Kowatli, Ibrahim, Curvers, Hans H., Stuart, Barbara, Yvette Sablesrolles, Henderson, Julian and...

31
A Pottery and Glass Production Site in Beirut (BEY 015) Ibrahim Kouwatli, Hans H. Curvers, Barbara Stuart, Yvette Sablerolles, Julian Henderson and Paul Reynolds Text submitted for Publication in BAAL 10 (2008): 103-130 Summary The concerted efforts of Paul Reynolds focusing on the pottery analysis of the Beirut assemblages (BEY 006, 015 and 045 and Julian Henderson and Yvette Sablerolles analyzing the glass finds of BEY 015, have encouraged the architectural and stratigraphic analysis of the site by its excavators. BEY 015 has provided clear evidence for primary production of glass in Beirut in the first half of the first century CE. Subsequently, between 50 and 150CE, among others the manufacture of the Beirut Amphora Reynolds type 3 and the so-called ‘carrot amphora’ took place in a kiln complex. Introduction In 1993 the excavations associated with the reconstruction of Beirut brought to light the first vestiges of its historical layers. More than a decade later, while excavations are still in progress, analysis of the results and in-depth research has yielded remarkable insights into the history of the site. Through the synergy of scholars specializing in the studies of pottery and glass, we are able to add yet another piece to the puzzle of Beirut’s history. The materials shedding new light on the manufacture of pottery and fabrication of glass were retrieved from BEY 015, situated at the modern Georges Haddad Avenue (formerly Georges Haddad Street). The excavated remains were in the right of way of the underpass designed to relieve the eastern part of the Beirut City Center from heavy traffic (Fig. 1). On February 27, 1995 Ibrahim Kouwatly started the excavations within the framework of the Unesco Leb 92/008 project. On the 1st of April, 1995 Camille Asmar, general director of Antiquities, committed the BCD Archeology Project (University of Amsterdam) to the supervision of the Infrastructure Project. Between April 1 and April 14, 1995 the southernmost part of the site became the focal point of a limited rescue excavation. Under the auspices of the Directorat Général des Antiquités (DGA) we defined the strategy of investigations in coordination with Philip Marquis (UNESCO).

Transcript of Kowatli, Ibrahim, Curvers, Hans H., Stuart, Barbara, Yvette Sablesrolles, Henderson, Julian and...

A Pottery and Glass Production Site in Beirut (BEY 015)

Ibrahim Kouwatli, Hans H. Curvers, Barbara Stuart,

Yvette Sablerolles, Julian Henderson and Paul Reynolds

Text submitted for Publication in BAAL 10 (2008): 103-130

Summary

The concerted efforts of Paul Reynolds focusing on the pottery analysis of the Beirut

assemblages (BEY 006, 015 and 045 and Julian Henderson and Yvette Sablerolles analyzing

the glass finds of BEY 015, have encouraged the architectural and stratigraphic analysis of

the site by its excavators. BEY 015 has provided clear evidence for primary production of

glass in Beirut in the first half of the first century CE. Subsequently, between 50 and 150CE,

among others the manufacture of the Beirut Amphora Reynolds type 3 and the so-called

‘carrot amphora’ took place in a kiln complex.

Introduction

In 1993 the excavations associated with the reconstruction of Beirut brought to light

the first vestiges of its historical layers. More than a decade later, while excavations are still

in progress, analysis of the results and in-depth research has yielded remarkable insights into

the history of the site. Through the synergy of scholars specializing in the studies of pottery

and glass, we are able to add yet another piece to the puzzle of Beirut’s history.

The materials shedding new light on the manufacture of pottery and fabrication of

glass were retrieved from BEY 015, situated at the modern Georges Haddad Avenue

(formerly Georges Haddad Street). The excavated remains were in the right of way of the

underpass designed to relieve the eastern part of the Beirut City Center from heavy traffic

(Fig. 1). On February 27, 1995 Ibrahim Kouwatly started the excavations within the

framework of the Unesco Leb 92/008 project. On the 1st of April, 1995 Camille Asmar,

general director of Antiquities, committed the BCD Archeology Project (University of

Amsterdam) to the supervision of the Infrastructure Project. Between April 1 and April 14,

1995 the southernmost part of the site became the focal point of a limited rescue excavation.

Under the auspices of the Directorat Général des Antiquités (DGA) we defined the strategy of

investigations in coordination with Philip Marquis (UNESCO).

Figure 1. Location of BEY 015 on the southern masterplan of the Beirut City Center.

The BCD Archeology Project excavation aimed at the continuation of the archaeological

work already undertaken on the site. Under the direction of Ibrahim Kouwatli the remains of

an industrial complex had been exposed, including several large pyro-technical features of

which some were circular, others rectangular. The rectangular features had been excavated

down to floor level and excavation of deeper deposits surrounding the floors was underway

(Fig. 2). Wasters suggested pottery manufacture and large chunks of melted glass suggested

glass production, Ibrahim Kouwatli reported the uniqueness and value of his discoveries in

reports to the DGA and Solidere (The Lebanese Company for the Development and

Reconstruction of Beirut Central District, s.a.l.). Realizing that the site was in the right of way

of important infrastructure works approved by the government, Hans Curvers and Barbara

Stuart focused their investigations on the southeastern part of the site. At the juncture

between the circular and the rectangular installations we hoped to obtain stratigraphic

information through systematic dismantling of the features already exposed (Fig. 3).

Figure 2. General view of BEY 015, looking SW.

When the details of the pottery sequences in Beirut and the manufacture of Beirut amphorae

were better understood, Paul Reynolds concentrated on the finds retrieved from BEY 015

(Reynolds 1997-98; 2005; Waksman et al. 2003). Reynolds typo-chronology of the Beirut

amphorae production puts the time range of the production of type 3 between 50 and 150 CE.

The presence of ceramic wasters and glass products on the site remained an enigma. With the

more precise date for the pottery established, the possibility of a sequence of ceramic

production and glass manufacture was considered. While working on this report with Paul

Reynolds, Julian Henderson examined the glass evidence, concentrating on fragments

previously left out of the typo-chronological report but recognized as products related to the

fabrication of glass (Fig. 4). Thus he was able to put together the bits and pieces concerning

the glass manufacturing process. The glass production was dated to a period before 50 CE

(Sablerolles 1995), a period for which Danièle Foy reported limited evidence for Roman glass

working from various sites in Beirut (Foy 2005).

Figure 3. Kiln Complex and Tank Furnace Complex 2 and 3, looking SW.

This report has been divided into the following sections: a description of the architecture and

stratigraphic analysis of glass tank furnace complex (Hans Curvers and Barbara Stuart), a

selective description of the typo-chronological aspects of the glass manufacture (Yvette

Sablerolles) and a first assessment of the technical details (Julian Henderson). Subsequently,

a brief summary of the pottery finds (Paul Reynolds) is followed by a description of the

architecture and stratigraphic analysis of the pottery kiln complex (Hans Curvers and Barbara

Stuart).

Figure 4. Raw highly colored glass fragments

The excavations

On April 1, 1995 the DGA handed over the supervision of site BEY 015 to the BCD

Infrastructure Project. In the previously established grid-system excavations continued down

to virgin soil. In the southern part of the site a smaller grid was laid out, allowing for detailed

investigations of the kilns. Sections were cut through the various cubicles and fire chambers

and at the same time the corner of a previously exposed glass covered surface and its robbed

wall were documented in detail.

The First Tank Furnace Complex (TFC 1, Figs. 5-6, 18)

The earliest phase of construction cannot be related unequivocally to the production of glass

or pottery. The orientation of the remains slightly differs from the later architecture, but

resembles the later complexes sufficiently to be described as the First Tank Furnace complex

(TFC 1). Wall remains of the northern fragment of the TFC 1 were found at ca. 10.65 m

above sea level, some 45 cm below the deepest floor (Tank 1) of the Second Tank furnace

complex (TFC 2).

Figure 5. Early wall/foundation separated from earliest tank floor (TFC1 or 2)

The southern fragment of the TFC 1 consists of a corner, delineating an area below the floor

of Tank 2 (TFC 2). A red brown soil deposit separates the lowest floor of Tank 2 and the top

of the wall (Fig. 5).An early wall assigned to TFC 1 at the same alignment as the later south

wall of Tank 2 (TFC 2) is also separated from this wall by a thin deposit of soil, pebbles and

broken stones.

Figure 6. Tank Furnace Complex 1 (small squares 2 x 2 m, larger squares 5 x 5 m).

The architecture of this phase is dated pre-50 CE. Its location outside the enclosure wall of

the Seleucid poleis suggests this complex was a workshop. The south wall of Tank 2 in the

TFC 2 and the south wall of the southern fragment of TFC 1 were rebuilt on almost the same

alignment, albeit with slightly different orientations. Their position suggests continuity from

the first phase into the second phase of the Tank Furnace Complexes. If we assume that the

southern and northern fragment of what we have coined the First Tank Furnace Complex

belong to the same building, the remains could represent the earliest phase of a complex for

the fabrication of glass in tank furnaces. However, the absence of plastered floors and

presence of soil deposits containing glass fragments clearly showing the impact of heat,

argues against this interpretation.

Another option is to attribute the wall to a pre-TFC phase and assume that the lowest crumbly

remains of a floor over soil showing the impact of heat do not belong to TFC 2 but instead

belong to the tank furnaces associated with TFC 1.

The Second Tank Furnace Complex (TFC 2, Figs. 7-10, 18)

A series of plastered surfaces characterizes the second building phase in BEY 015. A large

deposit of raw highly colored glass (large to small fragments) on these floors, layers of mortar

with melted fragments of glass between them and sandstones with melted glass attached to

them, suggested from the beginning that the residues at the site represent a phase in the long

cycle of glass production.

The plastered surfaces into which glass fragments still stuck were found at two elevations.

The upper plastered surfaces in Tank 4 were preserved in small patches at ca. 13.2 m above

sea level. In tanks 1- 4 the (lower) surfaces were situated at 12.60 m above sea level.

The walls associated with the surfaces were not found in situ. Occasionally one or two

sandstones seem to indicate the original walls which may have been rebuilt partially after

every melting process. Subsequently, the remaining stones may have been reused for the

construction of the later pottery kiln complex.

Figure 7. Tank Furnace Complex 2 (small squares 2 x 2 m, larger squares 5 x 5 m).

The plaster curving up at the edges of the robbed walls, however, provide clear indications for

a reconstruction of the tank furnace wall. In Tank 2 three surfaces were found associated with

plaster fragments still curving upwards.

One stretch of plaster was 14 m long and did not show any signs of a separation wall. The soil

belonging to this surface did not display the same impact from heat as did the soil below the

four other surfaces. We therefore consider this surface (14 x 2.2 m) to have been the floor of a

Service Corridor from which the tank furnaces were loaded and unloaded. The thick layer of

tiny glass fragments suggests that the breaking up of raw glass chunks took place in this area

and the tanks themselves.

Figure 8. Section through robbed wall between Tanks 1 and 2, the floor of the Service

Corridor is in the background with the floor of Tank 4 behind the corridor (TFC 2)

A fragment of a lower floor was found at elevation 12.28 m above sea level, whereas the

upper floor was situated at 12.55 m above sea level. The lower plastered floor of the Service

Corridor was situated at the same elevation as the lower Tank bottoms of Tanks 1, 2 and 4. In

the section cut through Tank 2 a lower layer with glass inclusions and red soil below it seems

to indicate an earlier presence of a tank furnace, to be related to the earliest phase of the TFC

2. The presence of a penultimate line of red brown soil deposited over the top of he wall

excludes assignment to TFC1. While the floor may be assigned to TFC1, the wall probably

belongs to a pre-TFC1 phase.

Figure 9. Section (N-S) through Tank 2 (TFC 2) with lower floor remains and still

unexposed earlier wall (Fig. 8) . Top right, TFC 2 remains: two Service Corridor floors and

the south wall of Tank 2

Tanks 1, 2 and 4 revealed a series of burnt layers (ca. 45 cm thick) below the highest

plastered surface of the TFC 2. We did not observe any fill between the plastered surfaces,

neither in the vertical, curving sections nor in the horizontal layers. In Tank 2 at least three

layers were observed in the upward curving plaster, suggesting the tank was relined after

every melt or more likely repaired and occasionally renewed.

A crumbly white layer over an ashy deposit represents the bottom of a tank possibly

belonging to the earliest architectural phase. This tank was built over or into red soil, most

likely the remains of another sequence of use. A deposit of red soil (ca. 30 cm) separated this

floor from the wall attributed to the TFC 1 (see above).

To establish the size of the tank furnaces we assumed symmetry in the design of the Tank

Furnace Complex and extrapolated the information on the layout collected in the preserved

fragments.

The SE corner of Tank 4 was exposed in the 2 x 2 grid excavations. The remains of the

(robbed) walls of this corner define the eastern limit of Tanks 4 and 3, as well as the southern

limit of Tanks 4 and 2. Furthermore, the robbed walls exposed at the edges of the plastered

floor surfaces allow for a reconstruction of the original width of tank 4, being ca. 4.8 m. The

length of Tanks 1 and 2 can be deduced from the traces of the northern wall of Tank 3, and

the northern edge of Tank 1 and the zone of the robbed partition wall between Tanks 1 and 2.

Tank 4 measures 6.6 m, whereas Tank 1 measures 6.2 m. The presence of the plastered

surface in the area of Tank 3 and its northern edge allows a reconstruction of its length -

similar to Tank- 1 of 6.2 m.

The height of the tanks remains unknown. The tanks excavated in Tyre, dated to the 11-12th

century, have a preserved height of .8 m (Aldsworth et al. 2002:62). From the evidence in

BEY 015 we were unable to reconstruct the original heights of the furnaces.

Figure 10. Schematic plan of Tank Furnace Complex 2 (small squares 2 x 2 m, larger

squares 5 x 5 m).

The surfaces of the tanks were rather clean, in contrast to the long corridor (14 x 2.2 m): large

deposits of raw glass had been heaped on its floor. We assume that this area was both the

loading and unloading area for all of the four furnaces before and after each melt. Aldsworth

et al. (2002:62) also observed that the floors of the tank furnaces at Tyre were clean.

In the architecture north of Tanks 1 and 3 and the Service Corridor one would like to

recognize the fire chambers. The walls assigned to TFC 1 in this area appeared at an elevation

below 10.60 m above sea level. The stones of these walls did not display any signs of heat

impact. Although we scrutinized the field notes for indications of the presence of a powerful

heat source, we did not find any evidence for a fire chamber in the areas north of Tanks 1 and

3. However, the chamber to the north of furnace 1 could well have acted as a firing chamber;

it is at a lower level than the floor of the furnace as is the case of the tank furnaces found at

Tyre, where the evidence survived (Aldsworth et al 2002, 55-59, Fig. 11). So whilst this is not

conclusive, it does seem to be a likely interpretation. Firing chambers for furnaces 2 and 4

would have been located at the opposite (southern) ends of those furnaces.

Similarly, we tried to locate fire chambers for Tanks 2 and 4. After overlaying the

archeological plan with the modern city plan of Beirut we concluded that the area south of

Tank 2 had been destroyed during the construction of a 19th

or 20th

century building. The area

south of Tank 4 had been re-used for the construction of Kiln 1 of the Kiln Complex. The

fire-chamber of Kiln 1 had been dug into partially constructed atop the surface of Tank 4.

The Third Tank Furnace Complex (TFC 3, Figs. 11-12, 18)

The plastered floor fragments over the Tank 4 floor of TFC 2 have been attributed to the

Third Tank Furnace complex (TFC 3). Observations of the section reveal that the soil below

the plastered surface both in Tank 3 and 4 does not display the same heat impact as the lower

surfaces belonging to tank 4 of the TFC 2. We assume that this plastered floor surface

belonged to a Service Corridor. We were able to trace this floor surface, referred to as `glass

pavement’ in the 1995 notes and reports, over a length of 10 m. The western wall of this

corridor had been provided with a plaster coating against two of its wall faces.

Figure 11. Tank Furnace Complex 3 (small squares 2 x 2 m, larger squares 5 x 5 m).

Figure 12. Schematic plan of Tank Furnace Complex 3 (small squares 2 x 2 m, larger

squares 5 x 5 m).

A horizontally placed ceramic tile indicating the bottom of the stoke hole in the northeastern

wall of Kiln 5 (see below) covered the remains of a foundation onto which the northern wall

of the cubicle had been constructed. We attribute this wall to the construction phase of TFC 3.

Remains of a robber trench were found at the same level as the floor of the Service Corridor.

The northern wall of this area was heavily truncated by the south walls of kilns 3 and 4. The

latter indicate that the structure was constructed and used before the Kiln Complex was built

atop the remains of the Tank Furnace Complexes 2 and 3.

Glass Finds (Pl. 1)

There is no doubt that the Syro-Palestinian coast played an important role in the production of

glass vessels in the later Hellenistic and early Roman period (for a summary, see Jennings

2006:28). The first indications that Beirut was involved in the production of glass vessels

during this period has come from site BEY 002 which yielded waste from an unidentified

glass workshop believed to have produced linear-cut and ribbed bowls (Foy 2005).

Until now it has been assumed that glass workshops in Beirut dating to the late

Hellenistic/early Roman as well as Islamic periods, were secondary glass workshops only,

depending on the import of raw glass for the production of vessels and other objects (Foy

2005:33). BEY 015 has yielded some firm evidence to the contrary, since the use of tank

furnaces implies that primary raw materials were used to fuse the glass.

Another assumption was that the glass industries on the Levantine coast did not follow the

fashion for strongly coloured glass vessels common in Italy in the first half of the 1st century

AD, with the exception of deep yellow (amber) glass for linear-cut cast bowls (Jennings

200:57). However, the range in glass colours retrieved from BEY 015 seems to indicate

otherwise. This site yielded huge amounts of raw glass, either still attached to fragments of

furnace floor or wall, or in the form of small, irregular lumps. These lumps occasionally

displayed concoidal fractures, indicating they had been struck off larger fragments (cf. Foy

2005:12, fig. 1). The quantification of the raw glass is pending.

Glass production waste from BEY 015 includes drawn rods and threads as well as pincered

glass fragments. It is interesting to note that the site yielded glass worked as flint, as shown by

the presence of at least one glass blade, a sickle blade, a blade-core and many fragments (pers.

comm. Mohammed Djaradat, Univ. of Yarmuk, Jordan). A similar practice was noticed on a

glass-making site in Raqqa (Syria) where glass was being made and worked in the late 8th

and early 9th c. AD.

The glass fragments found attached to furnace fragments are all translucent. Their colours are:

near colourless with a greenish tinge (and purple streaks), pale (blue-) green, pale yellowish

green, yellow-green (olive), pale-, medium- and deep blue, deep translucent purple, yellow-

brown (honey-colour) and orange brown (amber).

The strong colours suggest that linear-cut bowls and contemporary bowls with irregular ribs

were amongst the products of this glass workshop (cf. Foy 2005, figs. 5-7; Jennings 2006:37-

45). Grose (1989:247) dates linear-cut bowls between the late 1st century BC and the

beginning of the 1st century AD. Although the evidence from Beirut concurs with these dates,

Jennings (2006:44) believes that the production of linear-cut bowls may have continued until

the mid 1st century AD. She observed that fragments of linear-cut bowls were often found

with fragments of zarte Rippenschalen, as is indeed the case on the furnace site BEY 015 (see

below). These delicately ribbed free-blown bowls, when strongly coloured with opaque white

decoration, are thought to be imports from the west where they date from the late Augustan

period to the mid 1st century. Jennings (2006:44) also found that locally made plain zarte

Rippenschalen of a ‘natural’ greenish colour continued to be made into the third quarter of the

1st century AD, by which time they are no longer associated with linear-cut bowls, indicating

these bowls were no longer being made after the mid 1st century.

BEY 015 yielded two fragments of vessels, probably Italian imports, dating to the first half of

the 1st century AD. These are a neck fragment of a small, free-blown flask of translucent pale

blue glass with an opaque white festoon (Pl. 1:2) , and a body fragment of a free-blown bowl

with delicate ribs (Isings form 17), or zarte Rippenschale, of translucent medium blue glass

with an opaque white spiral trail (Pl. 1:1). The latter was found mixed in with the ‘glass

pavement’ of the corridor of the latest Tank Furnace Complex (TFC 3). This context suggests

that all phases of the Tank Furnace Complex have to be dated to a period before 50 CE.

A fragment of a twisted rod of translucent purple glass with four twists of an opaque white

trail (Pl. 1:3) can be dated to the 1st c. AD (cf. Grose 1989:370 fig. 670).

A rim fragment of an earlier date, a conical grooved bowl of near colourless glass with a

greenish tinge (Catalog of Glass Finds No. 4) belonging to Grose’s Group A (c. 150-50 BC)

(Grose 1979, 58) was found 40 cm below the floor of the TFC 3 corridor. This find, in a red

brown sand deposit, provides a terminus post quem for the TFC 3 remains. Unfortunately, its

stratigraphic relation to the preceding phases (TFC 1 and 2) cannot be established.

On the basis of the glass colours found on furnace fragments, the inferred production of linear

cut and ribbed bowls, as well as the vessel fragments, the use of the tank furnace complexes

can be firmly placed in the first half of the 1st century AD.

Catalog of glass finds

Pl. 1:1 002 (11-4-1995)

Fragment of a rim with shoulder and upper part of body with 2 pinched ribs; decorated with a marvered

opaque white (rightly turned) spiral trail, starting at the top of the rim, extending over the shoulder and

upper body, hardly visible between the ribs. An attempt of a 1st turn to the right at the top of the rim ends

in a big blob of white glass.

Technique blown, pinched ribs

Colour/quality translucent medium blue; opaque white spiral trail

Preservation dull, roughly weathered surface, esp. the opaque white glass is weathered, in some locations only the

'negative' (impression of the thread) is still visible

Type/date A zarte Rippenschale Isings Form 17, dating to the 1st half of the 1st c. AD (pre-Flavian).

Height 4.1 cm

Thickness Body 0.19-0.24 cm

Thickness Rim 0.2 cm

Diameter rim 9 cm

Pl. 1:2 004003-001

Small fragment of a neck, decorated with a marvered opaque white festoon

Technique blown

Colour/quality translucent light blue

Preservation dully weathered; pitted interior surface

Type/date Most likely an unguent flask dating to the 1st half of the 1st c. AD. Oliver 1980:49, cat.nr. 31

Height 1.7 cm

Thickness body 0.16-0.2 cm

Diameter neck 1.5 – 2 cm

Pl. 1:3 018007-009

Small straight section of a spirally twisted thick purple rod with 4 irregular turns/twists of an opaque

white cable; damaged

Technique twisted and possibly drawn

Colour/quality faintly translucent purple

Preservation medium iridescence all over

Type/date Probably a stirring-rod Isings Form 79, dating to the 1st-mid 3rd c. AD. (also interpreted as "kohl stick")

Meyer 1988:186, Fig. 5,N: "stirring rod, turquoise w/dk stripes, Roman"

Length 2.45 cm

Diameter rod c. 1 cm

Not ill. 004003-004

Fragment of a rim with 0.8 cm. below the rim 2 shallow, concave grooves

Technique mould-formed, interior wheel-polished, rim ground and wheel-polished, top 0.5 cm. of exterior wheel-

polished

Colour/quality colourless with a greyish tinge; some small bubbles

Preservation covered with a milky white weathering layer

Type/date Linear-cut bowl, dating to the 1st half of the 1st c. AD.

Height 5 cm

Thickness body 0.21-0.45 cm

Thickness rim 0.55 cm

Plate 1. BEY 015: Selected glass finds.

The technology of glass manufacture

The Levant has a very strong tradition of innovation in glass technology. This included glass

making and glass working- and even glass blowing. The discovery of these tank furnace

complexes appears to be the earliest known evidence for primary glass making in the ancient

Classical world. The furnaces find parallels in other tank furnaces in the Middle East, but all

of these are later. They include the discovery of a tank furnace at 4th

century Jalame

(Weinburg 1988), the bases of seventeen 6th

to 8th

century tank furnaces at Bet Eli’ezer

(Hadera) (Gorin-Rosen 1995 and 2000), the remains of 4 tank furnaces at Tyre, one of which

(furnace 1) has a tank wall surviving to 0.8M in height of a probable 11th-12th

century

(Aldsworth et al. 2002) and some 10 metric tons of destroyed tank furnaces at Raqqa dating

to the 8th

-9th

, 11th

and 12th

centuries (Henderson 1999, Henderson 2000, 77-83).

In the absence of primary evidence for glass production, it has been suggested that the Jalame

glass furnace was only used for secondary glass making (reheating raw glass). The Hadera

furnaces, and their proposed reconstruction, provide an idea of how tank furnaces for the

production of raw chunks of glass operated, and this is supported, with some structural

variations, by the better-preserved furnaces at Tyre. It is however worth noting that the glass

manufactured at Hadera was a (mineral-based) natron glass where as that made at Tyre was a

plant ash glass. Tank furnaces apparently had two main chambers. The evidence from Hadera

and Tyre shows that at one end there was a firing chamber divided into two longitudinally, in

which the fuel was placed. This chamber was separated from the main melting chamber by a

cross wall with an opening in it. Loading platforms were found at the opposite end from the

firing chambers of furnaces 1 and 2 at Tyre. Presumably at the opposite end of the main firing

chamber from the stoke hole there would have been a chimney through which the draught

would have been drawn, using the same principles as a domestic chimney. Whilst there is

archaeological evidence for the two main chambers in the furnaces excavated at Hadera, and

at Tyre (furnace 2), there is no archaeological evidence for the chimney. However, tank

furnaces could not have functioned without at least one chimney. There is in situ evidence

from Tyre that furnace 1 where traces of three re-linings were found (Aldsworth et al 2002,

62) indicating that the furnaces was used four times. Fragments of furnace base with raw

glass attached from Raqqa also show evidence of re-use.

The two most complete furnaces at Tyre measured 6.4 x 3.9 and 4.5m x 2.2m and those at

Hadera 4m x 2m. The excavators estimated that each furnace at Hadera could accommodate

eight tons of raw glass and at Tyre a minimum of 37 tons of raw glass in furnace 1. The

Beirut furnaces have the following dimension: tank 1 is 6.2 m long, tank 3 is 6.2 m long, tank

4: is 6.6 x 4.8 m. The furnaces therefore have comparable dimensions to those found at Tyre.

Preliminary chemical (electron microprobe) and isotopic (thermal ion mass spectrometry)

analyses of some raw glass samples from the furnaces have revealed several things. Firstly,

the highly coloured glass is of the anticipated soda-lime-silica composition, with low

magnesia and potassium oxide impurities and relatively high calcium oxide levels of c. 9%.

This indicates that the early Roman glass from Beirut was made from a mineral alkali source

(natron) and calcareous sand. Calcareous sands occur on the Levantine coast and were used to

make later Roman glass, such as at Jalame (Brill 1988). So obviously there is continuity in

glass production using a similar recipe as reflected in chemical compositions in the Levant for

at least 800 years. However, even more diagnostically, the determination of strontium

isotopes in three raw glass samples indicate that they fall within the field of isotopic

signatures for other natron glasses made in later glass factories on the Levantine coast. This is

even more conclusive evidence that a local silica source was used to make the glass. Further

detailed chemical and isotopic analyses of the raw glasses are planned in order to establish

whether there was any variations in the raw materials used and their sources. Since the glasses

are highly colored, a detailed investigation of the colorants will also be carried out.

It is debatable as to whether the raw materials would have been fritted before the batch was

loaded into the tank furnace for firing. It is only rarely that frit survives to be excavated,

particularly when it has been overheated (Henderson 2000:88-89). Had the raw materials

been fritted, a degree of purification would have occurred. The furnaces would have been

operated by filling them fully with the glass batch. Each time this occurred part of one of the

furnace walls would have been taken down. Evidently the bases of walls survived, revealing

the evidence of re-use. The most expensive raw material would have been the fuel and, in the

absence of environmental evidence, one can only speculate about what fuel types were used.

The furnaces would have been fired at least over night and probably for much longer at

temperatures of at least 1200C. In the process of firing the batch would contract leaving a

vitreous trail on the inside of the upper part of the furnace wall. Once the batch was fully

melted it would have attached itself as a large block to the floor of the furnace. Such a block

of glass measuring 3.4x1.95 was found at Bet She’arim (Brill and Wosinski 1965; Freestone

and Gorin-Rosen 1999). The glass would have been removed by breaking it with an

instrument such as a pick-axe. At Raqqa many fragments of glass with conchoidal fractures

were found which resulted directly form this percussive process. The raw glass would either

have been exported to secondary production centers where it would have been remelted and

blown or alternatively remelted and worked at nearby secondary production centers.

The pottery finds (Pls. 2-3)

The present stratigraphic analysis, supported by observations on the glass and ceramics of the

site, now help us to separate the pottery production (amphorae and cooking wares of the

second half of 1st century to early 2

nd century AD) from the glass production on the same site

of the second half of the 1st century BC to the first half of the 1

st century AD.

Plate 2. Beirut Amphora Typology: from late 2

nd century/100 BCE to 7

th century CE

A full report on the general range and date of the pottery associated with the BEY 015 kiln

site is being prepared (Reynolds et al. forthcoming). In short, as the chemical analyses prove

(Waksman et al. 2003; Roumié at al. 2004; Lemaître et al. 2005), the output of the workshop

comprised at least three, possibly four amphora forms, as well a range of cooking pots and

other kitchen wares.

The amphorae (Pls. 2-3)

1. The ‘Beirut amphora’. A typology for this form was published some years ago (Reynolds

1999, Appendix; 2000a; see also Reynolds 2005). Probably for wine, it was the most long-

lived of the amphorae produced in Beirut and other sites in its territory (Khalde and perhaps

Jiyé). A summary of its development, from the late 2nd

BC/100 BC to its final phase in the

first half/mid 7th

century is offered here (Pl. 2). ‘Ala’Eddine (2005) has recently provided

evidence for the earliest variants, pushing back the initial date somewhat, into the late

Hellenistic period, as well as providing drawings of complete examples of other stages in the

development of the type (some incorporated into Pl. 2). Quite exceptional for the Roman

Levant, the amphora was stamped, and in Latin (with reference to the Roman colonia

Berytus), but only during the early-mid 1st century AD (Pl. 2e and Pl. 3a).

BEY 015 confirms the mid 1st and early 2

nd century production of the amphora in Beirut on

this site (Pl. 3a-b), from both wasters (of late 1st-early 2

nd century variants of Reynolds

[2000] Beirut 2 and 3) and the chemical analysis of the clays. Earlier production of the mid 1st

century version close by, as well as contemporary kitchen wares and thin-walled wares, is

now attested by excavations on site GEM 003 (I thank Abdullah ‘Ala’Eddine and Fady

Beayno for showing me this material).

2. The ‘carrot amphora’/Schöne-Mau XV. The type, a small narrow-bodied amphora of

Phoenician tradition (with ring handles on the shoulder) is well known from exports (to Gaul,

the Rhine provinces and Britain) (Vipard 1995; Carreras Monfort and Williams 2002;

Reynolds 2005: 571) (Pl. 3c-e). Long suspected to be Levantine, production can now be tied

down unequivocally to Beirut. It would have carried locally-produced dates. There is some

variation in the clays, some examples being particularly lime-rich and thick-walled. Analyses

on finds of the carrot amphora in Gaul prove that some of them derive from the same Beirut

workshop (Lemaître et al. 2005).

3. Reynolds ‘AM 72’. The amphora type was defined by Reynolds (1999: ‘AM 72’; 2003;

2005, 568, Plate 9, Figs 59-60) (e.g. Pl. 3g). The form was produced in Beirut (BEY 015), but

has also been found in a dense, lime-rich fabric that indicates that the type was manufactured

also in northern Lebanon (Reynolds 1997-1998, BEY 006.11629.7, in fabric FAM 43). AM

72 dates into the early 3rd

century, but is absent in mid 3rd

century contexts. A large amphora,

with a wide shoulder, thicker-walled than the Beirut amphora, the deeply-scored handles and

a rim with a deep concave lid seat are distinctive features. The base ended with a long

cylindrical hollow toe. I have suggested, on the basis of the toe (Pl. 3i), that the form is a fish

sauce amphora.

This is the only type of amphora of this class I have encountered in Beirut deposits. However,

here on BEY 015, there are other variants that, as those illustrated, have equally the grooved

handle of AM 72, but lack the concave rim face. Pl. 3h, recalls another fish sauce

Plate 3. BEY 015 kiln complex products a) Beirut 2, b) Beirut 3a, c-e)

‘carrot’ amphorae, f-g) AM 72 variants, i) AM 72 base, j) 1st century cooking pot, k) Early 2

nd

century cooking pot, k) Early 2nd

century cooking pot, l) Carinated bowl, m) Lid, n-o) Stands

and p) Drain.

amphora, Dressel 7/11, as Baetican examples, whereas Pl. 3f seems to be a Dr 2-4 (wine

amphora). In fact, Pl. 3g could equally be a Dr 2-4.

4. Several examples of a short, hollow cone base, as that of the form Kingsholm 117

(Reynolds 2005, Fig. 136) and the Chhîm amphora (Reynolds 2004; 2005, Fig. 92) suggest

that there was another form produced on BEY 015.

Coarse and kitchen wares (Pl. 3)

The analyses are important confirmation of the Beirut production of certain kitchen wares

defined in previous work (Reynolds 1999). Pl. 3h) cooking pots with a tall collar rim (mid to

late 1st century AD), Pl. 3k) its successor (with a short, everted collar rim)

1, Pl. 3l) a

carinated bowl (a Hellenistic shape that was equally produced in Augustan Carthage), Pl. 3m)

a lid, and Pl. 3p) a drain. There were no thin-walled vessels (mugs-cups), jugs or flagons

found on BEY 015, as far as I know.

A large number of pot stands in the same fabric as the sampled material (i.e. produced on the

site) must have served as stands on which the potters rested vessels, the amphorae probably,

during the stages of manufacture (Pl. 3n-o).

Conclusions

The range of amphorae and their variants, including those of the Beirut amphora, is far wider

on BEY 015 than that of sites in the city itself. The definition of ‘AM 72’ is here not so

straightforward: whether the vessels Fig. 14f and h represent distinct forms is not yet clear:

they appear not to have been for the Beirut market or were short-lived (i.e. Pl. 3f and h). That

the carrot amphora is scarcely found elsewhere in Beirut, but occurs almost exclusively as an

export is a point of interest. Reynolds (et al. forthcoming) will present a fuller,

comprehensive illustration and discussion of all the amphora variants than is possible here.

The Kiln Complex (Fig. 13-18)

The exposed remains assigned to the Kiln Complex represent one main period. The complex

includes five cubicles, each with its own sequence of fire-chambers. The kilns had been

constructed into the ruins of the Tank Furnace Complexes, more specifically into the remains

of Tank 4 (TFC 2) and TFC 3. The fire chambers of four other kilns were found in four

adjacent cubicles, separated from the isolated kiln by a narrow space, previously the Service

Corridor of TFC 3.

The eastern part of the Kiln Complex consists of four cubicles, each containing a sequence of

fire chambers. The top of this part of the kiln complex had been exposed under the direction

of Ibrahim Kouwatli. Sections were cut to obtain information on the construction sequence of

three rooms and their fire chambers. At first sight they seemed to be similar in size and shape,

but in section different techniques of construction became visible.

As the stoke holes of all fire-chambers were oriented towards the northwestern wall, it is clear

1 For comment on the typology of early Roman cooking pots in Beirut, see Reynolds (1997-1998) and Reynolds

and Waksman (2007).

that the cubicles and their fire-chambers could not have been used at the same time. The

position of these features suggests that the cubicles in the southern part were built before the

cubicles in the northern half.

Kiln 1 had been constructed within and into the remains of TFC 3 and 2. The stoke hole left

its imprints in the floor of Tank 4 (TFC 2). The stones south and east of the packing

supporting the remaining circular fire chamber were partly exposed. The east and south wall

of the cubicle consisted of the old TFC 3 western Service Corridor wall. This wall still had its

wall plaster preserved on the eastern wall face. The south wall and southeast corner seem to

have been rebuilt for the construction of Kiln 1 or they were robbed including the plaster on

the wall face(s) after the Kiln Complex had been abandoned.

Remains of the west and north wall of Kiln 1 were not found. Since the dents of the back-hoe

are visible in the plastered floor of Tank 2, two possibilities remain; either the walls were

robbed, as the south wall, or truncated during the 1995 excavations for the construction of the

George Haddad Underpass.

A wall south of the cubicle with the fire-chamber of Kiln 1 abuts the western wall of the TFC

3 Service Corridor. This stratigraphic relation hints at two possibilities. Either the wall

represents an addition to the TFC 3, or the wall is related to kiln 1. No floors were exposed

related to this wall, suggesting the wall we exposed is a foundation. Another argument to

consider the wall or foundation to be late and belonging to the Kiln Complex phase is the fact

that west of the Service Corridor one would expect the presence of tank furnaces. However,

no remains of these tank furnaces were observed. If these assumptions are right the

foundation represents part of the superstructure of Kiln 1. The heat produced in the fire-

chamber would find its way into an upper chamber of the kiln built south of the fire-chamber.

The scant remains of a wall south of Kiln 2 and cutting into Kiln 3 represents a similar

method of construction (see below).

Figure 13. Plan of Kiln Complex (small squares 2 x 2, larger squares 5 x 5 m).

Kiln 2 has the largest fire-chamber in the Kiln Complex. Its diameter is over 3.5 m and the

stoke hole gave access to the chamber from the northeast. The walls of the cubicle consist of

rectangular blocks. The cubicle is almost square with walls of 4.8 m length.

The SW corner of the cubicle was included in our grid system of 2 x 2 m squares and

excavated. The space in this corner had been filled with compact brown soil and large

quantities of small to medium-sized stones (average size 20 cm). In addition to the wall of the

fire-chamber already exposed we found a second wall.

The circular fire chamber has its opening in the northeastern wall. The position of the cubicle

walls indicates that the use of kiln 2 could not have been contemporary with the use of Kiln 3.

The southwestern wall of Kiln 2 blocked the stoke hole in the northeastern wall of Kiln 3.

Close observation of the photographs revealed the presence of a central pillar in the latest fire

chamber of Kiln 2. The excavators under the direction of Ibrahim Kouwatli observed that the

whole cubicle and part of the cubicle of kiln 3 were covered by a deposit of red burnt soil.

This deposit could be the remainder of the upper structure of kiln 2, consisting of decayed

sintered clay bricks.

Kiln 3 was provided with a fire-chamber of 2.8 m in diameter. The fire chamber was

constructed in a cubicle of 4.2 by 4.2 m. The walls of the cubicle containing the fire-chamber

of Kiln 3 consist of a combination of rectangular blocks and rough, undressed stones.

The vertical wall of the circular fire-chamber consists of red, very compact clay, which had

been enforced with small stones and circular hypocaust tiles. The space between the fire

chamber and the outer wall of the cubicle had been filled with small to medium-sized stones

and soil.

Figure 14. East section of Kiln 3

The chamber had a central, plastered column, erected on a compact floor (Fig. 17). Atop the

fire chamber we observed a wall and associated lime plaster floors, separated by ash layers

from another phase. They could be related to the use and construction of Kiln 2. This

construction would represent the upper chamber into which the heat of the fire-chamber

would have been drafted. A situation we tentatively proposed in Kiln 1 (see above).

The floor of the fire chamber sloped down towards the opening in the northeast wall. We did

not observe a bond between the walls of Kilns 2 and 3, suggesting the two kilns were not

built at the same time. The fire chamber had been filled with reddish-brown debris containing

numerous small stones, sherds and fragments of roof tiles. This backfill, not observed in the

other fire-chambers of the Kiln Complex, is possibly related to the construction of the draft

chamber of the kiln.

One of the white floors is clearly associated with the E-W wall built atop the backfilled fire-

chamber of Kiln 3. Its relation to the wall(s) separating Kilns 2 and 3 has not been

investigated thoroughly, which remains a weak point in this interpretation. The wall and the

white floors could represent the remains of the draught chamber. The draught would enter the

kiln through a grill supported by the central column in fire chamber 2 and enter into the

southern part of the kiln.

Figure 15. Schematic plan of Kiln Complex (small squares 2 x 2 m, larger squares 5 x 5

m).

Another interpretation of the white plaster floors separated by ash layers could be that the

ruins of kiln 3 were used for the loading and unloading of kiln 2. If we assume that the

superstructure of the kilns had to be repaired after each firing, the white plaster floors

separated by ash layers might be the imprints associated with these repairs. The wall in this

interpretation would have been a mere retaining wall or wall separating the area around Kiln

2 from its wider surroundings, including more kilns to the south observed in the photographs

of the eastern section of the trench excavated mechanically for the Georges Haddad

Underpass.

Finally, we have to mention the interpretation of the wall and the floors as part of a feature

later than the kiln complex. The relation between the white floors and the wall(s) separating

cubicle 2 and 3 would have been crucial for a more comprehensive understanding. Now we

can only assume that the presence of the wall(s) suggests a relation of the wall and white

floors with Kiln 3 or Kiln 2.

Kiln 4 has a fire chamber with dimensions similar to the fire chamber of Kiln 3. The

remains of Kiln 4 consist of a round fire chamber wall placed in a square space. The space

between the circular fire chamber and the walls is filled in with medium-sized stones and soil.

The floor of the oven slopes down from east to west. The central column is conical in shape

(Fig. 19). The oven wall itself seems to have consisted largely of solid, compact red clay,

without much inclusions or additions.

The fire chamber sequence had two phases. The lower phase was preserved to a height of 40-

50 cm; the fire chamber build atop the remains of the earlier chamber had been preserved to a

similar height. The surface of the later chamber was of considerable thickness (15 cm). The

lower fire chamber of Kiln 4 had been filled with a homogeneous red-brown soil, whereas the

fill observed in the superimposed chamber consisted of a loose, light-colored ashy layer.

Since no opening was found related to the phase shown on the plan, it may well have been to

the north. If so, Kiln 5 represents a later addition to the Kiln Complex, due north of Kiln 4.

Kiln 5 had been cut by a recent foundation taking away part of the central column of

the fire chamber (Fig. 20). The vertical wall of the circular fire chamber consists of solid, red

and hard-baked clay. The wall may actually include more than a single phase, as two separate

walls were visible on the excavated surface. In contrast to the fire chambers of the other kilns,

the space between the vertical wall and the surrounding walls of the cubicle had been filled

with compact, red debris containing just a few stones.

Figure 16. North section of Kiln 4.

Figure 17. East section of Kiln 5

The stoke hole of the kiln had been in the northern wall. A horizontally placed ceramic tile

indicated the bottom of the stoke hole in the northeastern wall of cubicle 5. The tile also

covered the remains of a lower wall or foundation, which we attributed to the construction of

TFC 3. Kiln 5 may have been built subsequent to kiln 4 as two separated walls were visible

on the excavated surface.

The area adjacent to the building, directly east of the kiln complex, seems to have been empty

of constructions. A thick brown layer, including occasional thin layers of black ash or

yellow-brown sand was observed. The deposit contained numerous small sherds and stones,

and --in contrast to the Service Corridor of the TFC 3-- no glass fragments.

ADD FIG 18

Figure 18. Locations of photographic viewpoints and sections

The remains of Kilns 1-5 suggest that the Kiln Complex had been in use for several decades.

Every Kiln seems to have known a period of use in which the fire chamber had been rebuilt at

least twice. The Kilns in the northern half of the Kiln Complex had been integrated into the

remains of previous structures, either TFC 3 or Kilns 3 and 4. This interpretation allows us to

reconstruct a kiln with a lower fire chamber and a draught chamber. The central pillars in the

fire chambers supported a grill through which the heat was drafted into the upper chamber(s).

The TFC 3 wall to the southern limit of the excavation could have been used to create a

second draft chamber, but substantial evidence for this does not exist.

Conclusions

The increased insights into and knowledge of glass production and pottery manufacture in

and beyond Beirut since the excavation of BEY 015 in 1995 were completed, has helped

considerably to solve some stratigraphic enigmas. BEY 015 is situated on the eastern side of

the Seleucid poleis and the Roman colonia Julia Augusta Felix. The tank furnace complexes

(TFC 1-3, dated before 50 CE) were located outside the enclosure wall, the Kiln Complex

was situated at the outskirts of the settlement, being a buffer zone between settlement and

cemeteries. The site belongs to an industrial zone in which through the centuries a variety of

workshops for glass, pottery and metal production were located.

The excavators of the site are fully aware of the guidelines suggested for investigations of

kiln sites: A full catalogue of all the forms recovered from the site, a statistical breakdown of

their relative frequency and a full description of the fabric. It is also necessary to distinguish

intrusive material, which may have been used but not made on the site, and here fabric is

again crucial. Unfortunately, there are very few kiln reports which meet all these criteria in

full, or even in part, and this places a severe limitation on the use of kilns as a source of

evidence, although any information is better than none. (Peacock and William 1986:9; cf.

‘Ala’Eddine 2005:189-190). Off course we are to blame for the omissions and incorrect

interpretations and we can only hope that our first joint effort falls in the category ‘any

information is better than none.’

Bibliography

‘Ala’Eddine, A. (2005). The Development of Beirut Amphorae, A General Approach, in :

Morhange, Christophe and Saghieh-Beydoun, Muntaha, (Eds.) La Mobilité des Paysages

Portuaires Antiques du Liban. BAAL, Hors-Série II. Pp. 185-206.

Aldsworth, F., Haggerty, G., Jennings, S & Whitehouse, D. (2002). Medieval

Glassmaking at Tyre, Lebanon, Journal of Glass Studies 44: 49-66.

Brill, R.H. (1988) Scientific Investigations of the Jalame glass and related finds in:

Weinburg, G.D. Ed. (1988) 257-293.

Brill, R.H. and Wosinski, J.F. (1965). A huge slab of glass in the ancient necropolis of Beth

She’arim, in: Comptes rendu, VIIe Congrès International du Verre, Bruxelles, 28 juin- 3

juillet 1965, Section B, paper number 219, 219.1-219.11.

Foy, D. (2005). Une production de bols moulés à Beyrouth à la fin de l’époque hellénistique

et le commerce de ces verres en Méditerranée occidentale, Journal of Glass Studies 47 :11-

35.

Freestone, I. C. and Gorin-Rosen, Y. (1999). The great glass slab at Bet She’arim, Israel: an

Early Islamic Glassmaking Experiment?, Journal of Glass Studies 41:105-116.

Grose, D.F. (1979). The Syro-Palestinian Glass Industry in the Later Hellenistic Period, Muse

13: 54-67.

Gorin-Rosen, Y. (1995). Hadera Bet Eli’ezer, Excavations and Surveys in Israel 13: 42-43.

Gorin-Rosen, Y. (2000). The ancient glass industry in Israel: summary of the finds and new

discoveries, in M.-D. Nenna (Ed.) La Route du Verre, Lyon: Maison de l’Orient

Méditerranéen - Jean Pouilloux, 2000, pp. 49-64

Grose, D.F. (1989). The Toledo Museum of Art: Early Ancient Glass. Core-Formed, Rod-

Formed and Cast Vessels and Objects from the Late Bronze Age to the Early Roman Empire,

1600 BC to AD 50, New York.

Henderson, J. (1999). Archaeological and Scientific Evidence for the Production of Early

Islamic Glass in al-Raqqa, Syria, Levant 31:225-240.

Henderson, J. (2000). The Science and Archaeology of Materials, Routledge: London and

New York.

Isings, C. (1957). Roman Glass from Dated Finds, Archaeologica Traiectina 2, Groningen &

Djakarta.

Jennings, S. (2004-2005). Vessel Glass from Beirut. Bey 006, 007 and 045, Berytus 48-

49:27-292 ( issued 2006)

Oliver, A. (1980). Ancient Glass in the Carnegie Museum of Natural History, Pittsburg.

Pittsburg.

Peacock, D.P.S. and Williams, D.F. (1986). Amphorae and the Roman Economy, an

Introductory Guide. London and New York: Longman.

Sablerolles, Y. (1995). The Glass from Beirut Sites 015, 032 and 032. Report submitted to

the Directorate General of Antiquities, Lebanon.

Reynolds, P. (1997-98). Pottery production and economic exchange in second century

Berytus: ome preliminary observations of ceramic trends from quantified ceramic deposits

from the Souks excavations in Beirut, Berytus 43:35-110.

--- (2000). The Beirut amphora type, 1st century BC-7th century AD: an outline of its formal

development and some preliminary observations of regional economic trends, Rei Cretariae

Romanae Fautorum Acta (Ephesus, 1998) 36: 387-95.

--- (2003). Amphorae in Roman Lebanon: 50 BC to AD 250, Archaeology and History in

Lebanon (formerly National Museum News) 17: 120-130.

--- (2004). The Pottery, in R. Ortali Tarazi and B. Stuart. ‘Two rock-cut Roman tombs in

Chhîm’, BAAL 6: 107-134.

Reynolds, P. and Waksman, S.Y. (2007). Beirut Cooking Wares, 2nd

to 7th

Centuries: Local

Forms and North Palestine Imports, Berytus 50: 59-81.

Roumié, M., Waksman, S.Y., Nsouli, B., Reynolds, P. & Lemaître, S. (2004). Use of

PIXE Analysis technique for the study of Beirut amphora production at the Roman Period,

Nuclear Instruments in Physics Methods B, 215: 196-202.

Waksman, S.Y., Roumieh, M., Lemaitre, S., Nsouli, B. & Reynolds, P. (2003). Une

production d’amphores <carottes> à Beyrouth à l’ époque romain ?, Revue d’archéometrie

27 : 95 – 102.

Weinburg, G. D., Ed. (1988) Excavations at Jalame Site of a Glass Factory in Late Roman

Palestine, Missouri: University of Missouri Press.