NEW MYCENAEAN POTTERY PRODUCTION CENTERS FROM THE EAST

ERN PART OF CENTRAL OBTAINED BY NEUTRON ACTIVATION

ANALYSIS

& & #

H Mommsen A Hein D Ittameier J Maran and Ph Dakoronia

Institut fur Strahlen und Kernphysik Universitat Bonn Nuallee Bonn

&

Institut fur Ur und Fruhgeschichte Universitat Heidelb erg Marstallhof

th #

Heidelb erg and Ephorate of Antiquities Greece

November

Abstract Neutron activation analysis results of a set of more than sherds and whole

vessels including misred pieces from Lo cris and Southwestern Thessaly are

presented The material covers the time p erio d Middle Helladic to Late Helladic I I IC

Several new patterns are detected some of which can b e assigned with high probability

to that region A sp ecial result concerns several fragments of pictorial craters depicting

ships and sea battles found in LHI I IC Middle levels at LivanatesKynos As suggested

by our data they originate from a manufacturing place in Eastern Lo cris

Introduction

Since the midies our archaeometry group is working on a pro ject concerning the

classication and provenance determination of Aegean Bronze Age p ottery which has

b een a fo cus of archaeological interest for more than a years Of all the dierent

p ottery classes used during the Bronze Age sp ecial attention was paid to the ne

decorated p ottery of the Greek Mainland and adjacent regions The p erio d of the Late

Bronze Age the Mycenaean p erio d is esp ecially well investigated and this for the

following reasons

Mycenaean decorated p ottery shows a well dened range of vessel shap es and mo

tives so that even the smallest fragments can b e attributed with certainty to sp ecic

vase types It is very homogeneous in app earance over wide areas and it has a wide dis

tribution encompassing almost the whole Mediterranean The archaeological questions

arising immediately are

How centralized was the pro duction That is to say was this p ottery manufac

tured in only a few places and distributed by trade to the dierent regions where it

is found to day Or was its pro duction rather decentralized and organized by p otters

imitating the shap es and decorations on demand by the customers during this time

These questions esp ecially apply to Mycenaean vase painting of the highest quality

like the Pictorial Style

Was there some sort of internal ranking b etween pro duction places of only

regional and such of supraregional imp ortance If this were the cases could it b e

that the workshops connected with the Mycenaean palaces esp ecially in the Argolid

were the most imp ortant ones

To answer these questions by archaeological means tentative assumptions can b e

obtained for example by considering distribution patterns of p ottery shap es or styles

As is generally agreed to day comp ositional information obtained by natural scientic

metho ds will lead to new insights and either corrob orate or contradict these assump

tions but also add new questions for example

If in addition to a b etween sites comp ositional variation also a variation within

a site is detected i e a subgroup pattern can these subgroups stylistically or

typologically b e characterized and related to a technological tradition or change Are

they related to sp ecic pro duction techniques for dierent customers or use

Neutron Activation AnalysisNAA metho d and principles of chemical provenancing

To determine the comp osition of p ottery NAA is known to b e esp ecially well suited

since its prop erties fulll all the sp ecial demands very well Ab out elemental con

centrations p er sample are measured routinely in our lab oratory with the help of an

automated sample changer and an elab orate sp ectrum evaluation programMommsen

et al and references therein It is a mo died version of the former pro cedure of

Berkeley describ ed by Perlman and Asaro Our results can b e directly compared

to published Berkeley and Jerusalem data since we calibrated our Bonn standard

against the Berkeley p ottery standard Now we have ab out samples of Bronze

Age p ottery of Greece in our data bank Some more samples are still awaiting

analysis and the work will continue

First the principles of ceramic provenancing by elemental analysis will shortly b e

summarized p ointing out esp ecially what progresses have b een made in interpretation

of the data As long as the clay the raw material for pro duction was not traded

the elemental pattern measurable in ceramics is characteristic for a certain production

series of a p ottery workshop at a certain place Once this pattern is known e g by

measuring socalled reference material that is material which is pro duced lo cally with a

high probability all pieces of unknown pro duction place having this elemental pattern

can b e assigned to that origin

As members of a pro duction series we dene simply all pieces which analyse the

same This is a reductionist approach as Pollard and Heron wrote p but

it has the advantage that it is free of any p ossibly misleading interpretations and if

used as a rst working hypothesis results in given facts esp ecially if stringent rules

for chemical similarity are applied This similarity in many elemental concentrations

is an indication that the ancient p otters did not change anything during the manufac

turing pro cess or at least to b e precise that no change o ccurred which inuenced the

concentrations of the elements measured If a new clay or clay mixture was used or if

a new clay renement technique was applied a new pattern of elemental values may

show up a second pattern p ointing also to this pro duction place is pro duced Thus

there may b e many dierent patterns for each pro duction place Therefore the nding

of a dierent pattern in p ottery from the same site do es not necessarily mean that we

are dealing with pro ducts pro duced somewhere else we might have found just a second

pro duction series at the same place This is often exp erienced if material of dierent

type is analysed which was manufactured probably on purp ose with a dierent recip e

For instance in Roman vessels pro duced contemporaneously ab out years ago in

the same p ottery workshop in Bonn we detected dierent patternsMommsen et

al o ccurring in dierent vessel types We b elieve that the ancient p otters of

that workshop in Bonn used dierent recip es in the pro duction of these vessels they

already optimized their pro cedures to satisfy quality or market demands But with

certainty b oth patterns are characteristic for the pro duction in Bonn

One of the main problems in provenancing p ottery is the availability of reference

material since for these control samples the pro duction place has to b e known b efore

hand Rather secure are kiln wasters found in the vicinity of a p otters kiln But during

the Mycenaean p erio d only very few p otters kilns have b een detected on the Greek

Mainland and these kilns only rarely yielded misred pieces In such a situation one

has to rely on archaeological results which might introduce a bias from the b eginning

Therefore only the analysis of a very large number of samples will allow reaching well

founded conclusions

Imp ortance of precision of measurement and b est relative t

Another crucial p oint in provenancing is the quality of the measurements p erformed In

order to classify p ottery by its elemental comp osition as many elements as p ossible have

to b e measured with a high precision which is a dicult task for trace elements That

go o d measurements with high precision are needed is not astonishing since all ceramic

material contains more or less similar elemental abundances We will demonstrate

this on a newly detected group of sherds representing a pro duction series from a

p ottery workshop or district probably in Eastern Lo cris and quite p ossibly even in

the coastal settlement mound called Pyrgos Livanates which can b e identied with

Kynos according to Homer the home of Ajax Dakoronia a The group contains

to o all the samples from pictorial craters depicting Mycenaean war ships see b elow

In Tab the average concentrations M in p ercent or ppm are listed together with

the spreads sigma ro ot mean square deviations The third column expresses the

spreads in p ercent of the average values As can b e seen for many elements spreads of

less than are obtained As example for an element measured with such a precision

Fig shows the single raw data of La for the sherds of this group Per denition

ab out of all the values lie inside a range of M  sigma There is only one outlying

sample which for now is assigned to this group b ecause all the other elemental values

t but may b e found to b e a member of a subgroup if more samples are analysed

The discrimination can b e brought into still a b etter fo cus by a b est relative t

of the data This is similar to the consideration of elemental concentration ratios

instead of the concentrations directly and corresp onds to the reduction of the degrees

of freedom by which is not harmful b ecause of the many elements measured The

eect of such a b est relative t to the data of the group Lo crisl is shown in the

following columns of Tab and for the element La in Fig The average values M are

more or less unchanged but the spread values decreased enhancing the p ossibilities of

discrimination The reason for this is that on the one hand all exp erimental errors

lowering or raising all the values by constant amounts due for example to weighing

errors or to neutron ux inhomogeneities cancel and that on the other hand dilution

eects of the clays by additions of slightly varying amount of e g sand are corrected

at the same time Now we p erform always such a b est relative t during the group

formation to obtain sharp er patterns

In Fig the distribution of all the La values in our Greek databank is shown

together with the distributions of the group Lo crisl and of the most dominant group

in Greece called MycenaeBerbati The La values in the Greek p ottery analysed so far

cover only a rather small concentration range from ab out ppm ppm A similar

small range is found for many other elements To discriminate dierent pro duction

series the large number of measured elements helps but a precision of measurements

as high as p ossible is needed to o as demonstrated by the example of the La data

Groups of samples with similar comp osition are determined by our own search

metho d which is a lter metho d sorting out from a data bank of concentration values

all samples with similar comp osition A new similarity measure which we call mo died

Mahalanobis distance p erforms this task very eectivelyBeier and Mommsen

It is able to consider not only individual exp erimental errors and correlations but

corrects also for the constant shifts of the data so that diluted samples are not missed

during the ltering

Choice of samples from and NAA results

In general analytical data of Bronze Age and esp ecially Mycenaean ceramics from

the central part of Greece are rare Only a few analyses done by the Fitch Lab o

ratory of the British School at Athens existJones Our sampled vessels

and sherds which cover the time p erio d MHLHI I IC originate from excavations con

ducted by the th Ephorate of Antiquities under the direction of Ph Dakoronia

in Eastern Lo cris AchaiaPhthiotis and also in Southwestern Thessaly The bulk

of the samples however is Late Mycenaean in date and comes either from cham

b er tombs LivanatesKokkinonyzes LivanatesRema Pharmaki KolakaAgios Ioannis

MegaplatanosSventza Spartia TraganaDakoronia

b or the settlement site of LivanatesKynos in Eastern Lo cris Excavations in the

NW part of the hill of LivanatesKynos since yielded an extensive building com

plex of the LHI I IC p erio d which evidently was used for storaging purp oses In addition

to these storage facilities this area during LHI I IC Middle was also linked to p ottery

pro duction since in a p otters kiln was uncovered Apart from the architectural

remains the most conspicuous nds from the LHI I IC Middle levels at LivanatesKynos

are fragments of craters with a rich pictorial decoration consisting of ships and sea

battlesDakoronia since this is the earliest evidence for this artistic

genre in the history of Greece

The number of samples p er site and the ndsp ots themselves are listed in Tab

together with the grouping results The NAA concentration data average values and

spreads of the main groups are given in Tab As grouping result of the new set of

samples a picture already known to us emerges Besides ab out chemical loners

many dierent new groups some with only a few members can b e distinguished We

will not discuss all these sometimes still quite small groups The total number of

Bronze Age Greek groups distinguishable until now in our databank increased to over

but due to the lack of go o d reference material only a very limited number at the

time b eing only can b e assigned with high probability to denite pro duction places

Since the samples from Central Greece under investigation include several misred

pieces and sherds from the area of the kiln at LivanatesKynos we exp ected to nd

new groups assignable to that place In fact the hitherto unknown main group named

Lo crisl of pieces already used as example ab ove could b e added to the list of lo cated

patterns It is present not only in Kynos itself but traceable in several other sites in

Eastern Lo cris thus indicating a more than only lo cally limited imp ortance of this

workshop

The other smaller groups Lo crism and Lo crisn have wasters as members to o so

we assume that they are lo cally pro duced as well representing a dierent workshop or

a dierent pro duction series of the same workshop Compared to the group Lo crisl

the group Lo crism Lo crisn has much higher Co Fe and Ni La and Rb values

The main new groups Lo crisl and Lo crism having more than members are

depicted in Fig together with formerly detected patterns of Bronze Age Greece of

presumingly known lo cation Shown in Fig a is a result of a discriminant analysis of

samples assuming groups Taking only the samples of the cluster of groups in

Fig a the result Fig b is obtained As can b e seen the new Lo cris groups are well

separated from the other groups Still overlapping groups in the plots of discriminant

functions W and W are resolved in diagrams of the higher order functions The group

KnossosPhaistos KP published by Mommsen et al can b e distinguished

from the group Lo crisl mainly by its lower Hf and higher Ni values group KP also

tentatively assigned to Crete has lower Cs and Sc concentration values

Discussion

The dominant group in our set of vessels and sherds of Central Greece Lo crisl is of

sp ecial interest since all of the sherds of craters depicting ships b elong to it This

means that we are dealing with a distinct school of Pictorial Vase Painting which

was lo cated in Eastern Lo cris as suggested by our archaeometric data Up til now

the only workshop of Pictorial Vase Painting in Greece which was dened by nds of

wastersAkerstrom see b elow is lo cated in the Argolid near Mycenae exp orting

its pro ducts as could b e archaeometrically proven to Cyprus and the Levant Another

workshop known to have pro duced pictorial craters during that phase was situated most

probably in Kouklia in Cyprus as NAA data from Berkeley showKarageorghis

More archaeometric results concerning these high quality pictorial vessels should b e

obtained

Another noteworthy result concerns the analysis of mattpainted vessels mostly

b eaked jugs b elonging to the socalled Deltab eta type which is characteristic for the

Middle HelladicMH Period in Pho cis AchaiaPhthiotis and parts of ThessalyWace

and Thompson pp Maran a pp g

cf pls Although found at dierent nd sp ots in Southwest

ern Thessaly and the Sp erchios Valley they apparently are all pro duced at the same

workshop and therefore p oint to exchange networks op erating in these regions during

the MH Moreover since b eaked jugs with virtually identical features are found as far

south as Kirra in the region of Delphi the question arises whether these mattpainted

jugs were even exchanged over the high mountain barrier separating the Sp erchios Val

ley from Southern Pho cis To answer these questions samples of these jugs from Kirra

thanks to the courtesy of Dr D Skorda were included in the program in the

analysis of which is still underway In the absence of reference material the p oint of

origin of the Deltab eta type cannot b e established and thus the direction of the

exchange whether from north to south or vice versa is not certain However it has

to b e p ointed out that the p ossibility emerges that during the MH Period in Cen

tral Greece there existed p ottery pro duction centres distributing their pro ducts over

a wide area a situation comparable to the Aiginetan workshops pro ducing the Gold

Mica FabricsMaran b pp

Meanwhile not surprising to us is the presence of a large group of sherds MB

Lo cris having the well known pattern named MycenaeBerbatiMBMommsen et al

At many Greek sites this set of concentration data is detected in sherds spanning

the time p erio d from the MH to the LH I I IC phase and even to Corinthian sherds

of the classical p erio dBonn unpubl data This pattern is tentatively assigned to a

workshop in the Argolid near Mycenae b ecause of several wasters measured by Perlman

and Asaro having this pattern and found in the excavations of a Mycenaean kiln site

in BerbatiAkerstrom which also pro duced the pictorial vases mentioned ab ove

NAA detects the MB pattern to o in most of the pieces exp orted to the Eastern

Mediterranean and even in one sherd brought to Spain The NAA result therefore

here to o favors an assignment of the sherds found in Lo cris to an Argive origin

For most of the sherds of this group an imp ort to Lo cris would not b e contradicted

on archaeological grounds But a misred or secondary red vitried sherd of a large

pithos or even of a ro of tile from the kiln at LivanatesKynos a piece normally to b e

rated as lo cal has also pattern MB Furthermore a few members of this new set of

vessels e g mono chrome red burnished piriform jars do not have any comparison in

the Argolid This is mentioned since in a recent investigation of Bo eotian Late Bronze

Age ceramics by Tomlinson discussing old NAA results of the Berkeley group

the pattern MB was also found in p ottery which by archaeological reasoning was not

regarded as of Argive origin Further work on dierent ware types and from dierent

regions has to b e done to clarify this still op en question concerning the provenance of

the overabundant MB pattern in Greece The o ccurrence of a pattern which is similar

in so many elements at two dierent pro duction places is very improbable But a near

similarity of a pattern except for the element As normally not considered in provenance

studies has b een already detected in typical vessels from Troia which analysed like the

MB materialKnackeLoy and our own unpublished data We are eager to obtain

the results of the many samples taken already at Theb es due to the courtesy of Dr

Aravantinos to further investigate this question

Acknowledgement We thank the Greek Ministry of Culture and esp ecially Dr Yanis

Tsedakis for the p ermission to sample the material and for the supp ort of our work

We also wish to thank all the Greek colleagues and archaeologists who were willing

to supply samples for our pro ject The help of the sta of the reactor in Geesthacht

is thankfully acknowledged This pro ject is funded partly by the German Federal

Minister of Education Research Science and TechnologyBMBF under contract No

POBON

References

Akerstrom A Berbati The Pictorial Pottery Astroms Forlag Sto ckholm

Beier Th and Mommsen H Mo died Mahalanobis Filters for grouping

p ottery by chemical comp osition Archaeometry

Dakoronia Ph Arch Deltion Chron

Dakoronia Ph Arch Deltion Chron

Dakoronia Ph Warships on sherds of LHI I IC kraters from Kynos in Tzalas

Hed TROPIS I I nd Int Symp osium on Ship Construction in Antiquity

Delphi

Dakoronia Ph Arch Deltion Chron

Dakoronia Ph Arch Deltion Chron

Dakoronia Ph Arch Deltion Chron in print

Dakoronia Ph a Homeric towns in East Lo cris Problems of identication

Hesp eria

Dakoronia Ph b Arch Deltion Chron in print

Dakoronia Ph Kynos eet in Tzalas Hed TROPIS IV th Int

Symp osium on Ship Construction in Antiquity Athens

Jones R E Greek and Cypriot Pottery The British School at Athens Fitch

Lab oratory Occasional Paper

Karageorghis V Asaro F and Perlman I Concerning two Mycenaean sherds

from Kouklia Palaepaphos Cyprus Archaologischer Anzeiger

KnackeLoy O Isotop engeo chemische chemische und p etrographische Un

tersuchungen zur Herkunftsb estimmung der bronzezeitlichen Keramik von Troia

Heidelb erger Geowissenschaftliche Abhandlungen Universitat Heidelb erg Band

Maran J a PevkakiaMagula I I I Die mittlere Bronzezeit Hab elt GmbH Bonn

Maran J b Kiapha Thiti I I Ergebnisse der Ausgrabungen Jt v Chr

Keramik und Kleinfunde Marburger WinckelmannProgramm

Mommsen H Beier Th Heimermann D KesselringPoth L Gechter M and

Kaiser M Neutron Activation Analysis of sherds from Roman p ottery

st

workshops in Bonna in Europ ean workshop on archaeological ceramics

Burragato F Grub essi O and Lazzarini L eds Rom pp

Mommsen H Beier Th Hein A Ittameier D and Podzuweit Ch Ceramic

pro duction and distribution in Bronze Age settlements in Greece status rep ort

on NAA results in The Ceramics Cultural Heritage P Vincenzini ed Techna

Faenza

Perlman I and Asaro F Pottery Analysis by Neutron Activation Archaeom

etry

Pollard A M and Heron C Archaeological Chemistry The Royal So ciety of

Chemistry Cambridge

Tomlinson J E Statistical analysis of NAA data on Mycenaean p ottery from

Gla Theb es Eutresis Kallithea and Tanagra in Bo eotia preprint Dept Chem

University of Manchester

Wace A J B and Thompson M S Prehistoric Thessaly Cambridge Univer

sity Press

List of Tables

Tab Concentrations of elements averages M and spreads sigma in ppm if not

indicated otherwise The average statistical error of a single measurement is given for

comparison

Tab Overview of grouping results of sherds from Eastern Lo cris Phthiotis and

Southwestern Thessaly distribution of the sherds into the NAA groups Ll Lo crisl

Lm Lo crism Ln Lo crisn Phth MH mattpainted jugs MB Mycenae

Berbati s chemically ungroup ed single

Tab Concentrations of elements of the main groups of Eastern Lo cris Phthiotis

and Southwestern Thessaly and of the group MycenaeBerbatiMB averages M and

spreads sigma in ppm if not indicated otherwise corrected for dilution

List of Figures

Fig Distribution of La concentration values in ppm raw NAA data of group Lo crisl

of samples The average La concentration and the spread is given

Fig Distribution of La concentration values in ppm as in Fig NAA data of group

Lo crisl after b est relative t showing the decrease of the spread

Fig Distribution of La concentration values raw NAA data of total Greek databank

samples and of the groups Lo crisl samples and MycenaeBerbati

samples in ppm

Fig Discriminant analysis of a b samples of Greek Bronze Age Pottery

assuming a b groups Ll Lo crisl Lm Lo crism a only MB Mycenae

Berbati TA TirynsAsine Aig Aigina a only KP KnossosPhaistos

Me Melos a only Plotted are the discriminant functions W and W which

cover in a and and in b and of the b etween group variance

The ellipses drawn are the  b oundaries of the groups 14

12 25.7 +/- 2.4, 9.4 % samples 10

8

6

4

2

0 0 5 10 15 20 25 30 35 40 45 50

La ppm

Figure Distribution of La concentration values in ppm raw NAA data of group

Lo crisl of samples The average La concentration and the spread is given

14 25.4 +/- 1.5, 5.8 %

12 samples 10

8

6

4

2

0 0 5 10 15 20 25 30 35 40 45 50

La ppm

Figure Distribution of La concentration values in ppm as in Fig NAA data of

group Lo crisl after b est relative t showing the decrease of the spread

Table Concentrations of elements averages M and spreads sigma in ppm if not

indicated otherwise The average statistical error of a single measurement is given for

comparison

group Lo crisl group Lo crisl aver

samples samples stat

raw data after b est rel t error

M  sigma M  sigma

As  

Ba  

Ca  

Ce  

  Co

Cr  

Cs  

Eu  

  Fe

Ga  

Hf  

K  

La  

Lu  

Na  

Nd  

Ni  

Rb  

Sb  

Sc  

Sm  

Ta  

  Tb

Th  

  Ti

  U

W  

Yb  

Zn  

Zr   sums MB distribution jugs s Thessaly MB mattpainted MH Phth Southwestern Phth and Lo crisn Phthiotis Ln other groups NAA groups Lo cris Lo crism Eastern from Lm Ln sherds single Lo crisl Lm of Ll Ll results groups NAA grouping of the and SWThessaly into chamber tombs Kynos Overview of sherds Lo cris sherds no the archaeological groups

Table Eastern Atalanti Kolaka Livanates Megaplatanos Tragana Phthiotis and Achinos Neo Monastiri Perivoli Sp ercheiados total in NAA groups MycenaeBerbati s chemically ungroup ed of in  samples MBtotal M in  samples Southwestern Thessaly and of the MBLo cris M in  indicated otherwise corrected for dilution samples MHjugs M of Eastern Lo cris Phthiotis and in  sigma in ppm if not samples Lo crisn M spreads in  samples Lo crism M elements of the main groups in  samples Lo crisl M Concentrations of MycenaeBerbatiMB averages M and

Table As group Ba Ca Ce Co Cr Cs Eu Fe Ga Hf K La Lu Na Nd Ni Rb Sb Sc Sm Ta Tb Th Ti U W Yb Zn Zr El 225 all data 200 MB

samples 175 Locris-l 150 125 100 75 50 25 0 0 10 20 30 40 50 60 70

La ppm

Figure Distribution of La concentration values in ppm raw NAA data of total Greek

databank samples and of the groups Lo crisl samples and MycenaeBerbati

samples

L-m 50 MB 40 KP-2 KP-2 L-l W 2 (16.06 %) W 2 (24.77 %) 45 30 KP-1 L-l Aig MB 20 TA TA Me 40 KP-1

100 110 120 130 -10 -5 0

W 1 (71.22 %) W 1 (60.65 %)

Figure Discriminant analysis of a b samples of Greek Bronze Age Pottery

assuming a b groups Ll Lo crisl Lm Lo crism a only MB Mycenae

Berbati TA TirynsAsine Aig Aigina a only KP KnossosPhaistos

Me Melos a only Plotted are the discriminant functions W and W which

cover in a and and in b and of the b etween group variance

The ellipses drawn are the  b oundaries of the groups