Professional Documents
Culture Documents
The aim of this work is to present the results of the archaeometric characterization of
amphorae recovered at the Late Antique city of Emporiae, currently Sant Martí d’Empúries
(Catalonia, Spain). A total of 53 amphora samples have been analysed using a combination
of techniques, including X-ray fluorescence, X-ray diffraction and optical microscopy by
thin-section analysis. The integrated results shed light on the provenance of these materials,
providing evidence for the study of the commercial trends of Emporiae mainly for the fifth
and sixth centuries AD. The majority of the imports were related to Tunisian amphorae, for
which a diversity of compositions was revealed, evidencing a shift from northern to central
Tunisian production centres supplying the city from the sixth century. A number of Eastern
Mediterranean, South Hispanic and Italic products were also identified, associated with a large
variety of compositional groups. This diversity is also evident within the same production region
and for amphorae of the same typology, indicating the arrival of similar products from different
workshops, as has been observed especially for the Eastern Mediterranean LRA 1 amphora type.
KEYWORDS: AMPHORA, OM, XRF, XRD, HISPANIA, LATE ANTIQUITY, ROMAN POTTERY,
WESTERN MEDITERRANEAN
INTRODUCTION
The ancient city of Emporiae, located at the southern end of the Gulf of Roses (Catalonia, Spain) in
the north-east of the Iberian Peninsula, has evidence of Greek settlement from the sixth century BC
and later experienced development as a Roman city (Fig. 1). After a gradual abandonment of dif-
ferent urban areas that started from the first century AD, only the so-called Palaia Polis―in the
actual town of Sant Martí d’Empúries―continued to be occupied during Late Antiquity,
functioning as the episcopal see and administrative centre of a large territory (Aquilué 2012).
Archaeological excavations carried out between 1994 and 1996 uncovered a substantial body of
material of great relevance for understanding the settlement in this place during Late Antiquity
(Aquilué and Burés 1999). A continuity of occupation is clearly visible in the findings from several
walls, rooms, silos and rubbish dumps. In addition, intense port activity is reflected in the arrival of
all the most commonly traded products in the Western Mediterranean during this period, including
amphorae with their contents, fine wares, coarse wares and cooking wares (Aquilué et al. 2008).
Figure 1 The location of Emporiae and a plan of the archaeological site, indicating the area of the Palaia Polis, cur-
rently Sant Martí d’Empúries. The Neapolis and the Roman City were abandoned after the third century AD (Museu
d’Arqueologia de Catalunya-Empúries).
Amphorae constitute the majority of the ceramic materials from a rubbish dump dated to the
first half of the fifth century AD and a series of building structures, silos and burials dated to
the sixth and seventh centuries AD. These containers were essential for the transport of oil, wine,
garum and other foodstuff.
In order to analyse the diversity of amphora products in this Tarraconensian consumption
centre and to study their provenance and technology, in this paper we present the results of the
archaeometric characterization of the Late Roman amphorae recovered at the excavations.
Through the provenance approach, this analysis aims to offer an insight into the commercial
trends of Emporiae mainly for the fifth and sixth centuries AD.
Many other significant Late Roman amphora contexts are known in the Catalan area (Keay
1984; Remolà 2000; Cela and Revilla 2004; Macias et al. 2008); however, very few
archaeometric studies have been carried out so far. Samples from Tarragona were chemically
analysed by Remolà et al. (1993, 1996), although no petrographic or mineralogical techniques
were applied. Some amphorae from Guissona (Uscatescu and García 2005) and a few from Iluro
(Mataró, Barcelona; see Buxeda i Garrigós and Cau Ontiveros 2004) were also analysed;
however, their inclusion in broader studies along with different pottery classes (e.g., common
and cooking wares, fine wares) did not allow for detailed conclusions on the amphora materials.
On the contrary, the present paper is focused only on the archaeometric analysis of a large set of
Late Roman amphorae from the Catalan coastal site of Emporiae, through the integration of
petrographic, mineralogical and chemical evidence.
MATERIALS
A total of 53 amphora samples have been considered for archaeometric analysis (Table 1; see
also Appendices S1a–S1c). A first classification of the assemblage was made on the basis of
macroscopic observation, from which the prevalence of presumably African, mostly Tunisian,
amphora types was clear. These comprise 32 samples, belonging mainly to types Keay 27 (second
half of fourth century to mid-fifth century), Keay 35 and 36 (both especially from the fifth century)
and Keay 62 (mainly of variant A, more widespread during the sixth century). Possible Tripolitanian
amphorae include a sample similar to type Tripolitanian III as well as an individual of type
Benghazi LR7.
Eleven samples correspond to presumably Eastern Mediterranean amphorae, mostly related to
types LRA1 (only the early variant Kellia 169, from the fifth century to the early sixth century,
was identified) and LRA4 (including an example of the fifth-century variant A). Three samples of
types from the Aegean area were also included (LRA2, simile LRA3 and simile Samos Cistern
Type).
Other two samples can be related macroscopically to Baetican Keay 13A. The remaining
samples comprise unique individuals of different morphologies, including some rims and bases
of indeterminate type (Table 1).
ANALYTICAL METHODS
All the samples were analysed using a combination of techniques, including X-ray fluorescence
(XRF), X-ray diffraction (XRD) and optical microscopy (OM) through thin-section analysis,
except for two samples (EMP351 and EMP352) that were only subjected to OM, since the small
sample size was a limiting factor for XRF and XRD preparation.
Petrographic analysis was carried out on thin sections using an Olympus BX41 optical micro-
scope, working between × 40 and × 400 magnification. The ceramic fabrics were described fol-
lowing the system proposed by Whitbread (1989, 1995).
For the chemical analysis, XRF was performed using a PANalytical-Axios Advanced PW
4400/40 spectrometer. Major and minor elements were determined by preparing duplicates of
glassy pills, while powdered pills were used for the determination of trace elements (for the an-
alytical routine, see Buxeda i Garrigós et al. 2003). The quantitative determination of 29 chem-
ical compounds and elements (Fe2O3, Al2O3, MnO, P2O5, TiO2, MgO, CaO, Na2O, K2O, SiO2,
Ba, Rb, Th, Nb, Pb, Zr, Y, Sr, Ce, Ga, V, Zn, Cu, Ni, Cr, Mo, Sn, Co and W) was obtained, al-
though some of them were discarded: Mo and Sn because of their low counting statistics and Co
and W due to possible contamination from the tungsten carbide cell of the mill. The analytical
data were subjected to different statistical procedures using the S-PLUS 2000 software, after
an additive log-ratio (alr) transformation of the values obtained by XRF (Aitchison 1986; Baxter
1994; Buxeda i Garrigós 1999).
The mineralogical composition of the samples was studied by XRD through a PANalytical X-
ray diffractometer; spectra were taken from 5° to 80° 2Θ, using a step size of 0.026° and a step
time of 47.5 s. The evaluation of crystalline phases was carried out using the High Score Plus
software by PANalytical, including the Joint Committee of Powder Diffraction Standards
(JCPDS) data bank.
Chemical results
In order to explore the chemical variability of the data set, the compositional variation matrix was
calculated (Buxeda i Garrigós and Kilikoglou 2003). A total variation (vt) of 2.60 was obtained,
Figure 2 Microphotographs of thin sections from the main Tunisian petrographic fabrics (PF), all taken in
cross-polarized light at the same magnification (40×): (a) EMP345 (PF 1A), a large quartz–sandstone fragment with
Fe-rich cement can be observed; (b) EMP349 (PF 1A), with a large calcareous clay pellet and a large quartz–sandstone
fragment with Fe-rich cement; (c) EMP321 (PF 1B), a finer fabric showing quartz–sandstone fragments with calcareous
cement; (d) EMP332 (PF 1C), similar to the latter but with less fine fraction and a greater presence of quartz in the range
0.20–0.30 mm; (e) EMP340 (PF 1D), abundant microfossils and micritic inclusions, apart from quartz; (f) EMP300
(PF 1E), coarser aeolian quartz inclusions.
indicating the polygenic origin of the assemblage. Most of this variation is associated with the
variation of a number of elements, particularly Ni (τ.Ni = 12.59), Cu (τ.Cu = 8.76), Na2O
(τ.Na2O = 8.47), CaO (τ.CaO = 8.14), Cr (τ.Cr = 7.66), P2O5 (τ.P2O5 = 6.58), MnO (τ.MnO = 6.35)
and Zr (τ.Zr = 6.34).
Figure 3 Microphotographs of thin sections from different petrographic fabrics (PF), all taken in cross-polarized light
at the same magnification (40×): (a) sample EMP351 (PF 2D), coarse fraction of calcareous sand, with abundant ser-
pentine (srp) and some quartz (qtz), clinopyroxene (cpx) and chert inclusions; (b) sample EMP342 (PF 12), dominant
metamorphic rock fragments in addition to some garnet (grt) and quartz inclusions, in a micaceous groundmass; (c) sam-
ple EMP343 (PF 3), fine limestone inclusions in addition to some quartz and argillites; (d) sample EMP334 (PF 7), fine
quartz and limestone inclusions in a micaceous groundmass.
The analytical data obtained from XRF on 51 samples were additive log-ratio (alr) transformed
using Fe2O3 as divisor, since it is the element with the lowest variability within the data set. The
chemical data are presented in Appendix S4.
A cluster analysis (using the centroid agglomerative method and the squared Euclidean distance)
was performed, including the subcomposition Al2O3, MnO, TiO2, MgO, CaO, Na2O, K2O, SiO2,
Ba, Rb, Th, Nb, Zr, Y, Sr, Ce, Ga, V, Zn, Cu, Ni and Cr; P2O5 and Pb were excluded due to possible
contamination problems related to these elements. The resulting dendrogram is shown in Figure 4,
from which three major clusters can be distinguished. Cluster A is defined especially by its much
higher levels of Ni and Cr, apart from lower SiO2 percentages and CaO, MgO, MnO and Cu con-
centrations which tend to be higher than in other clusters; all these samples show high CaO percent-
ages (15.1–25.9%). Cluster B can be differentiated from Cluster C by its low calcareous composition
combined with higher contents of Fe2O3, Al2O3, MnO and Cu (Fig. 4 and Appendix S4).
Six chemical groups were defined (Fig. 4 and Table 2). The first two groups (CG I and II) are
related to Cluster A, while the others (CG III–VI) are included in Cluster C.
The much higher Cr content in CG II (corresponding to LRA1-type samples) is fundamental for its
differentiation from CG I, along with lower concentrations of Al2O3, K2O, Rb, Zr and Ce, and higher
concentrations of MgO, Na2O, Sr and―in most cases―CaO (Appendix S4 and Table 2).
Figure 4 The dendrogram resulting from cluster analysis (using the centroid agglomerative method and the squared
Euclidean distance) on 51 samples, including the subcomposition Al2O3, MnO, TiO2, MgO, CaO, Na2O, K2O, SiO2,
Ba, Rb, Th, Nb, Zr, Y, Sr, Ce, Ga, V, Zn, Cu, Ni and Cr; Fe2O3 was used as divisor in the log-ratio transformation of
the data. Chemical groups (CG) and petrographic fabrics (PF) for each sample are also indicated.
Regarding the chemical groups in Cluster C, CG III is characterized especially by its higher
Na2O, K2O and Rb concentrations. Samples EMP302 and EMP304 are more closely related
(CG IIIa), EMP350 also presenting some proximity but differing especially in its higher Zn
and Cu contents.
CG IV can be differentiated by its higher concentrations of TiO2 and Zr and lower Rb than in
other groups (Table 2). On the other hand, CG V is the only non-calcareous group in Cluster C
(CaO, 3.9%; Sr, 155–167 ppm).
The large group CG VI comprises almost all the samples that were petrographically de-
fined as Tunisian (PF 1)―except for EMP307 and EMP313 in CG V―along with only
three samples with other fabrics (PF 8 and 9). Seven chemical subgroups (CG VIa–VIg)
can be identified (Fig. 4). The variation matrix within CG VI yields a vt of 0.94, which
is too high to indicate a monogenic population (Buxeda i Garrigós and Kilikoglou 2003).
In the subgroup CG VIb (n = 9), the vt value is 0.26, low enough to assume that it is a mono-
genic population. On the other hand, CG VIe, VIf and VIg seem to be chemically interrelated
(Fig. 4), but if they are considered together vt = 0.49 is obtained, which is still indicative of a
polygenic population; however, their separate consideration allows us to work with monogenic
populations. The mean chemical composition of subgroups CG VIb and VIg (the only two with
n > 5) and the standard deviations for each element are given in Table 2.
In order to obtain a better definition of the chemical subgroups in CG VI, principal component
analysis was performed, without considering P2O5 and Pb; however, after this first PCA it was
decided to exclude Na2O and Cu as well, which allowed a better separation of the previously de-
fined petrographic subgroups. In this second PCA, now using Al2O3 as divisor, the plot of the
first two PCs (Fig. 5) supports the subgroup division in Figure 4, except for CG VIa, in which
some important chemical differences between these two samples (see Appendix S4) prove that
this subgroup may not have a real meaning (also in agreement with their different fabrics in thin
section). PC1 is dominated by CaO, MgO and Sr, whereas the main elements in PC2 are CaO,
MgO and Zr; together, these account for 63.3% of the total variance.
CG VIc is clearly more calcareous than the other subgroups (CaO, 16.6–17.8%), while the
lowest CaO percentages are found in CG VIe (5.6–5.7%) and VIf (6.8–7.3%). CG VIg is also re-
lated to the latter subgroups (Fig. 4), presenting a low PC1 score (Fig. 5), although its separation
can be explained mainly by its higher concentrations of CaO (8.5–10.1%) and Sr (232–296 ppm)
as well as a MgO content (2.3–2.7%) that is intermediate between those two subgroups. Diffe-
rences in CaO are often related to the calcareous content of the argillaceous paste used for manu-
facture or to processes of secondary calcite recrystallization (Heimann and Maggetti 1981;
Cau Ontiveros et al. 2002). In this case, the subgroups defined in CG VI are still consistent even
after the exclusion of CaO and Sr from the analysis. As significant problems with secondary cal-
cite have not been observed, these differences seem to be related to the original raw materials.
From the PCA plot (Fig. 5), a higher PC2 score in CG VIe is evident, due not only to its low
CaO but also to its Zr content (313–334 ppm); MgO (1.9%) is lower than in CG VIf and VIg.
Conversely, in CG VIf, the MgO concentration is the highest (3.2–3.5%) and that of Ba is the low-
est (225–233 ppm) of all the samples in CG VI; this high MgO and the relatively low Zr content
(217–230 ppm) account for its low PC2 score.
On the other hand, CG VIb, which is the larger subgroup in CG VI, can be clearly differ-
entiated from the other subgroups especially by its lower MgO content (1.0–1.5%), which is
reflected in its high PC1 score (Fig. 5), and also due to its relatively high CaO (9.7–13.9%)
and Sr (297–369 ppm) contents. The K2O and Rb concentrations are also lower than in the
other subgroups (1.6–1.8% and 50–65 ppm, respectively).
With regard to CG VId, it can be distinguished from the remaining subgroups by a lower PC2
score (Fig. 5), in association with its high CaO (11.3–13.9%) and low Zr (165–184 ppm)
contents. The Zr concentration is the lowest in CG VI—with the single exception of the outlier
EMP325 (Zr = 177 ppm)—while the V content is the highest (126–133 ppm).
CG I (n = 4)
m 6.38 14.88 0.12 0.39 0.70 3.24 17.66 0.76 2.93 52.78 398
sd 0.35 1.39 0.01 0.09 0.04 0.91 3.33 0.27 0.39 2.38 64
CG II (n = 4)
m 6.41 11.81 0.12 0.39 0.71 4.96 21.81 1.64 2.15 49.80 302
sd 0.70 0.79 0.03 0.10 0.10 1.14 2.78 0.25 0.46 0.81 44
CG III (n = 4)
m 5.53 15.61 0.10 0.44 0.70 2.50 10.83 1.25 3.08 59.79 464
sd 0.26 1.35 0.02 0.12 0.06 0.17 1.93 0.17 0.11 2.00 94
CG IV (n = 2)
m 6.31 11.65 0.11 0.20 1.24 2.43 13.01 0.89 2.09 61.91 412
sd 0.87 1.20 0.01 0.04 0.11 0.02 1.22 0.15 0.20 1.35 72
CG V (n = 2)
m 6.53 15.52 0.04 0.17 0.84 1.91 3.95 0.34 2.55 68.02 341
sd 0.39 0.32 0.00 0.01 0.05 0.08 0.01 0.08 0.39 0.23 87
CG VI (n = 32)
m 5.12 12.97 0.04 0.33 0.73 1.98 10.61 0.44 2.19 65.46 292
sd 0.67 1.52 0.01 0.20 0.07 0.65 3.18 0.17 0.38 3.54 55
Subgroup VIb (n = 9)
m 4.62 12.04 0.04 0.27 0.70 1.21 11.28 0.26 1.72 67.72 281
sd 0.49 0.93 0.00 0.07 0.05 0.15 1.56 0.07 0.10 2.69 24
Subgroup VIg (n = 6)
m 5.66 14.56 0.04 0.28 0.81 2.56 9.34 0.53 2.46 63.63 279
sd 0.23 0.72 0.01 0.09 0.03 0.12 0.55 0.12 0.20 1.00 30
DISCUSSION
The combination of petrographic, mineralogical and chemical evidence, along with the archaeolo-
gical information, enables us to characterize archaeometrically the several groups of Late Roman
amphorae found in the consumption centre of Emporiae and to shed light on their provenance.
African amphorae
The division of Tunisian amphora samples into five petrographic subgroups (PF 1A–1E: see
Appendix S2) finds a correspondence with the chemical data from XRF, since each of them is
associated with a different chemical subgroup (Fig. 5). A particular case is sample EMP335,
with a fabric (PF 1B) similar to those of the samples included in CG VIe (EMP321 and
EMP323) but chemically different from the latter, especially in their MgO, CaO, Sr and
Cu composition (Fig. 5 and Appendix S4). On the other hand, a number of samples that
are petrographic loners associated with Tunisian fabrics (PF 1) also tend to behave as
chemical outliers within CG VI (EMP308, EMP325, EMP336, EMP338 and EMP346) or
are grouped separately in CG V due to more remarkable chemical differences (EMP307
and EMP313).
Recent petrographic studies (Bonifay et al. 2002, 2010a, 2011; Sherriff et al. 2002; Capelli
2005a,b, 2007; Ghalia et al. 2005; Capelli et al. 2006; Capelli and Bonifay 2007; Gandolfi
Table 2 (Continued)
Rb Th Nb Pb Zr Y Sr Ce Ga V Zn Cu Cr
CG I (n = 4)
m 101 10 16 24 161 27 248 62 17 108 91 67 270
sd 20 2 2 2 20 4 38 6 1 21 15 19 41
CG II (n = 4)
m 56 5 12 16 122 21 369 40 13 113 78 54 570
sd 10 2 1 4 4 1 30 7 0 14 10 22 137
CG III (n = 4)
m 125 11 19 28 182 28 368 75 18 100 95 38 97
sd 25 6 5 8 56 5 90 27 2 8 18 18 7
CG IV (n = 2)
m 39 5 21 17 373 29 279 60 15 101 63 43 110
sd 3 0 1 5 38 1 3 2 2 16 2 3 3
CG V (n = 2)
m 79 8 20 21 231 23 161 78 18 114 81 22 105
sd 0 0 0 0 40 3 8 5 0 1 2 14 9
CG VI (n = 32)
m 70 7 18 17 241 23 279 65 16 101 70 17 107
sd 12 1 2 2 38 3 69 8 2 16 12 6 11
Subgroup VIb (n = 9)
m 58 6 17 16 245 22 321 59 15 85 59 18 112
sd 5 1 1 1 14 2 27 6 1 9 7 4 5
Subgroup VIg (n = 6)
m 78 8 21 17 250 25 257 74 17 113 77 13 110
sd 3 1 1 2 15 1 22 5 1 5 5 2 4
et al. 2010) have differentiated between a number of Tunisian amphora workshops. In contrast,
there is a general scarcity of chemical analyses on these production centres, apart from a
few―mainly unpublished―early studies (Liddy 1985; Taylor 1993; Taylor et al. 1997; Sherriff
et al. 2002) that have not been continued in recent years. From the chemical–petrographic infor-
mation obtained in this study and its comparison to the information available from different
workshops, along with the typological evidence, a provenance hypothesis can be proposed for
most of the analysed samples (Table 3).
A large subgroup, defined from both petrographic (PF 1A) and chemical (CG VIb) evi-
dence, is mainly related to type Keay 27 (second half of the fourth century to the fifth cen-
tury) and Keay 36 (late fourth century to the first half of the sixth century). A similar fabric
has been reported in other samples of these amphora types (Bonifay et al. 2011). No work-
shops are known for them and the archaeological evidence points to a production in the
western outskirts of Carthage (Bonifay 2004, 22; Bonifay et al. 2011).
Conversely, a certain provenance in Nabeul zone B (north-eastern Tunisia), particularly in Sidi
Zahruni, can be proposed for PF 1B, since it matches the petrographic definition of this workshop
(Capelli 2005a,b; Ghalia et al. 2005; Bonifay et al. 2010a; Gandolfi et al. 2010). This subgroup is
associated with the Keay 35B amphora type, dated to the last part of the fourth century and es-
pecially the fifth century. Chemically, these amphorae can show some variability—see, for exam-
ple, sample EMP335—but the petrographic evidence is solid, since the fabric from this workshop
is quite characteristic.
Figure 5 PCA of the alr transformed chemical data (using Al2O3 as divisor in the log-ratio transformation) from CG VI
samples (excluding loners EMP308 and EMP338): a bivariate plot with the first two principal components, PC1 and
PC2, which explain 40.9% and 22.4% of the total variance, respectively. PCA was performed on the covariance matrix.
The P2O5, Na2O, Pb and Cu concentrations were not considered in this plot.
Two other samples of type Keay 35 (EMP307 and EMP313) exhibit particular fabrics in thin
section (both ascribable to the Tunisian group PF 1) and present a low calcareous composition
allowing the formation of a separate chemical group (CG V). This evidence indicates that pro-
duction of this type—or at least of amphorae with a similar rim, since the body of these samples
was not found—may not be restricted to the Nabeul region, which is the only one that is well
attested archaeologically (Bonifay 2004, 135), although for the moment it is not possible to pro-
pose a provenance hypothesis for these loners.
With regard to sample EMP339, despite its morphological similarities with some Keay 35B
rims (e.g., Remolà 2000, Fig. 37, nn. 8–9), it cannot be considered as a characteristic rim
for this type. In addition, its analysis reveals a chemical and petrographic composition
related to the previously defined group PF 1A/CG VIb, with a north-western Tunisian prov-
enance. This suggests that this amphora could be classified as a different type—indetermi-
nate so far, since it is also not completely identical to well-known amphorae with this
fabric composition.
Samples EMP300 and EMP305 were attributed to type Keay 39/Sidi Jdidi 4 (Bonifay et al.
2011), dated to the late fourth century and the fifth century. Both individuals show a similar
EMP315 Keay 27B VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP316 Keay 27B VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP317 Keay 36 VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP322 Keay 27B VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP329 Keay 36 VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP330 Keay 36 VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP339 African VIb 1A Fabric not compatible with eastern North-western Tunisia?
indeterminate Tunisian workshops (Carthage region?)
rim?
EMP345 Keay 27 or 36 VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP349 Keay 36 VIb 1A Typology and fabric not compatible North-western Tunisia?
with eastern Tunisian workshops (Carthage region?)
EMP321 Keay 35B VIe 1B Typology and fabric typical of Sidi Nabeul zone B: Sidi
Zahruni Zahruni
EMP323 African VIe 1B Slight differences from Sidi Zahruni Nabeul zone B: Sidi
indeterminate fabric: other workshop? technology Zahruni?
spike (spike different from rim)?
EMP335 Keay 35B VI 1B Typology and fabric typical of Sidi Nabeul zone B: Sidi
Zahruni (but lower-fired) Zahruni
EMP312 Keay 62A VIg 1C Typology and fabric compatible with Henchir ech Chekaf
Henchir ech Chekaf
EMP314 Keay 62A VIg 1C Typology and fabric compatible with Henchir ech Chekaf
Henchir ech Chekaf
EMP319 Keay 62 VIg 1C Typology and fabric compatible with Henchir ech Chekaf
Henchir ech Chekaf
EMP332 Keay 62A VIg 1C Typology and fabric compatible with Henchir ech Chekaf
Henchir ech Chekaf
EMP337 Keay 62A VIg 1C Typology and fabric compatible with Henchir ech Chekaf
Henchir ech Chekaf
EMP348 Keay 62A VIg 1C Typology and fabric compatible with Henchir ech Chekaf
Henchir ech Chekaf
EMP333 Keay 62A VIc 1D Fabric typical of Moknine Moknine
EMP340 Keay 62A VIc 1D Fabric typical of Moknine Moknine
EMP300 Keay 39/Sidi VIf 1E Typology and fabric suggest a north- North-eastern Tunisia
Jdidi 4 eastern Tunisian production (Nabeul not excluded)
EMP305 Keay 39/Sidi VIf 1E Typology and fabric suggest a north- North-eastern Tunisia
Jdidi 4 eastern Tunisian production (Nabeul not excluded)
EMP328 African VIa 1 (loner) Fabric partially compatible with Leptiminus not
indeterminate Leptiminus, although with some excluded
rim doubts
(Continues)
(Continues)
EMP344 Keay 62? VIa 1 (loner) Typology may be compatible with Indeterminate (Tunisia)
several workshops, but no fabric
compatibility
EMP308 Spatheion 1 VI 1 (loner) Typology compatible with several Indeterminate (possibly
workshops, but no fabric northern or central to
compatibility northern Tunisia)
EMP325 Keay 62Q/ VI 1 (loner) Typology and fabric not compatible Indeterminate (Tunisia)
Albenga 11/12 with known workshops
EMP336 African VI 1 (loner) Coarse quartz fabric but different Indeterminate (Tunisia)
indeterminate from PF1D; no fabric compatibility
rim with any known workshop
EMP338 African VI 1 (loner) Morphological similarities to type Indeterminate (Tunisia)
indeterminate Tripolitanian III, but fabric not
rim Tripolitanian
(Tripolitanian
III?)
EMP346 Keay 55/56/57 VI 1 (loner) Type typical of Nabeul, but fabric Indeterminate (Nabeul
only partially similar to Nabeul zone not excluded)
B
EMP307 Keay 35 V 1 (loner) Type typical of Nabeul, but fabric not Indeterminate (not
compatible Nabeul)
EMP313 Keay 35 V 1 (loner) Type typical of Nabeul, but fabric Indeterminate (Nabeul
only partially similar not excluded)
EMP341 Benghazi LR7 Loner 16 Fabric and typology point to a Tripolitania?
Tripolitanian origin, although no
workshops are known
chemical–petrographic composition (CG VIf; PF 1E), indicating their production in the same
workshop. Some fabric similarities with Nabeul zone B (Bonifay et al. 2011), or possibly with
north-western Tunisia/eastern Algeria (Bonifay 2004, 129), have been reported for other samples
of this type. The fabric observed in the two samples from Emporiae (PF 1E), despite being dif-
ferent from those typical of Nabeul zone B, might in any case suggest a provenance in north-
eastern Tunisia (the northern Gulf of Hammamet or the Cape Bon area), although in an unknown
workshop, possibly―but not necessarily―in the same Nabeul region.
Several samples of type Keay 62A—the most widespread amphora in the sixth century—were
also analysed. Most of them (EMP312, EMP314, EMP319, EMP332, EMP337 and EMP348)
form a well-defined chemical–petrographic group (PF 1C; CG VIg), the fabric of which can be
associated with a provenance in Henchir ech Chekaf (Capelli 2005a,b, 2007), a workshop located
in central-eastern Tunisia. Two other Keay 62A samples (EMP333 and EMP340) are in a differ-
ent chemical–petrographic group (CG VIc; PF 1D) that, according to its fabric characteristics,
can be related with a provenance in Moknine (Gandolfi et al. 2010), another workshop in
central-east Tunisia not far from Henchir ech Chekaf. Another sample of a possible Keay 62 am-
phora (EMP344) behaves chemically and petrographically as a loner, the same being observed in
a number of samples related to various Tunisian amphora types: EMP308 (Spatheion 1),
EMP325 (Keay 62Q), EMP 346 (Keay 55/56/57) and EMP328 and EMP336 (both indeterminate
rims). So far, it is not possible to determine the provenance of these samples, although a possible
origin in a Leptiminus workshop for EMP328 and in the Nabeul region for EMP346 may be pro-
posed (Table 3).
Despite the morphological similarities with type Tripolitanian III, sample EMP338 seems to be
a Tunisian product as well, since its fabric is not compatible with those related to that type
(Peacock 1984; Bonifay 2004, 29; Bonifay et al. 2010a). On the contrary, fabric PF 16
(EMP341) matches the one found in type Benghazi LR7 (Bonifay et al. 2010b), dated to between
the late fifth century and the first half of the sixth century. A Tripolitanian origin has been pro-
posed for this type; however, no workshops are known to date.
Some fluctuations in the dominance of amphora imports from the Zeugitana or Byzacena re-
gions throughout Late Antiquity have been reported (Bonifay 2004, 457; Reynolds 2010a,
140). In the case of Emporiae, the present study indicates a main presence of amphorae from
northern Tunisia until the fifth century and, from the sixth century, the arrival of Keay 62A
amphorae only from the Byzacena region in central-eastern Tunisia. Nonetheless, further inves-
tigations on Tunisian workshops will be crucial for improving the definition of the reference
groups with which the Emporiae samples can be compared, in order to better examine the exact
provenance of some of them. From the results of this study, it is remarkable that for some of the
main amphora types (Keay 35 in the fifth century and Keay 62A in the sixth century), there is
clearly more than one production centre in each case. As for types Keay 27 and 36, they seem
to come from the same production region—most probably in north-western Tunisia—although
the possibility that more than one workshop may be represented cannot be ruled out.
EMP309 LRA 1 Kellia 169 IIa 2A Oriens: northern Syria, Cilicia or Cyprus
EMP326 LRA 1 IIa 2B Oriens: northern Syria, Cilicia or Cyprus
EMP310 LRA 1 Kellia 169 II 2C Oriens: northern Syria, Cilicia or Cyprus
EMP306 LRA 1 Kellia 169 II 2D Oriens: northern Syria, Cilicia or Cyprus
EMP351 LRA 1 Kellia 169 n/a 2D Oriens: northern Syria, Cilicia or Cyprus
EMP343 LRA 2 I 3 Macedoniae: possibly Kounoupi (Argolis)
EMP352 Simile LRA 3 n/a 4 Asiana: western Asia Minor (Meander Valley or
Ephesus region)
EMP320 LRA 4 IV 5 Oriens: Palestine
EMP331 LRA 4A IV 5 Oriens: Palestine
EMP327 Tipo Tardío A (Remolà 2000) I 6 Aegean (possibly Crete, but other possibilities not
excluded)
EMP303 Similar to Reynolds (2010b, Ia 7 Asiana: eastern Aegean islands or western Asia Minor
fig. 6k–l)
EMP334 Samos Cistern Type? Ia 7 Asiana: eastern Aegean islands or western Asia Minor
EMP301 Keay 13A VId 8 Baetica (Guadalquivir Valley?)
EMP318 Indeterminate rim (related to VId 8 Baetica (Guadalquivir Valley?)
Keay 13A?)
EMP324 Keay 13A VId 9 Baetica (Guadalquivir Valley?)
EMP304 VIL/2-215 (Remolà 2000) IIIa 10 Baetica? (not Málaga region)
EMP302 Indeterminate rim IIIa 11 Baetica? (other possibilities not excluded: i.e., north-
eastern Sicily?)
EMP350 Keay 52 III 12 Italia suburbicaria: north-eastern Sicily (Naxos)
EMP311 Keay 52? Loner 13 Italia suburbicaria: northern Sicily
EMP342 Imitation of Keay 25? Loner 14 Baetica? Carthaginensis?
EMP347 Indeterminate rim III 15 Italia suburbicaria: central Tyrrhenian coast
Production of the LRA2 type is evidenced in Kounoupi (Argolis), Chios, Knidos and Kos
(Megaw and Jones 1983; Opaiţ 2004; Poulou-Papadimitriou and Nodarou 2007). The close
petrographic similarities between a LRA2 sample found in Emporiae (EMP343, associated
with PF 3) and some others from Argolis (P. Day, pers. comm.) lead us to propose its pos-
sible provenance in the Kounoupi workshops.
Another Aegean petrographic group is PF 7, which is also defined by its particular chem-
ical composition (CG Ia). The fabric is similar to those described by Williams (1990) for
Samian amphorae, although, petrographically, other islands close to Samos as well as the
western Turkish coast cannot be excluded as possible source areas. This group comprises
one sample (EMP334) morphologically related to the Samos Cistern Type and one
(EMP303) that can be compared to a late sixth century Icarian amphora found in Butrint
(Reynolds 2010b, Fig. 6k–l).
A possible origin in Crete (Reynolds 2010a,2010b) or south-western Peloponnese (Yangaki
2014) was proposed for ‘Tipo Tardío A’ (Remolà 2000; Bonifay et al. 2013). Chemically, sam-
ple EMP327, ascribed to that type, presents some broad relationship with others from the Aegean
area (CG I). The fabric is less micaceous than the one described for Peloponnesian products by
Yangaki (2014) but, in any case, it does not provide conclusive evidence for assessing the hy-
pothesis of a possible Cretan provenance.
On the other hand, the fabric in sample EMP352 (PF 4) presents a general correspondence
with the one usually found in the LRA3 type (Peacock 1984; Capelli 1998; R. Sauer, in Bezeczky
2013), despite the fact that the sample is not morphologically equivalent to this type. The ac-
cepted production areas for the LRA3 type lie in western Asia Minor, particularly in the Meander
Valley and the Ephesus region, among others (Rautman 1995; Piéri 2005, 100; Bezeczky 2010,
2013), although no workshops have been identified so far.
Other amphorae
Ten samples from the data set can be associated with a provenance outside Africa and the Eastern
Mediterranean, including mainly southern Hispanic amphorae and a few Italic samples (Table 4).
Southern Hispanic amphorae represented one of the major imports in Tarraco and other
Tarraconensian centres until the mid-fifth century, arriving in lesser amounts until the first part
of the sixth century (Remolà 2000). However, Emporiae was an exception, as these are much less
represented than African and Eastern Mediterranean amphorae (Aquilué and Burés 1999).
The few Keay 13A samples analysed here (EMP301 and EMP324) reveal two different fabrics
(PF 8 and 9) that in both cases are compatible with an origin in the Baetican region, as is pro-
posed for this type (Remolà 2000); PF 8 is also associated with sample EMP318, an indetermi-
nate rim. Despite this petrographic diversity, the three samples are closely related chemically (CG
VId) and an interesting similarity is observed with Tunisian amphora composition, some trace
elements (Zr and V) being a key factor for their differentiation. It is not possible for the moment
to determine a more precise provenance for these samples from Emporiae. The Guadalquivir and
Genil Valleys seem a probable hypothesis, although archaeometric studies from workshops in
these areas are needed. From the small amount of published archaeometric information on the
Keay 13 production centres (Vigil et al. 1998), the Los Matagallares workshop—on the coast
of Granada—can be excluded, and the fabric is also different from those from the Málaga region
(C. Capelli, pers. comm.).
Sample EMP342 testifies to the arrival of an amphora with a spike similar to Keay 25 that, ac-
cording to its fabric (PF 14), does not have the Tunisian origin usual for this type. The production
of Keay 25 imitations is evidenced in some workshops near Cartagena (Berrocal 2012), although
the published fabrics (Arana 2007) do not seem to match with the sample found in Emporiae.
Other possibilities include some workshops on the Málaga coast (e.g., El Secretario), with reports
of Keay 25 type imitations (Serrano 2004) located near some outcrops of the Alpujarride com-
plex, in which the joint presence of metamorphic schists with some minerals such as garnet
and staurolite is not uncommon.
Samples EMP302 and EMP304 are similar chemically (CG IIIa) but each one presents a dif-
ferent fabric (PF 11 and PF 10, respectively). The first one, an indeterminate rim, might corre-
spond to a Baetican or north-eastern Sicilian product according to its fabric, but for EMP304
—of the VIL/2-215 type (Remolà 2000)—a Baetican provenance seems a probable hypothesis,
however excluding the Málaga region.
Sample EMP350 corresponds to a Keay 52 amphora and its fabric (PF 12) matches the ones
that Capelli (1998) describes for this type, particularly those with a provenance in Naxos
(north-eastern Sicily) rather than the southern Calabrian examples. Another similar concave base
(EMP311) could also be associated with the Keay 52 type, but its sedimentary fabric (PF 13) ex-
cludes production in either of those two areas, northern Sicily (west of the Naxos region) being a
possible hypothesis instead (C. Capelli, pers. comm.). On the other hand, the fabric in sample
EMP347 (PF 15) can be associated with a provenance on the central Tyrrhenian coast.
ACKNOWLEDGEMENTS
This study is part of the activities of the projects ‘Archaeology and archaeometry of Late
Roman cooking wares in the Western Mediterranean: a holistic approach’ (LRCWMED)
(ref. HAR2009-08290, sub-programme HIST) and ‘Late Roman pottery in the Western
Mediterranean: exploring regional and global trade networks through experimental
sciences’ (LRPWESTMED) (ref. HAR2013-45874-P), funded by the National Plan of I
+ D + i Ministerio de Economía y Competitividad, Secretaría de Estado de Investigación,
Desarrollo e Innovación, with contributions from the European Regional Development
Fund. This is part of the activities of the Equip de Recerca Arqueològica i
Arqueomètrica de la Universitat de Barcelona (ERAAUB), Consolidated Group (2014
SGR 845), thanks to the support of the Comissionat per a Universitats i Recerca del
DIUE de la Generalitat de Catalunya. The work of L. Fantuzzi was carried out thanks
to a scholarship, Ajut de Personal Investigador en Formació, of the University of
Barcelona.
We are indebted to the Museu d’Arqueologia de Catalunya-Empúries (P. Castanyer,
J. Tremoleda and M. Santos) for all the facilities for the study of the amphora assemblages.
The XRF and XRD analyses were undertaken at the Centres Científics i Tecnològics of the
University of Barcelona. We are grateful to P. Day and C. Capelli for their helpful observations
and suggestions in the petrographic analysis and to P. Reynolds for typological advice. We are
also most grateful to two anonymous referees for their helpful comments that have certainly
helped to improve the final version of this paper.
Aitchison, J., 1986, The statistical analysis of compositional data, Chapman and Hall, London.
Aquilué, X. (ed.), 2012, Empúries: municipium emporiae, Ciudades romanas de Hispania 6, L’Erma di Bretschneider, Roma.
Aquilué, X., and Burés, L., 1999, La ciutat en l’Antiguitat Tardana: Fase V, in Intervencions arqueològiques a Sant Martí
d´Empúries (1994–1996): de l´assentament precolonial a l´Empúries actual (ed. X. Aquilué), 389–422, Monografies
Emporitanes 9, MAC-Empúries.
Aquilué, X., Castanyer, P., Santos, M., and Tremoleda, J., 2008, L’evolució dels contextos ceràmics d’Empúries entre els
segles II aC i VII dC., in SFECAG—Actes du Congrès de L’Escala-Empúries, 2008, 33–62, Société Française d’Étude
de la Céramique Antique en Gaule, Marseille.
Arana, R., 2007, Estudio petrográfico de algunas cerámicas de El Mojón (Cartagena), in Estudios de cerámica
tardorromana y altomedieval (eds. A. Malpica and J. C. Carvajal), 315–8, Alhulia, Granada.
Baxter, M. J., 1994, Exploratory multivariate analysis in archaeology, Edinburgh University Press, Edinburgh.
Berrocal, M. C., 2012, Producciones anfóricas en la costa meridional de Carthago-Spartaria, in Cerámicas
hispanorromanas II: producciones regionales (eds. D. Bernal and A. Ribera), 255–77, Servicio de Publicaciones
de la Universidad de Cádiz, Cádiz.
Bezeczky, T., 2010, Trade connections between Ephesus and the Adriatic region, Histria Antiqua, 19, 351–8.
Bezeczky, T., 2013, The amphorae of Roman Ephesus, Verlag der Österreichischen Akademie der Wissenschaften, Wien.
Bonifay, M., 2004, Études sur la céramique romaine tardive d’Afrique, BAR International Series 1301, Archaeopress, Oxford.
Bonifay, M., Capelli, C., and Long, L., 2002, Recherches sur l’origine des cargaisons africaines de quelques épaves du
littoral français, in Vivre, produire et échanger: reflets méditerranéens. Mélanges offerts à Bernard Liou (eds. L. Rivet
and M. Sciallano), 195–200, M. Mergoil, Montagnac.
Bonifay, M., Capelli, C., and Moliner, M., 2011, Amphores africaines de la basilique de la rue Malaval à Marseille
e
(V siècle), in SFECAG—Actes du Congrès d’Arles, 2011, 235–54, Société Française d’Étude de la Céramique
Antique en Gaule, Marseille.
Bonifay, M., Capelli, C., and Muçaj, S., 2010b, Amphores tardives de tradition punique: observations sur le type
Benghazi LRA 7, in Oleum non perdidit. Festschrift für Stefanie Martin-Kilcher zu ihrem 65. Geburtstag (eds. C.
Ebnöther and R. Schatzmann), 151–9, Archäologie Schweiz, Basel.
Bonifay, M., Capelli, C., Drine, A., and Ghalia, T., 2010a, Les productions d’amphores romaines sur le littoral tunisien:
archéologie et archéométrie, Rei Cretariae Romanae Fautorum Acta, 41, 319–27.
Bonifay, M., Capelli, C., Franco, C., Leitch, V., Riccardi, L., and Berni, P., 2013, Les Thermes du Levant à Leptis Magna:
e e
quatre contextes céramiques des III et IV siècles, Antiquités Africaines, 49, 67–150.
Burragato, F., Di Nezza, M., Ferrazzoli, A., and Ricci, M., 2007, Late Roman 1 amphora types produced at Elaiussa Se-
baste, in LRCW 2. Late Roman coarse wares, cooking wares and amphorae in the Mediterranean: archaeology and
archaeometry (eds. M. Bonifay and J.-Ch. Tréglia), 689–700, BAR International Series 1662 (II), Oxford.
Buxeda i Garrigós, J., 1999, Alteration and contamination of archaeological ceramics: the perturbation problem, Journal
of Archaeological Science, 26, 295–313.
Buxeda i Garrigós, J., and Cau Ontiveros, M. Á., 2004, Caracterització arqueomètrica de les produccions tardanes d’Iluro,
Laietania, 15, 449–98.
Buxeda i Garrigós, J., and Kilikoglou, V., 2003, Total variation as a measure of variability in chemical data sets, in Pat-
terns and process: a Festschrift in honor of Dr. Edward V. Sayre (ed. L. van Zelst), 185–98, Smithsonian Center for
Materials Research and Education, Washington, DC.
Buxeda i Garrigós, J., Cau Ontiveros, M. Á., and Kilikoglou, V., 2003, Chemical variability in clays and pottery from a
traditional cooking pot production village: testing assumptions in Pereruela, Archaeometry, 45, 1–17.
Capelli, C., 1998, Il contributo delle analisi minero-petrografiche allo studio delle anfore Keay LII, in Ceramica in Italia:
VI–VII secolo. Atti del convegno in onore di John W. Hayes. Roma, 11–13 maggio 1995 (ed. L. Saguì), 335–42,
Edizioni All’Insegna del Giglio, Firenze.
Capelli, C., 2005a, Ricerche petrografiche preliminari sulle ceramiche ‘eoliche’, Antiquités Africaines, 38–9, 178–83.
Capelli, C., 2005b, Tunisian Fabric, in Roman amphorae: a digital resource (University of Southampton), Archaeology
Data Service, York; doi: 10.5284/1000021
Capelli, C., 2007, Appendice. Caratterizzazione mineralogico-petrografica della produzione ceramica di Henchir eck
Chekaf, in LRCW 2. Late Roman coarse wares, cooking wares and amphorae in the Mediterranean: archaeology
and archaeometry (eds. M. Bonifay and J.-Ch. Tréglia), 592–6, BAR International Series 1662 (II), Oxford.
Capelli, C., and Bonifay, M., 2007, Archéométrie et archéologie des céramiques africaines: une approche pluridisciplinaire, in
LRCW 2. Late Roman coarse wares, cooking wares and amphorae in the Mediterranean: archaeology and archaeometry
(eds. M. Bonifay and J.-Ch. Tréglia), 551–68, BAR International Series 1662 (II), Oxford.
Additional Supplementary material may be found in the online version of this paper on the pub-
lisher’s website:
Appendix S1a. Illustrations of the analysed amphora samples (from Aquilué and Burés 1999).
Appendix S1b. Illustrations of the analysed amphora samples (from Aquilué and Burés 1999).
Appendix S1c. Illustrations of the analysed amphora samples (from Aquilué and Burés 1999).
Appendix S2. Petrographic characteristics of the amphora samples analysed. PF: petrographic fab-
ric; Acc. comp.: accessory components; FF: fine fraction; CF: coarse fraction. Frequency of inclu-
sions (based on Whitbread 1995): xxx, predominant-dominant (>50%); xx, frequent-common
(15-50%); x, few-very few (2-5%); Tr. (trace), rare-very rare (<2%). Abbreviations for rock-
forming minerals (based on Kretz 1983): Bt, biotite; Cpx, clinopyroxene; Dol, dolomite; Ep, epi-
dote; Grt, garnet; Hbl, hornblende; Kfs, K-feldspar; Ky, kyanite; Ms, muscovite; Ol, olivine;
Opx, orthopyroxene; Pl, plagioclase; Qtz, quartz; Srp, serpentine; St, staurolite; Tur, tourmaline;
Zrn, zircon. Abbreviations for other inclusions: ARF, argillaceous rock fragments; Cal/Lms,
calcite/limestone; Cal.mf, calcareous microfossils; Fe.inc, ferruginous inclusions; MRF, meta-
morphic rock fragments; P.qtz, polycrystalline quartz; Qz.sds, quartz-sandstone; VRF, volcanic
rock fragments.
Appendix S3. Mineralogical composition and equivalent firing temperature (EFT) of the amphora
samples, determined from XRD analysis. PF: petrographic fabric. Abbreviations for minerals
(based on Kretz 1983): Qtz, quartz; Pl, plagioclase; Kfs, K-feldspar; Cal, calcite; Gh, gehlenite;
Px, pyroxene; Hem, hematite; Ill-Ms, illite-muscovite; Spl, spinel.
Appendix S4. Normalised chemical results of the amphora samples, determined by XRF.
Concentrations of major and minor oxides are in %, other minor and trace elements are in ppm.
Petrographic fabric (PF) is indicated for each sample.
Supplementary reference