objects are below 50 years. For both material categories, the data for the Late Imperial period and for Late Antiquity are less abundant, but overall chronological reliability remains relatively high.

Fig. 3. Chronological resolution of the white marble objects included in the study, as calculated using the midpoint-radius method. The size of the points represents the number of objects: (A) architectural objects; (B) sculptural objects. (D. Taelman.)

The problem of temporal diagnosticity remains pressing, however, when looking at medium- and long-term trends in the rise and fall of market shares of marble resources, and when aiming to compare the trade in marble with that of other (archaeological) proxies for the Roman economy. To deal with the issue of temporal uncertainty, and to enable graphing the variability in the objects' distribution over time with different levels of chronological resolution, a chronological distribution method has been applied to the objects in the marble dataset. For archaeological studies, several chronological distribution methods have already been proposed, the principal of which are midpoint chronology and aoristic weighting.¹⁷ The idea behind these methods is that they apply a probability distribution over an object's date range to assign the object to a particular temporal block. Both methods assign a temporal probability weight to a specific event of unknown exact date, based on its potential start and end dates. They differ in how they distribute date probability over the object's time span, with each approach having its own strengths and weaknesses (e.g., computational requirements and robustness to outliers, that is, objects with very low chronological resolution). Midpoint chronology assumes that an object was produced exactly in the middle of its presumed date range,¹⁸ while aoristic weighting applies a probability curve over the object's time span. Different probability curves can be applied; for example, a linear distribution, which assumes an equal probability over the object's date range, or a normal (bell-shaped) distribution, which assumes a higher likelihood in the middle of the object's date range than towards the start and end dates.¹⁹

For this article, marble objects were chronologically assigned to a particular period using aoristic weighting with a linear distribution. This method was chosen to reduce interpretational difficulties, as we generally have very little or no archaeological indication of the probability distribution of the production/distribution of marble objects within their date ranges. The timeline was divided into time blocks of 25 years, a choice that was based on and can be justified by the chronological resolutions of the objects under study (Fig. 2B). However, instead of summing the aoristic weighting results per time block, a simulation approach was applied to the distribution method with the goal of accounting for the probabilistic nature of the temporal distribution. This approach involves the creation of multiple possible “scenarios” of results, in this case 10,000 simulation runs, with the probability of occurrence for each trajectory being defined by the aoristic probabilities.²⁰ In this way, the difference between higher and lower temporal intervals (e.g., time blocks of 50 years instead of 25 years) has less influence on the results of the object chronology distribution. Raising or lowering the length of the time blocks will mainly affect the uncertainty level for that particular block. The chosen time interval, therefore, principally only affects the coarseness of the final curves.

White marble distribution trends and patterns

An initial exploration of the marble distribution graphs of Figure 4 shows a steady rise in the number of sites at which white marble has been attested in the Late Republic, with a minor peak in the first half of the 1st c. CE. A more pronounced peak can be observed for the 2nd c. CE, a period in which both architectural and sculptural marble had their widest distribution, but overall slightly more so for sculptural marbles: sculptural marbles are attested at more sites than architectural marbles. A gradual decrease in the distribution of architectural marbles followed during the Late Empire, and there was a stable period in Late Antiquity, with distribution figures similar to those of the early 1st c. CE. For sculptural marble, a significant decline set in at the beginning of the 3rd c. CE and distribution virtually halted in the mid-4th/5th c. CE.

Fig. 4. Number of sites per 25-year period at which white marble objects have been attested (bar height represents the mean number of sites, error bars the 95% confidence interval of the 10,000 simulations): (A) architecture; (B) sculpture. (D. Taelman.)

While Figure 4 shows a relatively straightforward pattern, with a peak in distribution for both architectural and sculptural marbles in the 2nd c. CE, the graphs in Figures 5, 6, and 7 display a more diverse and complex picture, with marble-, object-, and region-specific distribution patterns, and with significant overlap in the distribution areas of different marble varieties.²¹

Fig. 5. Number of sites per 25-year period and per marble variety at which white marble objects have been attested (site count represents the mean number of sites of the 10,000 simulations): (A) architecture, (B) sculpture. Quarry abbreviations: APH = Aphrodisias, CAR = Carrara, DOK = Dokimeion, EPH = Ephesos, GOK = Göktepe, NAX = Naxos, PAR-1 = Paros-1, PAR-2(3) = Paros-2(3), PEN = Pentelikon, PRO = Prokonnesos, THA-1(2) = Thasos-1(2), THA-3 = Thasos-3. The underlying data are provided in the Supplementary Materials accompanying this article. (D. Taelman.)

Fig. 6. Number of sites per 25-year period, marble variety, and region at which white architectural marble objects have been attested (site count represents the mean number of sites of the 10,000 simulations): (A) Italy, (B) Eastern Mediterranean, (C) Western Mediterranean. Quarry abbreviations: APH = Aphrodisias, CAR = Carrara, DOK = Dokimeion, EPH = Ephesos, GOK = Göktepe, NAX = Naxos, PAR-1 = Paros-1, PAR-2(3) = Paros-2(3), PEN = Pentelikon, PRO = Prokonnesos, THA-1(2) = Thasos-1(2), THA-3 = Thasos-3. The underlying data are provided in the Supplementary Materials accompanying this article. (D. Taelman.)

Fig. 7. Number of sites per 25-year period, marble variety, and region at which white sculptural marble objects have been attested (site count represents the mean number of sites of the 10,000 simulations): (A) Italy, (B) Eastern Mediterranean, (C) Western Mediterranean. Quarry abbreviations: APH = Aphrodisias, CAR = Carrara, DOK = Dokimeion, EPH = Ephesos, GOK = Göktepe, NAX = Naxos, PAR-1 = Paros-1, PAR-2(3) = Paros-2(3), PEN = Pentelikon, PRO = Prokonnesos, THA-1(2) = Thasos-1(2), THA-3 = Thasos-3.The underlying data are provided in the Supplementary Materials accompanying this article. (D. Taelman.)

The general chronological picture is one in which Carrara, Prokonnesos and, to a lesser extent, Pentelikon dominated the market for architectural marble, while other marble varieties were either not used or only occasionally used for architecture. Carrara had by far the widest distribution of all marbles during the Late Republic and Early Empire, with a peak throughout the 1st c. CE and the first half of the 2nd c. CE, followed by a relatively stable number of site attestations up to the early 3rd c. CE (Fig. 5A). This peak results from a rapid rise in distribution in Italy and to some degree the Western Mediterranean (Fig. 6). Similar patterns can be detected for Pentelikon, Paros-2(3), and Dokimeion marble, albeit with lower distribution levels. In the Eastern Mediterranean, there are almost no attestations of architectural applications of Carrara marble. From the second quarter of the 3rd c. CE onwards, the importance of Carrara for architectural purposes diminished. While Carrara was the dominant supplier for architectural marble in the Mediterranean, mainly Italy, up to the end of the 1st c. CE, this role was gradually taken over by Prokonnesos in the 2nd c. CE. In Italy, Proconnesian marble shows a clear co-occurrence with Carrara marble from the 1st c. CE. Prokonnesos consolidated its role as prime architectural marble in Italy and the Eastern Mediterranean throughout the Middle and Late Empire, and in Late Antiquity. Despite Prokonnesos being the dominant architectural variety, we can discern a reduction in distribution in the 4th c. CE, followed by a modest recovery in the 5th c. CE, particularly in the East (Fig. 6). At present, only very few attestations of Proconnesian architectural marble are known from the Western Mediterranean (i.e., in southern Gaul, the Iberian Peninsula, and western North Africa). Noteworthy for Late Antiquity is the rise, albeit rather limited, in the number of sites at which Thasos-1(2), that is, the calcitic marble variety from Thasos, has been attested for architectural purposes. The Late Antique rise is related to attestations of the material mainly in Italy and the Eastern Mediterranean.

While the market for architecture was dominated by essentially two white marble varieties, the market for sculpture was much more divided (Fig. 5B). Marbles from Aphrodisias, Ephesos, Göktepe, Naxos, and Paros-1 were almost exclusively used for sculpture, an observation that has already been made by some scholars.²² Overall, the entire Mediterranean shows a similar distribution pattern but with diverging chronologies, with a wide variety of marble being used for sculptural production. In the East, the practice of erecting statues was already underway in the 2nd/1st c. BCE. For Italy and the West, sculptural marble was introduced only in the Late Republic and the Early Empire respectively, but the phenomenon spread quite rapidly there. The peak of sculptural marble distribution in the East can be dated to the middle and late 2nd c. CE, while for Italy and the West, it had already occurred between the 1st c. CE and early 2nd c. CE (Fig. 7). Greek marbles (Pentelikon, Paros-1, Paros-2(3)), together with Carrara, prevailed in the market in the Late Republic, predominantly in Italy and the East. The late 1st c. BCE and the 1st c. CE show a sharp rise in the distribution of Carrara and Paros-1, followed by a period of slightly reduced popularity in the 2nd c. CE. At the peak of sculptural marble distribution, the market was essentially dominated by Pentelikon, Carrara, Thasos-3, Paros-2(3), Prokonnesos, and Dokimeion. Regional differences mainly relate to a focus on Prokonnesos, Pentelikon, and Dokimeion in the East, and Carrara and Paros-1 in Italy and the West. For example, the use of Paros-1 and Carrara in the Eastern Mediterranean has been attested in only a few, isolated cases. Worthy of mention is the almost complete disappearance of Carrara for sculptural purposes after the mid-3rd c. CE; this is in contrast with the more gradual decrease of Carrara marble in architecture. Particularly interesting is the curve of Göktepe marble, which had only a relatively short period of popularity around the middle and later 2nd c. CE, with a distribution essentially confined to Italy (mainly Rome). Two remarkable cases of Göktepe marble are its large-scale use for private and imperial statuary at the Roman villa of Chiragan (Martres-Tolosane, France) during the 2nd and 3rd c. CE,²³ and its use for a 4th-c. CE sculptural group found at the rural villa of Quinta das Longas in inland Lusitania.²⁴

Overall, the image of the Roman marble trade is that it started in the Late Republic and experienced its heyday in the course of the 2nd c. CE. The decline of the 3rd c. CE is particularly visible in sculptural marble applications, while it appears less pronounced for architectural purposes. The market for architectural marbles was dominated by Carrara and Prokonnesos, as well as Thasos-1 in Late Antiquity. For sculptural purposes, Carrara, Pentelikon, Paros-1 and Paros-2(3) were the main marble varieties until the middle of the 1st c. BCE. Most sculptural marbles (Aphrodisias, Dokimeion, Paros-2(3), Pentelikon, and Thasos-3) show similar distribution patterns, with a low peak in the early 1st c. CE and a second, higher peak around the middle and during the late 2nd c. CE. Regions of distribution for the marbles, however, diverged. Pentelic marble made its appearance in the Eastern Mediterranean during the Late Republic. In the late 1st c. BCE, it was introduced in Italy, first for sculpture. In the early 1st c. CE, it was widespread throughout the Mediterranean region, essentially for sculpture, while its architectural use was limited primarily to Italy. Well-known marbles like Ephesian and Naxian seem to have had only a restricted distribution.

Carrara appears to have been the dominant player in the market until the late 1st c. CE, with distribution mainly in Italy and the West. Its position as the leading architectural marble was taken over in the course of the 2nd c. CE by Prokonnesos, which remained the primary architectural marble until at least the end of the 5th c. CE. The decline of Carrara from the 3rd c. CE can be observed mainly for sculpture; for architecture, it is much less pronounced. It is often argued that the distribution and consumption of Carrara decreased principally from later 3rd and 4th c. CE onwards, mainly due to the gradual silting of the Luna harbor, the port from which the marble was shipped in Roman times, and the consequent transport difficulties.²⁵ The data presented here, however, show that the importance of Carrara marble had already decreased in the early 3rd c. CE, well before the silting of the Luna harbor, which has been dated to the 4th/5th c. CE by recent geoarchaeological studies.²⁶ These studies have shown that the coastal area near Luna was a geomorphologically very dynamic region, with frequent shifts in coastline controlled by changes in the sediment supplied by the rivers and streams descending from the Apuan Alps. Moreover, it has been demonstrated that this was a non-continuous process, already underway in the pre-Roman and Roman periods but becoming continuous only from the 5th–6th c. CE. Building activity in the 3rd/4th c. CE in Luna, the installment of a bishop's see in the town in 465 CE, and the role of the town as an important trading center in Byzantine times, in the latter half of the 6th c. CE,²⁷ all point to relative prosperity well after Carrara's marble output was reduced and suggest that the decrease in its distribution was not related to the silting of the town's harbor. Exhaustion of the marble resources does not explain the sudden fall in distribution either, given the later revival of extraction in the 12th c. CE using similar techniques to the Romans. These elements, together with different trends for Carrara architecture and sculpture, point towards market-driven processes behind its decline and the preference for Proconnesian marble in the Late Empire and Late Antiquity, especially in Italy. We should not ignore, however, that a reduction in the available labor force, resulting from a general population decrease in the region,²⁸ could also have played a role in the downfall of Carrara as a center of marble production.

Within this general image of quarry districts that exported on a supra-regional scale, we should not forget the importance of certain regional marbles. This is particularly the case for the Iberian Peninsula, which has several high-quality white marble resources, such as those from the Estremoz Anticline, Almadén de la Plata, and Mijas,²⁹ as well as the marble from Saint-Béat.³⁰ The presence of these high-quality white marbles resulted in distinct marble consumption patterns for the region, in which imported Mediterranean marbles from Italy and the East were used in combination with regional material.³¹ In the coastal parts of northeastern Tarraconensis, architectural objects were carved predominantly in Carrara, while Greek marbles (particularly from Paros, Pentelikon, and Thasos) together with Carrara were popular for statuary.³² Inland, mainly along the Ebro Valley, Pyrenean marble from Saint-Béat was the preferred material for architectural and sculptural purposes, with Mediterranean imports arriving there only in reduced numbers. For Baetica, the districts of Almadén de la Plata and Mijas, and to a lesser extent Estremoz, were the prime suppliers for both architecture and sculpture. Imported Mediterranean marbles, essentially from Carrara and sometimes Paros, are attested only in a few cases.³³ The situation was similar for western North Africa, where marble for architecture was imported from both Hispanic (Malaga, Almadén de la Plata, and Estremoz) and Mediterranean sources, while statuary used exclusively Mediterranean material (Parian, Thasian, and Pentelic).³⁴ For Lusitania, and northwestern and central Tarraconensis, the Estremoz Anticline provided the bulk of the marble. Imported white marbles (Carrara, Pentelikon, Aphrodisias, and Paros) are currently attested only at Emerita Augusta, the capital of Lusitania, and this is mainly for high-status sculptural objects and buildings that are linked to the Imperial family.³⁵ Overall, the Estremoz Anticline seems to have been the regional marble with the widest distribution. Exports have been attested along the entire Atlantic side of the Iberian Peninsula, parts of inner Iberia, as far east as Caesaraugusta (Zaragoza, Spain), and at Banasa (Sidi Ali bou Djenoun, Morocco) in Mauretania Tingitana.³⁶

The Roman white marble trade in a wider economic and societal perspective

It has long been the consensus among historians and archaeologists that the Roman economy experienced growth in the Late Republic and Early/Middle Empire, followed by decline in the Late Empire and Late Antiquity.³⁷ Recent studies, however, illustrate that this general pattern was locally much more diverse. There were clear regional differences, with decline setting in during the 3rd c. CE in the West (including Italy and Greece), while the East and North Africa experienced growth until the 4th–6th c. CE.³⁸ Much more central to today's scholarly debate is whether this growth was limited to aggregate growth or also included per capita growth. Positions in this debate largely align with academic background: archaeologists generally advocate for overall increased welfare and per capita growth, as evidenced by the unprecedented expansion of traded archaeological material around the Mediterranean between the 1st c. BCE and the 2nd c. CE; historians, on the other hand, tend to argue that the aggregate growth of the Roman economy, as illustrated by the archaeological evidence, was to a large extent outbalanced by demographic growth, resulting in only limited per capita growth or even decline.³⁹

As an object of trade and an indicator for urbanization, the marble data presented in this article can contribute to this debate about the performance and structure of the Roman economy. Roman culture is well known for its impressive architectural realizations. The use of stone in ever-increasing quantities was emblematic of Roman architecture and urbanization; monumental marble(-clad) architecture and statuary were an essential part of the Roman urban environment. While urbanization and economic performance might not be connected in a straightforward manner,⁴⁰ there is no doubt that town development had a considerable effect on the Roman economy. More and larger towns meant higher urban populations and more people to be active in non-agricultural labor, such as crafts, manufacturing, and services.⁴¹ Moreover, the foundation of new towns and the expansion of existing towns, with the erection and marble embellishment of monumental architecture, gave rise to large-scale building activity. This required efficient coordination and information exchange, and engaged workers with different professional backgrounds and specializations, including architects and builders as well as people involved in the production of building materials (quarrymen, carvers, sculptors, etc.) and in trade (merchants, shippers, transporters, etc.).⁴²

Marble distribution, of course, does not directly reflect the foundation or expansion of towns; it can, however, be used as an indicator of elite capital investment in the urban fabric. Higher levels of marble distribution suggest increased financial means among the elite for investing in urban embellishment and therefore indirectly illustrate economic growth.⁴³ The town was the typical setting wherein the Roman elite spent the income of surplus production from their estates to perform their social role, amongst other ways, in the form of architectural munificence ranging from gifting entire buildings to donations of individual architectural elements such as marble decoration, columns, or statuary.⁴⁴ In view of this, increased marble distribution can be seen as a reflection not only of increased trade (and the related growth of markets) but also of increased urbanization, greater enrichment of elites, shifts in capital and labor from agriculture to non-agricultural economic sectors, social complexity, and higher living standards. Whether this increased investment in marble decoration resulted from the upper elite investing more (indicating increased social inequality and thus extensive economic growth), or it resulted from a larger fraction of the elite having access to luxury material (indicating increased living standards for a larger part of the population and thus intensive economic growth) is less clear. Data from Pompeii and Herculaneum, however, seem to point in the direction of the latter. At these sites, a growing number of non-public buildings were being decorated with an increasing variety of marble in the 1st c. CE. This is particularly visible in marble insert and opus sectile pavements in residential structures, and in the marble-clad counters of bars, thus illustrating that increased marble consumption was not only restricted to the upper elite.⁴⁵ As such, increased marble distribution can point to real, per capita, intensive economic growth.

The distribution patterns of white marble in the Roman Mediterranean discussed in the previous section illustrate a start to the marble trade in the Late Republic, with its heyday in the 2nd c. CE. The decline that set in during the 3rd c. CE is particularly visible for sculptural marble. The distribution of architectural marble shows only a minor drop, followed by a relatively stable level of distribution throughout the Late Empire and Late Antiquity, with figures still higher than those for the Late Republic and Early Empire (that is, up to the 1st c. CE). The continued distribution of architectural marbles is mainly observed in Italy and the Eastern Mediterranean. The divergence between architectural and sculptural marbles suggests that reduced marble distribution in the Late Empire and Late Antiquity is not necessarily the result of a declining Roman economy, but rather points towards changes in the cultural practices related to statue erection in this period.⁴⁶ This is also illustrated by the fact that we can observe this phenomenon for all marble varieties and throughout the entire Mediterranean. The observations for marble data are very much in line with those for building inscriptions, which show a similar trend up to the end of the 2nd c. CE and early 3rd c. CE.⁴⁷ For the Late Empire and Late Antiquity, however, the building inscriptions show a drastic decline, which is not so marked in the marble distribution data. This might indicate that the urbanization decline after the 3rd c. CE was less pronounced than suggested by the building inscription record and is more related to changes in the epigraphic habit than to reduced elite investment in the urban fabric.

The decline in urbanization in the Later Roman period, albeit with notable regional chronological differences, that is advocated by several scholars⁴⁸ seems only partly supported by the marble distribution data. While it is certainly true that large-scale building activities in the Western Mediterranean in this period were less frequent, and several towns diminished in size or even disappeared, the reduced urbanization seems less valid for Italy and the East, or at least presented itself in a different manner. In the East, large-scale building projects and marble embellishment were still being undertaken in several regions.⁴⁹ In Italy, marble embellishment in urban centers seems to continue until Late Antiquity, albeit at a lower level than during the 2nd c. CE. This reduced investment in urban architectural marble seems to go hand in hand with a growing ruralization and reduced effort by the elite in using the town as the place to express their social and political roles. While urbanization levels were certainly lower in the Late Empire and Late Antiquity than in the Late Republic and Early Empire, the decline, as shown by the marble data, seems less pronounced or at least much more gradual than traditionally believed. And, it is not (only) the consequence of economic decline, but rather, at least partly, the result of changes in cultural practices and the way in which the Roman elite spent their money. The phenomenon of ruralization can also be seen clearly, for instance, in the province of Lusitania, where the 4th c. CE witnessed the development of several large and monumental villa estates, such as the villa of Quinta das Longas and Torre de Palma, at the expense of the main urban centers. Marble was common at these sites, sometimes even importing material from the East, as was the case for Quinta das Longas, which was decorated with statues carved from Göktepe marble.⁵⁰ These estates progressively acquired more land and gradually evolved into local centers where religious, funerary, social and economic aspects of society were concentrated. A focus on town contexts in marble provenance studies might thus partly explain the lower marble distribution numbers for this period.

Finally, we cannot ignore the importance of the reuse and recycling of marble architectural elements and marble sculpture in the Later Roman period. While this phenomenon was not uncommon already in Late Republican and Early Imperial times, and was not limited to marble,⁵¹ it is generally associated with Late Antiquity. The purchase of heavy and bulky stone material was an expensive affair in pre-industrial times, as it involved the costs both of the raw material and of the quarrying, carving and transport.⁵² For the Later Roman periods, it was often more economical to salvage the large quantities of marble that were available from demolished or abandoned pre-existing structures, thus reducing the costs related to long-distance transport. From this perspective, reuse can be considered simply as the use of an object because its properties are suited for its new function, beyond the one for which it was originally intended. Marble retained its value as raw material, making it unlikely that the resource would have been discarded if it could be usefully re-employed. The adoption of the practice of reuse can therefore not simply be considered an indicator of economic decline. On the contrary, the act of marble salvaging and reuse is a clear sign of continued investment in monumental building and of wider economic investment, albeit not at the scale of building activity in the Late Republic and Early to Middle Empire. Marble salvaging not only involved considerable organizational and infrastructural investments for material storage, but also required labor, including craftspeople for demolition and recarving, transporters (for local transport), and merchants of the secondhand material, as well as people involved in restoration or building.⁵³ Both growing ruralization and the practice of reusing/recycling marble elements fit into a wider phenomenon of the redefinition of urban public space in the Late Empire and Late Antiquity. It is within this frame that we need to interpret the marble distribution figures.

Because of the uneven geographical distribution of marble resources in the Roman Mediterranean, marble objects generally had to be traded over long distances. The traditional view is that marble exploitation and trade were associated with the emperor or the state. Most quarries are believed to have been part of the patrimonium Caesaris, especially after 17 CE, with the alleged confiscation of all principal mines and quarries by the emperor Tiberius.⁵⁴ The exploitation and distribution of the material were controlled by a centralized marble administration, the so-called ratio marmorum.⁵⁵ More recent studies however, have illustrated that this was not necessarily so, and that even at the “imperial” quarries, the presence of private and municipal ownership and production seems to have been the norm.⁵⁶ Even though the state was closely involved in the exploitation of (certain) marbles, most material entered the non-imperial market through market-focused trade and not through state redistribution. Looking at the enormous volumes of marble consumed in the Roman world, it seems unlikely that state redistribution was the main mechanism through which the material was exchanged. The overlapping distribution patterns of the different white marble varieties of the 1st and 2nd c. CE, particularly for sculpture, show a well-integrated, inter-regional market for Italy and the East. While imported material from Italian and Eastern quarries was not absent in the Western Mediterranean, the ample availability of regionally available marble resources, especially on the Iberian Peninsula, resulted in a much more complex picture. For Spain and Portugal, clear regionally distinct marble consumption patterns can be observed in which, in broad terms, sites in coastal Tarraconensis obtained regular imports from the Carrara quarries (and also from Greek quarries for sculpture), while inland Tarraconensis, Baetica, and Lusitania were oriented more towards regional marbles, with Mediterranean imports being used more scarcely and generally for prestigious projects.

Finally, comparing the marble distribution trends with the evidence obtained from shipwrecks on which stone cargoes were being transported further highlights some interesting points about the organization of the trade in white marble. The shipwreck evidence suggests that most distribution of marble in the Mediterranean dates to between the 1st and 3rd c. CE, with a peak in the 3rd c. CE.⁵⁷ The graphs provided in the current article, however, illustrate that the 3rd c. CE was a period with somewhat lower distribution, especially for sculpture but also to a lesser degree for architecture. Focusing on white marble, the somewhat later chronology of the shipwreck evidence probably has to do with a dominance of Eastern marbles in the cargoes. Most shipwrecks for this period have been encountered along Italian coasts, carrying Eastern marbles (Table 1⁵⁸). The discrepancy between the two datasets is related to the fact that in the 3rd c. CE, although less marble was distributed, Eastern marbles, in particular from Prokonnesos, were still popular in Italy. Provenance studies illustrate that large-scale building and restoration projects by the Severan emperors in Rome in the early 3rd c. CE relied heavily on Proconnesian marble, giving rise to intensified maritime transport between Rome and the East in this period.⁵⁹ Moreover, Proconnesian marble is also found at several other sites in Italy in architectural contexts dated to the late 2nd and 3rd c. CE.⁶⁰ The intensified trade, together with the longer voyages (especially in comparison to Carrara in the previous periods) and the related higher risk of wreckage, likely explain the slightly later chronology of the shipwreck evidence.⁶¹

Table 1. List of Roman shipwrecks with provenanced white marble cargoes. Cargo type abbreviations: ADC = architectural element, SCU = sculptural object, SRC = sarcophagus. Cargo provenance abbreviations: CAR = Carrara, DOK = Dokimeion, PAR-2(3) = Paros-2(3), PEN = Pentelikon, PRO = Prokonnesos, THA-1(2) = Thasos-1(2), THA-3 = Thasos-3.

The compositions of the wrecks with white marble cargoes suggest that at least two different trade systems, that is, direct and indirect trade, were in operation at the same time in the Roman period (Table 1). Cargoes consisting of material from a single quarry suggest that the traffic was part of direct trade or voyages, in which the cargo was loaded at one port for transport to another, pre-determined port. The material was transported to meet a specific demand of the final consumer or an intermediary workshop.⁶² Cargoes consisting of material from multiple quarries indicate an alternative system of indirect trade. Loading in this case would be either at various points along the voyage, that is, the respective quarries or their related ports, or at one or more entrepots or (re)distribution centers. The idea of such facilities for marble assumes the existence of stockpiles at centralized locations, where marble objects quarried for an undefined destination were stored.⁶³ The marble yards at Portus and the Emporium at Rome are generally advanced as evidence for such stocks intentionally amassed for (re)distribution purposes. More recent studies, however, critique the existence of such intentional marble stockpiles, even for the large marble yards of Portus and Rome.⁶⁴

Conclusions

The study of white marble distribution presented in this article is the first to approach the Roman white marble trade from a quantitative and diachronic perspective, based on a large dataset of reliable provenance data. The observed trends and patterns point towards increased trade and distribution of white marble in the period from the late 1st c. BCE to the end of the 2nd c. CE. The situation from the 3rd c. CE onwards is less straightforward and displays object-, material-, and region-specific trajectories. The decline in distribution starting in the 3rd c. CE seems to have been less abrupt and more gradual than traditionally suggested, and it can be discerned mainly for sculptural marble and much less so for architectural uses. The reduced distribution (or quasi-disappearance) of certain marbles in the Later Roman Empire, in particular Carrara, is at least partly countered by the continued widespread use of Eastern marbles, such as the material from Prokonnesos. Within this general view, the data shows a common market for Italy and the Eastern Mediterranean, while the situation in the Western Mediterranean (in particular, the Iberian Peninsula) is much more complex, with the joint usage of imported and regional marbles at many sites.

With this study, I argue for a more nuanced view of the complex situation of the Roman economy, particularly for the Later Roman periods, that is, from the 3rd c. CE onwards, for which quantified trends and patterns need to be interpreted within a frame of wider societal and cultural transformations. This is particularly true for the notable differences between architectural and sculptural marble distribution, which can be explained to a certain degree as resulting from changes in cultural practices related to the erection of statues and urbanization, and less in terms of a declining economic reality. However, it needs to be noted that the long-distance trade in white marble was not at the same level from the 3rd c. CE onwards as during the 2nd c. CE. Likewise, the decreasing figures for architectural marble distribution in the West and in Italy, albeit at a lower level than for sculptural marble, need to be seen in light of a growing ruralization and reduced interest among the Roman elite in using urban centers as a setting for capital investment, as well as a growing importance of reuse/recycling of marble elements.