- 1 We would like thank Maud Webster for proofreading the English text.
Fig. 1: Schematic map of Tunisian ARS-workshops and productions
(based on Bonifay 2004, p. 46, fig. 22; adapted by C. Hasenzagl)
- 2 Since Hayes’ typo-chronological system is mostly based on ARS vessel types exported to and found o (...)
- 3 A restricted geographical distribution was already assumed by Stern (1968, 149): “Sans doute, la d (...)
1The South Tunisian site of Sidi Aïch is located c. 35 km north-west of the city Gafsa (ancient Capsa) on the south-eastern foot of Djebel Sidi Aïch (fig. 1). It was one of the main producers of so-called continental2 African Red Slip Ware (hereafter ARS) made for intra-African distribution3. Like other main ARS centers, the site is located at the base of the hills of the Tunisian ridge (Dorsale Tunisienne), which runs from SW to NE, from the inland toward the coast at Cape Bon. The ridge is formed by sedimentary sequences locally rich in fine clays, suitable for good-quality pottery production with almost no need for modification. In particular, the hills near Sidi Aïch are mainly composed of Lower Cretaceous marls, clays, and sandstones (fig. 2). A few kilometers to the west of the site lies the wadi of Sidi Aïch.
Fig. 2: Geological map of the area around Sidi Aïch
Modified from Azzouz, A. and T. Lajmi (eds.) 1985-1987. Carte géologique de la Tunisie (éch. 1: 500.000).
(Tunis, Service de Géologie National)
- 4 Sidi Aïch was first mentioned by V. Guérin in his Voyages (1862, I, p. 290-294). See Saladin 1882- (...)
- 5 Cagnat 1886, p. 74; 1888.
- 6 Cagnat 1888; Stern 1968.
- 7 After Salomonson’s survey campaigns were concluded, the collected material was, authorized by the (...)
- 8 Nasr 1992 and 2005, unpublished surveys cited by M. Nasr in Bonifay et alii 2015, p. 147 and note (...)
2Ruins known as Henchir Sidi Aïch have been explored by travelers and scientists since the mid-19thcentury4. Due to a huge amount of sherds found at the site, a ceramic production was already identified by L. Cagnat in 18865. Further surveys and a first comprehensive analysis of finds, however, were carried out by J.W. Salomonson and E.M. Stern only in 1966, followed by a preliminary report published in 19686. Salomonson, who visited Sidi Aïch on the 2nd of October 1966, surveyed six small mounds (monticules), of which five are marked on a sketch plan (fig. 3) possibly indicating several workshop areas, with the two well-preserved mausoleums serving as landmarks. He collected 390 diagnostic pieces of ARS and Cooking Ware including a lot of wasters7. The most recent and most extensive archaeological activities in Sidi Aïch have been conducted by Mongi Nasr since 19908.
Fig. 3: Sketch plan by Salomonson of the surveyed area at Sidi Aïch showing five mounds (monticules) and two mausoleums serving as landmarks
A sixth mound is not illustrated in the plan.
(photo C. Hasenzagl)
- 9 See for example Capelli et alii 2016; Bonifay, Malfitana 2016.
- 10 The effectiveness of this combined approach was already tested on ARS and Cooking Ware samples fro (...)
3Linked to earlier chemical studies, the ongoing combined typological and petrographic characterization of kiln wasters and the creation of reference groups of African pottery workshops, in particular, ARS-ateliers now enable us to better identify the precise origin of most of the African wares found at consumption sites, and thus to significantly improve the reconstruction of ancient economic relations and trade routes9. This paper, following the combined methodological approach presented in the article on the pottery collected by Salomonson at the workshop of Sidi Khalifa, will focus on the petrographic characterization of 15 representative samples of Sidi Aïch’s ARS and Cooking Ware production by both thin section analysis under the polarizing microscope and standardized fabric description done with a stereomicroscope10. The selected samples cover the assemblage’s variability of fabrics, slip, and typology.
- 11 Bonifay, Capelli, Brun 2012, p. 54-55.
- 12 We thank M. Bonifay for this piece of information.
4The present study will integrate the preliminary petrographic characterization of Sidi Aïch’s ARS based on the thin section analysis11 of three kiln wasters (of two Stern XXXV and one Stern VI) selected among sherds from the Salomonson’s survey stored at the Centre Camille Jullian, given to R. Guéry by Salomonson himself12.
- 13 Bonifay et alii 2015, p. 144; Mukai et alii 2016, p. 158.
- 14 The comparison with that reference material also allowed the identification of several Sidi Aïch-A (...)
- 15 Cagnat 1888; Bonifay et alii 2015, p. 145, fig. 3; Mukai et alii 2016. The punch, which was alread (...)
5A previous preliminary archaeometric investigation focusing on the identification of pottery from Sidi Aïch was performed on ARS samples from the Aubert-Buès collection at the Museum of Gap (Hautes-Alpes, France). This collection, comprising a large quantity of ceramics from the Tunisian-Algerian inland, testifies to the efforts of Clément Aubert (1848-1932), who gathered the ensemble during his time (1877-1901) as an engineer and director of a railway company in Algeria and Tunisia, and of his nephew and heir Jean Buès (1884-1972)13. Some of the ARS was attributed to Sidi Aïch for typological and stylistic reasons. However, both petrographic and portable XRF chemical analyses of a few items neither confirmed that hypothesis nor yielded conclusive results for most of the studied cases, also because of the limited number of analyzed samples and the scarcity of reference material from Sidi Aïch (those stored at the Centre Camille Jullian) and other continental workshops available at that moment14. On the other hand, non-destructive portable XRF analyses contributed to ascertain the provenance of a double-punch included in the collection from Sidi Aïch15.
Fig. 4: Selected Sidi Aïch ARS samples from Salomonson’s survey
(drawings C. Hasenzagl)
Fig. 5: Selected Sidi Aïch ARS and Cooking Ware samples from Salomonson’s survey
(drawings C. Hasenzagl)
- 16 ARS material from Henchir es Srira that is also a producer of continental tableware and present in (...)
- 17 Moreover, Stern differentiated 17 base types (P1 – P17) and 10 types of stamp decoration (S1-S10). (...)
6The original typology of 42 different vessel types (Stern I – XLII, with subtypes) created by Stern and aiming at sorting and comparing the large amount of survey material from both Sidi Aïch and Henchir es Srira16 will be used in what follows17. Nevertheless, the repertoire of vessel types has similarities to the vessel types presented by Hayes or even equals them. Where applicable, an additional assignment to Hayes’s types will therefore be given.
- 18 A detailed typological study of all pieces from Sidi Aïch collected by Salomonson is part of the o (...)
7The ARS vessels documented in Salomonson’s survey collection18 are Stern Ia, c-f (Hayes 27, Hayes 27/31, Hayes 62, Hayes 1972, fig. 58a); Stern III; Stern IV (Hayes 1972, fig. 58b); Stern V; Stern VIa-b (Hayes 76, type Sidi Jdidi 3); Stern VIIa; Stern VIIb-c (Hayes 68); Stern Xd; Stern XI; Stern XII (Hayes 48B, Hayes 60, 1-2); Stern XIII; Stern XIV (Hayes 33,2); Stern XV (similar to Hayes 32); Stern XVI; Stern XVII; Stern XVIII (similar to Hayes 32); Stern XIX, Stern XX, Stern XXIa, Stern XXIII, Stern XXVII, Stern XXVIII, Stern XXIXa-c (Hayes 91); Stern XXXa-c; Stern XXXIII (Hayes 89); Stern XXXIV (Hayes 84); Stern XXXVa-b (Hayes 82, Hayes 87A); Stern XXXVI; Stern XXXVII; Stern XXXVIII (var. Hayes 76). A small amount of Cooking Ware collected by Salomonson includes casseroles Stern XLb (Hayes 183 and Hayes 184) and lids Hayes 195.
8The plain and stamp-decorated ARS of Sidi Aïch that was made from the 3rduntil the 6thcentury A.D. does not match the standard forms of exported African tableware. Sidi Aïch’s ARS is very heterogeneous in terms of the macroscopic appearance due to the great variation of slip color, which ranges from red (Munsell soil charts: 2.5 RY 4/6; 4/8; 5/6; 5/8), dark red (2.5YR 3/6), dark brown (7.5YR 3/2) to very dark gray (7.5YR 3/1) and black (7.5YR 2.5/1), and includes every nuance in between as well – all of which may appear on one fragment.
9Fifteen representative samples of varying fabric, typology, color, and slip were selected for thin section analysis, of which the results will be presented here. This selection was also informed by the binocular study of the standardized fabric description of all of Sidi Aïch’s samples from Salomonson’s survey collection that initially resulted in the identification of four main fabric groups. These representative samples include 14 specimens of ARS (Stern Ia, Stern Id, Stern Ie, Stern If, Stern VIa, Stern VIa/b, Stern VIIa, Stern XVIII, Stern XXIXc, Stern XXXIII, Stern XXXVa, see figs. 4 and 5), and one fragment of Cooking Ware (Stern XLb, see fig. 5).
10Under the polarizing microscope, the 15 samples evince a certain compositional and textural variability (figs. 6-7, tab. 1). However, they share various characteristics in common, in particular:
11a) an Fe-oxide-rich clay matrix with a very subordinate calcareous component; b) the presence in the groundmass of rather abundant calcareous microfossils (mostly foraminifera, with rare echinoid radioles, fig. 8), frequently decomposed by relatively high firing temperatures, forming micritic aggregates or voids with a yellowish reaction halo, and sometimes filled or pseudomorphosed by secondary carbonates; c) the relatively abundant and fine-grained inclusions, mostly <0.2-0.3 mm (never >0.5-0.6 mm); d) the dominant presence of quartz grains; they are angular in most cases, but rare rounded (aeolian) grains can be also found in the coarser fraction; e) the subordinate presence of frequently zoned feldspar; f) the minor presence of quartz-feldspar sandstone (fig. 6 with the photos of gr. 1.1 and 1.2 in the right column) or siltstone (fig. 6 with the photo of gr. 2 in the right column), whose grains are quite similar to the individual silty and sandy inclusions of the paste, pointing to one sedimentary source, shale (fig. 6 with the photo of gr. 2 in the right column) and Fe-rich clay nodules, and mica; g) the quite high quality of the Fe-rich slip, which is very pure and often rather regular in thickness, and generally redder/darker than the body matrix (fig. 10).
12However, the studied assemblage shows a clear variability in some textural features, in particular, the frequency and sorting degree of the siliciclastic inclusions (figs. 6-7), and the frequency of the planar parallel voids related to the wheel throwing (fig. 9). These differences led to the identification of five fabric groups (1-5) that can be distributed over two larger and less defined “families” (A-B).
Fig. 6: Representative fabrics of Sidi Aïch’s ARS
Left: thin-section microphotographs (crossed Nicols, real dimensions 1.3x1mm; C.Capelli). Middle: thin-section microphotographs (crossed Nicols, real dimensions: 1.3x1mm; C.Capelli). Right: macrophotographs (16×, real dimensions: 6.2×4.6mm; C.Hasenzagl)
13Group 1 (fig. 6) is characterized by the well-sorted, abundant, or moderately abundant inclusions, with a bimodal distribution (sand – mainly in the fraction 0.1-0.2 mm – and fine silt in comparable proportions). Mica is fairly frequent in the fine fraction. The rock fragments (especially the shales) are rather rare and included in the coarser fraction. The clay matrix is quite compact due to the scarce voids and silty fraction. Inclusions are slightly more numerous and coarser in sub-group 1.1, which also shows several silt-rich nodules, not completely mixed in the paste. It cannot be excluded that the sandy fraction was intentionally added as a temper.
14Group 2 (fig. 6), shows a silty fraction fairly comparable to group 1, but it is distinguished by the scarcer, poorly sorted, mostly fine-grained sandy inclusions and by the presence of relatively frequent clayey/limonitic nodules, especially in sample 13345. As in the previous group, planar voids are rare.
15Group 3 (fig. 6) is distinguished by poorly sorted, moderately abundant inclusions, with relatively frequent rock fragments, and by an oriented texture due to frequent planar parallel voids.
16Group 4 (fig. 7), rather coarse, is characterized by very abundant, poorly sorted siliciclastic inclusions. Like in group 3, rock fragments and planar voids are frequent. Shales and clayey/limonitic nodules are relatively frequent in both groups. Apparently, firing temperatures were higher here than in the other groups.
Fig. 7: Representative fabrics of Sidi Aïch’s ARS and Cooking Ware
Left: thin-section microphotographs (crossed Nicols, real dimensions: 1.3x1mm; C. Capelli). Middle: thin-section microphotographs (crossed Nicols, real dimensions: 1.3×1mm; C. Capelli). Right: macrophotographs (16×, real dimensions: 6.2×4.6mm; C. Hasenzagl).
17“Group 5” (formed by only one thin-sectioned sample, the only Cooking Ware analyzed, fig. 8) is similar to group 4, but inclusions are a little less coarse and microfossils are rarer.
Fig. 8: Sidi Aïch
Examples of calcareous microfossils: macrophotograph (above; 25×, real dimensions 2.6×1.9 cm; C. Hasenzagl); microphotographs (parallel Nicols, real area: 0.24×0.32 mm; C. Capelli). Foraminifer (middle), echinoid radiole (below).
18Particularly the sorting degree of the inclusions and the frequency of voids allow for the distinction of the two technological (?) families A and B.
19Group 1 (family A), characterized by a compact matrix and well-sorted inclusions, is different from groups 3, 4, and 5 (family B), which are distinguished by poorly sorted inclusions and frequent planar voids. Group 2 shows intermediate features (A/B), as it is rather similar to group 1, but the inclusions are poorly sorted.
20The slip thickness is variable from very thin (10-20 microns) to fairly thick (40-60 microns) (fig. 9). There are no specific slip variants related to groups or families. However, in a general way it can be noted that slips are rather thin in group 1 and thicker in groups 2 and 3, while they are more variable in groups 4 and 5.
Fig. 9: Sidi Aïch
Microphotographs (parallel Nicols, real area: 1×1.3mm; C. Capelli). Examples of textures of family A (left) and B (right).
21Finally, it is noteworthy that the previously analyzed reference samples from Stern’s surveys could be attributed to groups 1 and 2 (families A or A/B). On the contrary, no fabric typical of family B was observed.
Fig. 10: Sidi Aïch
Microphotograph (parallel Nicols, real area: 0.24×0.32 mm; C. Capelli). Example of slip.
22The considerable amount of ARS collected by Salomonson provide large numbers of comparable samples for a binocular study. All of the Sidi Aïch samples examined during the standardized fabric description by stereomicroscope are, first of all, characterized by the presence of numerous tiny, roundish, pale yellow to white carbonate inclusions (microfossils, essentially foraminifera: see the thin section paragraph) or micro-voids with a yellowish halo, which are especially discernable with samples of darker (= red) body clay and/or samples fired at relatively low temperatures (figs 6-7).
23The number and average size of these components are the primary microscopic features that allow for the distinction of four fabric types. Additionally, these types show varying percentages and sorting of quartz/feldspar (mostly angular in shape), rust-colored, brownish, and black particles (shale and Fe-rich nodules: see the thin section paragraph) as well as silver mica. In most cases, the types distinguished under the stereomicroscope match the fabric groups identified by the thin section analysis.
24Fabric group 1 (fig. 6) with the subgroups SA-ARS-1.1 and SA-ARS-1.2 can be included in thin section group 1, corresponding to subgroups 1.2 and 1.1. SA-ARS-1.1 has a medium-fine texture and is riddled with carbonate elements. There also exist frequent clear or gray quartz/feldspar particles, small black and small to medium-sized rust-colored inclusions as well as mica in small quantities. SA-ARS-1.2 is very similar to SA-ARS-1.1, but the inclusions tend to be larger, resulting in a generally coarser appearance of the fabric.
25SA-ARS-2 (fig. 6) equals thin section group 2. It shows fewer calcareous inclusions than SA-ARS-1.1 and SA-ARS-1.2. The petrographic relationship of groups 1 and 2, however, is visualized by the same nature of inclusions, their appearance, and their composition.
- 19 The planar voids with uniform orientation documented during thin section analysis were not conside (...)
26SA-ARS-3 (fig. 6) is analogous to thin section group 319. It is characterized by a generally bigger size of carbonate elements that can reach dimensions also visible to the
naked eye.
27SA-ARS-4 (fig. 7) includes the samples of thin section groups 4 and 5. It is the coarsest of all these fabrics, consisting of a large number of big, angular particles of quartz/feldspar, whereas carbonate inclusions are fairly scarce.
28The majority of the survey material consists of fabric types SA-ARS-1 – SA-ARS-2. Smaller quantities account for fabric SA-ARS-3. Representatives of SA-ARS-4 including the sample of Cooking Ware that corresponds to thin section group 5, are rare in Salomonson’s survey collection.
- 20 A differentiation of Sidi Aïch’s fabrics without optical aid is usually not possible, due to the g (...)
29This petrographic study of the samples from Salomonson’s survey has pointed out the presence of a combination of various discriminant markers in Sidi Aïch’s fabrics, in particular, the partially calcareous clay matrix, the abundance of microfossils, the fragments of quartz-feldspar sandstone/siltstone and shale, the relative abundance of feldspar and mica, the scarcity of aeolian quartz, and the pure, dark slip. Part of them is also easily recognizable with the binocular microscope or even a 10× lens. On the whole, the results of the standardized fabric description by stereomicroscope were confirmed by the thin section study20.
- 21 See Bonifay, Capelli, Brun 2012, p. 54-55: «la production de l’atelier de Sidi Aïch se reconnaît (...)
30These results support and integrate those deriving from the preliminary interdisciplinary research by Bonifay, Capelli, and Brun that evinced the peculiarity of Sidi Aïch’s production compared to other ARS productions, not only in terms of micro- and macro-features of the fabric and the slip, but also in terms of typology and chemical characteristics (in particular the high K-contents and the peculiar Si/Al ratios)21.
- 22 See Bonifay, Capelli, Brun 2012; Hasenzagl, Capelli 2019.
- 23 See Nasr, Capelli 2018b.
31The abundance of microfossils and a partially calcareous matrix are quite uncommon in all the other ARS productions, even if minor percentages of microfossils also characterize Djilma fabrics. The presence of quartz-feldspars and siltstone, the relative abundance of feldspar and mica, and the scarcity of aeolian quartz easily allow the distinction of Sidi Aïch, as well as the continental workshops of Henchir es Srira and Djilma, from the northern coastal ARS workshops22 and the southern one of Thelepte23.
- 24 The fabric classification is to be considered provisional, to be verified by the study foreseen fo (...)
32However, it is noteworthy that family B fabrics were not identified in the previously analyzed references. The petrographic composition of all Sidi Aïch fabrics shows direct relations to the raw materials (marls, clays, and sandstones) locally available in large quantities (fig. 2). The presence of several fabrics with similar characteristics24 (grouped in families) might be explained by “secondary” reasons, for example, the natural variability of the sedimentary layers, poorly standardized processes, or typological differences.
- 25 A detailed geological survey of the site area and the sampling of the potential raw materials woul (...)
33Nevertheless, the marked textural distinctions between the two major families A and B (more specifically, between group 1 and groups 3-5) depend on technological reasons, both in terms of choice/localization of the raw materials (not only in the sedimentary sequences but perhaps also in the alluvial deposits of the local wadis)25 and their processing.
34The high frequency of inclusions and their low degree of sorting point to the use of non- or poorly modified raw materials for family B fabrics. Among them is the one connected to (and appropriate for) Cooking Ware.
35On the contrary, the rather pure clay matrix and well-sorted sandy fraction of group 1 (family A) may be related either to the choice of very fine clay, to a levigation phase, or the use of a selected temper.
36The meaning of the presence of these two families is not completely clear at present. Neither a relationship with different workshop areas at the site nor the possibility that some samples are not kiln wasters but imports can be excluded. However, several clues seem to suggest that the different fabrics may relate to different chronological phases.
- 26 Whereas Stern VIIb and VIIc are the typical Hayes 68 versions with broad rim, curved wall and stro (...)
37In any case, a dependency of some fabric types and certain vessel types has been detected. Thus, the group of 5th-century plates and dishes, often decorated with chattering (see fig. 4), Stern XXXIII – XXXVI (one of the most common vessel types at Sidi Aïch and also represented in large numbers in Salomonson’s survey collection) is exclusively connected to fabric group 1 (SA-ARS-1.1 and SA-ARS-1.2). The 4th-5th century dishes Stern VI and VII26 as well as the flanged bowls Stern XXIX and Stern XXX were more often found in fabric group 1, but are also documented in group 2 (SA-ARS-2).
- 27 Paralleled with Hayes’ types, Stern I covers a typological development from the 3rdcentury forms H (...)
- 28 Whereas some fragments in group Stern XVIII are akin to Hayes fig. 58f, which itself is very simil (...)
- 29 It should be noted that fig. 4 and fig. 5 show only a small percentage of selected samples compare (...)
38On the other hand, Stern I, which is by far the most dominant vessel type in Salomonson’s survey, is connected to all fabric types. However, Stern I was produced in several variants from the 3rduntil the 5thcentury A.D.27. SA-ARS-3 is particularly connected to 3rd-4th century versions of Stern I, to Stern XVIII28 (3rd-4thcentury) as well as to Cooking Ware lids and, thus, perhaps to a specific production phase at Sidi Aïch. SA-ARS-4, which accounts for only a very small portion of the whole survey material, has a similar typological pattern, including a cooking ware casserole Stern XL (Hayes 184)29.
39The number of Cooking Ware samples in Salomonson’s survey is too small to make any conclusions about the numerical proportions of fabric types in the entire production. At the moment, however, it can be noted that there is an absence of Salomonson’s Cooking Ware samples in fabric group 1.
40In conclusion, at present, we cannot exclude the possibility of a chronological evolution of the local fabrics from family B to A connected to a typological evolution. In any case, the long period of production, as well as the existence of different fabric types, generally suggest the presence of several different workshops/districts in the same area of Sidi Aïch. Sidi Aïch is an exceptional case within Salomonson’s collection since it is the only site with detailed information on six distinct intra-site find spots (fig. 3).
41Considering the rather large number of samples, the scarcity of tools in Salomonson’s material is striking. Nevertheless, one single saggar from mound 5 and the misfired fragments found on all six mounds indicate activities connected to ARS production at each of them. The distributional analysis of fabric types in terms of their find spots, however, did not show any connection of individual fabrics to certain mounds since all fabrics were documented at all of them. If the location of different workshops at or close to the mounds is presumed, the occurrence of mixed fabric types could be explained by the usage of the same clay deposits by all workshops. Consequently, this would mean that the fabric types are not workshop-specific.
- 30 The comparative thin-section exam with the new references allowed us to more probably attribute se (...)
42Finally, the fact that family B fabrics were unknown until this study identified them has an important implication: in that only a part of Sidi Aïch products at consumption sites and in the Aubert-Buès collection had been recognized by previous petrographic analyses30. The new data from Salomonson’s survey increase the spectrum of the reference material and will allow to significantly improve the typological and distributional studies of Sidi Aïch’s ARS and Cooking Ware.