We would like to give special thanks to Michel Bonifay for his opinion and valuable input on the study of the Mejerda workshops. We are grateful to Maud Webster for proofreading the English text.
- 1 Carte géologique 1985-1987; Mackensen 1993, p. 54; see also Baklouti et alii 2014, p. 525, fig. 1.
- 2 Mackensen 1993, p. 479-486; 1998, p. 31-33.
- 3 The four phases represent a relative chronology mainly based on vessel types with stamp decoration (...)
- 4 Mackensen 1993, p. 490; see also Mackensen, Schneider 2002, p. 125; most recently Hasenzagl forthc (...)
1Located in the Mejerda Valley, Tunisia’s most fertile region producing cereals, fruits, and vegetables, the production centers of El Mahrine, Bordj el Djerbi, and Henchir el Biar lie within a distance of only a few kilometers (fig. 1) on the left bank of the Mejerda, which carries water year-round. Geologically situated in a sedimentary area, consisting of Tertiary carbonatic sequences, Paleogene and Neogene clays, and sandstones as well as quaternary alluvial deposits (fig. 2)1, the workshops are traditionally counted as producers of late Roman ARS D1. They are assumed to have been built on the same latifundium, to the south of the Roman city of Thuburbo Minus, around 320/330 AD2. A parallel production and organization of these workshops is suggested not only by the very short geographical distance but also by the similar and contemporaneous vessel types of the workshops that produced ARS, lamps, and cooking ware in four main and several sub-phases3. The first phase (1a-1c) of manufacture dates from the first third of the 4th to the mid-5th century AD. It was followed by phase 2, continuing until 470/480 AD, and phase 3, which lasted until 500 AD. The production ceased with the fourth and final phase (4a-4c) in the middle (or end) of the 7th century AD. The long-lasting assumption that the atelier at Henchir el Biar was closed down approximately 100 – or even 150 – years earlier than neighboring El Mahrine and Bordj el Djerbi was recently revised4.
Fig. 1: Schematic map of Tunisian ARS-workshops and productions
(based on Bonifay 2004, p. 46, fig. 22; adapted by C. Hasenzagl)
Fig. 2: Localization of Henchir el Biar, Bordj el Djerbi and El Mahrine in the geological map of the area
(modified from Carte géologique 1985-1987)
- 5 J.W. Salomonson, A. Ennabli, and L. Slim had been advised by J. Peyras to survey the Mejerda Valle (...)
- 6 The results of this survey have never been extensively published. See Deneauve 1972, p. 219-220; s (...)
- 7 Bordj el Djerbi was mentioned for the first time by L. Poinssot and R. Lantier (1923). They discov (...)
- 8 See Mackensen 1993. In 2006, M. Ben Moussa (2007, p. 79-80) revisited the site and discovered arch (...)
2The sites of Bordj el Djerbi and Henchir el Biar were surveyed by J.W. Salomonson, A. Ennabli, and L. Slim in 19685, followed by another short prospection done by J. Deneauve and A. Ennabli (1970)6, confirming their identification as ARS workshops, which was already indicated by earlier reports7. M. Mackensen’s surveys in 1987 (Henchir el Biar) and 1997-1999 (Bordj el Djerbi) represent the most extensive archaeological research that has been done at these sites since. El Mahrine, the best-studied8 workshop in the Mejerda valley, however, was never surveyed by Salomonson.
- 9 Bonifay, Capelli, Brun 2012, p. 48-49: « La distinction entre les trois ateliers repérés par M. Ma (...)
3From a typological and macroscopic point of view, the productions of El Mahrine and the neighboring workshops of Bordj el Djerbi and Henchir el Biar are difficult to distinguish from each other and form a relatively homogeneous group well separated from the other ARS workshops. Thus, they are usually collectively identified as imports from the “El Mahrine zone” at consumption sites9.
- 10 ARS from the Mejerda Valley is rich in Sr and Zr, in TiO2, K2O, Al2O3, Y, and Rb. Mackensen, Schne (...)
4However, chemical analyses showed that the three workshops are generally isolated from other ARS production sites and form individual clusters in most cases (some samples, particularly from Henchir el Biar, fall into the El Mahrine compositional field)10.
- 11 Bonifay, Capelli, Brun 2012, p. 48-49.
5Petrographically, only the workshop of El Mahrine has been characterized so far (C.C.). The thin section study is based on 12 representative samples of ARS and a few saggars selected from the reference material deposited at the Centre Camille Jullian, Aix-en-Provence11. El Mahrine’s paste is homogeneous and defined by well-sorted inclusions with a bimodal distribution. In the sandy fraction (<0.5-1 mm), frequent rounded/aeolian quartz grains are always associated with rarer fragments of quartz-silt/sandstone, shale, and limonite. The slip is generally rather thin (10-40µ) and its color is quite similar to the clay matrix.
- 12 These samples were given by M. Mackensen.
- 13 Bonifay, Capelli, Brun 2012, p. 48-49. Collection Jean Deneauve.
- 14 Capelli et alii 2016, Group 5; Capelli, Cardarelli 2020, Groups 9-13; in the latter case, however, (...)
6In contrast, before this study, only four samples12 from Henchir el-Biar and Bordj el Djerbi had been analyzed (C.C.) in thin section. The preliminary examination showed that three samples were different from El Mahrine, with coarser, less well-sorted fabric, whereas one sample from Henchir el Biar was quite similar to El Mahrine’s typical fabrics13. These reference samples were also used to analyze ARS from various consumption sites in Sicily and the Palatine hill at Rome, allowing us to assign (or exclude) the provenance of ARS fragments to the El Mahrine zone14. Most of them were attributed to El Mahrine itself, but for some marginal, a bit coarser fabrics, a possible origin from the surrounding workshops was considered.
- 15 Hasenzagl, Capelli 2019; 2020. Authorized by the INAA (Institut National d’Archéologie et d’Art), (...)
7In this paper, we will present a more extensive petrographic study of the workshops of Bordj el Djerbi and Henchir el Biar based on the analysis of Salomonson’s survey collection, and thus increase the number of available reference samples for further investigations. The effectiveness of the combined methodological approach of thin section analysis of selected samples under the polarizing microscope and standardized fabric description of all samples done with a stereomicroscope, integrated with the typological data, has been already demonstrated with two other workshops (Sidi Khalifa and Sidi Aïch) in Salomonson’s collection15.
- 16 In collaboration with Tunisian and French colleagues from the Centre Camille-Jullian (Aix-en Prove (...)
- 17 For example, Capelli et alii 2016. Distinguishing individual ARS workshops at selected Mediterrane (...)
8This research was carried out within the framework of an ongoing, more general characterization of individual ARS (as well as amphorae and common ware) production sites16, aiming for precise identification of the origin of African imports at Mediterranean consumption sites to improve the reconstruction of ancient economic relations and trade routes17.
- 18 The 65 pieces presented in Hasenzagl 2019, p. 32-38 were long thought to be the entire amount of c (...)
- 19 Mackensen 1993, p. 288-301, 366-374.
- 20 The main types mentioned by Mackensen are: Hayes 58B, 59A/B, 61A, 61A/B, 62A, 63, 67, 76A/B, 91A/B (...)
- 21 Hasenzagl forthcoming. See also Mackensen 1993, p. 463, note 7: Mackensen collected 69 (base) frag (...)
9In 1968, Salomonson collected 320 diagnostic pieces at Henchir el Biar18. The survey material includes a repertoire of 4th- to 6th-century vessel types consisting of Hayes 61A, Hayes 61A/B, Hayes 63, Hayes 67A-C, Hayes 73A/B, Hayes 76A/B, Hayes 91, Hayes 93, Mackensen 18, Mackensen 19, and Hayes 104A1, corresponding to production phases 1b-4a19. This chronological spectrum generally also complies with the main types documented by Mackensen during the survey in 198720. However, Salomonson’s survey also consists of vessel types connected to phases 4b and 4c (e.g. Hayes 104A3, Mackensen 39, or Mackensen 54), thus indicating that Henchir el Biar continued to produce ARS into the 7th century, like its neighboring potteries El Mahrine and Bordj el Djerbi. Most interesting, however, is the high number of stamp decoration (Mackensen décor I – III), of which evidence has been relatively scarce so far21.
10The collected ARS has a great variety of color, which ranges from light red (Munsell 10R 6/6 and 6/8) to red (2.5YR 5/6) and light reddish-brown (2.5YR 6/4), all of which can appear on one fragment. The quality of the slip, however, is very similar. It is generally very thin and matt, and usually covers the inside of the vessels and only parts of the outside wall. The surface often was not treated conscientiously because it is rough and granular, and the slip flakes off easily.
- 22 Unfortunately, all examples of production phases 3 and 4 were among the only recently re-discovere (...)
11The binocular study of all of Henchir el Biar’s samples from Salomonson’s survey collection initially resulted in identifying three main fabric groups. Eight samples were chosen for thin section analysis. They include three specimens of Hayes 67C, as well as one example of Hayes 73A, Hayes 73B, Hayes 76A, Hayes 76B, and the lamp type Atlante VIII A/C (fig. 3), and are representative of production phase 1 and 222.
Fig 3: Selected Henchir el Biar ARS samples from Salomonson’s survey
(drawings and photo C. Hasenzagl)
- 23 In general, all the studied samples from the two workshops share several petrographic features, wh (...)
12All the eight samples from Henchir el Biar (tab. 1, fig. 4) show a quite similar bimodal distribution of the inclusions that are composed of a well-sorted, fine-grained (silty) fraction (essentially quartz, mostly <0.1 mm), and a subordinate coarser (sandy) fraction (generally 0.2-0.4 mm); in the latter, rounded/aeolian quartz grains are relatively frequent, together with quartz-sand/siltstones (fig. 4c, f, h). They have been attributed to three different fabrics, essentially taking into account the frequency of the inclusions. Seven samples form two main groups (HB1-2), whereas no. 13316 is an outlier (HB3) in the selected material.
Fig. 4: Representative fabrics of Henchir el Biar’s ARS
Thin section microphotographs; a, b, g, i: parallel Nicols; c-f and h: crossed Nicols; real dimensions: 1.3x1 mm, except for i: 0,32x0,24 mm.
(photos C. Capelli)
Tab. 1: Henchir el Biar and Bordj el Djerbi
List of the samples selected for the thin section study with their main petrographic characteristics. B: bimodal distribution; = and > : slip color similar to or darker than that of the body. See also the Word format of the table.
- 24 Two reference samples (nos 11278 and 11279) of the DISTAV/CCJ database match HB2 and HB3, respecti (...)
13Whereas fabric HB1 (fig. 4a-d) presents rather scarce inclusions with a few relatively coarse (up to 0.9 mm) quartz grains, fabric HB2 (fig. 4e-g) is characterized by moderately frequent inclusions in both granulometric fractions. Fabric HB3 (fig. 4h-i) is distinguished by a very abundant fine fraction, whereas the coarser inclusions are relatively rare and small24.
14Henchir el Biar’s slip (fig. 4a, i) generally has a color similar to that of the vessel body and is rather thin (20-30µ). Only in a few samples (HB1, fig. 4b) can its thickness gradually increase to high values (60-100µ).
15In general, HB fabrics share several textural and compositional characteristics with El Mahrine-products. In particular, group HB2 is practically indistinguishable from typical El Mahrine-fabrics. Group HB1 slightly differs from the latter because of the minor percentages of the inclusions and the coarser dimensions of the sandy fraction, whereas HB3 shows a much more frequent fine fraction and rarer and smaller sandy inclusions. The slip is similar to that of El Mahrine in most cases, but some specimens of group HB1 show an anomalous thickness, never observed in samples from El Mahrine.
- 25 Hasenzagl 2019, p. 35-38.
16Under the stereomicroscope (fig. 5a-c), three fabric types had been established: HB-ARS-1, 2, and 325. That division is confirmed by the thin section study. They generally correspond to the groups HB1-3, even if no subgroups were distinguished.
Fig. 5: Representative fabrics of Henchir el Biar (a-c), Bordj el Djerbi (d-f), and ARS A (g; ignotum)
(macrophotographs at magnification 16X, C. Hasenzagl)
17All the fabrics appear quartz-rich with varying percentages and occurrences of brownish and black inclusions, rare argillaceous rock fragments (determined as shale and limonite fragments in thin section), infrequent calcareous elements as well as silver mica. Whereas the nature of the inclusions and their composition are identical in all fabrics, they can be distinguished by the number and average size of the particles, creating a slightly different texture of the fresh break with the transition often being smooth. Simplified, they can be described as fine/compact (HB-ARS-1, fig. 5b), fine/granular (HB-ARS-2, fig. 5b), and the most granular (HB-ARS-3, fig. 5c) of the ARS-series.
18The differences in texture in the division of the binocular fabric types result from the number and the average dimensions of quartz grains that gradually increase from HB-ARS-1 to 2 and 3. The overall percentage of the temper, as well as the porosity of HB-ARS-2 and 3, is generally higher than in HB-ARS-1. Whereas HB-ARS-1 is mainly characterized by its few distinctive inclusions, the most diagnostic particles in groups 2 and 3 are small to medium-sized black particles that look blurry against the granular texture.
- 26 For a more detailed evaluation, see Hasenzagl forthcoming.
19On the whole, the majority of the pieces (n=320) collected in 1968 are HB-ARS-1 (43%) and HB-ARS-2 (37%). A rather small share (12%) accounts for HB-ARS-326. The scarcity of group 3 is striking. However, representatives of group 3 are too numerous to suspect a non-local production. Like with the finds from Bordj el Djerbi, there was no chronological dependency between fabric and vessel – or stamp types within Salomonson’s survey material. Hence, it is assumed that all the fabrics are present during all four production phases from the 4th until the 7th century AD.
- 27 For a typo-chronological evaluation of all collected diagnostic ARS, see Hasenzagl 2019, p. 24-27.
- 28 The fragmentation of the pottery found at Bordj el Djerbi was also noticed by M. Mackensen during (...)
20Bordj el Djerbi’s ARS material (n=94)27 collected by Salomonson in 1968, partly showing heavy fragmentation28, is very heterogeneous in terms of macroscopic appearance and quality. The slip varies from reddish-yellow (Munsell 5YR 6/6) to light red (2.5YR 6/8) and red (10R 5/8), from (very) thin to thick, and from matt to lustrous. It usually covers the inside of the vessels and only parts of the outside wall and often has a rough and granular surface, which flakes off easily. Typologically covering the production phases 1a- 4c (4th-7th century AD), the forms Hayes 58B, Hayes 59A/B, Hayes 61A/B, Hayes 63, Hayes 67, Hayes 73, Hayes 76B, Hayes 91, Hayes 93A/B, and Hayes 104B, as well as the stamp styles A(i)-A(iii), A(iii)/E(i) and E(i), were determined.
- 29 Namely groups BD-ARS-1, BD-ARS-2, and BD-ARS-3, in Hasenzagl 2019, p. 27-32.
- 30 Moreover, sample 13312 (Hayes 59A) that showed a large amount of calcareous elements, and thus ano (...)
21Following the results of the observation with the binocular microscope of all Bordj el Djerbi’s samples from Salomonson’s collection, which initially led to the identification of three fabric groups29, ten representative samples (production phases 1-4a) were chosen for the thin section study (tab. 1). They consist of Hayes 59A, Hayes 61A, Hayes 67B (three samples), Hayes 67C (two samples), Hayes 76, Mackensen 18.2, and a stamped base showing a left-looking dove (fig. 6). Moreover, the material from Bordj el Djerbi contained a chronological outlier (Hayes 8, fig. 6) that was also petrographically analyzed30.
Fig. 6: Selected Bordj el Djerbi ARS samples from Salomonson’s survey
(drawings and photo C. Hasenzagl)
22The 11 samples (tab. 1) considered local can be attributed to two fabric groups (BD1-2, fig. 7), which are quite different in terms of texture and mineralogical-petrographic composition. Due to its typological anomaly, another sample (n. 13302, ARS A) was ascribed to a separate group (BD3).
Fig. 7: Representative fabrics of Bordj el Djerbi’s ARS
Thin section microphotographs; a, d, e, i: parallel Nicols; b, c, f-h: crossed Nicols; real dimensions: 1.3x1 mm.
(microphotographs C. Capelli)
- 31 The two reference samples (nos 11280, coarser; 11281, finer) of the DISTAV/CCJ database could be a (...)
23Group BD1 (fig. 7a-d) is characterized by well-sorted inclusions with bimodal distribution and scarce parallel voids. The fine fraction (<0.1 mm) is similar in texture and frequency to that of fabric HB1, and less rich than in the groups HB2-3. Conversely, the other fraction is rather abundant and coarse (generally 0.2-0.5 mm, up to 1 mm, including shales: fig. 7a-b), more than in the HB-fabrics. In this case, the presence of intentional temper cannot be excluded entirely. Larger quartz grains are frequently rounded/aeolian. In subgroup BD1.1, mica is less rare than in the other studied fabrics, and quartz-sand/siltstones almost absent in two samples31.
24Sample 13306 (BD1.2, fig. 7c-d) is slightly different from the first subgroup, mainly because of the relatively poorly sorted fine fraction and the relative abundance of mica.
25The slip, of a color similar to that of the body clay (BD1.1) or darker (BD1.2), is relatively thick (up to 40-100µ) and somewhat impure (fig. 7d), with micro-inclusions of quartz and mica.
26Group BD2 (fig. 7e-g) is characterized by abundant, relatively poorly sorted, mostly fine-grained (<0.2-0.3 mm) inclusions; they are essentially composed of quartz, with rare plagioclase feldspar, heavy minerals, and mica (sometimes up to 0.2 mm in length). The coarser fraction (up to 0.5-1 mm) is formed by quartz (often rounded/aeolian) with subordinate fragments of quartz-sand/siltstones (fig. 7f), shale (relatively frequent in subgroup BD2.1), and rather rare limonite/Fe-oxides.
27Slight granulometric variations in the inclusions’ dimensions and frequency led us to distinguish two subgroups and one intermediate sample. Subgroup 2.2 is slightly finer than 2.1 in both fine and coarse fractions. The apparent firing temperature is higher in the latter.
28Another distinctive feature of the group is the abundance of fine, planar parallel voids oriented by the wheel-throwing (fig. 7e), as well as the macroscopic color, which is a little lighter than in the other groups. The slip is generally thin (20-30µ), but in some cases, it is relatively thick and also somewhat impure (fig. 7e). Its color is similar to or darker than that of the body.
29Despite sample 13302 (BD3) being typo-chronologically different, its fabric is similar to that of group BD1 (fig. 7h-i), of which it could even be considered a subgroup (BD1.3). It is distinguished by the presence of more abundant shale fragments, either Fe-rich (red) or Fe-poor (yellow), the less coarse sandy inclusions, and the lighter (yellowish) color of the body close to the surface. The slip is relatively thick and darker than the body (fig. 7i).
30On the whole, it must be noted that not only can the production of Bordj el Djerbi easily be distinguished from those of both Henchir el Biar and El Mahrine because of the coarser fabrics, but also that it is composed of two quite different groups in terms of texture, technology, and possibly also raw material sources.
31On the one hand, group BD1 shares some textural features with Henchir el Biar and El Mahrine, in particular, the similar fine fraction and the bimodal distribution of the inclusions, but it differs in a coarser sandy fraction. The presence of an added temper could not be completely excluded in this case. On the other hand, group BD2 is notably distinct from the other groups/workshops because of the abundant, poorly sorted inclusions. Another feature distinguishing the production of Bordj el Djerbi from that of El Mahrine is the thick, rather impure slip evinced by a few samples, similar to some specimens from Henchir el Biar.
- 32 Hasenzagl 2019, p. 27-32.
32In the preliminary study under the stereomicroscope (fig. 5d-f), Bordj el Djerbi’s fabric types, sharing similar characteristics but differing in texture and particle sizes, were arranged in order from a compact (BD-ARS-1, fig. 5d) to a medium-granular (BD-ARS-2, fig. 5e) and very granular (BD-ARS-3, fig. 5f) matrix32.
33Group BD-ARS-1 matches thin section group BD1. It shows significant variations in particle size that are apparent at first sight. The partly very large inclusions mostly consist of quartz, rust-colored to reddish-brown and black particles, as well as lenticular argillaceous rock fragments (limonite and shale fragments), large to small, white and yellowish-beige calcareous elements, and rarely mica.
34Taking into account the thin section analysis, binocular groups BD-ARS-2 and BD-ARS-3 can now be considered as subgroups of the same fabric family (= thin section group BD2), with BD-ARS-2 (= BD2.2) being the finer version of BD-ARS-3 (= BD2.1).
35In contrast to BD-ARS-1, BD-ARS-2 and 3 have a far more granular texture. The inclusions of BD-ARS-2 (BD2.2) and BD-ARS-3 (BD2.1), whose nature and composition are consistent with those of BD-ARS-1, often remain unobtrusive due to the strongly and coarsely grained structure that is dominated (BD-ARS-2) or even riddled with (BD-ARS-3) small to medium-sized quartz. However, frequent rust-colored, argillaceous rock fragments, and white to light yellow calcareous elements are discernable from the matrix.
36Overall, the results of the analysis with the binocular microscope were confirmed by the thin section study. Some minor changes, i.e. the subgrouping of types, can foremost be explained by the varying degree of precision of the binocular analysis compared to that under the polarizing microscope but also by textural variations in different parts of a single vessel, the small size of the samples, the different sampling points for the two types of analysis, or by firing-related deviations or post-depositional processes that also changed the features of the sherd.
- 33 This fabric type has been labeled IG(ignotum)-ARS-2 in a previous preliminary study and is connect (...)
37The majority (48%) of the pieces collected in 1968 can be assigned to BD-ARS-1 (= BD1). BD-ARS-2 (= BD2.2) also accounts for a considerable amount (31%); BD-ARS-3 (= BD2.1) is the least represented (18%) of the fabric types. Whereas BD-ARS-1 and BD-ARS-2 cannot be differentiated from each other macroscopically, the appearance of BD-ARS-3 is often coarser, not due to the slip, but to the clay itself. The apparent correlation between some types and thin section groups/subgroups, however, could not be documented for the whole survey material that revealed no distinctive pattern of vessel and fabric types. Thus, judging by the material collected by Salomonson in general, any form could be produced with any fabric type during the workshop’s production period (4th-7th century AD). The fragment of the form Hayes 8 with a thick, glossy slip, which has already been typologically and macroscopically referred to as an outlier, also forms a separate fabric group (fig. 5g)33.
38The “El Mahrine zone” workshops present a common 4th-7th-century typological (known as ARS D1) repertoire with the main types Hayes 58, Hayes 59, Hayes 61A, Hayes 63, Hayes 67, Hayes 76, Hayes 73, and Mackensen 18. There are no general or clear macroscopic criteria to differentiate the products of El Mahrine and the neighboring potteries of Bordj el Djerbi and Henchir el Biar. The microscopic differences, as well as the chemical ones, however, are highly significant.
- 34 These fabrics are generally fine-grained (Sidi Khalifa) or very fine-grained (Oudhna); the coarse (...)
39This petrographic study of the samples from Salomonson’s survey has pointed out the presence of a combination of various discriminant petrographic/technological markers at Bordj el Djerbi, Henchir el Biar, and El Mahrine, differing not only from those of the continental and southern Tunisian workshops but also from the other known northern Tunisian production sites (e.g. Sidi Khalifa or Oudhna)34, even if they share common inclusions, especially fragments of shale, quartz/silt-sandstones, and limonite. Additionally and most interestingly, however, the results from the thin section analysis and fabric description showed that under the microscope, the two workshops of Bordj el Djerbi and Henchir el Biar could be easily distinguished from one another.
40An evident heterogeneity in the textural characteristics, as well as in the relative abundance of the various mineralogical-petrographic components, lead to the identification of several different groups and subgroups of fabrics. The pastes from Bordj el Djerbi and Henchir el Biar differ by their texture, which is much coarser in the BD-fabric series. Furthermore, Bordj el Djerbi has by far larger inclusions than Henchir el Biar’s fabrics. A typical feature of Bordj el Djerbi’s fabrics is several shale and limonite fragments easily discernable in the macroscopical observation. In contrast, they are hardly visible in Henchir el Biar’s finer-grained matrix.
41It is also noteworthy that, apart from group HB2, which matches typical El Mahrine fabrics, the other fabric groups, in particular from Bordj el Djerbi, are quite different from the previously studied El Mahrine-production. Several samples from both Bordj el Djerbi and Henchir el Biar show a thick, relatively impure slip; the slip of Bordj el Djerbi’s tableware in Salomonson’s survey material tends to be even thicker and glossier than that of Henchir el Biar.
42Whereas El Mahrine, HB1-3, and BD1 fabrics share textural/technological characteristics that point to the use of similar clay deposits, possibly belonging to the local Mio-Pliocene continental sequences (containing relics of the alteration of the Oligocene Numidian quartz-sandstones), group BD2 can be distinguished due to the use of poorly sorted sediments. These raw materials are clearly different from those used for the other groups. The reason could be the different geological location of Bordj el Djerbi, despite the geographical vicinity to the other workshops, inside quaternary alluvial sediments close to Numidian flysch outcrops. However, it is not clear at present why two different clay sources were used in that workshop, as, based on Salomonson’s survey material, they do not seem connected to different typologies or production phases.
43As for Bordj el Djerbi, it is also interesting to note that the comparative study with previously analyzed samples has raised a few issues that will need further investigation:
- 35 The term « Atelier X » was introduced by M. Bonifay (2004, p. 49). See also Mackensen 1998, p. 37; (...)
- 36 See e.g. Mackensen 1993, p. 441-448; 1998, p. 37; Mackensen, Schneider 2002, p. 128, 148-149; Boni (...)
- 37 Mackensen, Schneider 2002, p. 128, 139-140, 149-150.
44In the thin section, fabric BD2.2 shows similarities to the fabrics of the so-called Atelier X (also known as “large D2 pottery”)35, an unlocated and still ill-defined production of ARS D usually with a thin slip. It is connected to the 4th-7th-century vessel types Hayes 58-Hayes 61, Hayes 67, Hayes 80, and Hayes 91, but most frequently with Hayes 103, Hayes 104A–B and the stamp style Hayes E(ii), although the complete spectrum of vessel types is still unclear. It is assumed that Atelier X lies in the region north of the Djebel Zaghouan and east of the Oued Miliana, between El Fahs, Zaghouan, and Oudhna36. However, there are no archaeological traces to support this theory. Moreover, Atelier X is chemically distinguishable from the known workshops in the El Mahrine zone, as well as Oudhna37. Even though ruling out the possibility of Atelier X being located at Bordj el Djerbi, based on the petrographic similarities to the BD2 group, its location in the Mejerda Valley, in the same geological area close to Bordj el Djerbi, cannot be excluded.
- 38 Capelli et alii 2016, p. 311.
- 39 Reasons for trading firing aids are diverse. Both technological features (e.g. better temperature (...)
45Previous analyses38 evinced that the fabric of one saggar from the Centre Camille Jullian El Mahrine – reference group is very different from the typical local production (including other saggars with regular fabric) and compatible with Atelier X fabrics. The similarities with fabric BD2 now suggest an origin from Bordj el Djerbi for this as well39.
- 40 Imports or residual pieces are frequently found at production sites. Typological as well as petrog (...)
- 41 Bonifay, Capelli, Brun 2012, p. 44-46; Capelli et alii 2016, p. 299-301. A workshop of ARS A was r (...)
- 42 E.g. Mackensen 1993, p. 170, 465. See also Mackensen, Schneider 2002, p. 125; Bonifay 2004, p. 47; (...)
46Sample 13302, fabric BD3, collected by Salomonson at Bordj el Djerbi, corresponds to the late 1st/2nd century vessel type Hayes 8A, a chronological outlier quite uncommon in the workshop area and therefore to be considered as a residual piece, not locally produced40. Its fabric is comparable in terms of both texture and composition to the (less coarse variants of the) ARS A fabric, attributed to the region of Carthage41. Also, the thick, impure slip matches those of that production. However, apart from a higher percentage of shale fragments, this fabric is, interestingly, also similar to thin section subgroup BD1.1, which would not exclude a regional production of ARS A, even if no hypothesis can be proposed based on only one sample. The possibility of an early tableware production in the late antique workshops of the El Mahrine zone has already been frequently discussed; however, so far, there is no evidence of predecessor-workshops at Bordj el Djerbi, Henchir el Biar, or El Mahrine42.
47Even if the number of analyzed samples is still limited, the results of this study once again highlighted the importance of extending the archaeometric (both petrographic and chemical) characterization of the ARS workshops to cover all the fabrics of each of them, increasing the number of available reference samples. As with the already studied workshops of Sidi Khalifa and Sidi Aïch, we discovered new fabric types for Bordj el Djerbi and Henchir el Biar. Thus, a simple petrographic analysis can, in addition to providing a typo-chronological evaluation, contribute to identifying the provenance of further (previously undetermined) samples at consumption sites.
48The fact that the ARS from Bordj el Djerbi, Henchir el Biar and El Mahrine, collectively determined as the “El Mahrine zone”, can be microscopically (and chemically) distinguished offers new ways of specifying the provenance of ARS D1-tableware and will give new insights into the organizational structure and economic relation of the Mejerda potteries. Moreover, the new analyses raised some questions about other, still unknown production sites in the Mejerda Valley. These hypotheses, however, remain to be confirmed by further interdisciplinary investigations.