Vous êtes sur la page 1sur 36

Paléorient

Rethinking Halaf Chronologies


Stuart Campbell

Citer ce document / Cite this document :

Campbell Stuart. Rethinking Halaf Chronologies. In: Paléorient, 2007, vol. 33, n°1. pp. 103-136;

doi : 10.3406/paleo.2007.5209

http://www.persee.fr/doc/paleo_0153-9345_2007_num_33_1_5209

Document généré le 26/01/2017


Abstract
Résumé: Les archéologues sont constamment confrontés à des problèmes chronologiques, qu’ils
soient ou non le thème de leur recherche. Les chronologies sont utilisées pour ordonner le passé
et offrir un cadre permettant de discuter les changements sociaux observés. Devoir reprendre et
affiner les structures chronologiques au fur et à mesure que de nouvelles données apparaissent
est une évidence; c’est l’un des objectifs de cet article. Néanmoins, les chronologies sont aussi
des constructions artificielles créées par une compréhension qui reste partielle du passé faisant
appel à une série de propositions, souvent mal étayées. Même des termes aussi employés que
Hassuna, Samarra, Halaf, Obeid formatent notre appréhension du passé. Il devient alors
nécessaire d’élucider et de remettre en question les hypothèses et les modèles sur lesquels
reposent les chronologies. Cet article traite des 7e-5e millénaires cal. BC en Mésopotamie
septentrionale, période qui traditionnellement voit se succéder le Hassuna/ Samarra, le Halaf qui
en constitue le centre et l’Obeid. Il se focalise sur les idées sous-jacentes et les distorsions qui ont
affecté nos précédentes séquences chronologiques, en insistant sur les façons dont notre
connaissance des chronologies passées a pris forme puis a été utilisée. Une nouvelle approche
de la chronologie de cette période, moins rigide, est proposée ici qui prend en compte les données
les plus récentes et qui rend plus explicite le lien établi entre chronologie et interprétation des
changements sociaux.

Résumé
Résumé: Les archéologues sont constamment confrontés à des problèmes chronologiques, qu’ils
soient ou non le thème de leur recherche. Les chronologies sont utilisées pour ordonner le passé
et offrir un cadre permettant de discuter les changements sociaux observés. Devoir reprendre et
affiner les structures chronologiques au fur et à mesure que de nouvelles données apparaissent
est une évidence; c’est l’un des objectifs de cet article. Néanmoins, les chronologies sont aussi
des constructions artificielles créées par une compréhension qui reste partielle du passé faisant
appel à une série de propositions, souvent mal étayées. Même des termes aussi employés que
Hassuna, Samarra, Halaf, Obeid formatent notre appréhension du passé. Il devient alors
nécessaire d’élucider et de remettre en question les hypothèses et les modèles sur lesquels
reposent les chronologies. Cet article traite des 7e-5e millénaires cal. BC en Mésopotamie
septentrionale, période qui traditionnellement voit se succéder le Hassuna/ Samarra, le Halaf qui
en constitue le centre et l’Obeid. Il se focalise sur les idées sous-jacentes et les distorsions qui ont
affecté nos précédentes séquences chronologiques, en insistant sur les façons dont notre
connaissance des chronologies passées a pris forme puis a été utilisée. Une nouvelle approche
de la chronologie de cette période, moins rigide, est proposée ici qui prend en compte les données
les plus récentes et qui rend plus explicite le lien établi entre chronologie et interprétation des
changements sociaux.
CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 103

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
RETHINKING HALAF CHRONOLOGIES
S. CAMPBELL

Abstract: Archaeologists constantly work with chronologies, whether it is the explicit focus of study or not. Chronologies are used
to order the past and to provide a framework within which social change can be discussed. It is clearly necessary to review and refine
the chronological structures as new information becomes available, and that is one of the purposes of this article. Chronologies are
also, however, artificial constructs, created by a partial understanding of the past and incorporating a range of assumptions, often
unstated and unintended. Even the basic terms with which we use (Hassuna, Samarra, Halaf, Ubaid and so on) shape our view of the
past. It is also necessary, therefore, to understand and question the assumptions and models on which chronologies are based. This
article focuses on the 7th to 5th millennia cal. BC in north Mesopotamia, a period that, in traditional terms, has the Halaf at its
centre, Hassuna/Samarra earlier, and Ubaid later. It draws attention to the underlying assumptions and distortions behind our previous
chronological models, with explicit focus on how our understanding of past chronologies has been created and employed. A new,
more flexible approach to the chronology of this period will be proposed, which incorporates the most recent data and makes the link
between chronology and the interpretation of past social change more explicit.

Résumé : Les archéologues sont constamment confrontés à des problèmes chronologiques, qu’ils soient ou non le thème de leur
recherche. Les chronologies sont utilisées pour ordonner le passé et offrir un cadre permettant de discuter les changements sociaux
observés. Devoir reprendre et affiner les structures chronologiques au fur et à mesure que de nouvelles données apparaissent est une
évidence ; c’est l’un des objectifs de cet article. Néanmoins, les chronologies sont aussi des constructions artificielles créées par une
compréhension qui reste partielle du passé faisant appel à une série de propositions, souvent mal étayées. Même des termes aussi
employés que Hassuna, Samarra, Halaf, Obeid formatent notre appréhension du passé. Il devient alors nécessaire d’élucider et de
remettre en question les hypothèses et les modèles sur lesquels reposent les chronologies. Cet article traite des 7e-5e millénaires cal. BC
en Mésopotamie septentrionale, période qui traditionnellement voit se succéder le Hassuna/Samarra, le Halaf qui en constitue le
centre et l’Obeid. Il se focalise sur les idées sous-jacentes et les distorsions qui ont affecté nos précédentes séquences chronologiques,
en insistant sur les façons dont notre connaissance des chronologies passées a pris forme puis a été utilisée. Une nouvelle approche
de la chronologie de cette période, moins rigide, est proposée ici qui prend en compte les données les plus récentes et qui rend plus
explicite le lien établi entre chronologie et interprétation des changements sociaux.

Keywords: Halaf, North Mesopotamia, Chronology, Neolithic, Radiocarbon, Social change.


Mots-clés : Halaf, Mésopotamie septentrionale, Chronologie, Néolithique, Radiocarbone, Changements sociaux.

Chronology inevitably has a high profile in prehistoric but also whether they remain appropriate to the nature of the
archaeology. Its importance, however, needs to be seen not just questions to which we seek answers.
as a way of organizing the past but also as part of the process In very broad terms, this article addresses the classic
of understanding it. As the ways in which we try to understand problem of Halaf chronology. It attempts, however, to examine
the past have changed enormously over recent decades, it is it from a new point of view. The topic of Halaf chronology
reasonable to examine not only the dates we assign to past events is one with which I have tinkered intermittently for twenty
but also the nature of the chronologies we construct—what they years and I have gradually become dissatisfied with previous
are dating and how they are formed. We should be asking not approaches, my own included, because the chronologies they
only whether chronologies can be more accurate and precise create do not always seem the most useful to help us understand

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007 Manuscrit reçu le 11 décembre 2006, accepté le 26 juin 2007
CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 104

104 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
the changing society during this period. This dissatis- PREVIOUS APPROACHES
faction is mirrored by others both in print 1 and in numerous
conversations or conference papers. This article is not targeted In recent decades there have been many contributions to
narrowly on the Halaf, however, for reasons that I hope will the debate on Halaf chronology. Schemes have been revised
become clear. The time frame runs both earlier and later, and fine-tuned 2 and new schemes proposed 3. More localized
with a general time bracket between ca 6500 cal. BC and regional chronologies have been devised and inter-regional
4500 cal. BC, but has the traditional Halaf roughly at its centre. synchronizations developed. The challenges in understanding,
The geographical focus is broadly on north Mesopotamia. synthesizing or revising the intricacies of the chronologies of
The first section of this article will seek to understand why the period have grown to the point at which the phrase “terri-
previous approaches to the chronology of this period seem fying complexity of Halaf chronology” 4 has real meaning.
unsatisfactory, and can actually misdirect research. The This article is not intended to engage with these complexities.
second section will attempt to provide an alternative approach. Indeed, as one of those who has contributed to these complexi-
Traditional studies have emphasized ceramic sequences ties in the past, I have a suspicion that the complexities arise
and ceramics will eventually have a prominent place in the out of the nature of the approaches that have been used and
discussion here as well. There is an issue, however, in that any attempt to understand and refine the traditional
foregrounding ceramics in the initial construction of a approaches may carry a risk of ever-increasing complexity and
chronology. The ultimate aim in refining our chronological obscurification. Furthermore, it can be argued that many
understanding must be to allow us better to study social previous approaches to chronology building have been based
change. Change in the style of ceramics, although obviously on assumptions that, although they appear reasonable, can in
happening through time, is not directly the result of temporal fact be misleading.
change. Rather it is a result of the very social changes that Although it has certainly been criticized, the traditional
are the object of study. There is a dangerous element of circu- chronological model still shapes the way that archaeologists
larity if a chronological structure that is defined through approach northern Mesopotamia. Terms such as Hassuna,
change in material culture is then used as the framework within Samarra, Halaf and Ubaid still dominate and the chronology
which to explain social changes reflected in that material is almost entirely based on ceramics. Attempts to develop and
culture. The aim here is to attempt to distance the establish- refine dating schemes have generally focused on how these
ment of a chronological structure from any specific aspects of units are best sub-divided and what dates can be assigned to
material culture. The initial chronological structure will be their sub-divisions. Although there has been an increasing
approached mainly by using absolute dates and stratigraphy. awareness that these are rather arbitrary divisions, their use
This also has the potentially useful effect of highlighting how still encourages the implicit use of culture-historical models
thin some our chronological knowledge actually is during in which the entity labeled, for example, the Halaf, is treated
this time period. The final step will be to look at how as an historical actor. The acknowledgement of something
changes in material culture might be mapped onto the chro- called the Halaf generates questions such as “how long did it
nological structure, and here ceramics will be used as the last?” and “what was its origin?”. Its existence generates new
example. Inevitably, this will not result in clear-cut conclu- labels, such as Post-Halaf 5 and Proto-Halaf 6, that maintain
sions and we should also expect some of the conclusions to its centrality in the discussion.
be clarifications of changes already identified through more It is critical to recognize that these terms have their origins
traditional approaches. not in the distant past that we seek to understand but in the
history of 20th-century archaeology 7. Ultimately the traditio-
nal framework results from a remarkably small set of classic

2. e.g. WATKINS and CAMPBELL, 1987; COPELAND and HOURS, 1987;


MIYAKE, 1998; CRUELLS and NIEUWENHUYSE, 2004; CRUELLS, 2006b.
3. e.g. CAMPBELL, 1992; HIJARA, 1997.
4. CRUELLS and NIEUWENHUYSE, 2004: 49.
5. CAMPBELL et al., 1999, although it may be noted that that term has
since been abandoned in the Domuztepe ceramic phasing.
6. CRUELLS and NIEUWENHUYSE, 2004.
1. e.g. MIYAKE, 1998: 179; BERNBECK and POLLOCK, 2003: 36-37. 7. CAMPBELL, 1998 and 1999.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 105

Rethinking Halaf Chronologies 105

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
Key Type Sites Tepe Gawra in the 1930s 12. Finally, the Hassuna was added
following the excavation of Tell Hassuna in 1943-1944 13. This
Tell Arpachiyah

Tepe Gawra
Tell Hassuna

might suggest that the major chronological zones were


successfully defined in the early decades of prehistoric
archaeology in the Near East and that the first excavations in
each region examined sites that provided the most appropriate
definitions of a specific ceramic style.
Of course, the early excavators were not so fortunate.

Traditional Chronological Divisions


Figure 1 shows that the chronological divisions accord exactly
HASSUNA with the gaps in the excavated sequences at the key sites. The
significant changes in ceramic style that seem to define the
main periods are a product of the selection of sites that were
first excavated with the changes corresponding exactly to the
lacunae in the original site sequences. More recent work has
tended to highlight the gradual transitions that occur between
the earlier traditions and the Halaf 14 or between the Halaf and
HALAF
the Ubaid 15. Since our traditional chronological structure is
created by the arbitrary choice of early excavated sites then
questions that are generated by that structure start to lose their
apparent meanings. Thus the answer to the popular question
“when did the Halaf start as a cultural phenomenon?” 16 should
probably be “in the 20th-century AD”. Equally, the labels that
UBAID are generated by reference to the traditional periods, such as
Proto-Halaf or Post-Halaf or Halaf-Ubaid Transition, need to
be treated with care. At best, they are labels of convenience
and, at worst, draw us into conclusions that confuse the
sequence of archaeological discovery with real patterns from
Fig. 1: The origins of the traditional chronological sequence in north prehistory.
Iraq, showing the coincidence between gaps in the original excavated The use of the traditional terminology as the basis for our
sequences and the boundaries in the traditional chronology. chronology also carries a danger of drawing us into unsubtle
use of culture history models. The Halaf has frequently been
type sites, mainly in Iraq but fundamentally influencing the discussed as though it is a coherent cultural group. Sometimes
chronological divisions across northern Mesopotamia. The this has been done explicitly 17 but often it has been implicit.
sequences from the type sites have been superseded by more The term is used to describe an entity that is seen to have
comprehensive sequences excavated using more modern tech- common and distinctive characteristics, and is treated almost
niques but these original type sites still largely determine the as an historical actor in its own right. Although doubt has also
regional and chronological structure. The major ceramic frequently been cast on whether the Halaf has any internal
groupings were established by 1945 and have not been sub- coherence, as long as the term is used as a mainstay of our
stantially altered since. The Halaf was named after the type chronological structure, it is almost impossible to avoid this
site of Tell Halaf, excavated between 1911 and 1929 8 and the happening to some extent. Terminology matters and creates
definition elaborated through M.E.L. Mallowan’s excavations
at Arpachiyah in 1933 9. The definition of the Samarran came 11. Summarized in OATES, 1960.
12. TOBLER, 1950.
from E. Hertzfeld’s excavations at Samarra 10. The Ubaid was 13. LLOYD and SAFAR, 1945.
defined in southern Iraq in the 1920s 11 and in the north at 14. e.g. AKKERMANS, 1993 and 2000; CAMPBELL, 1998; CRUELLS and
NIEUWENHUYSE, 2004.
15. DAVIDSON, 1977; BRENIQUET, 1996.
8. OPPENHEIM VON, 1943. 16. e.g. BOGOSLAVSKAJA, 1972; MELLAART, 1978: 144; MUNCHAEV and
9. MALLOWAN and ROSE, 1935. MERPERT, 1981: 282; WATKINS and CAMPBELL, 1987; AKKERMANS, 1993: 293.
10. HERTZFELD, 1930. 17. e.g. WATKINS and CAMPBELL, 1987: 428-429.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 106

106 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
its own explanatory framework. An obvious solution to this model based on his own excavations at Arpachiyah 20. Hijara’s
might be to suggest a different terminology, better reflecting model has only had limited use and there are certainly some
our present understanding. This is not the intention in this potential problems with it 21. The original tripartite division
article for two reasons. Firstly, a new terminology would has continued to be influential and has been used with modi-
create confusion unless universally adopted, and the use of fications as an appropriate structure for the Halaf in several
terms such as the Halaf does at least have the advantage of more recent discussions 22. It remains embedded in the recent
familiarity. Secondly, any new terminology will tend to proposal of a six-stage chronology for the Halaf 23.
recreate exactly the same problems. Although we certainly T. Watkins and S. Campbell suggested a slightly different
have more data than we had when the initial periodisation emphasis, stressing the distinction between the original Early
emerged, it seems very clear that that data will prove to be and Middle Halaf and reducing the difference between Middle
inadequate in the future. and Late 24. This was developed further with the proposal of
The many attempts to construct and refine chronologies a sequence consisting of Halaf Ia (earlier than the original
for this period in northern Mesopotamia can perhaps be consi- Early Halaf and identified in the late 1980s at Tell Sabi
dered under three headings. The first approach might be Abyad 25 and NJP72 26) and Halaf Ib (the old Early Halaf),
termed “regional chronologies” and has sought to integrate followed by Halaf IIa (old Middle Halaf) and Halaf IIb (old
northern Mesopotamia within a single scheme. These chrono- Late Halaf) 27. For the purposes of the discussion later in this
logies have formed the traditional chronological structure for article, it is worth elaborating slightly on the definition of these
Near Eastern prehistory and it is out of this approach that the phases, although in many respects the specific labels used for
traditional terms such as Halaf and Ubaid originate. The them could be substituted by different terminologies.
second approach might be termed “trans-regional”, where the Halaf Ia was seen as representing the formation of an Halaf
different regional chronologies are pulled together to consider tradition within a longer sequence of ceramic change. In Iraq,
the Near East as a whole. The third approach, which has gained painted pottery came to dominate during the Hassuna III phase
greater emphasis in recent years, can be termed sub-regional as a contrast to the preceding Hassuna II phase 28, which had
and seeks to define a chronology first for a small area within predominantly incised decoration and, in Syria, as a contrast
which intensive research has taken place. It is useful to to largely undecorated ceramic Neolithic traditions. The begin-
highlight some aspects of these different approaches both to nings of the Halaf were thus associated with a spread of
identify ways in which chronologies have been approached in painted pottery across the north of both Iraq and Syria. It can
the past and to emphasize some key changes in the ceramic now be seen that this was indeed a widespread phenomenon
assemblages that have been previously highlighted. It is extending into southeast Turkey. The term Proto-Halaf has
certainly not my intention to provide a comprehensive analysis been reasonably introduced rather than Hassuna III because
of detailed ceramic sequences. of the need to be less oriented to north Iraq 29. More impor-
tantly than the modification of phase names is the growing
emphasis that this is a social phenomenon, where the intro-
REGIONAL CHRONOLOGIES duction of new ways of making and decorating pottery was
an important aspect of developing complexity in social inte-
Based on his excavations at Arpachiyah, Mallowan made raction, particularly in contexts of consumption 30.
the first proposal for a chronology of the Halaf period in
20. HIJARA, 1980 and 1997; MELLAART, 1981.
1935 18. His division was threefold, dividing the Halaf into
21. GUSTAVSON-GAUBE, 1981: 78-90; WATSON and LEBLANC, 1990:
Early, Middle and Late phases. With the addition of a scatter 52-55; CAMPBELL, 1992: 64-67.
of radiocarbon dates, this remained the accepted chronology 22. e.g. MIYAKE, 1998; CRUELLS et al., 2004; CRUELLS, 2006b.
into the 1970s, when T.E. Davidson re-examined the primary 23. CRUELLS and NIEUWENHUYSE, 2004: table 2.
24. WATKINS and CAMPBELL, 1987.
data in detail and largely confirmed Mallowan’s original divi- 25. AKKERMANS, 1989, 1993 and 2000.
sions, although with the addition of a Halaf-Ubaid Transitional 26. CAMPBELL, 1996 and 1998.
phase 19. Also working in the 1970s, I. Hijara rejected Mallo- 27. Ibid., 1992.
28. Again using the terminology of CAMPBELL, 1992. In the scheme
wan’s model and proposed a more complex, multi-phase proposed there, Hassuna I broadly equates to Archaic Hassuna, Hassuna II to
Standard Hassuna and Hassuna III was seen as chronologically successive
and equating to the Samarran.
18. MALLOWAN and ROSE, 1935; also formalized in PERKINS, 1949. 29. CRUELLS and NIEUWENHUYSE, 2004.
19. DAVIDSON, 1977. 30. JONES, 2004; NIEUWENHUYSE, in press.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 107

Rethinking Halaf Chronologies 107

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
Halaf Ib was introduced to define the Early Halaf phases seen The work that underlies the grouping into larger periods is
at both Arpachiyah and at Tell Aqab. These phases had a very thorough but leaves unresolved issues. It uses the classic
distinctive assemblage dominated by straight-sided bowls. At culture groups and generally accepts them (Natufian, PPNB,
Tell Aqab these made up as much as 70% of the total assem- Halaf, Ubaid, etc.) without critical examination of their vali-
blage 31 and in Hijara’s excavations at Arpachiyah more than dity. If the traditional culture groups have failings, these
60% of the vessels from his Phases I-III are made up of these failings will translate to the conglomerated periods. If the
straight-sided bowls 32. Again, it seems that the Halaf Ib assem- Halaf and Samarran, for example, are misleadingly simple
blage should be seen as reflecting changes in patterns of concepts founded in 20th-century archaeology, their combina-
consumption. tion into Period 6 will be equally artificial. Furthermore, even
Halaf II seemed to contrast markedly with Halaf Ib at both if the traditional cultures did remain sufficiently reliable to
Arpachiyah and at Tell Aqab. Since at neither site was there provide a chronological framework, they obviously need not
any clear evidence of a significant gap in occupation, it seemed start or finish at the same times. The beginning and end of a
reasonable to see this as a significant change. The main diffe- phase that combines several regional entities will not be coor-
rence from Halaf Ib was an increase in the diversity of vessels dinated and may be very ragged indeed. Ironically, this
forms, particularly the decline of straight-sided bowls and the problem is actually less acute where we have fewer dates
introduction of a more varied bowl assemblage 33. It was because the differences in date for the start of individual
suggested that the division into Halaf IIa and IIb was a subtle periods in different regions will be less than the precision
one, only seen in a few ceramic characteristics, such as small attainable using the scattered radiocarbon dates. As the number
quantities of polychrome paint and surface manipulation, that of available dates increases, the potential definition improves
had been greatly exaggerated in previous chronologies. It was and we should expect the poorly synchronized beginning and
also suggested that the division between Halaf IIa and IIb ending of periods to become a much greater issue. It is impor-
might be regionally variable. tant to realize that this issue is inherent and may only reveal
itself gradually as regional chronologies improve.
In many ways, this approach assumes that our understand-
TRANS-REGIONAL CHRONOLOGIES ing of the big picture is superior to our understanding of the
detail; that we may have problems with some of the local and
A sustained attempt to construct an overall chronology for short-term differentiation but that our general relative chrono-
Near Eastern prehistory has been made by a group of mainly logy is satisfactory. I wish to argue that, in fact, the true
French scholars 34, which was explicitly formulated in the Atlas position is almost the reverse. I would argue that our strongest
des Sites du Proche-Orient (ASPRO) 35. There are many strong information lies in the detailed understanding of the sequences
points in their analyses, not least of which is that they have at individual sites and sub-regions over shorter periods of time.
revisited the problem on different occasions as increasing Equally, I would suggest that our larger-scale chronological
quantities of dates have become available. The later portions and regional structures are, in fact, based on very shaky pre-
of their chronological framework directly overlap with the mises and cannot be assumed to be correct. We need, therefore,
material under consideration here. Their approach, however, to work from the bottom up rather than the top down if we
exhibits significant differences from that used here. These are are to make progress in disentangling chronologies, regiona-
not simply differences of interpretation and methodology but lisation and the social processes that underlie them.
also the result of different assumptions about how we should
structure our chronological knowledge of the prehistory. The
key to the approach formulated in ASPRO is that traditional SUB-REGIONAL CHRONOLOGIES
archaeological cultures across the entire Near East are grouped
together into a succession of periods, numbered 1 to 7. The third approach focuses on local chronological se-
quences. This has now been used successfully in several cases,
including the Balikh valley 36 and the Rouj basin 37, although
31. DAVIDSON, 1977: 111. it could be argued that the Amuq sequence is one of the first
32. CAMPBELL, 1992: 65; HIJARA, 1997.
33. CAMPBELL, 1992.
34. AURENCHE et al., 1987; HOURS et al., 1994; AURENCHE et al., 2001. 36. AKKERMANS, 1993.
35. HOURS et al., 1994. 37. IWASAKI and TSUNEKI, 2003.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 108

108 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
examples 38. By focusing on local sequences, there are several types of phenomena can be quite varied. In this article, the
strengths. Typically there is a considerable depth of knowledge emphasis is on periods where there was rapid change in cera-
through survey and excavation of more than one site so the mic assemblages. It could equally be on periods that saw
sequence is relatively complete. Furthermore it is not neces- stability in ceramics or changes in lithic technology or changes
sary to make assumptions about contacts over long distances in architecture. Obviously, if the emphasis did change, the
to justify synchronicities between different sites. In general, chronology might look rather different. In some ways, this
this article is addressing larger-scale issues and critiquing a may help liberate phase-based chronologies from the spiral of
local chronological approach is not the aim. Two points are ever more complex inter-regional synchronizations. If their
perhaps worth making, however. Firstly, although the local purpose becomes more explicitly limited to organizing and
density of information is generally much greater, the se- describing rather than explaining the past, the importance of
quences almost always rely on an implicit assumption that the accounting for every fine degree of variation decreases.
complete sequence is known. In fact, gaps in occupation at For a phenomena-based approach to chronology to be
sites and phases unrecognized in survey are probably much successful, chronologies need to be built in a slightly different
more common than often assumed. Secondly, because of the way than in previous approaches. The traditional approach has
desire to address larger-scale questions, local chronological been to take a set of phases derived from links that have been
sequences are generally linked into regional chronologies and made using relative dating techniques, partly stratigraphic but
thus ultimately interpreted in a context that shares many of particularly based on stylistic and technological parallels.
the problems raised by the regional chronological approach. Once these phases have been defined, radiocarbon dates are
then used to assign dates to boundaries that we already
“know”. These temporal boundaries are then presented as solid
lines across chronological charts. The approach used here is
TOWARDS A NEW APPROACH
to concentrate first on establishing the chronologies of indivi-
dual sites with as high a definition as possible and with limited
There seems little doubt that chronologies based on the reference to each other. Then, phases of apparent widespread
definition of phases, whether or not purely defined by ceramics change in ceramics will be examined, linking the individual
and whatever their regional scale, will persist and they do have site chronologies together explicitly as part of the process of
an important role. They allow archaeologists to place archae- investigating and interpreting these phenomena. The chrono-
ological material rapidly in a broad time frame and, for some logy suggested should not be seen as independent of the inter-
purposes, this is exactly what we wish to do. In survey, pretation of the phases of social change. Its validity is, in large
for example, the ability to place individual sherds into time part, as an interpretative device and the creation of overall
categories is critical. An important feature that allows phase- chronological schemes are pushed to perhaps exactly where
based chronologies to function in this way is that they are they should be—as the final stage of a process of interpretation
continuous. They seek to create a chronology where there are focused on how past societies functioned and interacted.
no gaps and all time periods are categorized and labeled. The discussion is deliberately concentrated on chronology
Where the traditional phase-based chronologies start to fail is within a rather loosely defined area of northern Mesopotamia
when they are used, not just to describe the past, but also as (fig. 2). There is undoubtedly further research to be done on
the means to interpret it. An alternative approach is to focus, the relations of the sites discussed with other sites in Cilicia
not on seeking to define and date entities that persist through and central Anatolia, the central and southern Levant and
time, but to use chronology to isolate particular phenomena southern Mesopotamia. It lies, however, outside the scope of
that have particular interpretational interest. This type of chro- this article. The database of radiocarbon dates that are used in
nology can be discontinuous. The only time at which inter-site this article is as comprehensive as possible. It has been assem-
and inter-regional synchronization becomes critical is when bled from various sources. At its core is a database that the
the archaeological interpretation demands it. This type of chro- author has built up over many years but also cross-checked
nology is also inherently flexible and variable. The nature of against publicly available databases 39 and supplemented in
the phenomena under study and the type of interpretation some cases by unpublished dates. Before discussing individual
sought will determine what the chronology focuses on. The
39. BÖHNER and SCHYLE, 2006; BISCHOFF, 2006; THISSEN, 2006;
38. BRAIDWOOD and BRAIDWOOD, 1960. BANADORA, n.d.; TAY, 2006.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 109

Rethinking Halaf Chronologies 109

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
Kenan Tepe
Domuztepe

Fıstıklı Höyük

Tell Kosak Shamali Tell Aqab


Tell Sabi Abyad
Tell Kurdu NJP72 Tell Arpachiyah
Tell ‘Abr
Tell Ziyadeh
Tell Yarim Tepe
Mashnaqa

Fig. 2: Sites with key chronological data.


Sites which provide key sequences with reliable absolute dates are indicated with filled circles.

sites, it is useful to examine the suitability of this database for To examine the evenness of our chronological knowledge,
studying a North Mesopotamian chronology in this period. a complete set of dates were selected without regard for cultu-
Although we may frequently complain of the quality of ral designations. All dates from northern Mesopotamia with
individual dates or the absence of groups of well stratified mean values between 8000 and 5500 BP were selected. This
dates from individual sites, the overall distribution of the dates gives a sufficiently wide bracket around the period of central
is rarely questioned. If we are to construct a chronological interest that the details of the dates that fall just within or just
framework, absolute or relative, for a lengthy period, it is outside this boundary can be reasonably ignored. This pro-
important to know whether we have significant gaps in the duces a sample of 321 dates. For simplicity, only the mean of
data. This is difficult to establish for relative dating since gaps each date will be used; since we are interested in the distri-
in occupation are archaeologically hard to establish and the bution of dates, using the centre of the probability distribution
length of a gap almost impossible to demonstrate. It is, how- of an individual date and ignoring the probability distribution
ever, more straightforward for a collection of absolute dates. itself is largely justifiable. To judge whether our sample repre-
An initial, basic question can be posed. Do we have sufficient sents an evenly distributed body of chronological knowledge,
dates to allow us adequately to resolve the chronology of the we need to compare the distribution of dates in the sample
time period? In one sense, the answer will inevitably be nega- with the radiocarbon determinations from a sample evenly
tive—more dates are always useful—but the level of negativity spaced in calendar years. Conveniently, there is a readily avai-
is also important. Are there time periods for which there is lable source of dates calculated at known intervals—the radio-
such a low density of radiocarbon dates that resolving the carbon calibration curve itself. The most recent version 40 has
chronology will be inherently problematic? Does this mean one radiocarbon date every five calendar years for the period
that the level of chronological certainly can vary through time,
depending on the density of dating information? 40. REIMER et al., 2004.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 110

110 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
under study. Again all the dates were used where the mean contribute 77 dates, sites in Syria 167 dates and sites in Turkey
falls between 8000 and 5500 BP, a total of 538 dates. Frequen- 167 dates. This is particularly significant because, as already
cies of dates per hundred years were calculated and, to make underlined, it is from the sites in Iraq that much of the tradi-
the distribution of dates directly comparable, the frequencies tional chronological framework is derived. This is compounded
of the dates derived from the curve were adjusted to normalize because the dates from Iraq are generally of poorer quality.
the data relative to the number of archaeological dates (fig. 3). Many of them were made in the early decades of radiocarbon
dating. They often have large standard deviations, some were
45
Archaeological dates
on samples kept in non-optimum storage for considerable
40
Dates at 5 calendar year intervals periods and sometimes the stratigraphic position of the radio-
carbon samples is imprecise. The traditional chronology,
35
therefore, rests on a poorly understood base. Although dates
Number of dates per 100 years

30
from further west and north may help to date developments in
25 north Iraq, this must rest on the assumptions of cultural conti-
20 nuity and synchronic change that rely on the interpretation of
15
the past that we seek to study.
10

0
THE CHRONOLOGICAL SKELETON
8000
7900
7800
7700
7600
7500
7400
7300
7200
7100
7000
6900
6800
6700
6600
6500
6400
6300
6200
6100
6000
5900
5800
5700
5600
5500

The approach that will be followed is to focus primarily on


Radiocarbon years BP
sites with excavated sequences and multiple radiocarbon dates.
Fig. 3: Comparison of the number of actual archaeological dates per The selection of sites is restricted to quite a tight group. Sites
century (radiocarbon years BP) with the number expected if there is on the periphery, such as Arjoune and Yumuktepe, have not
an even distribution of chronological knowledge (data for the latter been included in this study, although they obviously have great
from REIMER et al., 2004). relevance and good groups of dates. The dates from the Rouj
basin were not used because of their very large standard devia-
If the archaeological dates were evenly distributed through- tions. The limited selection of key sites does restrict severely
out the period, one would expect their distribution to be very the range of data but it has major advantages. It avoids any
similar to the distributions of the evenly spaced dates from emphasis on dates that may be old or poorly stratified. It places
the calibration curve. Differences between the distributions greatest emphasis on sites where our knowledge is also greatest.
should highlight time periods with more or fewer determina- The use of stratigraphic sequences is important because it makes
tions than expected. Figure 3 immediately makes it clear that maximum use of more recent approaches to calibration.
we have a very uneven selection of data. Two areas are The use of Bayesian statistics in radiocarbon chronologies
highlighted. The area in lighter gray shows the zone from has had a huge impact in recent years. The advantage that
about 7100 to 6700 BP (approximately 6100 to 5600 cal. BC). Bayesian statistics confer is that prior belief about the rela-
This directly reflects the number of dates from just four tionship of the dated samples, most frequently their stratigra-
sites—Tell Sabi Abyad, Flstlkll Höyük, Tell Kurdu and phic position, can be incorporated into the analysis and can
Domuztepe. The area in darker grey highlights a zone of produce a more exact posterior belief—the final calibrated
relatively poor data from 6300 BP onwards (approximately dates 41. The incorporation of stratigraphic information in cali-
5300 cal. BC). This is a period from which we have relatively bration not only improves the calibration process, it also makes
few dates and, for the period ca 6300 to 6100 BP, this lack it much easier to identify dates that are incompatible with their
is severe. Although this is only 200 radiocarbon years, it stratigraphic position and judgments about ill-fitting dates can
falls on a flat part of the curve and corresponds to at least be assessed probabilistically rather than subjectively. The use
300 calendrical years about which we have extremely poor of Bayesian statistics for calibration has become widespread
dating evidence. thanks to the availability of user-friendly programs. Although
Geographically, our chronological knowledge is even more
heavily skewed. Using the same sample as above, sites in Iraq 41. BUCK et al., 1991: 819; BUCK et al., 1992 and 1994.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 111

Rethinking Halaf Chronologies 111

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
there are good examples of its use in Near Eastern prehistory 42, The same general procedure was applied to all the sets of
it has not been systematically applied to the chronology of the radiocarbon dates. OxCal 4.0 was used throughout 46. The dates
period under study here. were initially screened for any that had known contamination
In almost all cases, the key dates below are relatively recent problems or were very clearly anomalous. A chronological
determinations made by major laboratories. Although there model was then constructed based on the excavator’s descrip-
may still be some issues with inter-laboratory variation, this tion. Appropriate use of the Sequence and Phase structures
is less of a problem than it is with older dates 43. Obviously, helped to define stratigraphic groupings in OxCal. A start
where older dates are referenced, inter-laboratory variability event and an end event were then added using the Boundary
is much more of a concern 44. In the overall database used statement to constrain the estimate of the beginning and end
here, there are determinations from 33 different laboratories of the sequence 47. Boundaries were not routinely used between
relevant to northern Mesopotamia while the dates from the phases within each site sequence in the diagrams presented
key sites in this discussion come from only eight. Probably at here, although they were frequently used during exploration
least as significant are issues relating to taphonomic processes. of different models. Although this additional constraint may
Samples may not be securely associated with the event they slightly improve the precision of the results, it was felt the
are claimed to date or may have arrived in their context of assumption that phases, and the contexts that constitute them,
discovery through redeposition. The sample may be from form a well constrained unit is not necessarily warranted. If
wood already old by the time it was buried. Although these the phase is indeed well defined and has several dates distri-
issues have been widely recognized, reporting of radiocarbon buted across its lifetime, there should be minimal difference
dates in Near Eastern archaeology is often inadequate. Con- in the results.
textual information is too often imprecise and the material This model was then run using OxCal 4.0. The agreement
dated is sometimes not specified or simply reported as wood index generated by OxCal was used to assess how satisfacto-
charcoal, without an identification that would aid assessment rily the radiocarbon dates fitted the stratigraphically derived
of potential “old wood” errors. model. An agreement index below 60% indicates empirically
The first step in Bayesian calibration is the creation of a that the fit between the dates and the model is equivalent to
formal chronological model for the specific site. This is a critical failing a conventional χ2 test at the 95% significance level 48.
step because the prior knowledge embodied in this model is This was calculated for both the overall model and the indi-
combined with the dates to produce the final dates. Within sites, vidual dates. Individual dates were removed from the model
the relative chronology proposed by the excavator has generally where they had very low agreement indexes and then the
been used to create the model, even when there are no direct probability of their belonging to their stratigraphic positions
stratigraphic links between phases. For example, Tell Sabi was assessed by rerunning the calibration in OxCal. In general,
Abyad has four separate stratigraphic sequences, each with dates removed in this way had less than a 10% chance of
multiple radiocarbon dates, from Operations I, II, III and IV. dating their excavated contexts, if the chronological model
Although there are no stratigraphic links between them, the site was correct, and individual examples are discussed below. The
phasing draws on other data to integrate them into one chrono- overall model was refined in this way until an agreement index
logical model 45. Although this does mean that the models used of more than 60% could be achieved for the model as a whole.
here for site chronologies sometimes still conflate chronological In a few cases, the overall model could be satisfactory even
and cultural interpretations, it is felt that this is justifiable though one or more individual dates had an agreement index
because it draws on the detailed knowledge of an excavator of of slightly less than the 60% level. In these cases, the indivi-
an individual archaeological site. Furthermore, it might also be dual dates were retained 49.
argued that basing a model on stratigraphic relationships also Most of the main sites used to construct the chronological
depends heavily on trusting the excavator’s own knowledge of framework have “good” sets of dates. In other words, they
a site; without very full publication, verification or reinterpre- come from well-stratified samples, often from short-lived
tation is often problematic. specimens and from major radiocarbon laboratories. Despite

42. e.g. PHILIP and MILLARD, 2000; CESSFORD, 2002. 46. BRONK RAMSEY, 1995 and 2001.
43. SCOTT, 2003. 47. cf. STEIER and ROM, 2000; BRONK RAMSEY, 2000.
44. AITCHISON et al., 1990. 48. BRONK RAMSEY, 1995: 428.
45. AKKERMANS et al., 2006: table 1. 49. cf. PHILIP and MILLARD, 2000.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 112

112 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
this, it is striking how many dates have been removed from but it is important to be clear that they will still be interpre-
the models in OxCal to obtain a good agreement index. From tations made as part of a process of constructing a chronolo-
the 136 dates used in the discussion of key sites, 23 dates gical narrative. Obviously different interpretations may be
(almost 17%) had a low probability of dating their apparent possible and, at times, optional interpretations will be proposed
contexts. Overall, it does seem to be the case that most errors and the alternative narratives they generate explored.
are in what M. Roaf and M. Killick termed the “respectable”
direction 50—that is, the dates seem to be earlier than the
contexts from which they originate, either due to redeposition TELL SABI ABYAD (fig. 4 and table 1)
or to the use of old wood. Nonetheless, there is a significant
number of cases where the dates are later than the context, There are 31 radiocarbon dates from Tell Sabi Abyad. Four
according to the chronological models. There may be two of these are from the north-east mound and their exact rela-
reasons that are worth taking into account because they reflect tionship with the remainder is unclear to the writer. The remain-
directly on generally unquestioned areas of archaeological ing dates can be placed in the published chronological model 53.
practice. Firstly, the chronological models may be wrong. Some of the individual dates fit badly in the wider chronolo-
Although stratigraphic analysis is a mainstay of archaeology, gical model. GrA-24219, GrA-24248, UtC-1009, GrA-26924,
and particularly Near Eastern archaeology, it often involves GrN-28244 and GrN-16801 all fit poorly in their apparent
unstated assumptions about the nature of deposition and site stratigraphic positions. Without removing them from the analy-
formation on archaeological sites. This is particularly the case sis, the agreement index of the model is well below 60% and,
where there are exposures of architecture that are judged to therefore, unacceptable. When removed and their association
be contemporary. Simply because they can be exposed at one with their apparent stratigraphic positions tested, none has a
time and drawn on a single plan does not necessarily mean probability of higher than 11.8% and the agreement index of
they were built and abandoned at the same time, and that the the overall model rises to a more acceptable level of 70%.
contents within them are contemporary. Archaeologists may Although three further dates have individual agreement indexes
need to reassess the logic that lies behind stratigraphic assump- less than 60%, these have been retained in the model presented
tions more explicitly. Secondly, we may need to question both in figure 4. If removed, they do not dramatically affect the
our understanding of site-formation processes and excavation result but the agreement rises to 90.2%. In two cases (GrA-
technique. Although results of excavations are generally stated 24219 and GrA-24248), the dates are on unidentified charcoal
in objective and clear-cut terms, the reality of excavation is and appear older than their contexts of excavation, possibly as
sometimes less clear. It is not always the case that deposits a result of old wood. In two cases, earlier dates than the context
are unmixed, with neither over-cutting or under-cutting that of discovery were obtained on carbonized seeds (GrN-16801
may introduce material into other contexts. Our understanding and GrN-2824) and redeposition might be suspected. The final
of site formation, deposition processes and artefact mobility two anomalous dates are also from carbonized seed. GrA-
is also generally assumed rather than demonstrated. 26924 is late for its attributed context in Operation II, level 2.
Each site will be discussed individually before there is any The other date from this context GrA-26925 fits the chrono-
attempt to place them relative to each other. No attempt will logical model much better, but it would also not be incompa-
be made to aggregate dates together to summarize the lengths tible with the slightly later date of GrA-26924. UtC-1009
of individual phases. The usefulness or clarity of phases, either similarly is late for its attributed context but associated with
at a regional or a site level, is open to dispute. Furthermore, another date from Operation I, level 8, which, while fitting the
the most obvious approach used elsewhere 51, the use of the overall model better, is also not incompatible with the anoma-
“summed probability distributions” in OxCal, seems possibly lous date. Although intrusive material is entirely possible, the
misleading in the current context 52. Although the chronologi- possibility remains that it is the chronological model itself that
cal models will be analyzed using OxCal 4.0, the results requires modification.
remain subject to interpretation. These interpretations will The results of the analysis suggest a reasonably clear picture.
obviously be based on the probabilities produced by OxCal 4.0 It is possible to reconstruct a chronology that is internally
consistent. The interpretation put forward in table 1 suggests
50. ROAF and KILLICK, 1987: 210.
51. e.g. WRIGHT and RUPLEY, 2001; AURENCHE et al., 2001; BLACKHAM, 52. BLACKWELL and BUCK, 2003: 233.
2002. 53. AKKERMANS et al., 2006: table 1.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 113

Rethinking Halaf Chronologies 113

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
Table 1: Tell Sabi Abyad chronological outline. The dates in the 68.20% and 95.40% columns
are from the calibration model produced using OxCal 4.0. The dates in the remaining columns are estimates.

Level
Phases 68.20% 95.40% Interpretation Span
length

6440-6380 BC
Start 6330-6260 BC 6492-6248 BC 6300
Op. III, level 4
Op. III, level 2 100 30

6318-6306 BC
Early Pottery Neolithic 6297-6294 BC
– Pre-Halaf 6249-6185 BC 6284-6113 BC 6200
Op. V, phase 2
Op. I, level 8 150 150
Pre-Halaf – Transitional 6052-6020 BC 6070-5998 BC 6050
Op. I, level 7
Op. I, level 6 125 31
Op. I, level 5
Op. I, level 4
Transitional – Early
Halaf 5946-5906 BC 5976-5891 BC 5925
Op. I, level 3
Op. I, level 2 75 25
Op. I, level 1

5964-5942 BC
End 5906-5830 BC 5922-5737 BC 5850

that levels were occupied for periods on the order of tion I, level 8 that is consistent with the model is at the end
25-30 years consistently across most of the sequence. This of the range for the dates from Operation V, Phase 2 and may
appears a reasonable estimate 54 and implies a relatively conti- suggest that it falls very late in this phase. There is a good
nuous sequence. The earliest dates for the Early Pottery Neoli- sequence in the succession of well-dated levels in Operation I
thic suggest a start to the sequence possibly as late as ca 6300 from level 8 to level 1. Although there is a relatively wide
cal. BC, although something closer to 6400 cal. BC is not bracket of possibilities for the date of the final occupation,
impossible. The start of the Pre-Halaf is better defined, a date earlier than 5865 cal. BC seems unlikely given that
although the dates also suggest that there are probably gaps in there are three levels to accommodate. The latest possible date
the sequence. This is not surprising because this part of the for the occupation seems unlikely to be much after 5800
Sabi Abyad sequence is not directly related by stratigraphy cal. BC.
but by relating ceramic changes across three different areas of The Transitional (or “Proto-Halaf”) Phase at Tell Sabi
excavation. The group of five very consistent dates from Abyad gains its designation because it is during this phase,
Operation V, Phase 2 has a longer span than is typical of other levels 7-4 in Operation I, that decorated Fine Wares appear
levels, possibly because the earlier end of their distribution is and come to dominate the assemblage 55. This is now clearly
not effectively constrained by the dates from Operation III, documented but equally clearly is a gradual change and
level 2. This might suggest a gap in the occupation that is perhaps one that needs to be seen as a series of changes rather
represented by radiocarbon dates. The single date from Opera-

but a maximum figure of 45-90 years use-life for buildings has been suggested
for Çatalhöyük, CESSFORD, 2002.
54. Estimates for lengths of occupation of levels are poorly understood 55. CRUELLS and NIEUWENHUYSE, 2004; NIEUWENHUYSE, in press.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 114

114 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCal v4.0. (2006); r:5 IntCal04 atmospheric curve (REIMER et al., 2004)

Sequence Sabi Abyad

Boundary Start

Phase Early Pottery Neolithic


Phase Early Pottery Neolithic
Op.IV
GrA-24247 (104.5)
Sequence Early Pottery
Neolithic Op.III
Phase Op.III Level IV

GrA-26927 (101.3)

GrA-26928 (53.4)

Phase Op.III Level II

GrN-28851 (110.4)

GrN-28855 (133.4)

Phase Pre-Halaf

Phase Op.I Level 8

GrN-16805 (66.4)

Phase Op.V Level 2

GrN-28237 (109.8)

GrN-28238 (104.3)

GrN-28239 (88.4)

GrA-24245 (104.4)

GrA-24246 (104.3)

Phase Transitional

Phase Op.II Level 2

GrA-26925 (111)

Sequence Transitional Op.I

Phase Op.I Level 7

GrA-24218 (65.5)

GrN-28240 (131.4)

Phase Op.I Level 6

GrN-19367 (107.1)

GrN-19368 (124.6)

Phase Op.I Level 5

Phase Op.I Level 4 Calendar Date (BC)

GrN-16803 (108.9)

UtC-1008 (55.6)

Phase Early Halaf

Sequence Early Halaf Op.I

Phase Op.I Level 3

GrN-16802 (92.7)

Phase Op.I Level 2

GrN-16800 (116)

Phase Op.I Level 1

GrN-16804 (123.2)

Boundary End

7000 6800 6600 6400 6200 6000 5800 5600

Fig. 4: Probability distributions for the radiocarbon dates from Tell Sabi Abyad.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 115

Rethinking Halaf Chronologies 115

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
than a single transition 56. The extremely low quantities of fine from the model, none has a probability of relating to their
wares present in levels 7 and 6 must have had a very different actual contexts of discovery of greater than 5%. The model
role in society than the moderately common quantities in with the remaining 14 dates produces a satisfactory agreement
levels 5-4. The jump in quantity of fine decorated pottery in index of 94.4%. There are two occasions where dates were
level 3 (“Early Halaf” or “Halaf Ia”) is striking, and again combined prior to calibration. In both cases the determinations
suggests a different context of use. It has previously been had been made on individual cereal grains that were very
pointed out that the overall transition takes places over a short securely stratified in a deposit that related to a single event.
period of time 57. This is confirmed by this more detailed chro- This was not done for other sites because the same level of
nological analysis, which suggests the gap between the start understanding of the depositional context was unavailable.
of level 7 and the end of level 4 to be most probably in the The results from the Domuztepe chronological model need
order of 125 years. OxCal calculates the span of Operation I, to be read slightly differently from that of most of the other
levels 7-4 as 62-115 years (68.2% probability) or 32-140 years sites because stratigraphy has not been analyzed as a simple
(95.4% probability) so this estimate might already be consi- sequence of levels, but as a more dynamic process of parallel
dered slightly high. If the transition was to be seen as taking processes of deposition, construction and abandonment in
place in two stages, each may have been only ca 50-60 years different parts of the excavated areas. Nonetheless the outcome
long. Change within a human lifetime opens up a range of of OxCal analysis is reasonably clear and is presented in
exciting interpretations, but interpretations that focus on the table 2 and figure 5. The earliest excavated deposits come
social not the chronological. from a ditch in Operation I. There is then a gap in the known
deposits in Operation I that is filled by a deep series of proba-
bly short-lived building levels in Operation II. The sequence
DOMUZTEPE (fig. 5 and table 2) picks up in Operation I and continuous activity is documented
until the abandonment of the site. Because of the detailed
There are 21 radiocarbon dates from Domuztepe, 19 of stratigraphic record, which reflects the writer’s familiarity with
which come from the prehistoric occupation of the site. Of it as much as any intrinsic quality, it is possible to use addi-
these, two are omitted here because they appear to have suf- tional information on the length of some of the episodes in
fered from contamination—they are burnt bone samples from the sequence. It would be possible to incorporate this along
near the surface and are several hundred years later than any with the other prior information into the chronological model
likely date. When a chronological model is constructed using for Bayesian analysis but, for reasons of simplicity, this has
the remaining dates, a further three determinations seem to be not been done on this occasion.
incompatible with their stratigraphic positions. When omitted

Table 2: Domuztepe chronological outline. The dates in the 68.2% and 95.4% columns are from the calibration model
produced using OxCal 4.0. The dates in the remaining columns are estimates.

68.20% 95.40% Interpretation Span

Start 5910-5850 BC 5985-5818 BC 5875

Op. I, late ditch 5848-5770 BC 5878-5724 BC 5800


42-154 years (68.2%)
Interval 125 0-186 years (95.4%)

Op. II, Start 5696-5639 BC 5722-5626 BC 5675

Op. II, Late 5668-5622 BC 5707-5609 BC 5630

Op. I, Pre-Death Pit 5622-5588 BC 5632-5568 BC 5590

Op. I, Death Pit 5592-5562 BC 5628-5552 BC 5575

Op. I, final phase 5582-5509 BC 5622-5472 BC 5525

End 5503-5376 BC 5539-5226 BC 5475

56. CRUELLS and NIEUWENHUYSE, 2004: table 1.


57. Ibid.: 63.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 116

116 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCal v4.0.1 Bronk Ramsey (2006); r:5 IntCal04 atmospheric curve (REIMER et al., 2004)

Sequence Domuztepe

Boundary Start

Sequence Ditch

UCIG-12392 (111.8)

UCIG-12394 (122.3)

UCIG-12391 (112.2)

UCIG-12393 (97.3)

R_Combine Op.II Room 2 dates (91.2)

Phase Op.II Pit

AA-44409 (104.2)

UCIG-12389 (106.8)

R_Combine Op.I Death Pit 4 dates (63.6)

Phase Op.I Final Phase

AA-44414 (116.7)

SUERC-3546 (92.6)

Boundary End

7000 6500 6000 5500 5000 4500

Calendar Date (BC)

Fig. 5: Probability distributions for the radiocarbon dates from Domuztepe.


Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

The earliest dated occupation in Operation I has been given “Post-Halaf” for much of the Domuztepe sequence before we
a relatively wide span in the interpretation in table 2. All of received radiocarbon dates 58. The next key date is the Opera-
this occupation is associated with ceramics of the Halaf Ia tion I, Death Pit. Apart from its interest as a large-scale fune-
variety. On stratigraphic grounds, a date later than 5800 rary context, this is an excellent reference point for dating,
cal. BC is certainly possible, almost to the extent of the latest since the four dates come from a single event and were all
date in the 95.4% range of 5724 cal. BC. This, however, raises dated on very well-stratified, individual cereal grains. The final
issues of interpretation with other sites (see discussion below). phase is less well constrained, however. There is certainly a
The gap before the sequence continues in Operation II has a gap between the Death Pit and the deposits for which we have
potentially wide range but the earliest occupation in Opera- dates, although not more than a generation or so, and there is
tion II is very unlikely to be much earlier than 5700 cal. BC also evidence for slightly later occupation badly disturbed by
and is associated with what appear to be Halaf II ceramics. It ploughing. The final date for the sequence can only be approxi-
is worth noting that the date described in table 2 as “Opera- mate. The suggestion of 5475 cal. BC in table 2 seems reaso-
tion II, late” relates to a pit securely stratified within the Opera- nable in light of the quantity of building phases. Certainly
tion II sequence that contained sherds that were considered
typologically Ubaid, and resulted in the now-rejected label of 58. CAMPBELL and FLETCHER, forthcoming.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 117

Rethinking Halaf Chronologies 117

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCal v4.0.1 Bronk Ramsey (2006); r:5 IntCal04 atmospheric curve (REIMER et al., 2004)

Sequence Fıstıklı Höyük

Boundary Start

Phase IIIc

AA-37096 (82.8)

AA-37093 (105.6)

Boundary IIIc-b

Phase IIIb

AA-40381 (73.9)

AA-40382 (61.5)

Boundary IIIa-b

Phase IIIa

AA-37091 (117)

AA-37092 (117)

AA-37094 (102.8)

AA-37095 (65.2)

AA-37097 (126.9)

AA-40374 (110.9)

AA-40376 (118.9)

AA-40379 (112.4)

AA-40380 (126.3)

Boundary End

6800 6600 6400 6200 6000 5800 5600

Calendar Date (BC)

Fig. 6: Probability distributions for the radiocarbon dates from Flstlkll Höyük.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

anything later than 5400 cal. BC is unlikely and the extension FΙSTΙKLΙ HÖYÜK (fig. 6 and table 3)
of the end boundary for the Domuztepe model beyond this
date is almost certainly due to the lack of later constraining There are 20 radiocarbon dates from Flstlkll Höyük, all
dating evidence. From the lowest dated context of Operation II from prehistoric contexts 59. The excavators divided the Halaf
until the abandonment of Operation I, the ceramics would be stratigraphy into five phases and dates are available from all
categorized as Halaf II in general terms. There are distinct but the first and last of these. This phasing was produced
changes but they are gradual and probably reflect local deve- through comparative stratigraphy that also made use of a cera-
lopments. They certainly cannot be usefully or safely corre- mic seriation to correlate different contexts. When placed in
lated with terms such as Middle or Late Halaf. the chronological sequence that the excavators propose, two

59. BERNBECK and POLLOCK, 2003: 19-23.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 118

118 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
Table 3: Flstlkll Höyük chronological outline. The dates in the is rather early (AA-40370). When all five problematic dates
68.2 % and 95.4 % columns are from the calibration model produced are removed from the model, none has a probability of dating
using OxCal 4.0. The dates in the remaining columns are estimates. to their excavated contexts of more than 4.8%. If boundaries
are inserted between adjacent phases, something that has gene-
68.20% 95.40% Interpretation Span
rally not been done because of the assumptions it introduces,
Start 5969-5814 BC 6087-5766 BC 5850
the remaining dates have an acceptable agreement index of
IIIC 35
94.5%, despite the possible issues with the two remaining
IIIC-B 5855-5783 BC 5902-5757 BC 5815 dates from Phase IIIB.
IIIB 35 Because of the absence of earlier phases, the beginning of
IIIB-A 5803-5755 BC 5832-5744 BC 5880 Phase IIIC is poorly constrained. Even by placing a boundary
IIIA 30 at the start, the true beginning of the phase is unlikely to be
End 5776-5731 BC 5811-5711 BC 5750 as early as the results from OxCal suggest is possible. If
Phase IIIC is of a similar length to Phases IIIB and IIIA, and
there is no strong evidence to the contrary, a more realistic
of the dates from Phase IIIB seem to be incompatible with estimate for the start of the sequence might be ca 5850 cal. BC.
their stratigraphic positions, both being rather too late This is very close to the end of the occupation at Tell Sabi
(AA-40377 and AA-40378). Given that the other two dates Abyad. Even if the dates for Flstlkll Höyük are pushed as late
from this phase also cause some problems (they have agree- as is feasible, its occupation is likely to start within decades
ment indices of 68.1% and 56.0%), it is possible that there is of the end of that of Tell Sabi Abyad. The end is better defined
an issue with the chronological model. The remainder of the because of the quantity of dates from Phase IIIA. A short
dates does fit the published chronological model, with the occupation is very likely, perhaps only 100 years and certainly
exception of three dates from Phase IIIa, two of which are less than 150 years, with occupation ending around 5750
uncomfortably late (AA-40371 and AA-40372) and one that cal. BC or soon after.

Table 4: Tell Kurdu chronological outline. The dates in the 68.2 % and 95.4 % columns are
from the calibration model produced using OxCal 4.0. The dates in the remaining columns are estimates.

68.20% 95.40% Interpretation


Main phase start 6046-5877 BC 6156-5806 BC 5850
Main phase end 5783-5660 BC 5847-5612 BC 5700

5624 BC (92.1%) 5465 BC


Room 39 lower 5436 BC (1.1%) 5424 BC
AA-52957 5561-5476 BC 5405 BC (2.2%) 5386 BC

5466 BC (2.9%) 5436 BC


Room 39 Upper 5426 BC (1.5%) 5406 BC
AA-52958 5326-5222 BC 5382 BC (90.9%) 5208 BC

000 (61.8%) 235 years


291 (0.8%) 296 years
340 (0.9%) 346 years
Interval 356 (4.7%) 394 years 0-488 years

4879 BC (95.2%) 4238 BC


Later occupation end 4816-4612 BC 4232 BC (0.2%) 4220 BC

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 119

Rethinking Halaf Chronologies 119

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al.,2004)

SequenceTellKurdu

B oundary S tart

PhaseMaingroup

AA-52965 (91.2)

AA-52963 (100.1)

AA-52962 (100.7)

AA-52961 (100.3)

AA-52959 (103.6)

SequenceRoom41

AA-52956 (102.3)

AA-52964 (113.6)

AA-52960 (105.5)

SequenceRoom39

AA-52957 (102.6)

AA-52958 (88.4)

B oundary E nd of main group

PhaseLaterKurduDates

G rN-22964 (100.3)

G rN-22965 (100.1)

B oundary E nd

8000 7000 6000 5000 4000

CalendarDate(BC)

Fig. 7: Probability distributions for the radiocarbon dates from Tell Kurdu.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

TELL KURDU (fig. 7 and table 4) came from what was thought to be a broadly contemporary set
of contexts, with one (AA-52960) slightly later, the dates
There are 12 available dates from Tell Kurdu. Ten come suggested that the use of Room 39 may be significantly later
from Amuq C contexts 60, traditionally equated with the Halaf, than expected 62. This is confirmed by the OxCal model. When
and two from Amuq E contexts 61, generally equated with Ubaid. the two dates from Room 39 are included in their expected
As noted in the publication of the dates from the Amuq C stratigraphic position, the agreement index is extremely low
contexts, there are problems reconciling the expectations from (1.6%). When the dates from this context are excluded, the rest
the excavation with the dates. Although most of these dates of the model has a satisfactory agreement of 98.4%. Since the
dates from Room 39 are internally consistent (and in their
60. ÖZBAL et al., 2004: 85.
61. YENER et al., 2000: 171. 62. ÖZBAL et al., 2004: 51-52.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 120

120 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
correct stratigraphic order), a further possibility is to consider belong to a later phase or the association of the date of the
the possibility that Room 39, or at least its use, is later than all samples with the structure may be erroneous, perhaps through
the other activity in the excavated part of the site; this was intrusion of later material or contamination of the sample.
considered in the original report, where the excavators noted it There is obviously a large gap until the occupation dated by
causes other problems of interpretation 63. While acknowledging the two latest dates from Tell Kurdu. The extent of the gap
the problems, this remains the only way to allow the dates to partly depends on the view taken on the dates of Room 39.
fit into a single chronological model. If this is done, the OxCal Obviously, if they do not date that structure, then the gap will
model has an agreement index of 102.2%. This is the model be correspondingly larger. Although the end boundary is also
illustrated here. elongated because of lack of later constraints, it seems likely
In some ways, the results of the Tell Kurdu analysis seem that this phase dates within 100 years or so of 4750 cal. BC.
clear but in others they raise significant issues. As in other
examples, the start of the sequence is elongated by the lack of
earlier dates to constrain it. The excavators argue that the main YARIM TEPE (fig. 8 and table 5)
phase of architecture is broadly contemporary and, if that is the
case, one might expect the length of the phase to be relatively There are 11 radiocarbon dates published from Yarim Tepe,
short. It seems unrealistic, on the basis of the chronological all from the Leningrad laboratory 64. They only cover a small
model, to pull the start much later than ca 5850 cal. BC; 5806 part of the sequence. Two come from Yarim Tepe I and none
cal. BC is the extreme of the 95.4% distribution and might be from Yarim Tepe III, and even those from Yarim Tepe II are
considered the absolutely latest feasible date. An end date for concentrated in levels 6 to 8. Clearly this will not permit abso-
these main constructions might be around 5700 cal. BC. lute dating of the entire long sequence but it should have the
Although a later date is possible, if this is a single phase of potential to allow us to tie down some key points within it.
architecture, it may be more likely to be slightly earlier. This Although the dates were largely determined over a short period
confirms the problems identified by the excavators of of time in a single laboratory, they are the oldest group of
Room 39. The two dates from here suggest a gap of some 150 determinations considered amongst the key sites in this study.
to 300 years. This clearly remains a problem that can only be They also have large errors associated with them. When placed
explained in a limited number of ways. Room 39 may actually in a chronological model in OxCal, two of the dates have very

Table 5: Yarim Tepe chronological outline. The dates in the 68.2 % and 95.4 % columns
are from the calibration model produced using OxCal 4.0. The dates in the remaining columns are estimates.

68.20% 95.40% Interpretation


Main phase start 6046-5877 BC 6156-5806 BC 5850
Main phase end 5783-5660 BC 5847-5612 BC 5700

5624 BC (92.1%) 5465 BC


Room 39 lower 5436 BC (1.1%) 5424 BC
AA-52957 5561-5476 BC 5405 BC (2.2%) 5386 BC

5466 BC (2.9%) 5436 BC


Room 39 Upper 5426 BC (1.5%) 5406 BC
AA-52958 5326-5222 BC 5382 BC (90.9%) 5208 BC

000 (61.8%) 235 years


291 (0.8%) 296 years
340 (0.9%) 346 years
Interval 356 (4.7%) 394 years 0-488 years

4879 BC (95.2%) 4238 BC


Later occupation end 4816-4612 BC 4232 BC (0.2%) 4220 BC

63. Ibid.: 52. 64. MERPERT et al., 1976; MUNCHAEV and MERPERT, 1981.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 121

Rethinking Halaf Chronologies 121

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al,2004)

SequenceYarimTepe

B oundary S tart

PhaseYarimTepeI,Level7

Le-1070 (104.1)

Le-1086 (89.1)

PhaseYarimTepeII,Level8

Le-1291 (101.3)

Le-1011 (116.3)

PhaseYarimTepeII,Level7

Le-1212 (85.4)

PhaseYarimTepeII,Level6

Le-1172 (108.6)

Le-1174 (109.1)

Le-1211 (106.4)

PhaseYarimTepeII,Level3

Le-1015 (92.2)

B oundary E nd

8000 7000 6000 5000 4000

CalendarDate(BC)
Fig. 8: Probability distributions for the radiocarbon dates from Yarim Tepe I and II.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

low agreement indexes (Le-1173 and Le-1012). When these consistent. Without the two dates, the model has a good level
are removed from the model, the overall agreement rises to of agreement and the probable time differences among levels
101.1% and both the removed dates have extremely low proba- are reasonable. The absolute dates, however, cause systematic
bilities of relating to their stratigraphic positions. problems in comparison to other sites. Yarim Tepe I, level 7 is
A little more can be gleaned from the archaeology that has only mid-way through the stratigraphy of that mound and sees
not been taken into account in the OxCal scenario. All archae- the first appearance of Hassuna II ceramics, characterized by
ological levels at Yarim Tepe II actually encompass a degree quantities of incised pottery. It is very difficult to push the
of time. There are alterations to buildings and often individual apparent date for level 7 earlier than 6050 cal. BC and it may
buildings are replaced or rebuilt within an archaeological level. well be later than that. Looking at the other end of the sequence,
Dating solutions that suggest short periods for individual levels an apparent date for Yarim Tepe II, level 3 would be around
are, therefore, inherently unlikely. The results of the calibration 5200 at earliest; any earlier will squeeze the space for two full
suggest that a reasonable estimate for the length of time repre- levels of architecture to an unfeasible extent. Level 3, however,
sented by each level might be between 65 and 100 years. is still characterized by ceramics comfortably within the Halaf II
The results of the OxCal analysis are internally reasonably style and there are two more levels to accommodate before any

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 122

122 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al.,2004)

SequenceTellKosakShamali

B oundary S tart

PhaseLateNeolithic-Level18

TKa-11668 (104.7)

PhaseLevel11

TK-999 (106.7)

PhaseLevel10

TK-1000 (113.1)

TKa-11656 (114)

TKa-11860 (113.9)

TKa-11859 (100.5)

PhaseLevel1

TKa-11660 (127.6)

PhaseSectorBLevel6

TKa-11664 (100.1)

TKa-11678 (83.4)

PhaseSectorBLevel5

TKa-11655 (84.8)

PhaseSectorBLevel3

TKa-11674 (80.4)

B oundary E nd

10000 8000 6000 4000 2000 1BC/1AD

CalendarDate(BC/AD)
Fig. 9: Probability distributions for the radiocarbon dates from Tell Kosak Shamali.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

sort of transition to the Ubaid is noted in the lowest levels of 200-300 years, an amount consistent with known variation
Yarim Tepe III. It is true that it will be argued below that we among laboratories of the time 66. Unfortunately, since the
are poorly informed about the chronology of this phase but it exact level of error cannot be determined, the results from
seems unfeasible to pull it this late. Much the most likely Yarim Tepe must largely be disregarded.
explanation is that there was a systematic error in the dates
being determined in the 1970s at the Leningrad laboratory 65.
This may have made the dates consistently too late by
print by V. Timofeev. The article referred to has not been located by this
65. An email in the Radiocarbon Mailing List from I. Antanaitis, Vilnius writer.
University, dated 1. Nov. 1999 suggests that this has been acknowledged in 66. AITCHISON et al., 1990.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 123

Rethinking Halaf Chronologies 123

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
TELL KOSAK SHAMALI (fig. 9) vated (TK-997 and TK-998) and were taken out of the OxCal
model. Omitting these dates produces an overall agreement index
There are 13 radiocarbon dates from Tell Kosak Shamali of 102.8%. The dates from the earliest and latest strata were
with excellent information on their nature and stratigraphic posi- retained because they are consistent with the stratigraphic model,
tions, although some of them relate to periods slightly beyond although it should be noted that the excavator argues convincin-
the scope of this article 67. The details of the ceramic assemblages gly that both are too late for their cultural context 68. Y. Nishiaki’s
from the site are, however, not yet published in detail so the discussion, however, also highlights the fact that several of these
position of the relative sequence at Tell Kosak Shamali is entirely dates are not in stratigraphic order or appear to be too early or
based on the excavators’ summary. Two of the dates have very late for their contexts 69. In fact, the dating scenarios used here
low agreement indexes for the locations in which they were exca- suggest the former is not necessarily a problem and the latter

OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al.,2004)

SequenceTellMashnaqa

B oundary S tart

PhaseStratumI

R _Date AAR -1005 (101.2)

R _Date AAR -1006 (112.6)

R _Date AAR -1092 (109.5)

PhaseStratumIII

R _Date AAR -1093 (112)

R _Date AAR -1008 (114.9)

R _Date AAR -1090 (114.3)

R _Date AAR -1091 (99.5)

B oundary E nd

7500 7000 6500 6000 5500 5000 4500 4000 3500

CalendarDate(BC)
Fig. 10: Probability distributions for the radiocarbon dates from Tell al-‘Abr.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

68. Ibid.: 153, 156.


67. NISHIAKI, 2001: 154. 69. Ibid.: 155-156.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 124

124 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
not necessarily true. To a considerable extent, this is a pattern TELL AL-‘ABR (fig. 10)
generated by the scattering of the means of radiocarbon dates
that have relatively large standard deviations. Five radiocarbon dates are available from Tell al-‘Abr,
For present purposes, the key dates relate to the middle from three of the Ubaid levels 70. The dates come from
section of the phasing of Tell Kosak Shamali. The problematic levels 7-5, and level 7 represents the earliest level at the site.
single date from level 18 will be ignored. Level 11 is the OxCal produces an acceptable agreement index of 69.7%
earliest level with a good date and level 10 has four consistent when all dates are included with the individual dates; all are
dates. There is rather a wide range of possible dates for these above the 60% agreement threshold. The start of the Tell ‘Abr
levels. If the slightly questionable assumptions are made that sequence is associated with Ubaid ceramics, although the
the stratigraphy is relatively continuous and that the building excavators suggest that it is very early in the northern Syrian
levels last no longer than 100 years at most (the excavator Ubaid tradition 71. Level 7 seems unlikely to date before 5200
estimates 30-50 years), a time frame of ca 5100 cal. BC to cal. BC and may date as late as 5000 cal. BC (5208-5002
ca 4900 cal. BC would be possible but this could move by cal. BC at the 95.4% level). Although either date could be
100-150 years in either direction. These levels are associated consistent with the chronological model, the earlier date is
with Ubaid pottery. Ubaid style pottery, however, is already harder to accommodate if the occupation at the site is con-
present in level 17 when it is assumed to be very early in the tinuous and the levels of short or moderate length.
north Syrian phasing. Levels 11 and 10 are, therefore, indicators
of a date at which Ubaid pottery was already a major feature.

OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al.,2004)

SequenceTellal-'Abr

B oundary S tart

Phase7

Hd-17471-17510 (69.4)

Phase6lower

Hd-17470-17919 (89)

Hd-17469-17928 (117.9)

Phase6upper

Hd-17468-17781 (102.7)

Phase5

Hd-17467-17772 (70.9)

B oundary E nd

6500 6000 5500 5000 4500 4000 3500

CalendarDate(BC)
Fig. 11: Probability distributions for the radiocarbon dates from Tell Mashnaqa.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

70. HAMMADE and YAMAZAKI, 2006: 431.


71. Ibid.: 432.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 125

Rethinking Halaf Chronologies 125

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al.,2004)

SequenceTellZiyadeh

B oundary S tart

PhaseEarlierUbaid

AA-24863 (94.5)

AA-24866 (105.1)

AA-30490 (114.2)

AA-30495 (115.3)

B eta-84582 (116.5)

AA-30496 (71.7)

PhaseLaterUbaid

B eta-84579 (105.9)

B eta-84581 (105.9)

AA-24862 (104.2)

AA-24865 (112.4)

AA-24860 (111.3)

AA-24859 (108.9)

AA-24861 (108.9)

B eta-84580 (108.6)

AA-30494 (107.5)

PhasePost-Ubaid

AA-30491 (88)

AA-24858 (102.8)

AA-24856 (102.3)

AA-24857 (101.5)

AA-24855 (105.5)

AA-30492 (105.6)

B oundary E nd

5600 5400 5200 5000 4800 4600 4400 4200 4000

CalendarDate(BC)

Fig. 12: Probability distributions for the radiocarbon dates from Tell Ziyadeh.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 126

126 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
TELL MASHNAQA (fig. 11) phases 73. Although the dates come from different trenches that
were connected on the basis of ceramic style rather than stra-
There is a group of seven dates from Tell Mashnaqa, on tigraphy, the three groups make a coherent set. All but four
the Khabur in northeastern Syria, three from stratum I and dates come from the same laboratory so internal consistency
four from stratum III 72. Stratum I is the earliest at the site and should be good. Only two dates fit poorly into their stratigra-
appears to have some of the earliest northern Ubaid ceramics. phic positions (AA24864 and AA30493) and have been
All seven dates can be satisfactorily accommodated in a simple removed from the final chronological model. Without those
OxCal model, with an agreement index of 128%. The start of dates, the OxCal model produces an agreement index of
the sequence is poorly constrained. Although the range for the 119.1%.
start of the sequence is 5455-5004 cal. BC at the 95.4% level, The dates are sorted into three main phases. The start of
the three dates from Stratum I all have modeled calibrations the Earlier Ubaid phase is almost certainly after 4900 cal. BC
of between ca 5300 and ca 4990 cal. BC. This seems to suggest and probably no later than ca 4800 cal. BC. The range for the
a starting point for the sequence no earlier than 5300 cal. BC end of this phase is 4864-4714 cal. BC at the 95.4% level.
and possibly as late as ca 5000 cal. BC. The end of the Late Ubaid phase falls between 4698 cal. BC
and 4472 cal. BC, again at the 95% level. Although there is
information on the ceramics from the site 74, it is not possible
TELL ZIYADEH (fig. 12) to correlate this fully with the contexts of the radiocarbon
dates. As the phase designations suggest, however, the two
There is a large set of 23 radiocarbon dates from Tell earlier phases are associated with Ubaid ceramics.
Ziyadeh, which have been divided into three successive

OxCalv4.0.1BronkRamsey(2006);r:5IntCal04atmosphericcurve(REIMERet al.,2004)

SequenceKenanTepe

B oundary S tart

PhaseUbaid

SequenceKenanTepe

PhaseStructure1

B eta-176372 (100.5)

B eta-176373 (86.6)

B eta-180241 (98.9)

Phaselater

B eta-1? (107.6)

B eta-2? (102.4)

B oundary E nd

6000 5500 5000 4500 4000 3500

CalendarDate(BC)
Fig. 13: Probability distributions for the radiocarbon dates from Kenan Tepe.
Agreement levels for individual dates are in brackets after the laboratory code. Generated by OxCal 4.0.

73. HOLE, 2001: 74.


72. THUESEN, 2000. 74. ARZT, 2001.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 127

Rethinking Halaf Chronologies 127

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
KENAN TEPE (fig. 13) DISCUSSION: PERIODS OF CERAMIC
CHANGE (fig. 14)
There are five radiocarbon dates from the Ubaid levels at
Kenan Tepe 75. Three dates come from one building, Struc- The focus on sites with sets of stratified dates emphasizes
ture 1, with a further date from an overlying structure. The the gaps in our knowledge. There are certainly other radiocar-
final date comes from another structure not yet stratigraphi- bon dates that can be considered, but there are major areas of
cally connected to the other two 76. A chronological model ignorance. This means that when the different sequences are
using all five dates produces a good overall agreement index integrated, some of the periods highlighted will be ones in
of 97.7%. The date of Structure 1 is very unlikely to be before which key questions can be posed and potentially answered
4800 cal. BC and, if the latest of the three dates (Beta-17637) but others will be periods of which we currently have too little
relates directly to the structure and if that structure did not detailed chronological knowledge to be so specific. It is almost
stand for an exceptionally long period, it may be unlikely to impossible to construct an overall chronological structure at
date before 4700 cal. BC. The date for Structure 2 seems more this point—and attempting to do so could be subject to the
consistent with the latter suggestion (4609-4500 cal. BC at same criticisms made of previous schemes. Instead the discus-
68.2%, 4662-4460 cal. BC at 95.4%). sion will be structured around key phenomena, where there

Fig. 14: Overall chronological model from the key sites, highlighting key phases of ceramic change discussed in the text. Note that the time
ranges for individual sites are only indicative; reference should be made to the text, figures and tables discussing individual sites for precise
dates. The phenomena discussed in the text are cross-hatched, while traditional phases are labeled in the gaps between them.

75. PARKER and DODD, 2005; PARKER et al., 2006.


76. PARKER and DODD, 2005; Parker and Dodd, personal communication.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 128

128 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
seem to be periods of significant change over relatively short that they fit with the beginnings of the phase at Tell Sabi
intervals at different sites or where there are major ambiguities Abyad; that is contemporary with the Transitional/Proto-Halaf
that challenge our accepted view of the past. In these examples, phase at the site in levels 7-4, albeit possibly late in the
the phenomena will be focused around ceramics, not because phase 80. The radiocarbon dates, in fact, fit better with the
it has any primacy but because its development is best docu- following phase at Tell Sabi Abyad, levels 3-1. Placed in the
mented and its previous dominance in chronological studies Transitional/Proto-Halaf phase in the chronological model
makes it useful to build a contrasting emphasis in this discus- used above, the two dates have low agreement indexes of
sion. The phenomena selected are only one set of examples. 46.3% and 19.3% for Beta-174042 and Beta-174043 respec-
The phenomena could equally be change in ceramic techno- tively. Placed in the Early Halaf Phase, the agreement indexes
logy or change in any other aspects of material culture. are 84.5% and 53.8%. Although the latter agreement index is
still poor, it is because the date is later rather than earlier. This
is a comment based on the radiocarbon evidence alone,
THE SPREAD OF PAINTED POTTERY although it should be noted that the ceramic sample at Chagar
Bazar is relatively small.
As this has already been discussed in detail recently 77 and
mentioned more briefly above, the discussion here can be
limited. Nonetheless, it is worth re-emphasizing that the spread THE END OF THE FIRST PHASE OF PAINTED
of painted pottery at the end of the 7th millennium BC marks POTTERY
a significant development in the way in which ceramics were
used and the complexity of symbolism associated with them. The sub-title of this section is a rather coy attempt to avoid
The development is a useful chronological marker but it is using the word, “Halaf”, and illustrates why the traditional
much more significant in terms of social change. The way in terminology retains a role in promoting communication. What
which painted ceramics were incorporated into the wider is really meant is, “what is the relative position of ceramic
ceramic assemblage is only completely documented at Tell assemblages that have been variously termed Halaf Ia, Early
Sabi Abyad, where it can also be given a fairly precise Halaf and Balikh IIIB to assemblages that have been termed
date. It has been apparent for some time that the emergence Halaf Ib, Traditional Early Halaf and Balikh IIIC?” 81 For
of this new painted pottery style happened over a wide area brevity, the terms Halaf Ia and Halaf Ib will be used here, as
in a very similar way 78. This generates immediate questions will the term Halaf II to refer to the traditional Middle and
concerning the nature of the phenomenon. Was it associated Late Halaf style assemblages. Nonetheless, it must be empha-
with rituals of consumption, or even with new cuisines? Did sized that these terms are increasingly unsatisfactory. As with
the style carry common meanings or reference common much of the terminology of the Halaf, they are derived from
beliefs? Did it develop in one region and then spread through northern Iraq and their use in northern or northwestern Syria
a process of emulation or is it the outcome of multiple parallel and southeastern Turkey is problematic.
developments within an existing network of interaction? It was Compared to the focus on the beginnings of the Halaf, the
suggested above that, in fact, it may not be a case of asking processes of later ceramic development have received almost
a single set of questions because this is not a single event but no recent attention. This is partly due to the accidents of
a process that may have had several stages. excavation (the main periods present at Tell Sabi Abyad have
Unfortunately, there are currently no other dates relating been seminal in drawing attention to the slightly earlier changes)
to the initial appearance of this painted pottery tradition. No and partly also due to the apparent attraction of seeking the
dateable samples were recovered from NJP72 and no dates are origins of an entire period. Nonetheless, the two assemblages
published from Tell Boueid II or the relevant levels of Tell exhibit significant differences. The purpose of this article is not
Halula. There are two dates from Chagar Bazar from the to present a detailed discussion of ceramic shapes, fabrics and
earliest part of the sequence at that site 79. It has been suggested decoration, so generalisations have to suffice.

77. CRUELLS and NIEUWENHUYSE, 2004; NIEUWENHUYSE, in press. 80. CRUELLS and NIEUWENHUYSE, 2004: 54; CRUELLS, 2006a; Ibid., 2006b:
78. CAMPBELL, 1998; AKKERMANS, 2000; CRUELLS and NIEUWENHUYSE, 128-129.
2004. 81. CAMPBELL, 1992; CRUELLS and NIEUWENHUYSE, 2004; AKKERMANS,
79. CRUELLS, 2006a. 1993 respectively.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 129

Rethinking Halaf Chronologies 129

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
It was suggested above that the end of sequence in Opera- ca 5850-5800 cal. BC (fig. 14). This is perfectly conceivable.
tion I at Tell Sabi Abyad might be placed somewhere between The basic technology does not seem to have changed. Instead
5860 and 5800 cal. BC. At Domuztepe, the Operation I ditch it is a stylistic change reflected to an extent in vessel forms
material is generally comparable to levels 3-1 at Tell Sabi but particularly in decoration. There could be many explana-
Abyad. There is a gap before the occupation in Operation II, tions that will not be developed here but I would point again
which is characterized by ceramics that are actually more simi- to the possible questions they might generate. Are we looking
lar to Halaf II in style than either Halaf Ia or Halaf Ib. The at a pattern of change within the existing network of interaction
assemblages are still being analyzed but decoration in parti- or a more definable process of innovation and emulation?
cular seems to change, with the disappearance, for example, What mechanism may have led to a significant stylistic reor-
of the striking horizontal cross-hatch that is also very charac- ganization at this time? Is the association with contexts of
teristic of painted pottery of Tell Sabi Abyad. There is also consumption important in understanding these changes? How
some shift in vessel shape. Obviously, because of the gap in is it associated with community identities? And so on. We
the sequence at Domuztepe, it is impossible to know the way should recognize, however, that this is not the only line of
in which this change occurred or its speed. It was suggested interpretation. There may not have been a universal change
that the excavated Operation II deposits are unlikely to start over this short period. Instead it is possible that different settle-
much before 5700 cal. BC. The latest date of the ditch material ments, or even groups within settlements, may have been using
in Operation I is less precise and could be several decades different styles of ceramics contemporaneously, perhaps
either before or after 5800 cal. BC. To this writer’s eye, the denoting different identities. Since Flstlkll Höyük is situated
limited range of published ceramics from Flstlkll Höyük do between Domuztepe and Tell Sabi Abyad, this might suggest
not appear to share the same characteristics as the material that any differentiation along these lines is more complex than
from Tell Sabi Abyad’s levels 3-1, although the excavators a simple regional one.
suggest that the material most closely parallels Level 3 82. In Whatever interpretation we place on it, the relative speed of
part, this difference of opinion may be due more to the lack the transition seems clear in northern Mesopotamia to the west
of comparative data than anything more fundamental. None- of the Balikh valley. This is surprising because it leaves no room
theless, I would suggest that the published material from to accommodate Halaf Ib or the traditional Early Halaf. Where
Flstlkll Höyük is significantly more similar to Domuztepe this ceramic assemblage has been documented, it seems to be
Operation II than it is to the ditch assemblage of Operation I; very distinctive, yet there are no clear examples of it further
in other words, it is closer to a Halaf II type of assemblage west than Tell Aqab. The only possible conclusion must be that
than a Halaf Ia one. This is significant because it was suggested these assemblages do not occur across all of northern Mesopo-
above that the start of the sequence at Flstlkll Höyük might tamia but, on present knowledge, are restricted to the upper
be ca 5850 cal. BC and certainly within a few decades of both Khabur and northern Iraq.
the end of the Tell Sabi Abyad sequence and the end of the There must be significant social processes underlying the
ditch in Operation I at Domuztepe. Although Tell Kurdu has a emergence of these assemblages. The Halaf Ib ceramic assem-
very different type of assemblage with a very low percentage blage is significantly different from the Halaf Ia one. It is well
of painted pottery, a similar pattern emerges. The published represented at Arpachiyah, Yarim Tepe II and Tell Aqab. The
painted pottery is certainly not similar to the type of Halaf Ia most obvious feature is the dominance of straight- and convex-
assemblage seen at Tell Sabi Abyad and its affinities are more sided bowls, to the extent that they make up over 60% of the
with the later material at Domuztepe 83. Very crudely, it is an total range of shapes. There seems to be a technological change
assemblage with Halaf II affinities rather than Halaf Ia, although as well. At NJP72, the Halaf Ia ceramics generally have mode-
the percentage of painted pottery is very low. The chronology rate grit tempering but there is a small number of straight-sided
suggested for Tell Kurdu would place the main building phase bowls with almost no visible temper at all. This much finer
before 5800 cal. BC and possibly as early as 5850 cal. BC. fabric dominates in the Halaf Ib assemblages, and it is this
This is striking in several ways. It appears that the Halaf Ia type of fabric that came to be seen as typically Halaf. There
type of assemblage may have been replaced by a related but is also a change in decoration as horizontal cross-hatching
distinctly different assemblage over a very short time period, disappears and large areas on the exterior of the bowls are
covered with single rows of lozenges or with cross-hatching.
82. BERNBECK and POLLOCK, 2003: 23. All of these changes suggest a significant correlation with
83. ÖZBAL et al., 2004. changing social patterns.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 130

130 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
The change seems to have taken place relatively quickly. poor resolution but generally suggest a date sometime after
At Arpachiyah, it is very obvious in Hijara’s ceramic analy- ca 5800 cal. BC for the Halaf II assemblage 87. Although its
sis 84. The ceramics of Phases I-III are dominated by straight- stratigraphic position is ideal, the single date from layer 25 at
sided bowls while Phase IV is not. Although the limited size Arpachiyah has too broad a spread to help resolve this ques-
of the excavation makes it hard to be certain, it seems that tion 88. In general, these dates suggest that the Halaf Ib assem-
little time elapsed between the phases. The change can also be blage had been replaced by a Halaf II assemblage by ca 5700
seen at Yarim Tepe II. A detailed breakdown of vessel shape cal. BC (fig. 14). It is impossible, however, to suggest a date
by depth is available for a sample of 1,004 sherds in one with any reliability for the emergence of the Halaf Ib tradition.
10 x 10 m square 85. This provides information on the relative It may be that the earlier Halaf Ia assemblage developed at
quantities of eight major shapes in 34 levels of 20 cm each. roughly the same date as it did further west, although that is
Figure 15 uses this data in reworked form to show the quantity conjecture. It may well be the Halaf Ia tradition was replaced
of straight-sided and convex-sided bowls by depth (shapes earlier in the east than the west, but that is also guesswork, as
SVS2 and SVS3 in the original analysis). Although this cannot is the possibility that the Halaf II tradition may have developed
be correlated exactly with the stratigraphic divisions used in in the west and only later spread eastward.
other publications of the site, it is clear that these bowl types There are several possible areas where we might seek to
dominated for rather a short period and then declined rapidly. identify the significance of the change from Halaf Ia to
The lower numbers in levels 34 and 33 may document a period Halaf Ib in northern Iraq and northeastern Syria. Once again,
preceding their dominance, or it may simply be an artefact of we can point to the association with consumption. The change
sample size. in vessel form might suggest that a specific way of eating or
drinking became more prominent, perhaps associated with
70 processes of prestige and emulation. The absence of Halaf Ib
P e r c e n ta g e o f s tr a ig h t- s id e d b o w ls

60
assemblages further west might suggest a breakup of earlier
interaction networks across northern Mesopotamia and, possi-
50
bly, the development of regional identities marked by different
40 styles of ceramics and social behaviours. Since the Halaf II
30 assemblages have a greater level of similarity once again
across north Mesopotamia, it may be that this regional frag-
20
mentation was later reversed. It would, however, be very
10
dangerous to make sweeping conclusions at this point. Much
0
work remains to be done to resolve a sufficiently fine detail
3 2 -3 1

2 2 -2 1

1 2 - 11

6 -1
3 4 -3 3

3 0 -2 9

2 8 -2 7

2 4 -2 3

1 8 -1 7

1 4 -1 3
2 0 -1 9

8 -7
1 0 -9
2 6 -2 5

1 6 -1 5

to understand the processes involved.


Layer

Fig. 15: Straight-sided bowls as a percentage of the main vessel types


at Yarim Tepe II. The stratigraphy is divided into 34, 20 cm thick THE HALAF-UBAID TRANSITION
spits. Data from AMIROV and DEOPEAK, 1997.
The previous examples have focused on short periods in
It is difficult to place a date on either the emergence of the which significant change seems to have occurred and where
Halaf Ib type of assemblage or its decline. Although there are taking a chronology that concentrates on episodes of change
no long sequences with large numbers of dates, there are some invites a more subtle level of interpretation. The final example
relevant sites with smaller numbers of dates associated with to be discussed here instead involves a period that traditional
Halaf II assemblages. At Umm Qseir, three dates suggest that phase-based chronologies have highlighted and suggests that
such an assemblage was present by ca 5700 cal. BC at the latest we may actually know less than we have traditionally thought.
and possibly several decades earlier 86. The small group of dates It has long been argued that the Halaf period came to an
from Banahilk were determined in the early 1970s and have end when it was replaced by the Ubaid, in an episode in which
a cultural tradition from southern Iraq expanded to dominate
84. HIJARA, 1997.
85. AMIROV and DEOPEAK, 1997. 87. LAWN, 1973: 373.
86. KITAGAWA and VAN DER PLICHT, 1998; HOLE, 2001. 88. HIJARA, 1980: 144.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 131

Rethinking Halaf Chronologies 131

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
northern Mesopotamia. Mallowan sought to explain this in is clear in figure 3. Other than an unsatisfactory group of dates
terms of invaders and population movement 89. Davidson from Tepe Gawra 96, there are actually no dates at all that
demonstrated that the change was not, in fact, abrupt and clearly relate to the HUT. To define it, we are dependent on
labeled the period of transition the Halaf-Ubaid Transitional dating the end of the Halaf and the start of the northern Ubaid.
phase (HUT) 90. This was further elaborated to the status of a The latest Halaf is also poorly served by radiocarbon dates.
final stage of the Halaf during the 1980s 91 and it has remained There are few good sequences that we can use and we must
a firm feature of chronological charts of northern Mesopotamia make rather unsound assumptions about the contemporaneity
since then. The most extensive attempt to provide an interpre- of the final stages of the Halaf. The occupation at Domuztepe
tation of the HUT was carried out by C. Breniquet 92. Most seems to have been abandoned before a major Ubaid impact.
recently, that transition from the Halaf to the Ubaid in northern It was suggested above that the site was abandoned ca 5475
Mesopotamia has been examined by G.J. Stein and R. Özbal 93 cal. BC and that anything later than 5400 cal. BC was extrem-
and by M. Frangipane 94. These approaches to the HUT have ely unlikely. Given that Domuztepe is at the northwestern
all shared certain characteristics 95. The first key characteristic extreme of the Halaf region, there must be doubt whether this
that is most relevant here is that the HUT can be usefully is a useful date for northern Iraq, for example. There are very
conceived as a transitional phase between the two larger enti- few well-corroborated dates from Halaf contexts that date
ties, the Halaf and the Ubaid; it is a way to move from one much after this. A single date from a bone artefact from the
to the other with a single interpretation relating to changes in TT6 Burnt House at Arpachiyah (OxA-4982, 6390 ± 110)
society. The second characteristic implication is that, as a might date a little later. Doubt may be cast on that date, how-
transitional phase, it is relatively short compared to the more ever, not only because it is a single date but also by comparison
substantive phases on either side and this is reflected in most with the date from a comparable level from Tell Aqab
chronological charts. (AA-30498 6575 ± 55) which has a calibrated range of 5626-
The dating evidence, however, can be used to suggest that 5469 cal. BC at two standard deviations or the single date
things may not be so straightforward. Leaving the traditional from apparently very late Halaf levels at Tell Kashkashok I
nomenclature unchallenged for the moment, it is difficult to (AA-30497 6845 ± 75) which has a calibrated range of
define either the end of the Halaf or the beginnings of the 5896-5622 cal. BC, again at two standard deviations 97. In
northern Ubaid. In many cases, there are gaps in the sequence summary, it is very difficult to argue that there is clear
where the HUT would be (e.g. Arpachiyah), or the sequence evidence that the Halaf continued, in well documented form,
stops before one would expect the Ubaid to have an effect as late as 5400 cal. BC. It is certainly possible to argue that
(e.g. Domuztepe) or only starts once the Ubaid is established this date could be pushed as early as 5500 cal. BC.
(e.g. Tell Kosak Shamali, Tell Mashnaqa, Tell al-‘Abr, Tell The early Ubaid in the north is better defined with se-
Ziyadeh). At Tepe Gawra, the HUT is probably present but quences associated with a series of radiocarbon dates. Tell
the excavated sequence does not run back to the Halaf continu- Kosak Shamali, Tell Mashnaqa, Tell Ziyadeh, Tell al-‘Abr and
ously. At Yarim Tepe, the transition from the Halaf to the Kenan Tepe have the most useful sets and their absolute
Ubaid may be present in its entirety but too little is published chronologies have been discussed above. It is clear from the
to understand what is happening. The best documented exam- dates from these sites that, by ca 4900 cal. BC, Ubaid ceramic
ple of a relatively continuous sequence across the HUT proba- assemblages were widespread. Tell Mashnaqa, Tell Kosak
bly remains Tell Aqab, although, even here, the exposure is Shamali and Tell al-‘Abr have the earliest dates. The dates
relatively small and is not fully published. from Tell Kosak Shamali do not date the beginning of the
Even more problematic is the absence of dates from the Ubaid there, because there are several deeper levels without
key contexts. The general decline in dates during the period dates. It would be possible to argue that the start of the
from ca 5600 cal. BC and particularly from ca 5300 cal. BC sequence might be as late as ca 5000 cal. BC or perhaps as
early as ca 5200 cal. BC. The earliest level at Tell Mashnaqa
89. MALLOWAN and ROSE, 1935.
seems to fall between ca 5300 and ca 5000 cal. BC. Similarly,
90. DAVIDSON, 1977. it was suggested that the lowest level of Tell al-‘Abr might be
91. e.g. WATKINS and CAMPBELL, 1987. as late as ca 5000 cal. BC or as early as ca 5200 cal. BC.
92. BRENIQUET, 1996.
93. STEIN and ÖZBAL, 2006.
94. FRANGIPANE, 2007. 96. LIBBY, 1955: 52-53; LAWN, 1973: 371-372.
95. Critiqued in more detail in CAMPBELL and FLETCHER, forthcoming. 97. HOLE, 2001: 73.

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 132

132 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
Clearly these chronologies allow considerable debate. One periods of a few generations starts to open up increasingly
could argue that there is a gap about which we know little subtle explanatory models.
archaeologically and almost nothing chronologically, which At times in past discussions, oral as much as in print, the
might be as long as 5500-5000 cal. BC or perhaps as short as major problem with traditional phasing has sometimes seemed
5400-5200 cal. BC (fig. 14). to be not whether it is problematic, but what we can replace
If this gap is closer to the longer period of time, then it it with. It has persisted almost by default and, although schol-
would start to cast considerable doubt on the traditional ars have sometimes been able to propose more subtle and
approaches to the HUT. Firstly, it would be difficult to regard dynamic interpretations 98, I would suggest that this is often in
a period this long as transitional; it might easily be as long spite of the traditional phasing and awkwardly constrained by
as the two periods it is supposedly transitional between. it. The solution to the problem may simply be that there is no
Secondly, the density of knowledge that we have from the need for a replacement. Instead of concentrating on new and
HUT would drop from its already sparse level to the point at better phase-based chronologies, it may be useful to restrict
which it might be questioned whether we know enough to their function to simply ordering the past. Debates about
make any general conclusions. The gap is obviously not neces- whether, for example, Halaf Ia (or whatever name we attach
sarily this long but it remains a matter of interpretation and to it) should be called Halaf can be restricted to how we order
the traditional interpretation of a relatively short transition is time rather than how we interpret past social phenomena. This
clearly not the only possibility. is important because a great deal of archaeology is concerned
I would suggest that one major reason why the HUT has with how we order material assemblages in the past but it can
been approached as a short transitional period is a direct result very usefully be distanced from interpretation—in other
of interpretation being driven by a phase-based chronology. It words, how we understand these past events.
is because the Halaf and Ubaid were conceived as defined In the examples used here, it may superficially seem that
phases within Near Eastern prehistory that they generate the the traditional divisions are being recreated through a rather
question of how the transition between them should be explained tortuous process. There are, in fact, very significant differences.
and relegate that transitional period to a lesser status. Although they share a focus on ceramics, a phenomenon-based
chronology is essentially an inverse of the phase-based one.
The emphasis is not on discrete and relatively stable phases,
between which we seek to date boundaries. Instead it is on
CONCLUSIONS
phases of change—essentially the boundaries themselves. It
makes few assumptions about what is happening between the
The change in emphasis from phases to phenomena in this phenomena so it handles the endemic discontinuities in our
approach is productive in several ways. Because the initial archaeological knowledge more effectively. The approach can
chronologies were based on individual site sequences, chro- easily be extended across traditional boundaries. It is only
nological construction is distanced from interpretation, shortness of space, for example, that means that an examina-
although this is, strictly speaking, only partially true because tion of the decline of painted decoration during the northern
it still rests on the stratigraphic interpretation of each site and Ubaid has not been included here, although this might be seen
sometimes links made on stylistic grounds between different, as a logical extension of a discussion of the painted pottery of
unconnected areas of excavation. Individual site sequences northern Mesopotamia. It also becomes easier to study indi-
need only be linked together as an explicit part of the inter- vidual traits, such as particular motifs or technologies, as
pretation of past phenomena and the links can be very different phenomena with a chronological dimension if they are no
depending on the focus of study. Critically the interpretational longer seen as tied to particular phases.
and chronological logic that justifies the links should be The approach to interpretation and chronology proposed
closely aligned and much more explicit. The emphasis is here is also inherently flexible and multi-faceted. The relevant
placed on periods of change, not as boundaries on a chrono- phenomena can and should vary, depending on the nature of
logical chart that lie between the dominant phases, but as the specific research. Different phenomena may relate to differ-
processes that are important in themselves. The finer chrono- ent aspects of material culture. It seems likely that architecture
logical definition should allow us to move away from explan- or subsistence strategies, for example, may have had very
atory frameworks that, consciously or not, draw on contentious
culture-historical models. The ability to talk about change over 98. e.g. FRANGIPANE, 1996.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 133

Rethinking Halaf Chronologies 133

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
different patterns of change to those of ceramics, perhaps as more data becomes available, hopefully simplifying and
because they are related to very different social stimuli 99. The clarifying as much as complicating. It also gives emphasis to
phenomena of interest can be defined in various ways. They specific aspects of social change as well as highlighting areas
might relate to periods of rapid change or to periods of stabi- where traditional explanations seem unsatisfactory.
lity. They will certainly vary with the scale of the project
because the relevant phenomena are simply those that are
relevant at that scale. A small region may see developments ACKNOWLEDGEMENTS
that are local in extent. A study such as the present one may
look for phenomena that are widespread in space. A significant This article has benefitted from discussions in a wide range of
contribution is simply the need to unpackage the different conferences, seminars, excavations, cafes and pubs over the many
elements of material cultural subsumed within the old culture- years in which it has been pondered, although too many individuals
historical models that persist implicitly in almost all study of have influenced the development of this paper to name them all.
later Mesopotamian prehistory. Thanks are particularly due to colleagues who have provided infor-
In the end, it is not the time periods that necessarily matter mation on radiocarbon dates in advance of publications and, in this
at all. Indeed one could argue that the effort that has been article, specifically B. Parker and L. Swartz for allowing me to
include the dates from Kenan Tepe. The dates from Domuztepe were
directed to assigning dates to chronological phases has actually
funded through the University of California, Los Angeles and the
distracted attention from interpreting past social processes. To
University of Manchester, and thanks are due to E. Carter and
enhance our ability to focus on social interpretation, we need
B. Damiata for their role in organizing them.
to isolate processes of social change and place them in contexts
of time and space. Whether they form part of more general Stuart CAMPBELL
entities is, perhaps, irrelevant. The approach advocated here School of Arts, Histories and Cultures
The University of Manchester
does not present a complete solution and certainly does not
Oxford Road
remove the vast number of chronological uncertainties that Manchester M13 9PL – UNITED KINGDOM
bedevil Near Eastern prehistory. It does, however, offer a stuart.campbell@manchester.ac.uk
flexible way of approaching chronologies that can be extended

BIBLIOGRAPHY

AITCHISON T., SCOTT E., HARKNESS D., BAXTER M. and COOK G. AMIROV S. and DEOPEAK D.
1990 Report on stage 3 of the International Comparison 1997 Morphology of Halafian painted pottery from Yarim Tepe 2,
Programme. Radiocarbon 32,3: 271-278. Iraq (using computerised shape analysis). Baghdader Mittei-
lungen 28: 69-85.
AKKERMANS P.M.M.G.
1989 Excavations at Tell Sabi Abyad. Oxford (BAR Int. Ser. 468). ARZT J.
1993 Villages in the Steppe: Later Neolithic settlement and 2001 Excavations at Tell Ziyadeh, Syria: The northern Ubaid
Subsistence in the Balikh valley, northern Syria. Ann Arbor reconsidered. Unpublished PhD. New Haven, Mass.: Yale
(International Monographs in Prehistory). University.
2000 Old and new perspectives on the origins of the Halaf Culture.
In : ROUAULT O. et WÄFLER M. (éd.), La Djéziré et AURENCHE O., ÉVIN J. et GASCO J.
l’Euphrate syriens de la Protohistoire à la fin du IIe millé- 1987 Une Séquence chronologique dans le Proche-Orient de 12 000
naire av. J.-C. : Tendances dans l’interprétation historique à 3700 BC et sa relation avec les données du radiocarbone.
des données nouvelles : 43-54. Turnhout : Brepols. In : AURENCHE O., ÉVIN J. et HOURS F. (éd.), Chronologies
du Proche-Orient : 21-37. Oxford (BAR Int. Ser. 379).
AKKERMANS P.M.M.G., CAPPERS R., CAVALLO C., NIEUWENHUYSE O.,
NILHAMN B. and OTTE I. AURENCHE O., GALET P., RÉGAGNON-CAROLINE E. and ÉVIN J.
2006 Investigating the Early Pottery Neolithic of northern Syria: 2001 Proto-Neolithic and Neolithic cultures in the Middle East
New evidence from Tell Sabi Abyad. American Journal of – The Birth of agriculture, livestock raising, and ceramics:
Archaeology 110: 123-156. A Calibrated 14C chronology 12,500-5500 cal. BC. Radio-
carbon 43,3: 1191-1202.
BANADORA
99. It is worth noting that HIJARA (1980 and 1997) used very different n.d. BAnque NAtionale de DOnnées RAdiocarbone pour l’Europe
sequences to describe the ceramic and architectural developments at et le Proche-Orient.
Arpachiyah. http://www.archeometrie.mom.fr/banadora/

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 134

134 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
BERNBECK R., POLLOCK S., ALEN S., CASTRO-GESSNER A.G., Iraq. In: LEBEAU M. (ed.), Subartu IV, Vol. I, Landscape,
COSTELLO R., FOREE M., GLEBA M.Y., GODWIN M., LEPINSKI S., Archaeology, Settlement: 39-52. Brussels: Brepols.
NAKAMURA C. and NIEBUHR S. 1999 Archaeological constructs and past reality on the upper
2003 The Biography of an Early Halaf village: Flstlkll Höyük Euphrates. In: DEL OMO LETE G. and MONTERO FENOL-
1999-2000. Istanbuler Mitteilungen des Deutschen Archäoo- LÓS J.-L. (eds), Archaeology of the Upper Syrian Euphrates:
logischen Instituts 53: 9-77. The Tishrin Dam Area: 573-583. Barcelona: Editorial Ausa.
BISCHOFF D. CAMPBELL S. and FLETCHER A.
2006 CANeW 14C databases and 14C charts: Upper Mesopotamia Forthcoming Questioning the Halaf-Ubaid Transition. In: CARTER R.A.
(SE Turkey, N Syria and N Iraq) 11,000-5000 cal. BC. and PHILIP G. (eds), The Ubaid and Beyond: Exploring the
http://www.canew.org/data.html Transmission of Culture in the Developed Prehistoric Socie-
BLACKHAM M. ties of the Middle East. Proceedings of the International
2002 Modeling Time and Transition in Prehistory: The Jordan Conference on the Ubaid, Durham, 20-22 April 2006.
Valley Chalcolithic (5500-3500 BC). Oxford (BAR Int.
CAMPBELL S., CARTER E., HEALEY E., ANDERSON S., KENNEDY A. and
Ser. 1027).
WHITCHER S.
BLACKWELL P. and BUCK C. 1999 Emerging complexity on the Kahramanmaras Plain, Turkey:
2003 The Late Glacial human reoccupation of north-western The Domuztepe Project 1995-1997. American Journal of
Europe: New modelling approaches to space-time model- Archaeology 103: 395-418.
ling. Antiquity 77: 232-240.
CESSFORD C.
BOGOSLAVSKAJA N. 2002 Bayesian statistics and the dating of Çatalhöyük East. In:
1972 On the problem of the origin of the Halaf culture. Sovets- GÉRARD F. and THISSEN L. (eds), The Neolithic of Central
kaya Archeologia 2: 3-16. Anatolia. Internal Developments and External Relations
BÖHNER U. and SCHYLE D. during the 9th-6th Millennia cal. BC: 14-31. Istanbul:
2006 Radiocarbon CONTEXT database 2002-2006. Yayinlari.
http://context-database.uni-koeln.de/
COPELAND L. and HOURS F.
BRAIDWOOD R. and BRAIDWOOD L. 1987 The Halafians, their predecessors and their contemporaries
1960 Excavations in the Plain of Antioch I. Chicago: University in northern Syria and the Levant: Relative and absolute
of Chicago (Oriental Institute Publications 50). chronologies. In: AURENCHE O., ÉVIN J. and HOURS F. (eds),
BRENIQUET C. Chronologies in the Near East, Relative chronologies and
1996 La Disparition de la culture de Halaf : les origines de la absolute chronology 16,000-4,000 BP: 401-426. Oxford
culture d’Obeid dans le Nord de la Mésopotamie. Paris : (BAR Int. Ser. 379).
Éditions Recherche sur les Civilisations. CRUELLS W.
BRONK RAMSEY C. 2006a Chantiers E et F. Les Datations C14. In : TUNCA Ö. et
1995 Radiocarbon calibration and analysis of stratigraphy: The BAGHDO A. et la collaboration de CRUELLS W. (éd.), Chagar
OxCal program. Radiocarbon 37,2: 425-430. Bazar (Syrie) I. Les Sondages Préhistoriques (1999-2001) :
2000 Comment on “The use of Bayesian statistics from 14C dates 117-119. Louvain : Peeters.
of chronologically ordered samples: A Critical review”. 2006b Chagar Bazar Préhistorique. Esquisse de la séquence d’évo-
Radiocarbon 42,2: 199-202. lution et la périodisation. In : TUNCA Ö. et BAGHDO A. et
2001 Development of the radiocarbon calibration program OxCal. la collaboration de CRUELLS W. (éd.), Chagar Bazar (Syrie)
Radiocarbon 43,2A: 355-363. I. Les Sondages Préhistoriques (1999-2001) : 121-142.
BUCK C., LITTON C. and SCOTT E. Louvain : Peeters.
1994 Making the most of radiocarbon dating – Some statistical CRUELLS W. and NIEUWENHUYSE O.
considerations. Antiquity 68: 252-263. 2004 The Proto-Halaf period in Syria. New sites, new data.
BUCK C., LITTON C. and SMITH A. Paléorient 30,1: 47-68.
1992 Calibration of radiocarbon results pertaining to related archae-
CRUELLS W., MOLIST M. and TUNCA Ö.
ological events. Journal of Archaeological Science 19,5:
2004 Tell Amarna in the general framework of the Halaf period.
497-512.
In: TUNCA Ö. et MOLIST M. (éd.), Tell Amarna (Syrie) I :
BUCK C., KENWORTHY J., LITTON C. and SMITH A. 261-283. Louvain : Peeters.
1991 Combining archaeological and radiocarbon information –
A Bayesian approach to calibration. Antiquity 65: 808-821. DAVIDSON T.E.
1977 Regional Variation within the Halaf Ceramic Tradition.
CAMPBELL S. Unpublished PhD. Edinburgh: University of Edinburgh.
1992 Culture, Chronology and Change in the Later Neolithic of
North Mesopotamia. Unpublished PhD. Edinburgh: Univer- FRANGIPANE M.
sity of Edinburgh. 1996 La Nascita dello Stato nel Vicino Oriente. Rome: Laterza.
1996 Rescue excavations at Ginnig and Khirbet Garsour in the 2007 Different types of egalitarian societies and the development
north Jazira. A summary report. Sumer 48: 6-10. of inequality in Early Mesopotamia. World Archae-
1998 Problems of definition: The Origins of the Halaf in North ology 39(2): 151-176.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 135

Rethinking Halaf Chronologies 135

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
GUSTAVSON-GAUBE C. revisited. In: TSUNEKI A. and MIYAKE Y. (eds), Excavations
1981 Shams ed-Din Tannira; The Halafian pottery of Area A. at Tell Umm Qseir in the Middle Khabur Valley, North
Berytus 29: 9-182. Syria: 177-188. Tsukuba: University of Tsukuba.
HAMMADE H. and YAMAZAKI Y. MUNCHAEV R. and MERPERT N.
2006 Tell al-’Abr (Syria): Ubaid and Uruk Periods. Louvain: 1981 Earliest Agricultural Settlements of Northern Mesopotamia.
Peeters. Moscow: Nauka.
HERTZFELD E. NIEUWENHUYSE O.
1930 Die Ausgrabungen von Samarra V. Berlin: Dietrich Reimer. In press Plain and Painted Pottery. The Rise of Late Neolithic cera-
mic styles on the Syrian plains. Brussels: Brepols.
HIJARA I.
1980 Arpachiyah 1976. Iraq 42: 131-154. NISHIAKI Y.
1997 The Halaf Period in Northern Mesopotamia. London: Nabu 2001 Radiocarbon dates and the absolute chronology of Tell
Publications. Kosak Shamali. In: NISHIAKI Y. and MATSUTANI T. (eds),
Tell Kosak Shamali, volume I: 153-157. Tokyo: The Univer-
HOLE F.
sity Museum.
2001 A Radiocarbon chronology for the Middle Khabur, Syria.
Iraq 63: 67-98. OATES J.
1960 Ur and Eridu, The Prehistory. Iraq 22: 32-50.
HOURS F., AURENCHE O., CAUVIN J., CAUVIN M.-C., COPELAND L. et
SANLAVILLE P. OPPENHEIM M. VON
1994 Atlas des Sites du Proche-Orient (14 000-5700 BP). Lyon 1943 Tell Halaf. Vol. 1. Die Prähistorische Funde. Berlin: Walter
(Travaux de la Maison de l’Orient méditerranéen 24). de Gruyter and Co.
IWASAKI T. and TSUNEKI A. ÖZBAL R., GERRITSEN F., DIEBOLD B., HEALEY E., AYDIN N.,
2003 Archaeology of the Rouj Basin. Vol. I. Tsukuba: University LOYETTE M., NARDULLI F., REESE D., EKSTROM H., SHOLTS S., MEKEL-
of Tsukuba. BROBOV N. and LAHN B.
2004 Tell Kurdu excavations 2001. Anatolica 30: 37-107.
JONES S.-A.
2004 Taxonomies of symbolism and style; Re-considering the PARKER B. and DODD L.
social significance of incipient decorated pottery from the 2005 The Upper Tigris Archaeological Research Project
sixth millennium in the Near East. Unpublished PhD. (UTARP): A Preliminary report from the 2002 field season.
Manchester: University of Manchester. Anatolica 30: 69-110.
KITAGAWA H. and VAN DER PLICHT J. PARKER B., DODD L., CREEKMORE A., HEALEY E. and PAINTER C.
1998 AMS 14C results. In: TSUNEKI A. and MIYAKE Y. (eds), 2006 The Upper Tigris Archaeological Research Project
Excavations at Tell Umm Qseir in the Middle Euphrates (UTARP): A Preliminary report from the 2003 and 2004
Valley, North Syria: 214. Tsukuba: University of Tsukuba. field seasons at Kenan Tepe. Anatolica 32: 72-151.
LAWN B. PERKINS A.
1973 University of Pennsylvania radiocarbon dates XV. Radio- 1949 The Comparative Archaeology of Early Mesopotamia.
carbon 15: 367-381. Chicago: University of Chicago Press.
LIBBY W. PHILIP G. and MILLARD A.
1955 Radiocarbon Dating. Chicago: University of Chicago Press. 2000 Khirbet Kerak Ware in the Levant: The Implications of
radiocarbon chronology and spatial distribution. In :
LLOYD S. and SAFAR F.
MARRO C. et HAUPTMANN H. (éd.), Chronologies des pays
1945 Tell Hassuna – Excavations by the Iraq Government Direc-
du Caucase et de l’Euphrate aux IVe-IIIe millénaires :
torate General of Antiquities in 1943-1944. Journal of Near
279-296. Paris : Institut français d’études anatoliennes
Eastern Studies 4: 255-289.
d’Istanbul, diff. De Boccard (Varia Anatolica XI).
MALLOWAN M.E.L. and ROSE J.C.
REIMER P., BAILLIE M., BARD E., BAYLISS A., BECK J., BERTRAND C.,
1935 Excavations at Tall Arpachiyah, 1933. Iraq 2: 1-178.
BLACKWELL P., BUCK C., BURR G., CUTLER K., DAMON P., EDWARDS R.,
MELLAART J. FAIRBANKS R., FRIEDRICH M., GUILDERSON T., HOGG A., HUGHEN K.,
1978 Earliest Civilisations of the Near East. London: Thames KROMER B., MCCORMAC G., MANNING S., BRONK RAMSEY C.,
and Hudson. REIMER R., REMMELE S., SOUTHON J., STUIVER M., TALAMO S.,
1981 The Prehistoric Pottery from the Neolithic to the Beginning TAYLOR F., VAN DER PLICHT J. and WEYHENMEYER C.
of EB4. In: MATTHERS J. (ed.), The River Qoueiq, Northern 2004 IntCal04 terrestrial radiocarbon age calibration,
Syria and its Catchment: 131-326. Oxford (BAR Int. 0-26 cal kyr BP. Radiocarbon 46,3: 1029-1058.
Ser. 98).
ROAF M. and KILLICK M.
MERPERT N., MUNCHAEV R. and BADER N. 1987 A Mysterious Affair of Styles: The Ninevite 5 pottery of
1976 Investigations of the Soviet expedition to Iraq 1973. northern Mesopotamia. Iraq 49: 199-230.
Sumer 32: 25-62.
SCOTT E.
MIYAKE Y. 2003 The Third International Radiocarbon Intercomparison
1998 New light on the Middle Halaf Period: Halaf chronology (TIRI) and the Fourth International Radiocarbon Intercom-

Paléorient, vol. 33-1, p. 103-136 © CNRS ÉDITIONS 2007


CNRS Éditions - Paléorient-num 33-1 - - 22 x 28 - 30/11/2007 - 16 : 22 - page 136

136 S. CAMPBELL

• Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS • Tirés à part CNRS ÉDITIONS •
parison (FIRI), 1990-2002. Results, analyses, and conclu- IIe millénaire av. J.-C. : Tendances dans l’interprétation
sions. Radiocarbon 45,2: 135-408. historique des données nouvelles : 71-80. Turnhout :
STEIER P. and ROM W. Brepols.
2000 The Use of Bayesian statistics for 14C dates of chronologi- TOBLER A.
cally ordered samples: A Critical analysis. Radiocar- 1950 Excavations at Tepe Gawra. Vol. 2. Philadelphia: University
bon 42,2: 183-198. of Pennsylvania Press.
STEIN G.J. and ÖZBAL R. WATKINS T. and CAMPBELL S.
2006 A Tale of Two Oikumenai: Variation in the expansionary 1987 The Chronology of the Halaf culture. In: AURENCHE O.,
dynamics of ‘Ubaid and Uruk Mesopotamia. In: STONE E.C. ÉVIN J. and HOURS F. (eds), Chronologies in the Near East,
(ed.), Settlement and Society: Ecology, urbanism, trade Relative Chronologies and Absolute Chronology
and technology in Mesopotamia and Beyond (Robert
16,000-4,000 BP: 427-464. Oxford (BAR Int. Ser. 379).
McC. Adams Festschrift): 356-370. Los Angeles: Cotsen
Institute of Archaeology. WATSON P. and LEBLANC S.
1990 Girikihaciyan: A Halafian Site in Southeastern Turkey. Los
TAY
2006 TAY-Project database (Türkiye Arkeolojik YerleŞmeleri). Angeles: Institute of Archaeology, University of California.
www.tayproject.org WRIGHT H. and RUPLEY E.
THISSEN L. 2001 Calibrated radiocarbon age determinations of Uruk-related
2006 CANeW 14C databases and 14C charts: Central Anatolia and assemblages. In: ROTHMAN M. (ed.), Uruk Mesopotamia and
Cilicia: 10,000-5000 cal. BC. its Neighbours: Cross-Cultural Interactions in the Era of
http://www.canew.org/data.html State Formation: 85-122. Santa Fe: School of American
Research Press.
THUESEN I.
2000 Ubaid expansion in the Khabur. New evidence from Tell YENER K., EDENS C., HARRISON T., VERSTRAETE J. and WILKINSON T.
Mashnaqa. In : ROUAULT O. et WÄFLER M. (éd.), La Djéziré 2000 The Amuq Valley Regional Project, 1995-2000. American
et l’Euphrate syriens de la Protohistoire à la fin du Journal of Archaeology 104: 163-220.

Paléorient, vol. 33.1, p. 103-136 © CNRS ÉDITIONS 2007

Vous aimerez peut-être aussi