OLWEN WILLIAMS-THORPE

A THOUSAND AND ONE COLUMNS: OBSERVATIONS ON THE

ROMAN GRANITE TRADE IN THE MEDITERRANEAN AREA

Summary. Provenancing and archaeological information on Roman granite

columns in the Mediterranean area has been collated from a range of published
papers by the author and others, together with new analyses for Rome, to
produce an integrated dataset comprising 1176 columns. This dataset allows
an overview of Roman granite trade in seven regions across the Mediterranean
area. Examination of the data indicates that columns made from Troad
(Turkish) granite are the most numerous observed overall (compatible with
Lazzarini’s earlier (2004) observation that this is the most widely distributed
type), followed by Aswan, then Elba and Giglio, and Kozak Dağ (Marmor
Misium). In the city of Rome, Mons Claudianus columns predominate. In
geographically peripheral parts of the Roman world (Spain, Israel), granite
columns are mainly from local sources, and are generally of smaller sizes than
those seen in Rome and Tuscany. Analytical data can be used to suggest
multiple extraction sites within some quarries, and have the potential for
identification of specific intra-quarry provenance. Dating evidence for primary
use of columns from the quarries considered is relatively sparse, but suggests
early (first century BC) exploitation of Spanish and Elba granites, while
column production at Aswan and Troad persisted into the fourth century
followed by reuse within later antiquity, in the fifth and seventh centuries AD.

introduction
Granite columns are key components of many notable and impressive Roman buildings,
and tracing the quarry sources of the granite raw material offers opportunities for understanding
this aspect of the Roman stone trade. There is now a large body of evidence concerning Roman
granite quarries and the distributions of their products, evidence stemming from continuing
projects both in Italy (in particular from Lazzarini, e.g. 2004 and see references therein), and
the UK (e.g. Peacock et al. 1994; Peacock and Maxfield 1997). The UK work incorporated
non-destructive scientific provenancing studies carried out largely at the Open University and
reported in a series of regional papers (e.g. Williams-Thorpe et al. 2000; Williams-Thorpe and
Potts 2002).
The provenancing evidence and publications include geochemical, magnetic and
mineralogical data on columns and quarries, together with measurements of column sizes. In
OXFORD JOURNAL OF ARCHAEOLOGY 27(1) 73–89 2008
© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK
and 350 Main Street Malden, MA 02148, USA. 73
 A THOUSAND AND ONE COLUMNS

addition, some information on the dates of column use (and, therefore, of quarry exploitation) is
available. By examining this evidence all together, we can take a Mediterranean-wide view of the
Roman granite trade and address the following questions:
1. Which quarry sources were numerically dominant in the Roman Mediterranean?
2. How do source distributions vary geographically, for example from west to east, and
comparing central and peripheral parts of the Roman Mediterranean?
3. How big are the columns, and is there a relationship between column size and quarry source
or region of use?
4. How much analytical variation is there within individual quarry sources, and can the data
throw light on how individual quarries were exploited in antiquity?
5. At what dates were Roman granite quarries producing these columns?
In order to consider these questions, we need to collect the provenancing and other
evidence into a single dataset. This has been done for the present paper, and the resulting dataset
comprises 1176 columns (details in the Appendix, full dataset available as noted there). The
dataset includes the columns for which there is numerical (provenancing and size) data, and
covers only areas where sufficient columns are reported to allow reasonable interpretation of the
questions listed above (we have omitted, for example, single and small numbers of column
provenances reported by Peacock et al. (1994) for Turkey and France respectively).
The columns are from 126 sites of Roman and post-Roman (secondary) use within
37 locations in Spain, Tuscany, Sardinia, Rome and its environs, Libya (one site only, Leptis
Magna), Cyprus and Israel. They were manufactured from a wide range of granitoid rocks and
originated at varied quarry sources across the Mediterranean area. Table 1 summarizes the
locations of the studied columns and published references, and Figure 1 shows the positions of
sites where columns were studied, together with the 16 relevant quarry sources.
Care was taken to ensure representative, and as full as possible, coverage of the range
of different granitoid column types seen at each site. However, the geographical coverage within
the Mediterranean as a whole necessarily reflects practical limitations on fieldwork, and thus,
gaps in our distribution, for instance in Greece and North Africa, remain to be filled.
The columns and their raw materials include many different petrologies ranging from
granites to meta-gabbros, but for simplicity they will be referred to here as ‘granites’ or
‘granitoids’. The strict geological definitions of the source rocks are given by Lazzarini (2004).

provenance summary
The methodologies used to provenance the columns, analytical data for columns and
quarry sources, and provenancing interpretations, have been set out for most of the studied areas
in previous papers (references given in Table 1; additional geochemical data and some
provenancing interpretations for columns in Rome published for the first time here). Within the
dataset of 1176 columns that we shall discuss in the present paper, 1051 have been provenanced
to a quarry or outcrop source region. For the remaining 125 columns, the available provenancing
information is insufficiently diagnostic or unambiguous to be sure of source, though in many
cases a general area can be suggested. In a small number of cases, the characteristics of the
columns differ markedly from all the source areas that we have studied, suggesting derivation
from as yet unknown quarrying areas.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
74 Journal compilation © 2008 Blackwell Publishing Ltd.
 table 1
Summary of the locations and numbers of columns studied

© 2008 The Author

Site number on Figure 1 Location/site Parts where columns were studied Number of References/data sources
columns studied
Spain Williams-Thorpe and Potts 2002
1 Mérida Museo Nacional de Arte Romano (2), Theatre (19), 90
Amphitheatre (36), Casa del Anfiteatro (23), Temple of
Diana (4), Casa del Mithraeo (4), Plaza del Toro (2)
2 Alange Roman baths gardens 3
3 Fuente de Cantos Church of Nuestra Senora de Granada porch 2
4 Mulva ? 2

Journal compilation © 2008 Blackwell Publishing Ltd.

OXFORD JOURNAL OF ARCHAEOLOGY
5 Écija Palacio de Valldemoza (5), town (3), Church of Jesus Soga 17
(5), gardens next to river (4)
6 Seville Cathedral (outside) (28), Calle Marmoles (3), Avenida 33
Isabel la Catolica (2)
7 Itálica Casa de los Pasaros (Pajaros) (3), and site entrance (1) 4
8 Gerena Church, on wall outside front 1
Sardinia Williams-Thorpe and Rigby 2006
9 Porto Torres Zona Archeologica (8), near baths (2), cemetery (modern) 34
(2), Piazza XX Settembre (1), Cathedral of S Gavino
(21)
10 Cornus Paleo-Christian Basilica 1
11 Oristano Santa Giusta Basilica 4
OLWEN WILLIAMS-THORPE

12 Cagliari Museum Porto Cristina (4), Duomo S Elena (1), Piazza 7

Indipendenza (2)
13 Olbia Archaeological store on Via Argentina (18), San Simplica 50
Basilica (5), villa garden on main street (27)
14 Porto Rotondo Harbour west side 1
15 Capo Testa Cala Romana (1), Petri Taddati (1), Capicciolu quarry 7
areas (5)
Tuscany Williams-Thorpe and Rigby in press
16 Pisa Outside Leaning Tower (5), Ufficio delle Opere hall (1), 72
Duomo (58), Baptistry (8)
17 Fiesole Duomo garden 1

75
76
table 1
continued

Site number on Figure 1 Location/site Parts where columns were studied Number of References/data sources
columns studied
18 Florence Baptistry (12), Ponte Vecchio (1), Piazza Republica (1), 18
Piazza Trinita (1), Borgo S Jacopo (3)
19 Arezzo Amphitheatre (1), Churches of Badia (1) and of S Maria 27
in Piave (14), Duomo (11)
20 Vallebuia Quarry area 1
21 S Piero and S Churches of S Nicolo (S Piero in Campo) (4) and of S 14
Ilario Ilario in Campo (10)
22 Giglio Castello Church at top of hill inside castle 8
23 Grosseto On town walls above Via Manetta 1
Rome and Environs Williams-Thorpe and Thorpe 1993;
Williams-Thorpe et al. 1996;
portable gamma-ray spectrometry
data (not previously published)
for 141 columns (see dataset)
24 Rome Mercati Traiani (2), Basilica Ulpia (6), Colle Oppio (3), 371
Colosseum (12), Pantheon (16), Baths of Diocletian and
of Caracalla (37 and 18), Roman Forum (45), Palatine
(2), Temples of Saturn and of Venus and Rome (11 and
30), Domus Flavia (3), Domus Augustana (5) and
Domus Transitoria (1), Palazzi Farnese and Cancelleria
(12 and 39), Churches of S Prassede (28), S Giovanni in
A THOUSAND AND ONE COLUMNS

Laterano (26), S Agnese (2), S Constanza (27), S Maria

Nova (3), S Maria Maggiore (10) and S Martino ai
Monti (1), Via S Martino ai Monti (2), Piazza Navona
(3), Piazzale del Colosseo (2), St Peter’s Basilica (25)
25 Tivoli Hadrian’s Piazza d’Oro (9), Imperial palace (1), 3 exedras building 29
Villa (13), Teatro Maritimo (1), Triclinio Estivo (3), Grandi
Termi (2)

© 2008 The Author

Journal compilation © 2008 Blackwell Publishing Ltd.
OXFORD JOURNAL OF ARCHAEOLOGY
© 2008 The Author
26 Ostia Antica Site entrance (1), Decumanus Maximus (2), Via Fontana 50
(1), Theatre (1), main Forum (24), House of Cupid and
Psyche (3), Forum of Statue Eroica (3), outside Curia
(3), harbour area (4), Via del Foce (2), outside museum
(6)
Libya Williams-Thorpe et al. 2000
27 Leptis Magna ? 30
Cyprus Williams-Thorpe and Webb 2002

Journal compilation © 2008 Blackwell Publishing Ltd.

OXFORD JOURNAL OF ARCHAEOLOGY
28 Paphos Agia Kyriaki/Khrysopolitissa (22), Saranda Kolones (58), 89
Odeon (5), Roman fortifications (beach area, 2),
Limeniotissa Basilica (2)
29 Kourion Christian Basilica 3
30 Nicosia Folk Art Museum courtyard 3
Israel Williams-Thorpe and Henty 2000
31 Nazareth Churches of the Annunciation (12) and St Joseph (5) 17
32 Tiberias Park 4
33 Susita Cathedral 18
34 Beth Shean Theatre (22) and main site (30) 52
35 Caesarea Crusader town (harbour, promenade, 27), citadel (12), 54
OLWEN WILLIAMS-THORPE

Maritima Oraqivah (4), Roman theatre (6), near porphyry statue

to east of Crusader walls (5)
36 Ashkelon Main site and park (16), Nike area (5), Bouleterion (19), 46
St Mary the Green (5), sea wall fortifications (1)
37 Jerusalem Rockefeller Museum (2), Old City near Aftimos Fountain 12
(2), Church of the Holy Sepulchre (8)
Total number of columns included in the study 1176

77
 A THOUSAND AND ONE COLUMNS

18
Elba 16 17
20 19
21
23
Spain I, II, III Corsica 22 24 25 26
1 CapoTesta 15 Giglio
9 14 Porto
3 2 10 13 Troad
4 11 Rotondo Kozak Dag
8 Sardinia Olbia
7 5 12
6

30

28 29
32
31 33
35
34
36
27
37

Wadi Barud
Mons
Claudianus
Quarries Wadi Semnah
Columns 500km Wadi
Fawakhir
Aswan

Figure 1
Map of the Mediterranean region showing the positions of column sites included in this paper, and of the relevant quarry
areas. Sites are numbered: (Spain) 1 Merida, 2 Alange, 3 Fuente de Cantos, 4 Mulva, 5 Ecija, 6 Seville, 7 Italica, 8
Gerena; (Sardinia) 9 Porto Torres, 10 Cornus, 11 Oristano, 12 Cagliari, 13 Olbia, 14 Porto Rotondo, 15 Capo Testa;
(Tuscany) 16 Pisa, 17 Fiesole, 18 Florence, 19 Arezzo, 20 Vallebuia, 21 S Piero and S Ilario, 22 Giglio Castello, 23
Grosseto; (Rome and environs) 24 Rome, 25 Tivoli Hadrian’s Villa, 26 Ostia; (Libya) 27 Leptis Magna; (Cyprus) 28
Paphos, 29 Kourion, 30 Nicosia; (Israel) 31 Nazareth, 32 Tiberias, 33 Susita, 34 Beth Shean, 35 Caesarea Maritima, 36
Ashkelon, 37 Jerusalem. Many of these sites include several, or many, locations where columns were studied, e.g.
Seville, Jerusalem, Paphos, Rome, cf. Table 1. Quarry areas are labelled.

A convenient method of illustrating the major provenance groupings of the columns is

to combine geochemical and magnetic data in the form of a plot of magnetic susceptibility
against Thorium (Th) (Figs. 2a and 2b) for grey and red granite columns respectively.
The column provenances thus summarized on Figure 2 provide the basis for the
discussion that now follows.

discussion

Which quarry sources were numerically dominant in the Roman Mediterranean?

Figure 3 illustrates the frequency of occurrence of columns from the different quarry
sources, firstly including all the columns studied (Fig. 3a), and then in turn by different regions
(Figs. 3b–3h).
Overall, Troad columns dominate as the most numerous observed in our studies
(Fig. 3a); this is in accord with Lazzarini’s statement (2004) that this granite is the ‘piu diffusa

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
78 Journal compilation © 2008 Blackwell Publishing Ltd.
 OLWEN WILLIAMS-THORPE

a
100
v
Kozak Dag
Magnetic Susceptibility (SI units x 10 )

Wadi Semnah
-3

10 Troad
Corsica

Mons
Claudianus Spain III
1

Spain II
Elba and
Giglio
0.1
Spain I

0.01
-20 -10 0 10 20 30 40 50 60 70 80 90 100
Th (ppm)

b
16
Wadi Fawakhir
14
Magnetic Susceptibility (SI units x 10 )
-3

Sardinia,
12
Olbia area

10

6
A S W A N C O L U M N S
4
Sardinia,
Capo Testa area
2
Sardinia,
Olbia area
0
-10 -5 0 5 10 15 20 25 30 35
Th (ppm)

Figure 2
Plots of magnetic susceptibility against Thorium (Th) for: (a) grey granite columns, and (b) red and pink granite columns.
Magnetic susceptibility for grey granites (Fig. 2(a)) covers a large numerical range and therefore is shown more
conveniently on a logarithmic scale. The groupings (surrounded by lines for clarity) on the graphs reflect the different
quarry sources, which are named on the figures (note that the Kozak Dağ group (Fig. 2(a)) includes a small number
believed to originate at outcrops close to Kozak, named Edremit and Evciler). Olbia columns shown (Fig. 2(b)) are of two
geochemical types (probably two quarries), cf. Williams-Thorpe and Rigby 2006). The data plotted are columns for which
both geochemical and magnetic data are available, and so are fewer than total numbers of columns studied. Elba and
Giglio granites are not always differentiated in provenancing papers, so are grouped together here. The error bar shown
for Th is analytical precision at 1 standard deviation (Williams-Thorpe et al. 2000; typical repeatability for duplicate
analyses of a single column is about the same). Instrumental precision for magnetic susceptibility is 0.2% rsd at 7 ¥ 10-3
SI (Williams-Thorpe and Thorpe 1993), and the typical rsd on averaged measurements for a single column is between
about 2% and 10% (cf. Williams-Thorpe et al. 1996; 10% shown on the figures to represent worst case).

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd. 79
 A THOUSAND AND ONE COLUMNS

350 100 50 100

a b c d
Troad 90 45 90
300
80 40 80
Elba and
250 70 35 70 Giglio
(Tuscany)
Frequency

60 60

Frequency

Frequency

Frequency
200 Kozak 30 Sardinian
v
Dag 50 25 sources 50
150 Elba and 40 40
Aswan Spanish 20
Giglio
sources 30
100 30 15
20 10 20
50
10 5 10
0 0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

10
11
12
13
14
15
10
11
12
13
14
15

1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9

10
11
12
13
14
15
1
2
3
4
5
6
7
8
9
Quarry Source Quarry Source Quarry Source Quarry Source
All Columns Spain Sardinia Tuscany

120 20 100 160

e f 90
g h
18 140
100 Mons 80 Troad
Kozak 16 Troad
v Claudianus 120
Dag
14 Troad 70
80
Elba
Frequency

100
Frequency

12 Aswan 60

Frequency

Frequency
and
Troad

60 Giglio 10 50 80
Aswan
8 40
60
40 Kozak
30
Wadi 6 Dag
v
Aswan
40 Kozak
v
Fawakhir 4 20 Dag
20
20
2 10
Aswan
0 0 0 0

10
11
12
13
14
15

10
11
12
13
14
15
1
2
3
4
5
6
7
8
9

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
10
11
12
13
14
15

1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9

Quarry Source Quarry Source Quarry Source

Quarry Source
Rome and Libya Cyprus Israel
environs (Leptis Magna)

Figure 3
Histograms showing the numbers and proportions of granite columns from different sources, for: (a) all columns
studied, (b) Spain, (c) Sardinia, (d) Tuscany, (e) Rome and its environs (Ostia and Tivoli Hadrian’s Villa), (f) Libya
(Leptis Magna columns), (g) Cyprus, and (h) Israel. The sources are numbered: 1 Spain group I, 2 Spain group II, 3
Spain group III, 4 Sardinia Porto Rotondo, 5 Sardinia Capo Testa area, 6 Sardinia Olbia area, 7 Corsica, 8 Elba and
Giglio, 9 Turkey Troad, 10 Turkey Kozak Dağ area, 11 Mons Claudianus, 12 Aswan, 13 Wadi Fawakhir, 14 Wadi Barud,
15 Wadi Semnah (11–15, Egyptian sources). Selected sources discussed in the text are also marked on the histograms.

dell’antichita’. On Figure 3a Troad is followed in frequency by the well-known Aswan type, then
by Elba and Giglio, and (nearly as numerous) Kozak Dağ (Marmor Misium); the widespread use
of Kozak Dağ rock has been increasingly recognized in recent years (Lazzarini 2004, 109, and cf.
de Vecchi et al. 2000). The large proportion of Elba/Giglio columns reflects to some degree the
concentration of studied sites in Italy, where Elba/Giglio granite was relatively locally available.
The picture of column use varies within individual areas. Spain, Sardinia and Tuscany
(Figs. 3b, 3c and 3d) are each dominated by their own local sources. In Rome and its environs
(Fig. 3e), columns from the extensive Egyptian quarries of Mons Claudianus predominate,
underlining earlier observations of Imperial control of this granite (e.g. Peacock 1993). Still
considering Figure 3e, Italian Elba and Giglio, and then Aswan, columns are most numerous
after Mons Claudianus, closely followed by Kozak Dağ (Marmor Misium) and only then by
Troad products. Other sources are present in Rome in smaller numbers (and it should be noted
that several additional columns of probable Fawakhir origin were observed in Rome that we did
not have the opportunity to measure and so do not appear on Figure 3e). Cyprus and Israel
(Figs. 3g and 3h) used proximal Egyptian and Turkish sources, and those same sources appear
in Libya at Leptis Magna (Fig. 3f).

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
80 Journal compilation © 2008 Blackwell Publishing Ltd.
 OLWEN WILLIAMS-THORPE

EAST

15 Wadi Semnah
14 Wadi Barud
EGYPT

13 Wadi Fawakhir
12 Aswan
11 Mons Claudianus
TURKEY

10 Kozak Dag
SOURCE AREA

9 Troad
TUSCANY

8 Elba & Giglio

7 Corsica
6 Olbia
SARDINIA

5 Capo Testa
4 Porto Rotondo
3 Spain III
SPAIN

2 Spain II
1 Spain I
SP

TU
WEST

SA

LI EP

IS
RO NV

YP
B T

R
A

(L
SC

E
R

M IRO

YA IS

A
IN

RU
D

EL
E N
A
IN

WEST EAST
an S
N

S
IA

M
AG
N
A
)
REGION OF USE

Figure 4
Diagram showing Roman granite quarry sources (y axis) plotted against the regions in which those sources were used
for columns (x axis). The sources and the regions of use are arranged on the plot in approximate geographic order, from
west to east as marked. The wide variety of sources used in Rome is evident, while at the peripheries of the Roman
Mediterranean, geographically proximate sources predominate.

How do source distributions vary geographically, for example from west to east,
and comparing central and peripheral parts of the Roman Mediterranean?
A useful way of examining geographical variations within the source distributions is
to plot the regions in which we have studied columns from west to east, against the sources
similarly plotted according to their locations along an approximate west–east line within the
Roman Mediterranean.
Thus, on Figure 4, Rome lies in the middle section of the diagram, and peripheral parts
of the Roman world are at the left and right hand edges. Clearly, and as we might expect, the city
of Rome enjoyed access to the greatest variety of sources. At the peripheries of the Roman
Mediterranean, geographical proximity of source remained the dominant influence in granite
selection. Thus, Spanish sources were used mainly in Spain, and are not found east of Rome.
Similarly, Tuscan and Sardinian sources have distributions limited to the central and western part
of the Mediterranean. Columns from Egyptian and Turkish sources transcend such limitations
and are found all across the Mediterranean – though we should recall from Figure 3 (above) that
the frequency of such columns is, apart from Rome, generally much lower in the west of the
Mediterranean than in the east.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd. 81
 A THOUSAND AND ONE COLUMNS

160

140

120
Diameter (cm)

100

80

60

40

20

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

W
SP

SP

SP

SA or

SA ap

SA lb

EL

TR

KO

W
O

SW
O

A
A

A
(P II

(C IA on

(O IA a)
A

BA

O
R to R

R oT

R ia)

ZA
R

D
N

D
IN

IN

IN

A
D

D e

SI

IS
A
S

IF

IB
D

K
IN ot

IN s

IN

&

N
C
I

II

AW

EM
A
DA
IA

LA

RU
IG

N
G

A
D
LI
t

H
O

IA

IR
do

N
U
)

S
Quarry Source

Figure 5
Diameters of studied columns in the Roman Mediterranean area as a function of their quarry sources. Sources are
named on the figure. Note we have no diameter measurements for Wadi Barud columns (source no. 14).

How big are the columns, and is there a relationship between column size and quarry source
or region of use?
Figure 5 illustrates the diameters of the studied columns as a function of quarry source.
The largest columns (diameters of greater than 140 cm) were produced at the Egyptian
quarries of Aswan and Mons Claudianus. Diameters of over 100 cm are also observed within
columns produced on the Italian islands of Sardinia, Elba and Giglio. One column made of
Spanish granite is larger than 100 cm, but in general the Spanish sources produced smaller
columns, 60 cm or less in diameter.
Roman standardization of column sizes (cf. Dodge 1991, 37) is the likely cause of the
clustering of points apparent on Figure 5; Troad and Kozak Dağ in particular have concentrations
of columns of specific sizes, distributed around 50 cm and 40 cm respectively.
In Figure 6, we consider column diameters by region of use, including here all the
quarry sources used in that region.
In Spain (Fig. 6a), the columns used were mainly of a relatively modest size, peaking at
only 40 cm diameter. In Sardinia (Fig. 6b) rather larger columns (c.50 cm) are frequent, while in
Tuscany and Rome (Figs. 6c and 6d respectively) many columns are seen with diameters of
70 cm and 100 cm or greater. Further south and east in the Mediterranean, columns at Leptis

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
82 Journal compilation © 2008 Blackwell Publishing Ltd.
 OLWEN WILLIAMS-THORPE

60 30 40
a b 35 c
50 25
30
40 20
25
Frequency

Frequency

Frequency
30 15 20

15
20 10
10
10 5
5

0 0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Diameter (cms) Diameter (cms) Diameter (cms)
Spain Sardinia Tuscany
12
45
d e f
40 3 10
35
8
30
Frequency

2
25
Frequency

Frequency
6
20
4
15 1
10 2
5
0 0
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Diameter (cms) Diameter (cms) Diameter (cms)
Rome Libya Cyprus
(Leptis Magna)
20
18 g
16
14
12
Frequency

10
8
6
4
2
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Diameter (cms)
Israel

Figure 6
Histograms comparing diameters of columns in the regions studied. (a) Spain, (b) Sardinia, (c) Tuscany, (d) Rome and
environs (Ostia and Tivoli Hadrian’s Villa), (e) Libya (Leptis Magna), (f) Cyprus, (g) Israel.

Magna (Fig. 6e) are large (c.60 to 90 cm), while in Cyprus and Israel (Figs. 6f and 6g
respectively) columns are mainly around 60 and 50 cm in diameter respectively.

How much analytical variation is there within individual quarry sources, and can the data
throw light on how individual quarries were exploited in antiquity?
Consideration of the provenance of columns within a particular quarry source can throw
light on the way in which the quarry was exploited, or was developed over time. For example,
at Mons Claudianus, variations in magnetic susceptibility levels were used to demonstrate coeval
use of different parts of the extensive quarry area (Williams-Thorpe et al. 1996).
By examining analytical data for our thousand plus columns all together, it becomes
clear that the analytical spread for column types from individual sources often greatly exceeds
the expected analytical precision (this point is illustrated for Thorium (Th) values on Figure 2

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd. 83
 A THOUSAND AND ONE COLUMNS

3 PALF4
PALF2
Higher Th and
2 U group
U (ppm)

-1
Lower Th and U
group
-2

-3
-10 -5 0 5 10 15 20 25 30 35
Th (ppm)

Figure 7
Graph of Uranium (U) plotted against Thorium (Th) for columns assigned to the Aswan quarries, suggesting two
compositional groups within these data. Negative values occur because concentrations cluster around lower limits of
detection for the method (5 ppm for U, 8 ppm for Th; Williams-Thorpe et al. 2000). PALF2 and PALF4 probably belong
to the low U and Th group.

above). This wide spread of column data suggests exploitation of many different parts within
most quarry areas. This is, of course, consistent with the need to extract enough material for
thousands of columns.
The range of Th values for Kozak Dağ and Troad columns, reflected by analogous
variations in magnetic susceptibility (again, cf. Fig. 2a), raises the possibility of relating
individual columns to the different extraction sites that have been identified in the field (see De
Vecchi et al. 2000 for Kozak Dağ, and Cook 1973 for Troad). Such a progression would require
fieldwork at the quarries to establish geochemical and magnetic variations or trends, after the
manner achieved for Mons Claudianus (Williams-Thorpe et al. 1996).
The question of column provenance within a source area has recently been explored for
Elba, where a combination of microprobe, portable gamma-ray spectrometry, and magnetic
analysis has been used to suggest a shift away from Roman workings (at Cavoli on the south
coast of the island), towards quarries further north and east in Elba, during later (Pisan, eleventh
century AD) exploitation.
The dataset prepared for the present paper includes the first substantial body of
geochemical data on columns of Aswan granite. Examination of Uranium (U) and Thorium
data for these columns (Fig. 7) allows an, albeit tentative, suggestion of the presence of two
geochemical types, one with higher U and Th concentrations, the other with very low
concentrations (at and below detection limits, though clustering of data gives some confidence in
the magnitude of values obtained).

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
84 Journal compilation © 2008 Blackwell Publishing Ltd.
 OLWEN WILLIAMS-THORPE

The Aswan columns that we measured include dark red and lighter (pink) examples (cf.
especially Lazzarini 2004, 78, on colour and textural variability of Aswan granite), but these
colour variations are not reflected in the putative analytical groupings on Figure 7, each group
containing columns of both colour types. The low U and Th type has been identified so far only
in Rome (reused columns at the Palazzo Farnese and Palazzo Cancelleria). The higher U and Th
type is present in Rome, Cyprus and Israel. We do not have sufficient U and Th data for the
Aswan quarries to relate the two column groups to different parts of Aswan, except only to
comment that the low U and Th group is more like source sample LG10 reported by Galetti et al.
(1992; LG10 contains only 7 ppm of Th).

At what dates were Roman granite quarries producing these columns?

Secure dates for the manufacture and primary use of columns are rare because of the
very extensive reuse of Roman decorative stone particularly in the Renaissance period. Figure 8
summarizes the available information on primary dates, and also includes evidence for secondary
use within the first millennium AD.
Earliest dates for our columns are first century BC examples from Spain (columns in
Merida), and from the Elba/Giglio source area (Elba columns in Rome). The period of most
intensive column production was during the second century AD, declining in numbers in the
third century. Troad and Aswan columns persist most frequently in primary use into the fourth
century.
This evidence accords well with Lazzarini’s and others’ observations on dates (see,
especially, Lazzarini 2004), though the first use of Troad granite columns within the Roman
period is uncertain. Lazzarini (2004) puts this at the beginning of the second century AD, while
Figure 8 shows a small number of examples in the first century AD. These are columns in the
Roman theatre at Caesarea Maritima in Israel – a structure that dates from the first century AD,
and was rebuilt during the second–third century, but using materials from the original building.
The evidence for export of this rock is thus ambiguous, though it is known that the Troad granite
was already in local use well before the first century AD, in the Hellenistic period.
Turning to the information in Figure 8 on reuse of columns within later antiquity, Aswan
and Troad columns were incorporated into Christian basilicas in the fifth century AD, and into
Byzantine sites of the seventh. Pisan exploitation (eleventh century onwards) of Elba and Giglio
granite columns included both reuse of Roman products and newly quarried and manufactured
columns. Subsequent to the period shown on Figure 8, there is a wealth of Roman granite
columns to be seen as spolia in churches, cathedrals and palaces all across the Mediterranean and
into northern Europe and even in the Ukraine (see, for example, Gnoli 1971, and (for Ukraine,
columns in Kiev) Ainalov 1920).

conclusions
1. A new dataset for 1176 Roman granite columns, including provenancing criteria (magnetic
susceptibility measurements, geochemical analyses, and mineralogical observations)
together with column diameter measurements and information on dating and archaeological
contexts, has been constructed from previously published regional papers by the author and
others together with new information for columns in Rome. The dataset allows conclusions
to be drawn regarding the Roman granite trade in the Mediterranean area.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd. 85
 A THOUSAND AND ONE COLUMNS

40 3.5 30
a b c
35 3
25

Much Pisan reuse/new use

30
2.5
Frequency

Frequency

Frequency
25 20
2
20 15
1.5
15
10
1
10

5 0.5 5

0 0 0
-2 -0.5 1 2.5 4 5.5 7 8.5 10 -2 -1 0 1 2 3 4 5 6 7 8 9 10 -2 0 2 4 6 8 10
BC CENTURY AD BC CENTURY AD BC CENTURY AD
SPAIN I II III CORSICA ELBA AND GIGLIO

30 16 70
d f
e
25 14 60

12
57 re-used in Cyprus

50
20
Frequency
Frequency

Frequency
10
40
15 8
30
6
10
20
4
5 10
2

0 0 0
-2 -0.5 1 2.5 4 5.5 7 8.5 10 -2 -1 0 1 2 3 4 5 6 7 8 9 10 -2 -1 0 1 2 3 4 5 6 7 8 9 10
BC CENTURY AD BC CENTURY AD BC CENTURY AD
TROAD KOZAK DAG MONS CLAUDIANUS

8
g
7

6
Frequency

0
-2 -0.5 1 2.5 4 5.5 7 8.5 10
BC CENTURY AD
ASWAN

Figure 8
Histograms showing dates of primary use (solid bars), and of secondary reuse within the first millennium AD (dotted
lines) of Roman granite columns from seven quarry areas: (a) Spain, (b) Corsica, (c) Elba and Giglio, (d) Troad, (e)
Kozak Dağ, (f) Mons Claudianus and (g) Aswan. The dates are in centuries BC and AD, with entries for ‘2.5’ and ‘4.5’
indicating dates given as ‘second–third century AD’, and ‘fourth–fifth century AD’ respectively.

2. Within the seven regions studied (Spain, Sardinia, Tuscany, Rome, Libya, Cyprus and Israel)
columns from the Turkish Troad source are most numerous overall.
3. Local granite sources dominate in Spain, Sardinia and Tuscany, with a minority of imports
from Turkey and Egypt.
4. In regions where granite column quarries are absent (Cyprus, Israel, and Libya (Leptis
Magna)), proximal sources in the eastern Mediterranean were used exclusively.
5. In Rome and its environs, Mons Claudianus columns are the most frequently occurring type
followed by Elba and Giglio, Aswan, Kozak Dağ, and then Troad.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
86 Journal compilation © 2008 Blackwell Publishing Ltd.
 OLWEN WILLIAMS-THORPE

6. Rome contains the greatest variety of granite column types (at least 12 granite sources
noted), while three and five source areas, respectively, provided all the columns observed in
geographically peripheral regions of the Roman Empire (Israel, Spain).
7. The largest columns observed (>140 cm diameter) are from Aswan and Mons Claudianus.
8. Column sizes (including all quarry sources) vary by region, with Spain containing the
highest frequency of small (c.40 cm diameter) columns, and Rome and Tuscany containing
most of the largest columns.
9. Aswan granite columns comprise two compositional types, probably from two different
parts of the Aswan area.
10. Granite columns from quarries in Spain and Elba were in use by the first century BC. Most
intensive exploitation of Roman granite quarries occurred during the second century AD,
with Troad and Aswan the most frequent types observed in the fourth century. Reuse of
Aswan and Troad columns is noted in the fifth and seventh centuries AD.

Acknowledgements
My involvement in Roman granite columns stems from research initiated by Professor David
Peacock then at the University of Southampton, and I am grateful to him for many valuable discussions and
for loans of samples and thin sections. Professor L. Lazzarini has expressed kind support for the work and
drawn my attention to new publications. For their help and invaluable collaboration on fieldwork and in
other aspects of the research, I thank my colleagues at the Open University and elsewhere: Peter Webb, Phil
Potts, Chris Jones, Mike Henty, Andy Tindle, and Ian Rigby. The late Richard Thorpe played a key role in
the early days of our joint granite research. For financial support I thank the Leverhulme Trust (grant
F.269P). Archaeological permits relevant to the work are acknowledged in the series of regional papers
from the Open University that are given in the reference list of this paper. I thank Isabel Jones for a nice
title suggestion, and Ian Rigby for comment on early drafts and diagrams and producing the much
improved figures included in the paper. The paper has also benefited from helpful comment made by John
Humphrey and colleagues.

High Gable House

Belmangate
Guisborough TS14 7BD1

APPENDIX: DETAILS OF THE DATASET OF COLUMNS

The dataset of 1176 columns on which this paper is based is in Excel™ format and is available
at http://www.blackwell-synergy.com/doi/abs/10.1111/j.1468-0092.2007.00297.x. It includes:
• 2911 measurements of magnetic susceptibility for 999 columns (data averaged for each column).
• 290 geochemical analyses for 252 columns (potassium (K), Uranium (U) and Thorium (Th)
concentrations from portable gamma-ray spectrometry, averaged where duplicate or multiple analyses
were done on a single column).
• Comments on the colour, mineralogical and petrographic characteristics of columns (mainly from
examination in the field; occasionally thin sections were available to supplement by-eye and hand lens
examination).

1 Formerly at Department of Earth Sciences, The Open University, Milton Keynes, UK, MK7 6AA.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd. 87
 A THOUSAND AND ONE COLUMNS

• Diameters for 922 columns (measurements were made at about the mid-point of each column, and
diameters were preferred to lengths as a measure of size, because lengths are often incomplete).
• Dating evidence (primary and/or secondary use) for 450 columns.

references
ainalov, d.v. 1920: History of Ancient Russian Art. Izvestiya of the Tauric Archaeological Commission
57, 136–248.
de vecchi, g., lazzarini, l., lünel, t., mignucci, a. and visonà, d. 2000: The genesis and
characterisation of ‘Marmor Misium’ from Kozak (Turkey), a granite used in antiquity. Journal of Cultural
Heritage 1, 145–53.
cook, j.m. 1973: The Troad: an archaeological and topographical study (Oxford).
dodge, h. 1991: Ancient marble studies: recent research. Journal of Roman Archaeology 4, 28–50.
galetti, g., lazzarini, l. and maggetti, m. 1992: A first characterization of the most important granites
used in antiquity. In Waelkens, M., Herz, N. and Moens, L. (eds.), Ancient Stones: quarrying, trade and
provenance (Leuven, Acta Archeologica Lovaniensa Monograph 4), 167–78.
gnoli, r. 1971: Marmora Romana (Rome).
lazzarini, l. 2004: Pietre e Marmi Antichi (Padua).
peacock, d.p.s. 1993: Rome in the Desert: a symbol of power (Southampton).
peacock, d.p.s. and maxfield, v.a. 1997: Mons Claudianus Survey and Excavation 1987–1993. Vol. I
Quarries and Topography (Cairo).
peacock, d.p.s., williams-thorpe, o., thorpe, r.s. and tindle, a.g. 1994: Mons Claudianus and the
problem of the granite del foro: a geological and geochemical approach. Antiquity 68, 209–30.
williams-thorpe, o. and henty, m.m. 2000: The sources of Roman granite columns in Israel. Levant 32,
155–70.
williams-thorpe, o. and potts, p.j. 2002: Geochemical and magnetic provenancing of Roman granite
columns from Andalucía and Extremadura, Spain. Oxford Journal of Archaeology 21, 167–94.
williams-thorpe, o. and rigby, i.j. 2006: Roman granites of Sardinia: geochemical and magnetic
characterisation of columns and quarries, and comments on distributions in the Mediterranean area.
Marmora 2, 83–112.
williams-thorpe, o. and rigby, i.j. in press: Roman granite columns in Elba, Giglio and Tuscany: a
contribution to provenancing and distribution studies. Marmora.
williams-thorpe, o. and thorpe, r.s. 1993: Magnetic susceptibility used in non-destructive
provenancing of Roman granite columns. Archaeometry 35, 185–95.
williams-thorpe, o. and webb p.c. 2002: Provenancing of Roman granite columns in Cyprus using
non-destructive field portable methods. Report of the Department of Antiquities, Cyprus 2002, 339–63.
williams-thorpe, o., jones, m.c., tindle, a.g. and thorpe, r.s. 1996: Magnetic susceptibility variations
at Mons Claudianus and in Roman columns – a method of provenancing to within a single quarry.
Archaeometry 38, 15–41.
williams-thorpe, o., webb, p.c. and thorpe, r.s. 2000: Non-destructive portable gamma ray
spectrometry used in provenancing Roman granitoid columns from Leptis Magna, North Africa.
Archaeometry 42, 77–99.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
88 Journal compilation © 2008 Blackwell Publishing Ltd.
 OLWEN WILLIAMS-THORPE

supplementary material
The following supplementary material is available for this article:

Dataset of 1176 columns

This material is available as part of the online article from:

http://www.blackwell-synergy.com/doi/abs/10.1111/j.1468-0092.2007.00297.x

(This link will take you to the article abstract).

Please note: Blackwell Publishing are not responsible for the content or functionality of any
supplementary materials supplied by the authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.

OXFORD JOURNAL OF ARCHAEOLOGY

© 2008 The Author
Journal compilation © 2008 Blackwell Publishing Ltd. 89