journal of archaeological science · journal of archaeological ... doi:10.1016/j.jas.2009.05.029...

15
Ceramic production and provenience at Gordion, Central Anatolia Peter Grave a, * , Lisa Kealhofer b , Ben Marsh c , G. Kenneth Sams d , Mary Voigt e , Keith DeVries f a Archaeology & Palaeoanthropology, University of New England, C02 Building, Armidale NSW, Australia b Anthropology/Environmental Studies Institute, Santa Clara University, Santa Clara, CA, USA c Geography and Environmental Studies, Bucknell University, Lewisburg, PA, USA d Department of Classics, University of North Carolina, Chapel Hill, NC, USA e Anthropology, College of William and Mary, Williamsburg, VA, USA f University of Pennsylvania Museum of Archaeology & Anthropology, Philadelphia, PA, USA article info Article history: Received 2 February 2009 Received in revised form 22 May 2009 Accepted 28 May 2009 Keywords: Turkey Political economy NAA Anatolian Iron Age ceramics project abstract Phrygian Gordion was the political center of an influential Iron Age polity that extended across west central Anatolia during the first half of the 1st millennium BC. Though the borders of this polity remain vague a characteristic of the Phrygian ‘‘footprint’’ is the distribution of highly distinctive ceramics. The extent to which Gordion potters were the originators of these wares remains uncertain. In this paper we use Neutron Activation Analysis (NAA) to establish the local signature of predominantly Iron Age ceramics for this site by combining samples from several decades of excavation with an extensive regional sediment sequence. We also compare previous NAA work at Gordion to suggest that the formative stages of the Phrygian state appears to have involved a more extensive network of non-local specialist producers than previously thought. Crown Copyright Ó 2009 Published by Elsevier Ltd. All rights reserved. 1. Introduction Understanding the political and economic organization of the Iron Age state of Phrygia in central Anatolia requires data related to local production as well as regional exchange patterns. In this paper we seek to characterize local ceramic production at Gordion, the capital of the Phrygian state, so that we can better understand the complexity of both local and regional exchange patterns in relation to the Phrygian political economy (Fig. 1). To achieve this we use Neutron Activation Analysis (NAA) to compositionally compare a relatively large sample of sediments from the Gordion region with excavated ceramics from the site. Establishing ‘local’ production for ceramics (or additive tech- nologies) is often not as straightforward as establishing prove- nience for raw materials like obsidian. Social, economic, and technological variables combine to alter the geo-chemical finger- print of the clay sources potters used. In addition, the identification of the original quarries – in this case clay beds – is often impossible, given both erosional and depositional processes in what are commonly highly altered landscapes. In this paper, we define ‘local’ production through a systematic sampling of the baseline geology, a ‘bottom up’ ceramic sampling strategy, and through assumptions based on relative group sizes. A previous NAA study of Gordion ceramics (Henrickson and Blackman, 1996) defined not only local production, but provided a set of interpretations about the changing production and economic trajectories at Gordion from the Late Bronze Age into the Iron Age. Here, we re-analyze these datasets in relation to our geologically established ‘local.’ Combining these two large NAA datasets provides a new perspective on the nature of production and exchange during the Iron Age in Phrygia. 2. Background 2.1. Gordion The archaeological site of Gordion (modern Yassıho ¨yu ¨ k), 100 km SW of Ankara in central Turkey, has a long sequence of occupation, from at least the Early Bronze Age through to the Medieval period (Fig. 1). By the early 1st millennium BC, Gordion became the political base for the emerging state of Phrygia, which controlled much of inland western Anatolia over the first half of the 1st millennium BC (Sams, 1995; Voigt and Henrickson, 2000). The largest scale settle- ment at the site occurred during the Iron Age and the subsequent Hellenistic period. Best known historically as the seat of King Midas, the Gordion landscape includes an impressive array of more than 100 burial mounds of Phrygian and later Hellenistic elites. Gordion lies on the floodplain of the Sakarya River within a broad valley system (Fig. 1b). Tertiary evaporites and pale silts * Corresponding author. E-mail address: [email protected] (P. Grave). Contents lists available at ScienceDirect Journal of Archaeological Science journal homepage: http://www.elsevier.com/locate/jas 0305-4403/$ – see front matter Crown Copyright Ó 2009 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2009.05.029 Journal of Archaeological Science 36 (2009) 2162–2176

Upload: trananh

Post on 09-Jun-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

lable at ScienceDirect

Journal of Archaeological Science 36 (2009) 2162–2176

Contents lists avai

Journal of Archaeological Science

journal homepage: ht tp: / /www.elsevier .com/locate/ jas

Ceramic production and provenience at Gordion, Central Anatolia

Peter Grave a,*, Lisa Kealhofer b, Ben Marsh c, G. Kenneth Sams d, Mary Voigt e, Keith DeVries f

a Archaeology & Palaeoanthropology, University of New England, C02 Building, Armidale NSW, Australiab Anthropology/Environmental Studies Institute, Santa Clara University, Santa Clara, CA, USAc Geography and Environmental Studies, Bucknell University, Lewisburg, PA, USAd Department of Classics, University of North Carolina, Chapel Hill, NC, USAe Anthropology, College of William and Mary, Williamsburg, VA, USAf University of Pennsylvania Museum of Archaeology & Anthropology, Philadelphia, PA, USA

a r t i c l e i n f o

Article history:Received 2 February 2009Received in revised form22 May 2009Accepted 28 May 2009

Keywords:TurkeyPolitical economyNAAAnatolian Iron Age ceramics project

* Corresponding author.E-mail address: [email protected] (P. Grave).

0305-4403/$ – see front matter Crown Copyright � 2doi:10.1016/j.jas.2009.05.029

a b s t r a c t

Phrygian Gordion was the political center of an influential Iron Age polity that extended across westcentral Anatolia during the first half of the 1st millennium BC. Though the borders of this polity remainvague a characteristic of the Phrygian ‘‘footprint’’ is the distribution of highly distinctive ceramics. Theextent to which Gordion potters were the originators of these wares remains uncertain. In this paper weuse Neutron Activation Analysis (NAA) to establish the local signature of predominantly Iron Ageceramics for this site by combining samples from several decades of excavation with an extensiveregional sediment sequence. We also compare previous NAA work at Gordion to suggest that theformative stages of the Phrygian state appears to have involved a more extensive network of non-localspecialist producers than previously thought.

Crown Copyright � 2009 Published by Elsevier Ltd. All rights reserved.

1. Introduction

Understanding the political and economic organization of theIron Age state of Phrygia in central Anatolia requires data related tolocal production as well as regional exchange patterns. In this paperwe seek to characterize local ceramic production at Gordion, thecapital of the Phrygian state, so that we can better understand thecomplexity of both local and regional exchange patterns in relationto the Phrygian political economy (Fig. 1). To achieve this we useNeutron Activation Analysis (NAA) to compositionally comparea relatively large sample of sediments from the Gordion region withexcavated ceramics from the site.

Establishing ‘local’ production for ceramics (or additive tech-nologies) is often not as straightforward as establishing prove-nience for raw materials like obsidian. Social, economic, andtechnological variables combine to alter the geo-chemical finger-print of the clay sources potters used. In addition, the identificationof the original quarries – in this case clay beds – is often impossible,given both erosional and depositional processes in what arecommonly highly altered landscapes. In this paper, we define ‘local’production through a systematic sampling of the baseline geology,a ‘bottom up’ ceramic sampling strategy, and through assumptionsbased on relative group sizes.

009 Published by Elsevier Ltd. All

A previous NAA study of Gordion ceramics (Henrickson andBlackman, 1996) defined not only local production, but provideda set of interpretations about the changing production andeconomic trajectories at Gordion from the Late Bronze Age into theIron Age. Here, we re-analyze these datasets in relation to ourgeologically established ‘local.’ Combining these two large NAAdatasets provides a new perspective on the nature of productionand exchange during the Iron Age in Phrygia.

2. Background

2.1. Gordion

The archaeological site of Gordion (modern Yassıhoyuk), 100 kmSW of Ankara in central Turkey, has a long sequence of occupation,from at least the Early Bronze Age through to the Medieval period(Fig.1). By the early 1st millennium BC, Gordion became the politicalbase for the emerging state of Phrygia, which controlled much ofinland western Anatolia over the first half of the 1st millennium BC(Sams, 1995; Voigt and Henrickson, 2000). The largest scale settle-ment at the site occurred during the Iron Age and the subsequentHellenistic period. Best known historically as the seat of King Midas,the Gordion landscape includes an impressive array of more than100 burial mounds of Phrygian and later Hellenistic elites.

Gordion lies on the floodplain of the Sakarya River withina broad valley system (Fig. 1b). Tertiary evaporites and pale silts

rights reserved.

Fig. 1. a: map of Turkey showing location of Gordion (Yassihoyuk); b: composite map of geology and topography for Gordion and hinterland with locations of sediment samples andcompositional group attributions as discussed in text and presented in Table 2a and b.

Fig. 2. Histogram of the Gordion sample population by chronological phase (EP¼ EarlyPhrygian:- 10th–9th c. BCE; MP¼Middle Phrygian:- 8th–mid 6th. c. BCE; LP¼ LatePhrygian:- mid 6th–mid 4th. c. BCE; Hellenistic:- mid 4th–early 2nd c. BCE; Roman:-1st BCE–3rd c. CE).

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–2176 2163

dominate the landscape near the site and into the inhospitablelands to the west (Erentoz, 2002). Toward the east, the geologyshifts to Tertiary continental clastics, and later basalts in theuplands. Locally some sediments are heavily altered by hydro-thermal processes. Heavily weathered soils and pediments overlaythe bedrock. Much of this material was removed by erosion,beginning by the Early Bronze Age, that redeposited sedimentswithin the small stream floodplains. The Sakarya River has aggra-ded deeply since the Bronze Age with pale silts.

The site has had three major phases of excavation. The earliestwas by the Korte brothers at the turn of the 20th century (Korte andKorte, 1904). The second phase followed World War II, whenRodney Young re-initiated excavations both of the Citadel Moundand of adjacent tumuli (Young,1951). On the mound, Young focusedon the ‘Palace Area’, and ultimately exposed a ‘‘Destruction Level’’that he believed dated to the time of the Kimmerian invasion andcaused the collapse of Phrygia. Young’s excavations ended abruptlyafter his death in 1974. The third and latest phase began in the late1980s, when excavations and survey resumed at Gordion underProject Director G. Kenneth Sams and Field Director Mary Voigt(Voigt, 1994; Voigt, DeVries, et al., 1997). Voigt’s goals were tobetter define the stratigraphic sequence at the site, establish a morerefined absolute chronology, explore non-elite areas of the site, andgenerally gain a greater understanding of the domestic andindustrial side of Gordion’s occupations.

The most recent phase of excavation included systematic studyof local ceramic production from the Late Bronze Age through to theHellenistic period (Henrickson, 1993, 1994, 1995; Henrickson andBlackman, 1996; Henrickson, Vandiver, et al., 2002). An importantcomponent of that work was the comparison of large scale ceramicproduction at the site for the Late Bronze Age (YHSS 9-8, c. 1400–1200 BC) and the Early Phrygian period (YHSS 6B, 950–800 BC).This was done using elemental data (NAA) for a comparatively largesample of ceramics excavated by Voigt from well defined archae-ological contexts as well as samples of clay from the local region(Henrickson and Blackman, 1996). Henrickson and Blackmandocumented a major shift in resource use between the two periodsand flagged the character of local production at Gordion as highlycomplex. One of the more surprising aspects of their study was theidentification of compositional groups as either Late Bronze or EarlyPhrygian with little overlap between the two periods.

In 2003 we commenced a large scale assessment of non-localceramics at Iron Age sites across Western Anatolia (Anatolian IronAge ceramics project (AIA): http://aia.une.edu.au). The project goalsfocused on understanding exchange and emulation during the IronAge, as new political economies emerged after the collapse of theLate Bronze Age empires in the eastern Mediterranean, particularlyAnatolia. Gordion, as one of the best excavated Iron Age sites in theregion, provided a foundation for developing a more nuancedunderstanding of emerging regional polities. In addition, we wereable to build on a settlement and landscape survey project whichprovided a detailed geological and geomorphological backgroundfor defining local sediments and their ancient distribution(Kealhofer, 2005; Marsh, 2005).

This paper has two aims. The first is to present the results of ourfirst phase of ceramic analyses for Gordion (2003–2005), andspecifically to define the pattern of non-local and locally madeceramics. A broad methodological goal of the AIA project is toincorporate legacy NAA datasets to extend the scale and research

Table 1NAA results for three standard reference materials (SRM 697, 2711 and 1633b), National Institute for Standards and Technology, Washington D.C. Table shows experimentalresults for replicates measured during the analysis of the Gordion ceramic sample presented in this paper. Results are given as mean values with % coefficient of variation (C.V.)alongside certified/published values for each element and the deviation of the experimental mean from the certified/published values (% recovery). Elements reported as partsper million (ppm) unless otherwise indicated.

SRM 1633b (n¼ 4) SRM 679 (n¼ 4) SRM 2711 (n¼ 5) ppm SRM 1633b SRM 679 SRM 2711

Avg. C.V. Avg. C.V. Avg. C.V. Cert/pub % Recovery Cert/pub % Recovery Cert/pub % Recovery

Ba 667.50 9.02 442.09 8.08 684.67 3.31 Ba 709.00 94.15 432.20 102.29 726.00 94.31Ca% 1.95 22.18 – – 2.81 11.42 Ca% 1.51 128.97 0.16 – 2.88 97.71Ce 181.40 2.07 102.48 0.95 72.38 5.30 Ce 190.00 95.47 105.00 97.60 69.00 104.89Co 48.94 3.56 25.77 2.19 10.07 7.87 Co 50.00 97.88 26.00 99.12 10.00 100.68Cr 203.90 4.59 109.06 2.53 47.58 6.62 Cr 198.20 102.87 109.70 99.42 47.00 101.23Cs 10.31 2.68 9.65 7.22 6.57 4.29 Cs 11.00 93.75 9.60 100.52 6.10 107.67Eu 3.96 1.71 1.82 6.67 1.09 3.39 Eu 4.10 96.52 1.90 95.92 1.10 99.27Fe% 7.79 2.43 9.06 1.01 2.88 2.62 Fe% 7.78 100.10 9.05 100.08 2.89 99.72Hf 6.78 2.58 4.18 3.72 7.61 1.41 Hf 6.80 99.71 4.60 90.82 7.30 104.27K% – – 2.01 35.02 2.81 18.07 K% 1.95 – 2.43 82.61 2.45 114.78La 87.86 1.27 50.55 0.97 37.63 2.69 La 94.00 93.47 – – 40.00 94.08Lu 1.02 8.98 0.53 7.70 0.45 4.65 Lu 1.20 84.58 – – – –Na 0.20 0.00 0.13 0.00 0.88 55.98 Na 0.20 99.50 0.13 99.69 1.14 77.19Rb 132.09 18.72 177.09 10.57 109.02 9.19 Rb 140.00 94.35 190.00 93.20 110.00 99.11Sb 5.07 3.47 0.78 1.18 19.61 3.64 Sb 6.00 84.50 – – – –Sc 40.08 1.13 22.42 9.21 9.27 1.52 Sc 41.00 97.76 22.50 99.64 9.00 102.98Sm 18.26 1.62 9.03 0.76 5.93 2.42 Sm 20.00 91.28 – – 5.90 100.47Ta 2.25 26.01 1.31 22.54 1.59 23.67 Ta 1.80 125.00 – – 2.47 64.29Tb 2.71 4.54 1.25 12.89 0.81 12.48 Tb 2.60 104.23 – – – –Th 25.13 1.44 13.92 3.02 13.41 2.81 Th 25.70 97.78 14.00 97.60 14.00 95.79U 8.05 11.27 2.12 16.06 2.58 7.80 U 8.79 91.52 – – 2.60 99.31Yb 7.26 2.21 3.63 3.72 3.00 4.61 Yb 7.60 95.56 – – 2.70 111.04Zn 186.88 14.57 112.81 8.36 349.17 4.26 Zn 210.00 88.99 150.00 69.50 350.40 99.65

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–21762164

depth of our analytic program. Given the previous extensive NAAdataset of Henrickson and Blackman (1996), a further aim of thepresent study is to compare and contrast the analysis and inter-pretations of both NAA datasets. Together this extensive corpus ofmaterial, from both Young (this study) and Voigt excavations(Henrickson and Blackman, 1996 and this study), provides theplatform for more detailed, future analyses of exchange andemulation during the Phrygian periods.

3. Methodology

Understanding trade and exchange is predicated on the possi-bility of differentiating between local and non-local production. Toachieve this we employ two strategies. First, we work closely withsite ceramicists with long experience with the local and non-localassemblages to target ceramics defined as local, non-local, andunknown based on typological and fabric criteria. Second, thegeomorphology of the region is studied, and the regionalsedimentary sequences are differentiated. In this case, prior workby project geomorphologist, both in relation to the site and thefloodplain and in relation to the regional catchments, provided anin-depth understanding of the ancient regional environments andsediment sequences (Marsh, 1999, 2005).

In 2003, the ceramics available for analysis included a limitednumber of sample bags saved from specific Young excavation contextsthat had not been registered, as well as a much larger array of samplesfrom the more recent Voigt excavations (although a smaller array ofPhrygian period material). With the aid of Keith DeVries, Ken Sams,and Robert Henrickson an assemblage of ceramics from these twoexcavations was sampled (Fig. 2). Of this set, 146 Young excavationsamples and 133 Voigt excavation samples were chosen for analysishere (n¼ 279). Most samples were chosen specifically because theywere thought to be non-local. Some samples, however, were includedas a measure of what ceramicists identified as local.

A total of 73 sediment samples were also chosen for analysis, 24from stream sediment cores and 49 from different geologicalsediments around the region (Fig. 1b, Table 2a). Rather than

attempting to find ancient clay beds, most likely buried in thisheavily eroded landscape, the focus here is on characterizing theentire geological and sedimentary variability in a w20 km radiusaround the site.

Elsewhere we have detailed our sampling, processing, andanalysis procedures (Grave, Kealhofer et al., 2008; Kealhofer, Grave,et al., in press). In summary, ceramic and sediment samples arephotographed and recorded in the field, and prepared at theUniversity of New England for NAA. Comparatively large (1 g)samples are submitted for NAA, with the advantage of minimizingmeasurement distortions due to sample heterogeneity, a particularconcern in the analysis of sediments and coarse ceramics. The NAAdataset, composed of twenty three elements with good countingstatistics are then processed through an iterative multivariateroutine of Principal Components Analysis (PCA) and CanonicalVariates Analysis (CVA), identifying and removing outliers, identi-fying and removing non-local groups, and ultimately differenti-ating the range of locally produced geo-chemical groups.

3.1. Standards

We routinely include replicates of three NIST standards (SRM679, SRM 1633b, and SRM 2711) in the NAA sample runs both asquality control checks for individual datasets and for individualelements, and as an important element in published NAA datasetsto enable long term comparison and correction by future studies(Table 1).

3.2. Sediment fitting

Our approach assumes that geologically comprehensive sedi-ment collection will contain samples that approximate theelemental profile of the local ceramic signature (see also Kealhofer,Grave, et al., in press). However, sediments are not the same asclays; minimally, they will have a coarser rock component andexhibit a greater degree of compositional heterogeneity. Typically,summed NAA results for each sediment sample are up to 30%

Table 2aDescription, and UTM locations for Gordion sediments used in this study and identified in Fig. 1b.

Grp# AIA # Description Field numbers UTM zone UTM E UTM N

i 305 Silty core @ 80 cm Suluklu floodplain 01-3-1 36 416842 4388284i 310 Silt bank @ 45 cm, marl & basalt sed 01-5-1 36 416853 4382199i 318 Silty core @ 225 cm basaltic sed near river 01-7-7 36 413966 4396129i 319 Silty core @ 295 cm basaltic sed near river 01-7-10 36 413966 4396129i 320 Sandy core @ 250 cm near basalt mts 01-8-5 36 421170 4388782i 2341 Surface sample, BA mound wash 06_34 36 422164 4387735i 2342 Sandy silt sed below pediment 06_35 36 422263 4387101i 2346 Sandy loam basaltic pediment sed, dark 06_39 36 420728 4386516i 2361 Dark clay surface loam Sabanozu plain 06_54 36 416681 4396202i 3972 Sand-sized basalt small stream sed @ Sabanozu G6 36 422898 4397314

ii 307 Silty core @ 308 cm Suluklu floodplain 01-3-11 36 416842 4388284ii 308 Gravelly bank, @ 85 cm basaltic sed 01-4-4 36 420953 4387884ii 321 Sandy core @ 530 cm near basalt mts 01-8-11 36 421170 4388782ii 2350 Pediment sed in low-basalt catchment 06_43 36 417099 4383634ii 2353 Silt-loam basalt-rich sed below Uçpinar 06_46 36 421455 4380644ii 2354 Light-colored silt-loam sed below Uçpinar 06_47 36 421634 4380797

iii 306 Silty core @ 165 cm Suluklu floodplain 01-3-2 36 416842 4388284iii 322 Sandy core @ 775 cm near basalt mts 01-8-15 36 421170 4388782iii 323 Silt-sand core @ 255 cm, near tumulus MM 01-9A-1 36 415166 4390094iii 324 Silt-sand core @ 360 cm, near tumulus MM 01-9A-4 36 415166 4390094iii 325 Silt-sand core @ 470 cm, near tumulus MM 01-9A-8 36 415166 4390094iii 326 Pale silt core @ 165 cm below marl banks 01-12-4 36 408846 4392693iii 2332 Red expansive clay in Suluklu fan at river 06_25 36 412977 4389945iii 2338 Fresh sandy basaltic sediment below Çekerdeksız 06_31 36 418663 4386573iii 2343 Sandy basaltic pediment sed 06_36 36 422164 4386594iii 2352 Weather basalt near Dua Dag Rd. 06_45 36 419407 4384078iii 2355 Silty BA mound material at Çekerdeksız 06_48 36 418889 4388170iii 2359 Silt-clay stream bottom below Sabanozu 06_52 36 419865 4396518

iv 311 Silt bank @ 155 cm, marl & basalt sed 01-5-7 36 416853 4382199iv 312 Silt bank @ 230 cm, marl & basalt sed 01-5-11 36 416853 4382199iv 313 Silty core @ 160 cm deep silty floodplain 01-6-3 36 416460 4392478iv 314 Silty core @ 245 cm deep silty floodplain 01-6-6 36 416460 4392478iv 315 Silty core @ 330 cm deep silty floodplain 01-6-9 36 416460 4392478iv 316 Silty core @ 560 cm deep silty floodplain 01-6-18 36 416460 4392478iv 317 Silty core @ 75 cm basaltic sed near river 01-7-1 36 413966 4396129iv 2351 Clay-rich soil, continental clastics 06_44 36 417293 4383902

v 2331 Brick wash in Middle Phrygian layer, Citadel mound 06_24 36 412266 4389533v 2356 Silty surface sed Suluklu tributary 06_49 36 419793 4389132v 2357 Silt from deep historic sedimentation T4 area 06_50 36 418410 4390761v 2358 Pale silt low pediment SW Sabanozu 06_51 36 418395 4395482v 2360 Silty abandoned plain Sabanozu stream 06_53 36 417169 4394758v 2370 Silty M Phrygian brick wash S edge Citadel mound 06_63 36 412428 4389361

vi 327 Pale silt core @ 265 cm below marl banks 01-12-8 36 408846 4392693vi 328 Pale silt core @ 360 cm below marl banks 01-12-12 36 408846 4392693vi 329 Silty core @ 308 cm Suluklu floodplain 01-3-7 36 416842 4388284vi 2333 Silt Sakarya R. sed N of mound @ 200 cm 06_26 36 412246 4389805vi 2334 Silt Sakarya R. sed N of mound @ 400 cm 06_27 36 412246 4389805vi 2335 Gleyed silt-clay in Sakara R. dredge pile 06_28 36 412012 4389442vi 2336 Silt Sakarya sed S. of mound @100 cm 06_29 36 411799 4389049vi 2367 Clay Porsuk floodplain dredgings at Kiranharman 06_60 36 411645 4392510vi 2369 Silt Sakarya sed 300 cm 06_62 36 412279 4389836vi 3969 Marly slope wash G2 36 413799 4392952

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–2176 2165

below summed values for ceramic samples. This difference reflectsthe diluting effects of dominant but non-measured elements (i.e.silicon and magnesium in basalts and clastics, and sulphur inevaporites). The compositional differences between ceramic andsediment profiles can be expressed as the % difference betweenelemental averages (Table 3). The majority of elements are moreconcentrated in the ceramics but are not equally weighted. Toaccommodate this variability the multivariate sediment centroid ismade equivalent to the multivariate ceramic centroid1:

1 Where Ceramicavg is the average of all of the dataset’s value for an individualelement, and Sedimentavg is the average of all of the sediment dataset’s values forthe same element. Sedimentsample is the value of a single sample for the sameelement.

¼ Ceramicavg=Sedimentavg *Sedimentsample

� �

This fitting technique, effectively aligning both datasets inmultivariate space facilitates matching likely sediment sources forlocal ceramics around a common multivariate centroid withoutdistorting multivariate structural differences between sedimentand ceramic groups.

4. Results

Prior to comparing the ceramics with sediments both datasetswere analyzed separately using a standard multivariate protocoldetailed elsewhere (Grave, Kealhofer et al., 2008). For the initialceramic dataset two broad groups were identified with different

Table 2bNAA results for Gordion sediments used in this study organized by PCA/CVA identified groups i–vi with group average value and % coefficient of variation (C.V.).

Grp# AIA# Ba Ca% Ce Co Cr Cs Eu Fe% Hf K% La Lu Na% Rb Sb Sc Sm Ta Tb Th U Yb Zn S

i 305 350 6.3 43 21 152 5.2 1.2 3.9 2.6 1.8 23.3 0.26 0.94 64 0.5 12.8 3.98 1.4 0.6 6.3 0.8 1.7 48 752i 310 430 6.1 40 27 197 4.4 1.4 5.4 4.4 1.6 23 0.29 1.22 48 0.5 16.1 4.34 1.3 0.7 5.2 1.1 1.9 70 891i 318 300 5.6 40 26 128 4.6 1 4.68 3 1.7 22 0.27 0.66 47 0.5 15.4 3.86 1 0.6 5.3 1.2 1.7 50 664i 319 310 4.9 43 26 133 4.6 1.2 4.84 3.2 1.6 23.5 0.28 0.67 53 0.5 16.3 4.02 0.9 0.5 5.6 1.1 1.8 61 702i 320 320 5.4 41 26 139 4.8 1 4.51 2.9 1.7 22.6 0.26 0.71 57 0.5 15.6 3.96 1 0.5 5.7 1.3 1.7 55 712i 2341 380 6 44 21 216 4.5 1.1 4.14 3 1.7 21.2 0.24 0.81 52 0.5 14.4 4.04 1 0.5 5 0 1.6 70 853i 2342 320 6 45 27 215 5.6 1.3 5.01 3 1.8 24.6 0.29 1.1 61 0.7 16.4 4.4 1 0.7 6.1 0.9 1.9 63 812i 2346 300 6.7 43 23 170 4.6 1.2 4.42 3.4 1.7 21.1 0.28 0.89 41 0.5 15 4.08 1 0.6 5 0.8 1.8 60 710i 2361 330 6.2 45 27 149 3.8 1.1 4.88 3.2 1.8 22.6 0.27 0.71 46 0.5 16.8 4.11 0.8 0.6 5.1 0.9 1.7 57 729i 3972 320 6.3 47 30 177 4.5 1.3 5.42 3.2 2 25.1 0.26 0.83 58 0.4 21.1 4.46 1.5 1 5.5 0.8 1.9 100 818YH i (n¼ 10) Avg. 336 5.95 43.1 25.4 168 4.66 1.18 4.72 3.19 1.74 22.9 0.27 0.85 52.7 0.51 16 4.13 1.09 0.63 5.48 0.89 1.77 63.4 764

C.V. 12.2 8.72 5.3 11.3 19.6 10.3 11.2 10.5 15 6.75 5.71 5.79 22 13.9 14.5 13.4 4.94 20.9 23.7 8.24 40.5 5.99 23.4

ii 307 360 5.4 47 39 197 6.3 1.1 4.24 3 2.2 26.1 0.28 0.74 74 0.8 14.7 4.41 1 0.5 8.2 1.8 1.8 53 853ii 308 450 5.7 45 20 338 4.4 1.2 3.93 3.5 2.2 25.3 0.28 0.98 61 0.6 13.1 4.41 1.1 0.6 6.8 1.2 1.8 52 1043ii 321 420 5.4 48 22 225 5.8 1.1 4.2 3.4 2 26.2 0.29 0.95 72 0.8 14.5 4.5 1.2 0.7 8.1 1.2 1.8 59 928ii 2350 360 5.1 48 22 216 8.3 1.1 4.33 3.8 1.6 23.4 0.27 0.71 71 0.9 14.5 4.08 0.9 0.6 7.4 1.1 1.7 57 854ii 2353 360 6.2 52 24 222 4.2 1.3 4.68 3.4 1.6 26.9 0.28 0.73 62 0.7 15 4.7 1.1 0 7.5 1.2 1.9 66 867ii 2354 430 6.6 53 20 301 3.5 1.1 3.81 3.5 1.5 26.4 0.28 1 54 0.8 12.2 4.56 0.8 0.7 7.4 1.3 1.8 65 1000YH ii (n¼ 6) Avg. 397 5.73 48.8 24.5 250 5.42 1.15 4.2 3.43 1.85 25.7 0.28 0.85 65.7 0.77 14 4.44 1.02 0.52 7.57 1.3 1.8 58.7 924

C.V. 10.4 9.85 6.27 29.6 22.4 32.4 7.28 7.33 7.52 17.3 4.85 2.26 16.2 12 13.5 7.85 4.69 14.5 51.1 6.82 19.5 3.51 10

iii 306 420 5.4 37 18 213 4.3 1 3.37 2.8 1.9 22.3 0.23 0.96 63 0.6 11.8 3.72 1.1 0.5 6.4 0.9 1.6 57 877iii 322 440 5.6 45 19 297 5.2 1.1 3.46 3.1 2 23.5 0.26 0.98 70 0.6 12 4.04 1.1 0.4 7.2 1.5 1.6 44 989iii 323 450 6 44 18 348 4.1 1 3.05 3.4 1.5 23.4 0.26 1.16 60 0.6 10.7 4.05 0.7 0.6 7 1.6 1.6 47 1038iii 324 420 4.7 38 18 132 7.8 0.92 3.42 3 1.9 21.4 0.23 0.88 73 0.7 11.7 3.61 1 0.6 7.3 1.2 1.5 57 810iii 325 300 5.9 36 20 172 5.3 0.94 3.59 3 1.7 19.9 0.24 0.74 56 0.5 12.8 3.53 0.8 0.6 5.7 1.5 1.6 46 698iii 326 260 6.8 37 17 130 6.5 0.89 3.08 2.9 1.7 21 0.22 0.55 64 0.6 10.6 3.39 0.7 0.5 6.7 2.2 1.4 49 627iii 2332 300 5.6 41 18 132 4.6 0.94 3.53 2.7 1.7 21.4 0.24 1.6 50 0.6 11.2 3.71 0.8 0 5.9 1.4 1.6 41 650iii 2338 400 5.9 44 16 203 11 0.93 3.49 3.2 2.3 22.4 0.24 0.87 65 0.7 12.2 3.85 0.9 0 6.8 0.9 1.6 44 849iii 2343 230 7.1 39 24 198 6.9 0.92 4.45 3.1 1.6 18.8 0.23 0.37 73 0.8 14.7 3.54 0.6 0.5 5.5 0 1.4 62 697iii 2352 410 9 40 20 182 2.9 1 4.05 2.8 2.2 18.5 0.22 1.2 51 0.9 12.9 3.89 0.6 0.5 5.8 1.4 1.5 52 824iii 2355 350 8.1 40 19 228 5.2 1 3.6 2.8 2 20.9 0.24 0.83 48 0.6 11.7 3.6 0.6 0 5 1.5 1.6 56 810iii 2359 310 7.4 40 23 150 4.7 1.1 4.41 2.9 0 19.7 0.23 0.83 52 0.5 14.9 3.68 0 0.7 5 1 1.6 53 697YH iii

(n¼ 12)Avg. 358 6.46 40.1 19.2 199 5.71 0.98 3.63 2.98 1.71 21.1 0.24 0.91 60.4 0.64 12.3 3.72 0.74 0.41 6.19 1.26 1.55 50.7 797C.V. 21.1 19.3 7.41 12.2 33.9 37.3 6.98 12.5 6.74 34.4 7.81 5.5 34.5 14.9 18.1 11.2 5.52 40 63.1 13.2 42.4 5.15 12.8

iv 311 270 5 30 24 159 7.7 0.89 4.36 2.6 2.9 18 0.22 0.96 73 1 15.3 3.16 1.1 0.6 5.6 0.7 1.5 69 697iv 312 250 3.8 31 23 157 8.9 0.8 4.27 2.6 3.7 17.8 0.19 1.17 92 1.1 13.7 2.98 1.1 0.4 6.9 0.8 1.3 64 689iv 313 380 5.3 30 24 145 7.7 0.91 4.36 2.8 2.6 18.8 0.23 1.11 68 0.9 15.4 3.27 1.2 0.5 5.2 0.7 1.5 64 783iv 314 250 4.7 38 22 139 8 1 4.24 3 2.5 21 0.24 0.82 73 0.9 13.5 3.61 1.3 0.8 7.3 2.1 1.6 44 643iv 315 250 5.1 39 23 133 8.1 1 4.3 3 2.5 22 0.25 0.77 76 0.9 13.6 3.71 1.2 0.6 7.6 2.3 1.6 51 651iv 316 240 4.5 38 24 134 8.2 1 4.62 2.9 2.5 21.9 0.26 0.85 70 0.8 14.6 3.81 1.1 0.4 7.1 2 1.7 62 646iv 317 240 4.3 39 20 116 8.5 0.88 4.08 2.9 2.5 22.7 0.24 0.76 75 0.8 13 3.68 1 0.4 7.7 2.2 1.6 60 627iv 2351 190 2.7 31 22 181 7.8 0.71 4.62 2.7 2.6 16.8 0.18 1.5 86 1.1 14 2.77 0.7 0 6.1 1.4 1.2 61 638YH iv (n¼ 8) Avg. 259 4.43 34.5 22.8 146 8.11 0.9 4.36 2.81 2.73 19.9 0.23 0.99 76.6 0.94 14.1 3.37 1.09 0.46 6.69 1.53 1.5 59.4 672

C.V. 20.9 19.1 12.5 6.1 13.7 5.16 11.6 4.25 5.84 15.3 11.5 12.5 25.8 10.7 12.7 6.19 11.4 16.6 50.3 14.1 46.5 11.3 13.5

v 2331 320 8.9 35 15 176 5.9 0.66 2.53 2.4 2.6 17.5 0.18 1.1 54 0.7 8.3 2.91 0.7 0.6 5.6 1.6 1.2 58 721v 2356 300 10 38 17

1185.7 0.91 3.18 2.4 1.2 19.7 0.22 0.41 61 0.7 10.5 3.34 0.8 0 6.1 1.9 1.4 45 647

v 2357 280 7.1 29 15150

5.2 0.68 2.67 2.1 1.6 15 0.18 0.53 43 0.5 9.2 2.63 0.6 0.6 4.6 1.6 1.2 43 616

v 2358 260 7.2 32 15 120 7.3 0.69 2.75 2.4 1 16.6 0.17 0.43 53 0.7 8.9 2.67 0.6 0 5.4 2.9 1.2 38 579v 2360 270 8.2 37 17 138 7.7 0.72 3.22 2.4 1.1 18.6 0.2 0.43 59 0.7 10.2 2.97 0.9 0 6.1 3.1 1.4 46 635v 2370 300 11 38 16 147 7 0.82 2.83 2.4 3.2 18.4 0.19 0.48 50 0.6 9.4 3.12 0.6 0 5.8 1.9 1.3 57 677

P.Grave

etal./

Journalof

Archaeological

Science36

(2009)2162–2176

2166

YH

v(n¼

6)

Avg

.2

88

8.7

33

4.8

15

.81

42

6.4

70

.75

2.8

62

.35

1.7

81

7.6

0.1

90

.56

53

.30

.65

9.4

22

.94

0.7

0.2

5.6

2.1

71

.28

47

.86

46

C.V

.7

.73

17

.81

0.5

6.2

11

5.2

15

.51

3.1

9.7

55

.21

51

9.4

39

.42

47

.31

2.1

12

.98

.69

9.1

71

8.1

15

51

03

0.6

7.6

61

6.7

vi3

27

41

01

03

31

82

60

9.3

0.6

42

.49

2.5

1.1

19

.50

.19

0.3

25

51

8.8

2.9

0.6

0.4

7.6

1.7

1.3

36

88

2vi

32

82

50

10

34

20

28

41

10

.72

.86

2.4

1.4

19

0.2

10

.33

63

0.9

10

30

.70

.47

.22

.11

.34

97

74

vi3

29

26

09

.23

42

33

84

14

0.6

62

.98

2.7

1.1

20

0.2

20

.38

69

1.1

10

.33

.14

0.7

0.4

8.1

2.2

1.4

49

89

8vi

23

33

32

09

.44

41

81

97

13

0.8

63

.38

31

.72

2.3

0.2

40

.46

71

11

1.6

3.7

0.8

07

.41

.81

.55

47

86

vi2

33

42

40

11

38

15

29

11

30

.77

2.5

73

.11

.41

9.8

0.2

0.6

96

30

.89

.13

.35

0.7

06

.72

.31

.44

07

64

vi2

33

52

60

11

34

12

28

81

10

.59

2.0

92

.71

.21

7.4

0.1

90

.64

49

0.7

7.2

2.8

40

06

.22

1.2

38

74

8vi

23

36

31

09

.44

02

02

73

9.4

0.8

3.3

62

.72

.12

0.5

0.2

30

.81

66

0.9

11

.43

.49

0.7

0.6

7.2

2.8

1.5

58

84

5vi

23

67

25

07

.54

22

04

31

22

0.7

73

.09

2.8

1.5

21

.80

.21

0.4

37

41

10

.93

.27

0.6

0.5

8.9

3.2

1.4

55

96

2vi

23

69

26

01

33

61

51

88

22

0.7

92

.64

2.4

01

8.3

0.2

0.4

76

60

.99

.43

.09

0.6

06

.71

.81

.45

17

00

vi3

96

92

80

10

39

15

12

61

80

.73

2.5

72

.21

.52

0.5

0.1

60

.31

76

1.3

8.4

12

.94

10

.47

.44

.91

.15

46

73

YH

vi (n¼

10

)A

vg.

28

41

0.1

37

.41

7.6

27

21

4.3

0.7

32

.82

.65

1.3

19

.90

.21

0.4

86

5.2

0.9

69

.71

3.1

70

.64

0.2

77

.34

2.4

81

.35

48

.48

03

C.V

.1

8.1

14

.31

0.2

18

.83

33

3.6

11

.41

4.6

10

.64

27

.47

11

.13

5.7

12

.71

7.2

14

.48

.77

39

.88

9.1

10

.43

9.2

9.4

16

Table 3To compensate for a systematic offset between sediments and the ceramicscompositions they most closely match (i.e. and therefore presumed to be local)a fitting factor is calculated as discussed in the text. This table shows the combinedmeans of the general ceramic groups YH A and YH B, and of sediment groups (YH i–vi)(note the lower S for the sediment mean), and the factor ((Ceramicavg/Sed-imentavg)� Sedimentsample)) to combine the sediments and YH A/B ceramics overthe same (ceramic sample) centroid.

YH A & B YH sed. (i–vi) (Ceramicavg/Sedimentavg)� Sedimentsample

Avg. Avg. Fitting factor

Ba 374.10 320.21 1.1683Ca% 8.02 6.89 1.1632Ce 49.58 39.79 1.2459Co 28.40 20.88 1.3607Cr 286.48 195.90 1.4624Cs 7.17 7.44 0.9633Eu 1.16 0.95 1.2275Fe% 4.80 3.76 1.2761Hf 3.70 2.90 1.2766K% 2.70 1.85 1.4561La 25.77 21.19 1.2162Lu 0.27 0.23 1.1712Na% 0.98 0.78 1.2572Rb 83.63 62.32 1.3419Sb 0.92 0.74 1.2294Sc 16.99 12.59 1.3495Sm 4.47 3.63 1.2330Ta 0.98 0.88 1.1109Tb 0.58 0.41 1.4056Th 8.74 6.48 1.3492U 1.69 1.60 1.0519Yb 1.90 1.54 1.2303Zn 96.71 54.72 1.7672P

1009.73 767.69

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–2176 2167

trajectories in the multivariate projections. These were thenanalyzed iteratively with outer groups and outliers identified andremoved until only a core cluster of 106 samples remained. Thisgroup could be decomposed into two major subgroups (YH A and B)that in turn contained smaller internal clusters and outliers(Table 4).

4.1. Sediments – establishing ‘local’

Of the 73 sediment samples analyzed, 21 were removed fromfurther analysis due to their relatively impoverished compositionalprofiles. The remaining 52 samples separated into six composi-tional groups, YH i–vi (Fig. 3a; Table 2b). An important methodo-logical step in the sediment analysis is to establish overallcompositional trajectories in multivariate projections that relatesthe sediment groups to landscape locales.

Compositional distinctions between these six groups reflecteda discontinuous trajectory from basalt to marl. This can be mostreadily expressed in a bi-variate plot of calcium and iron showinghow groups vary between calcium rich marl-type compositions toiron rich basalts (Fig. 3b). YH i and ii are basalt-derived sediments,with YH ii revealing more chemical weathering. YH iii and v aredominated by pediment soils derived from the clastic lake depositsrocks, with YH iii located closer to the basaltic uplands. YH iv and viare floodplain deposits from deep, silt-rich sections. YH vi is heavilymarl-dominated. The basalt groups are defined by relatively higherconcentrations of iron, scandium, zinc, cobalt and are depleted incalcium, rubidium, and cesium. YH vi has more calcium, cesium,thorium, uranium, and chromium. While a clear trajectory frommarls to basalts is evident in the YH sediments, mapping individualmembers of these groups back into the landscape reveals the

Table 4Summary statistics for the local component of the Gordion NAA dataset identified organized by the two major compositional groups: marl (YH A) and basaltic (YH B) and theirsubsets giving group identification, number of samples in each group, average value and % coefficient of variation (C.V.). Below detection limit measurements marked with ‘‘–’’.

YH A 1 (n¼ 20) 1.1 (n¼ 17) 1.2 (n¼ 4) 1.3 (n¼ 3) 1.4 (n¼ 3) 1.5 (n¼ 3)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 317.50 21.51 447.47 24.95 227.50 29.46 299.67 25.22 266.00 34.03 284.67 10.26Ca% 10.55 15.89 8.98 29.60 11.50 11.23 14.47 8.89 8.09 5.34 5.81 6.90Ce 44.09 11.12 49.16 10.03 31.00 7.90 41.83 5.39 54.75 4.26 40.57 5.96Co 24.88 17.04 19.27 21.73 24.25 11.36 16.10 10.26 30.15 3.99 27.43 56.43Cr 273.80 27.67 266.29 33.63 341.25 18.27 222.33 12.76 298.50 3.55 231.00 13.93Cs 7.45 24.76 7.11 36.23 6.08 19.91 11.21 44.78 12.14 28.66 6.76 16.37Eu 1.02 14.41 1.05 12.18 0.79 10.49 0.74 17.57 1.15 19.14 1.15 19.39Fe% 4.08 10.85 3.44 10.91 3.56 10.38 2.79 8.92 4.62 4.59 4.55 19.27Hf 3.56 18.75 3.71 15.92 2.03 16.32 3.02 8.41 4.09 6.06 3.21 12.15K% 2.45 24.20 2.73 26.08 3.15 12.83 2.09 34.71 2.31 11.96 2.05 15.34La 22.95 9.63 26.38 10.13 16.55 3.61 20.73 6.07 28.85 8.58 21.07 3.56Lu 0.26 8.58 0.25 9.44 0.20 8.65 0.14 86.90 0.26 0.00 0.32 9.45Na% 1.06 27.79 0.96 20.83 0.81 22.87 0.65 6.97 0.78 2.74 1.38 24.98Rb 65.50 27.44 65.81 20.18 53.50 6.38 66.80 16.48 104.25 24.08 34.60 87.06Sb 0.82 20.73 0.86 22.15 0.70 11.66 0.90 22.22 0.97 6.59 3.13 67.53Sc 13.52 13.20 11.78 8.56 13.43 5.88 9.97 8.99 16.90 5.02 21.20 12.68Sm 4.01 8.72 4.19 9.12 2.87 3.08 3.47 1.80 4.59 5.55 3.89 8.88Ta 1.11 48.98 0.90 71.30 0.73 68.85 0.80 106.80 – – – –Tb 0.35 104.65 0.58 50.64 0.15 200.00 0.42 88.19 0.75 9.43 0.75 5.53Th 7.52 21.18 8.41 9.09 5.43 10.89 7.28 10.99 11.05 4.48 8.96 15.18U 1.66 81.52 1.89 63.35 2.50 36.22 1.30 108.51 1.34 141.42 0.47 173.21Yb 1.77 10.64 1.64 9.71 1.33 7.23 1.43 10.66 1.97 0.36 2.37 2.32Zn 88.56 17.60 73.44 13.39 97.00 16.47 73.50 30.60 94.55 11.14 87.53 6.58P

898.48 1006.29 846.27 801.64 948.03 792.86

YH B 2 (n¼ 30) 2.05 (n¼ 3) 2.1 (n¼ 8) 2.3 (n¼ 3)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 432.63 27.24 414.00 9.40 294.75 48.22 305.00 9.97Ca% 5.96 17.01 5.79 41.76 6.29 10.25 2.54 15.82Ce 57.00 6.83 51.50 2.39 47.70 5.37 49.33 3.10Co 34.57 17.65 37.63 15.53 30.04 9.77 39.53 21.14Cr 300.03 10.89 338.33 13.33 266.88 5.66 364.00 20.53Cs 6.14 14.15 8.06 11.02 8.87 8.52 5.01 25.69Eu 1.39 13.13 1.13 10.89 1.15 7.21 1.36 17.54Fe% 5.94 8.00 6.07 7.28 5.47 5.65 6.33 8.29Hf 3.94 9.57 4.06 5.49 3.87 8.86 4.17 1.46K% 2.92 20.17 3.00 24.35 2.56 22.04 2.51 37.64La 29.22 7.07 25.93 6.79 24.45 3.23 26.67 1.32Lu 0.30 7.56 0.31 1.84 0.35 5.62 0.32 6.25Na% 0.98 20.79 0.90 7.12 0.99 15.34 1.19 17.22Rb 104.13 23.47 109.83 18.34 108.66 23.24 78.77 20.49Sb 1.08 15.06 1.12 23.77 0.84 20.92 0.61 26.56Sc 20.90 7.67 21.97 3.92 21.58 4.14 22.57 8.08Sm 5.25 6.87 4.76 3.18 4.36 7.60 4.80 1.70Ta 1.18 53.23 0.39 173.21 0.54 149.20 1.68 15.85Tb 0.76 40.57 0.54 87.21 0.58 62.18 1.07 21.04Th 9.83 8.08 10.02 7.74 9.71 2.87 7.67 4.91U 1.60 86.83 1.58 86.99 2.01 45.59 0.97 93.87Yb 2.06 9.92 2.18 7.12 2.36 9.97 2.10 8.25Zn 113.74 15.09 110.67 7.52 97.84 16.32 122.10 22.04P

1141.57 1159.76 941.85 1050.29

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–21762168

sedimentary complexity of the local region, both in space and overtime (Fig. 1b).

To define the local component of the ceramic assemblage, thesediment groups are compared with the larger ceramic assemblage.Prior to applying the sediment fitting procedure both sedimentsand ceramics are analyzed together. Reassuringly, this combinedanalysis showed that the large core ceramic group of 106 sampleswas the most proximate to the sediments. Following sedimentfitting to the multivariate centroid of the core ceramic group,a partial match between sediment and ceramic groups highlightedthe fact that not all sediments in the environs of the site would havebeen suitable for ceramic production (Fig. 3c). The clastic-derivedYH sediments iii and v match with the YH A ceramics, while basaltic

YH sediments i and ii match with the YH B ceramics. The remainingfloodplain sediments (iv and vi) are without ceramic matches.

4.2. Defining the non-local ceramics

Correlation of local sediments with ceramics was used toidentify the most probable local component of the ceramic sample.Beyond the local ceramics, multivariate analysis identified a struc-turally complex suite composed of small groups and singletons.When compared with other Iron Age sites in the region, the PCAprojection of the non-local component of the Gordion sample(Fig. 3d, Table 5) is both large and highly diverse reflecting their

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–2176 2169

compositional and geological heterogeneity (see Grave, Kealhofer,et al., 2008; Kealhofer, Grave, et al., in press).

4.3. Characterising the ceramic assemblage

A total of 10 local (YH 1-2.3 and local outliers; n¼ 106) and 29non-local groups (YH 3-2000 and non-local outliers; n¼ 173) wereidentified (Tables 4, 5 and 6). Outliers include 12 local and 17 non-local samples. All outliers were removed from further analysis.

4.3.1. Local ceramics (n¼ 106)Local ceramics are defined both by their abundance and their

matches with local sediment samples (Fig. 3c). YH A includes YH 1and its subgroups, while YH B is composed of YH 2 and itssubgroups. Of these two clusters, YH 1 and its subgroups aremarginally larger (n¼ 50). The subgroups, within both local clus-ters, may reflect shifts in clay source location within a similargeology or shifts in processing over time. In general, both local YHclusters are common in all periods. The local YH 2 cluster (includingsubgroups) is proportionally about one-half as common as thelocal YH 1 cluster over time. The one main exception is during theLate Phrygian (LP), when YH 2 is nearly as abundant as YH 1 (seeTable 6). This suggests considerable continuity in local ceramicproduction.

Both of the main local groups have different chronologicaltrajectories (Table 6). YH 1 is first used in the Middle Phrygian (MP)period, and is most common at that time; however its usecontinues into the Hellenistic period. YH 1.1 and 1.2 begin in theEarly Phrygian (EP) period, but 1.1 follows a similar trajectory to YH1, while 1.2 is last used in the LP. YH 1.4 and 1.5 were only used inthe MP and LP.

YH 2 is used throughout the periods studied, but is mostcommon during the LP, overshadowing any other group at anyother time. In the Roman period it is the only local YH group rep-resented. YH 2.1 and 2.3 are used only during the Phrygian periods,and only YH 2 and 2.3 date earlier than the MP. Additional, rare,local sources are used throughout the site’s occupation (singletons).However, these patterns are based on small group sample sizes andtherefore may not be representative.

The trajectories of YH 1 and 2 suggest a shift in importancebetween YH 1 and YH 2, which could relate either to shifting sourceavailability or to changing clay preferences (in relation to forms andstyles).

In the MP, the greatest number of local groups was represented(9), while eight were found in LP. All of the groups found in the LPwere used in the MP, suggesting a very strong continuity in bothclay sources and in clay processing. Both before and after theseperiods four different local clay types were in use, and there is lessoverlap in clay types (only 2 shared: 1.1 and 2).

4.3.2. Non-local ceramics (n¼ 173; 29 groups plus 17 outliers)While the number of local groups is large, the number of non-

local groups is even greater2 (Table 6). We expect that observedcompositional diversity will increase with increasing group size(Rhode, 1988) but relative to their size, the Late Phrygian samplehas an unusually high number of non-local groups and the Helle-nistic sample a comparatively low number of non-local groups(Fig. 4 inset; Table 6). The LP includes not only the most non-localsamples, but also the greatest number of non-local groups (20).Four groups (YH 10, 600, 900 and 2000) account for>5% each of the

2 The high proportion of non-local groups as well as of non-local ceramics in thesample reflects the well-informed selection of non-local samples by siteceramicists,

non-local assemblage, while the remainder (16 groups) have <5%each of the non-local samples. YH 3, and its two subgroups (n¼ 8),are in a different, closer orientation to the local clusters than theother non-local groups, but are compositionally and typologicallydistinctive and considered non-local.

The EP, MP and MP/LP each account for 10% of the non-localassemblage with a range of 6–11 non-local sources. In the LP boththe percentage of the non-local assemblage (22%) and the numberof non-local sources is doubled (n¼ 20). In the Hellenistic period,the percentage of non-local ceramics (27%) is the highest of anyperiod, but the number of groups decreases to levels comparable tothe earlier Phrygian periods.

The low frequency of any one group makes the significance ofpatterning somewhat difficult to interpret. All non-local groupswith >2% of the non-local assemblage are represented in at leasttwo periods. All of the groups with 5% or more of the non-localsample have samples from at least as early as the MP. Nine groupsbegin in or after the LP (all 2–4% of non-local samples). In generalterms, there appears to be considerable continuity in exchangepatterns through the Phrygian period, although a significantnumber of new sources were added in the LP. Non-local outliers arepresent in low frequencies in every period. There is one majorexception to this. During the Hellenistic period, one source standsout as the most dominant non-local source for any period (YH 10).This group includes mainly black glazed wares, and first occurs inthe Middle Phrygian period.

5. Henrickson and Blackman 1996

Based on NAA of 289 samples from Gordion including bothexcavated ceramics and clays, Henrickson and Blackman (1996)identified five compositional clusters (here labeled as HB A–E) thatcould be statistically decomposed into 13 ‘‘sub-clusters’’ (Table 7).Explicitly aware of the probabilistic character of inferring ceramicprovenance from compositional data (Henrickson and Blackman,1996, p73 note 35) they defined a set of empirical criteria fromwhich to infer local production: a) compositional matches betweenclay and pottery samples; b) compositional groups with a large n;c) large n groups that also contain a wide range of types; d)inclusion of heavy, large or cumbersome types most readilyproduced locally; e) compositional groups that contain samplesfrom multiple periods (1996: p. 76). They considered the presenceof only a small number of large groups in their dataset indicative oflarge scale local production both for the LBA and EP periods atGordion. The largest group in this dataset, HB B (n¼ 159), repre-sents more than 50% of the total sample and is largely restricted toLate Bronze Age samples. While considerably smaller, HB A (n¼ 69)is predominantly Early Phrygian. HB C–E are even smaller (Table 7).Henrickson and Blackman matched HB B samples with local sediments.While Henrickson and Blackman did not find a local sedimentmatch for HB A groups they assumed that these groups were alsolocal because they included comparatively large subgroups anddominated the entire Early Phrygian sample.

With the likely signatures of local and non-local samples for ourGordion Iron Age sample now identified we can compare ourdataset with that of Henrickson and Blackman. There are two majorconcerns for this type of comparison. In order to integrate the twodatasets each must be reduced to a common set of elements (in thiscase 21). Reduction of the number of elements used in a multivar-iate analysis has a direct impact on the level of resolution that canbe achieved (Grave, Lisle, et al., 2005). We therefore relax theresolution requirements of the comparison by restricting it togroup mean values for the local YH A and B, non-local YH groups,and the published BH mean values. The combined dataset is also‘‘decompressed’’ for multivariate analysis by removing the more

Fig. 3. Multivariate analysis of the present study’s (YH) and Henrickson and Blackman (1996) (HB) NAA datasets. Point projections are combined with conventional normaldistribution ellipses (2s) but also with non-parametric density contours to highlight the probabilistic complexity inherent in multivariate modeling of this dataset: a) PCA projectionshowing multivariate relationship for sediment groups YH i–YH vi. In this projection the YH sediment centroid has been fitted to the centroid of the two clay groups of Henricksonand Blackman (1996) (labeled HB Clay 1 and Clay 2) as discussed in text to indicate a high level of correspondence between the basaltic and marl sediments and clays of the twodatasets; b) bi-variate plot of calcium % and iron % for the sediments showing the negatively correlated compositional trajectory (arrow) moving from marl (YH vi) to basaltic (YH i)compositions. Discontinuous and non-linear character of data (consistent with the discrete geological origins of the samples) highlighted by non-parametric density contours (eachcontour accounts for 5% of the sample) and comparison of linear and exponential fits; c) PCA of the Gordion ‘‘local’’ dataset composed of two general ceramic groups YH A and YH Band the sediment groups that have been fitted to the YH A and B multivariate centroid (note this project excludes sediment groups that remain outliers after fitting (YH iv and vi); d)PCA of the ceramic dataset showing orientation of YH local and YH non-local compositions – outliers removed. The local compositional trajectory is indicated by the yellow arrow

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–21762170

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–2176 2171

compositionally exotic non-local groups identified in the separateYH and the HB analysis (HB C–E).

A second potential problem is the possibility of an offsetbetween NAA measurements from different facilities. As the stan-dards used in the Henrickson and Blackman (1996) paper are eitherno longer available (SRM 1633) or published in a way that is notamenable to offset calculations, we could not directly identify orcorrect potential differences between the NAA datasets. However,their published standards data indicate an overall high level ofabsolute precision and accuracy for the HB dataset comparable toours. In addition, from our previous experience in combiningdifferent uncorrected NAA datasets of ceramics for a single site NAAmeasurement offsets are small, particularly at the relatively coarselevel employed here of distinguishing between local and non-localelemental profiles (Grave, Kealhofer, et al., 2008).

Multivariate analysis (PCA, CVA, Hierarchical Cluster) of thecombined dataset shows it to be composed of three groups (Fig. 3e& f). The first of these captures all groups identified as local (YHA,B and BH B, C and Clays). The second group is composed of non-localYH500, 600, 800 and 900 and it is with these that the Henricksonand Blackman groups HB (‘local’) A1-4 most closely correspond.A third group is exclusively composed of YH non-local groups. Thepresence of non-local groups from YH that are not replicated in theHB dataset is likely to reflect the very different archaeologicalcontexts from which the samples were taken (Voigt vs. Youngexcavations) as well as the sampling regime of Henrickson andBlackman focusing on what were thought to be predominantlylocal ceramics.

This apparent contradiction in the interpretation of the HB Aseries allows three possible explanations: the HB A seriesrepresents a local compositional profile not sampled for ouranalysis; HB samples collected from the destruction level weresignificantly compositionally altered by fire; HB A samples arenon-local. We suggest that a non-local origin offers a ‘‘best-fit’’interpretation.

The first alternative, that we did not identify a local composi-tional profile, seems the least likely. Our sediment sampling wascomprehensive and designed to represent the geological range inthe Gordion catchment within a 15–20 km radius. All of theretained sediment samples fall within the local YH A and B clusterof ceramics suggesting that we had captured the range of localcompositional variability. Our ceramic sampling was also a broadspectrum approach designed to encompass the range of likely localand non-local wares present in the Iron Age levels. From thecomparison of the sediment and ceramic data we could clearlyidentify both local and non-local elemental packets.

The second alternative, in-situ compositional alterationthrough burning, while possible, does not match the elementaldata profiles. Notwithstanding the explicit selection criteriaemployed by Henrickson and Blackman to exclude obviously fire-affected samples (1996: p. 70 note 19), the high temperatureconflagration of the Destruction Level (one source for the Hen-rickson and Blackman Early Phrygian samples) provides thepotential for systematic volatilization of temperature sensitiveelements (Grave, 2009). However, this scenario is not supportedas HB A is enriched in several elements with particularly low

running from marl (YH A) to basaltic (YH B) compositions. Note the relatively compact charareflecting the geological heterogeneity of the non-local samples; e) PCA projection of grouColored circles mark the local group means for YH A (red) and YH B (green); colored squareceramic group means. Note only the major YH local and non-local groups are used for this cred demarcation line shown in f; f) Two dimensional projection (Hierarchical Cluster Analysiconfined to the YH local groups (red); HB group A sub-clusters are confined to a subset of Yseparating local and non-local groups equates with 3D red ellipse of e.

melting and boiling points relative to its potentially unalteredequivalent HB B (i.e. Cs, Rb, K, and Zn – which melt below 500 Cand boil below 1000 C).

The third alternative, that, HB A1-4 are non-local, is supportedby the ‘matches’ between them and the YH non-local groups whosemultivariate trajectory suggests a distinct geological origin relativeto the local YH A and B packet.

6. Discussion

Through the definition of local vs. non-local ceramics we haveidentified several unusual patterns in ceramic production atGordion, particularly during the Early Phrygian period. While theceramic sample chosen for our analyses was focused on non-localsamples, at most other Anatolian sites locally produced (stylisti-cally non-local) ceramics usually dominate our sample. AtGordion, however, the non-local sample was substantially largerthan the local. This partly reflects the expertise of the siteceramicists in non-local sample selection and their long experi-ence at the site, and partly the prominence of imports at thisinland site, particularly in the early stages of Phrygian politicaldevelopment. Non-local groups are not strongly patterned byperiod in our sample (except the Hellenistic period black glazedsamples of YH 10).

The importance of our definition of the local and non-localsample at Gordion is more strongly highlighted when wecompare it to the Henrickson and Blackman (1996) dataset thatexplicitly targeted local ceramics. When re-analyzed in relation toour local sediment groups their Early Phrygian assemblageappears to be ca. 80% non-local and belongs to one of the threebroad packets of non-local samples in our analysis. The archaeo-logical context of the HBA samples, within the Palace Area of theCitadel Mound, undoubtedly plays a strong role in this distribu-tion. When compared to their dominantly local LBA sample, thiseven more strongly supports a dramatic change in the organiza-tion of the political economy at Gordion than originally suggestedby Henrickson and Blackman. While these contexts, and thesesamples, provide no data about the range of (non-ceramic) goodsthat may or may not have been imported to the site in the EarlyPhrygian period, the diversity of non-local proveniences and therange of jugs, jars, and bowls represented in the EP assemblagesuggests that elite food practices, potentially feasting, usingvessels brought from a broad region, were a significant factor inEarly Phrygian political dynamics. The inflow of vessels into thePalace Area runs counter to a view that political centers were alsoplaces for production of exports into the larger region. The natureof economic relationships and production is substantiallydifferent from LBA patterns at Gordion.

The comparison of the two NAA datasets underscores thedifficulties in obtaining archaeological ceramic samples that areboth well contextualized and broadly representative. Whilerelatively large in number, the limited geographic extent of thesediments used by Henrickson and Blackman constrained theirability to differentiate non-local groups, with their conservativeinterpretation erring on the side of an undefined ‘local’. Thoughspecific, their selected archaeological contexts, as noted above,

cter of the YH local compared to the multivariate extent of the two YH non-local groupsp means for the YH dataset and the HB NAA dataset based on 21 common elements.s mark clay means for HB Clay 1 (blue) and HB Clay 2 (red). Solid circles mark the HBomparison. Red ellipse separating local from non-local groups is a 3D rendering of thes - Ward’s algorithm) of the same dataset and symbols of e. HB group B sub-clusters areH non-local groups (green), with the remaining YH non-local groups (yellow). Red line

Table 5Summary statistics for the non-local component of the Gordion NAA dataset giving group identification, number of samples in each group, average value and % coefficient of variation (C.V.). Below detection limit measurementsmarked with ‘‘–’’.

YH3 (n¼ 3) YH3.1 (n¼ 3) YH3.2 (n¼ 2) YH10 (n¼ 29) YH11 (n¼ 5) YH11.5 (n¼ 2) YH12 (n¼ 2) YH100 (n¼ 6) YH150 (n¼ 2) YH200 (n¼ 5)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 221.33 12.56 388.00 7.70 278.00 63.08 512.24 26.54 439.60 12.98 495.50 3.28 360.00 7.86 646.00 24.13 730.50 23.13 937.60 22.06Ca% 5.96 15.12 5.99 33.61 14.45 0.49 5.64 16.45 4.30 25.84 5.71 2.48 3.90 47.14 3.64 31.48 4.15 5.29 2.63 63.97Ce 44.53 7.32 51.70 2.98 47.95 0.44 60.75 4.64 68.52 3.12 67.85 2.81 68.00 0.00 90.17 5.27 102.00 0.00 110.60 4.59Co 56.50 12.78 49.73 5.97 13.65 2.59 35.18 13.78 40.20 10.92 51.00 4.99 41.00 6.90 28.98 17.17 24.60 9.20 29.76 23.41Cr 669.33 19.20 932.67 15.93 160.50 3.08 297.48 10.92 526.80 4.70 559.50 5.18 586.00 2.65 233.17 31.73 132.00 0.00 134.60 15.90Cs 4.60 18.08 18.13 30.43 15.40 5.51 9.70 12.17 12.50 7.63 12.50 21.50 11.00 12.86 13.64 31.17 10.35 4.78 11.34 17.33Eu 1.20 27.36 1.20 25.49 1.01 5.60 1.31 10.27 1.17 13.80 1.54 0.46 1.05 6.73 1.54 5.29 1.92 6.28 1.84 10.99Fe% 6.35 7.29 6.38 21.17 4.18 1.19 5.55 5.23 5.53 5.73 5.82 4.62 5.08 8.63 5.75 3.05 5.93 0.24 5.04 9.11Hf 3.04 14.04 3.78 7.35 2.68 8.44 4.31 9.36 4.34 9.52 4.04 16.65 3.85 1.84 6.18 8.40 4.09 23.71 6.63 9.15K% 1.90 16.55 2.50 24.33 2.31 26.33 2.42 15.75 2.63 13.89 2.57 12.41 3.40 12.48 3.38 40.04 3.46 5.94 3.55 15.24La 23.07 7.81 27.97 5.06 24.40 0.58 30.72 3.95 33.52 3.69 34.15 1.04 34.20 1.65 45.58 7.65 49.80 3.69 59.40 6.43Lu 0.28 10.19 0.32 23.62 0.27 2.67 0.37 5.65 0.36 2.31 0.38 3.72 0.34 2.11 0.47 8.26 0.42 8.52 0.48 5.37Na% 1.01 31.76 1.02 8.01 0.21 0.00 0.82 11.33 0.58 8.51 0.50 10.00 1.41 22.07 0.92 33.10 0.65 4.35 0.67 35.37Rb 60.60 56.39 72.33 8.33 105.55 9.98 123.79 11.98 140.40 7.94 143.50 1.48 97.00 4.37 149.67 9.32 169.50 7.93 188.60 9.73Sb 1.30 50.87 1.62 8.18 4.88 0.15 1.40 15.25 1.44 22.57 1.45 9.75 0.90 31.43 2.43 60.72 2.21 7.04 1.93 54.55Sc 23.17 14.98 24.90 26.52 12.60 2.24 21.59 6.17 21.58 4.80 22.20 2.55 18.00 7.86 21.68 3.47 20.70 1.37 19.62 9.36Sm 4.27 13.34 4.76 11.61 4.57 5.11 5.69 4.47 6.29 3.37 6.85 2.06 5.37 3.16 7.57 8.06 9.29 1.14 8.68 4.52Ta 1.66 33.65 1.40 18.90 0.52 141.42 1.30 68.02 0.68 144.62 1.55 10.95 1.15 6.15 1.61 19.17 1.39 21.95 2.03 40.19Tb 0.79 27.33 0.83 18.94 0.68 8.32 0.92 7.27 0.77 57.96 1.04 2.72 0.85 24.96 1.06 18.59 1.31 9.72 1.24 10.00Th 6.90 19.48 8.63 19.11 11.00 3.86 11.79 5.04 11.42 5.52 11.75 3.01 11.00 12.86 16.57 11.38 16.80 1.68 23.16 14.78U – – 0.33 173.21 2.70 1.05 1.76 71.37 1.91 92.99 – – 1.35 5.24 2.97 21.90 2.49 6.25 4.32 20.59Yb 2.06 8.99 2.20 18.18 1.83 0.00 2.68 5.14 2.70 5.06 2.81 2.77 2.00 7.07 3.22 4.79 2.74 2.06 3.40 12.08Zn 104.47 9.17 122.67 26.56 86.60 19.11 110.13 9.37 115.66 17.15 122.50 14.43 115.00 6.15 114.62 14.62 156.00 5.44 107.62 20.94P

1244.32 1729.06 795.92 1247.53 1442.91 1554.69 1371.85 1400.81 1452.27 1664.74

YH300 (n¼ 7) YH400 (n¼ 5) YH500 (n¼ 5) YH550 (n¼ 2) YH600 (n¼ 10) YH700 (n¼ 2) YH800 (n¼ 3) YH900 (n¼ 9) YH1000 (n¼ 2) YH1100 (n¼ 6)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 748.00 17.99 682.00 30.18 616.80 21.66 510.00 11.09 572.36 33.02 817.50 9.95 616.67 26.76 457.89 22.70 286.00 15.82 633.50 41.77Ca% 3.14 31.83 2.08 83.92 3.10 2.50 5.70 17.37 3.22 21.07 0.70 1.02 2.63 78.23 5.87 31.62 4.64 3.96 1.98 66.99Ce 98.31 3.25 103.80 3.94 83.38 7.56 81.50 6.07 84.71 5.95 103.40 13.13 81.33 2.56 77.19 10.02 72.90 4.27 86.33 5.28Co 24.96 13.96 31.80 35.81 24.48 3.89 18.50 3.82 30.42 22.39 13.25 18.68 19.33 11.95 24.22 17.28 22.85 1.55 39.82 33.78Cr 137.71 18.66 167.20 16.21 167.80 17.48 161.00 7.03 219.09 25.66 66.55 11.58 152.67 22.50 147.97 24.71 182.50 3.49 264.67 16.54Cs 17.00 19.64 9.04 7.92 8.10 21.30 21.00 6.73 7.58 14.26 10.70 9.25 11.40 29.98 8.37 18.56 11.90 0.00 10.13 39.93Eu 1.60 12.41 1.78 9.23 1.43 10.56 1.35 5.24 1.65 10.03 1.62 6.98 1.37 11.18 1.43 15.36 1.31 10.30 1.64 8.86Fe% 5.37 5.62 5.26 3.22 5.04 5.04 4.55 2.18 5.36 5.79 3.77 7.32 4.08 2.14 4.51 6.23 6.07 4.78 5.72 3.66Hf 5.50 13.18 5.98 9.34 5.94 8.54 5.85 3.63 6.19 11.99 7.68 10.68 6.60 13.64 5.15 9.59 5.05 11.20 5.95 10.12K% 4.12 16.41 3.36 5.80 3.20 14.91 3.25 10.88 3.09 14.91 3.88 11.86 3.07 9.96 2.81 16.22 3.88 1.46 2.95 16.93La 54.20 3.53 53.94 2.53 42.80 9.46 43.25 5.07 44.07 5.15 54.40 7.28 41.80 5.52 40.37 10.11 39.05 1.27 45.13 3.80Lu 0.46 5.95 0.42 14.48 0.42 4.98 0.45 1.59 0.43 3.64 0.61 3.51 0.50 7.62 0.36 10.40 0.40 1.79 0.41 7.45Na% 0.95 24.71 1.05 30.97 0.85 24.53 0.47 19.77 1.13 15.86 1.02 16.64 0.94 34.17 0.87 39.13 0.27 2.67 1.24 7.16Rb 180.29 12.84 142.00 7.71 146.20 14.93 130.00 10.88 137.91 9.53 159.50 8.42 130.00 15.38 119.70 25.97 177.50 4.38 153.83 17.41Sb 2.87 11.54 1.14 65.23 1.27 53.25 0.80 17.68 0.83 29.00 2.32 17.07 2.03 38.20 1.12 27.12 0.98 17.32 0.61 85.06Sc 20.50 7.67 19.36 5.04 19.38 4.12 18.75 3.39 20.72 5.96 18.30 1.55 14.97 6.82 16.86 6.67 20.40 2.77 20.62 10.41Sm 8.13 4.13 8.28 1.75 6.74 5.74 6.54 1.95 6.93 7.58 8.89 4.14 7.16 6.74 6.40 9.58 5.72 3.21 7.07 2.88Ta 1.99 28.11 2.26 42.74 0.81 94.10 1.75 28.28 1.12 70.52 2.11 26.14 1.87 38.01 1.47 22.46 1.94 37.91 1.66 80.50Tb 1.22 12.94 0.96 13.98 0.99 7.10 1.05 20.20 1.11 12.33 1.40 0.00 1.30 23.08 0.82 20.08 0.91 4.66 1.11 11.43Th 21.69 5.76 17.00 7.20 15.44 4.73 16.50 12.86 15.51 6.83 22.30 3.81 17.00 11.76 13.50 4.37 15.55 4.09 15.83 9.37U 3.38 12.05 2.12 87.58 2.67 28.04 3.15 20.20 1.96 51.93 3.51 23.77 2.83 20.07 2.55 19.59 3.66 23.57 0.23 244.95Yb 3.13 5.10 2.88 12.37 2.86 3.95 2.80 0.00 3.09 5.63 4.06 2.09 3.23 7.14 2.54 6.68 2.79 3.30 2.96 7.54Zn 124.71 14.75 120.00 8.33 106.70 5.28 130.00 10.88 109.70 11.54 168.50 15.53 100.33 19.44 102.94 11.71 121.00 7.01 122.33 8.23P

1469.23 1383.71 1266.41 1168.20 1278.18 1475.95 1223.10 1044.93 987.25 1425.71

P.Grave

etal./

Journalof

Archaeological

Science36

(2009)2162–2176

2172

YH1200 (n¼ 2) YH1300 (n¼ 3) YH1400 (n¼ 5) YH1500 (n¼ 4) YH1600 (n¼ 3) YH1700 (n¼ 3) YH1800 (n¼ 6) YH1900 (n¼ 5) YH2000 (n¼ 15)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 831.00 12.08 676.33 7.47 657.40 58.50 642.50 22.46 428.00 31.39 614.33 10.90 449.00 23.68 454.00 5.07 784.00 13.91Ca% 4.71 5.70 0.90 22.14 2.45 111.89 2.60 47.73 5.72 41.67 4.53 10.06 9.11 31.56 5.94 11.52 2.62 20.35Ce 91.50 3.86 112.33 8.27 84.72 11.51 98.75 19.24 71.17 5.21 98.67 3.10 83.92 4.67 65.60 7.96 89.64 8.73Co 12.00 23.57 31.50 35.35 24.90 24.74 53.00 21.07 28.80 34.98 15.47 11.30 19.80 24.58 75.40 19.55 31.93 18.75Cr 78.70 12.04 67.77 3.36 100.82 47.65 127.25 7.16 219.33 10.54 104.67 4.90 112.00 9.47 304.00 5.83 164.81 7.33Cs 19.55 3.26 13.40 6.84 4.32 27.68 10.43 30.62 46.83 25.13 17.53 7.33 25.82 21.64 8.26 25.40 21.42 28.92Eu 1.77 5.21 1.64 6.42 1.46 16.98 1.58 13.09 1.35 11.06 1.24 4.45 1.37 18.62 1.42 10.45 1.85 6.94Fe% 3.22 11.86 3.92 1.88 3.93 9.72 5.29 14.87 4.96 14.20 4.25 0.62 4.13 9.90 6.44 5.89 6.94 5.02Hf 8.76 12.27 8.00 6.57 6.44 11.76 10.05 13.14 5.68 15.73 5.57 13.43 4.95 9.19 10.60 8.44 4.92 13.49K% 2.73 3.89 4.75 6.14 3.33 20.96 3.50 8.41 2.85 12.46 4.59 13.97 2.83 13.42 2.34 58.02 3.49 19.29La 51.95 3.13 58.67 6.73 47.12 10.28 52.55 17.33 37.63 7.59 52.30 0.96 44.28 9.68 32.54 3.03 45.48 8.01Lu 0.48 1.49 0.67 2.29 0.23 56.15 0.41 18.94 0.41 9.78 0.57 1.01 0.33 6.20 0.34 11.59 0.48 7.07Na% 1.10 13.56 0.93 6.59 1.30 43.16 0.81 69.03 0.44 5.76 0.31 17.58 0.80 45.99 0.82 9.76 1.16 23.45Rb 138.50 8.68 236.33 3.60 98.02 32.29 160.00 13.50 181.00 34.11 228.33 3.34 134.50 11.82 120.00 10.21 151.06 13.78Sb 4.80 8.70 2.11 1.80 0.52 63.01 1.63 45.88 2.74 64.89 3.38 8.40 1.79 25.17 0.90 33.33 4.89 26.70Sc 15.30 11.09 20.73 2.66 12.26 30.56 18.05 27.39 20.00 3.04 16.47 4.60 13.23 10.68 22.92 3.77 24.68 4.67Sm 7.67 0.65 8.90 6.73 6.06 19.81 7.44 29.05 6.48 4.03 8.60 2.59 6.23 4.80 5.74 3.20 8.39 7.73Ta 1.36 5.74 3.75 36.83 2.10 47.89 5.35 20.62 1.17 97.15 1.88 15.56 1.04 85.72 4.88 46.72 1.32 59.42Tb 0.97 24.79 1.38 11.26 0.91 34.42 0.30 200.00 1.00 12.70 1.52 15.58 1.01 15.39 0.72 91.81 1.26 16.27Th 17.85 6.73 26.10 4.00 16.44 31.49 19.10 18.37 16.67 23.05 25.30 5.82 17.23 3.51 10.40 8.60 14.32 8.68U 4.03 32.64 5.03 15.74 2.48 25.47 2.58 34.79 1.49 87.85 3.63 3.91 2.77 53.09 2.44 34.24 2.40 47.09Yb 3.10 8.91 4.68 5.76 1.99 18.27 3.10 20.91 2.98 11.61 3.85 2.40 2.43 4.09 2.44 8.50 3.36 7.92Zn 92.70 7.17 198.00 10.44 77.26 16.58 130.00 10.88 100.93 15.63 84.63 12.68 89.53 23.71 124.00 9.19 132.53 13.68P

1393.71 1487.81 1156.48 1356.24 1187.64 1301.63 1028.11 1262.14 1502.95

P.Grave

etal./

Journalof

Archaeological

Science36

(2009)2162–2176

2173

Fig. 4. Chronological comparison of YH local and non-local samples by chronolog-ical phase (EP¼ Early Phrygian:- 10th–9th c. BCE; MP¼Middle Phrygian:- 8th–mid6th c. BCE; LP¼ Late Phrygian:- mid 6th–mid 4th c. BCE; Hellenistic:- mid 4th–early2nd c. BCE; Roman:- 1st BCE–3rd c. CE). Note for comparison both local and non-local components recalculated to sum to 100. Inset shows the % of non-local groupsby % sample size as a measure of import diversity for each chronological phase. TheLate Phrygian sample has an unusually high number of non-local groups and theHellenistic sample a comparatively low number of non-local groups relative to theirsample size.

Table 6Counts of YH local and non-local groups by chronological phase.

YH local EP MP MP & LP LP LP/Hell. Hell. Roman UNIDP

YH A 1 10 2 3 3 2 201.1 2 4 1 4 2 3 1 171.2 1 1 2 41.3 1 1 1 31.4 3 31.5 1 1 1 3

YH B 2 1 4 3 12 4 2 3 1 302.05 1 1 1 32.1 5 3 82.3 1 1 1 3

Outlier 1 3 2 1 3 2 12

P6 33 10 28 13 8 3 5 106

YH non-local EP MP MP & LP LP LP/Hell. Hell. Roman UNIDP

3 1 1 1 33.1 2 1 33.2 2 210 1 1 2 25 2911 1 4 511.5 2 212 1 1 2100 1 1 2 1 1 6150 1 1 2200 2 2 1 5300 1 5 1 7400 2 1 2 5500 1 1 2 1 5550 2 2600 2 5 1 1 1 10700 1 1 2800 1 1 1 3900 2 1 1 2 1 2 91000 2 21100 3 1 1 1 61200 1 1 21300 3 31400 1 2 1 1 51500 1 3 41600 1 2 31700 3 31800 2 1 3 61900 2 3 52000 1 4 6 1 3 15Outlier 3 1 2 3 3 3 2 17

P18 18 18 38 12 46 12 11 173

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–21762174

did not provide a systematic sample of the overall range ofceramics during the Early Phrygian period. While the strength oftheir sampling was both the contextual and chronologicalcontrol of the archaeological contexts, this also limited thesample assemblage to a highly specific and specialized EarlyPhrygian sample. The diversity of proveniences and styles in theEarly Phrygian assemblage, suggests that Henrickson andBlackman’s arguments for standardization and local massproduction in both the LBA and Early Phrygian periods at Gor-dion may need to be reevaluated.

7. Conclusion

In this paper we specifically aimed at characterizing localproduction and distinguishing imports during the Iron Age atGordion. In conjunction with a comparison with earlier NAA workof Henrickson and Blackman we could address two issues, onemethodological and one substantive. Methodologically, incorpo-ration of legacy NAA datasets into recent NAA analyses allows us,

and others, to integrate a wide field of data, leading not only toreinterpretations, as suggested here, but also to much morecomprehensive understandings of individual sites and theirregional contexts. Care must be taken, however, to considerdifferences in sampling framework and analytic regime whenevaluating differences in interpretation. Cross comparison of thenumerous legacy datasets for Turkey and the wider Aegean offersthe opportunity for a more appropriate large scale understanding ofproduction and exchange dynamics across the region.

Substantively, the combined datasets from the Henrickson andBlackman study and the current AIA work demonstrate theremarkable transition in political economic relationships thatoccurred during the formation of the Phrygian state at Gordion.This is revealed in the apparent shift from locally made, yet highlystandardized ceramics of the LBA identified by Hendrickson andBlackman, to the import (at least into the elite Palace Area) of(relatively?) standardized ceramics from a geographically widerange of non-local sources during the Phrygian periods. Despite theloss of political hegemony during the later Middle and Late Phry-gian period (Lydian, Persian incursions), the ceramic assemblagessuggest a strong continuity in the use of sources/technologies.Preferences shift among local sources (from YH 1 to YH 2), but bothremain in use. The strength of external influences at Gordion in theLP is seen both in the abundance of non-local groups and thediversity of groups represented. This economic and exchangeflorescence, in the midst of political change, suggests that while thecontrol of Phrygia may have been wrested from Gordion, the cityremained an important center. Subsequent increases in thefrequency of imports, alongside a decline in the number of importsources, suggest a substantial change in both the political andeconomic composition of the site during the Hellenistic period. Thetwo Hellenistic phases at the site (YHSS 3a and 3b), however, needto be more carefully evaluated. While ceramic stylistic evidencesuggests a similar trajectory, documenting the changing pattern

Table 7Henrickson and Blackman (1996, appendix 2, p 83) NAA data, giving group identification, number of samples in each group, average value and % coefficient of variation (C.V.).Below detection limit measurements marked with ‘‘–’’.

HB A-1 (n¼ 27) HB A-2 (n¼ 11) HB A-3 (n¼ 9) HB A-4 (n¼ 22) HB B-1 (n¼ 7) HB B-2 (n¼ 8) HB B-3 (n¼ 78) HB B-4 (n¼ 49)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 477 24.5 490 21.5 600 22 458 27.3 536 21.3 595 15.9 367 21.8 371 31Ca% – – – – 6.16 38.4 6.05 23 6.12 25 7.7 22.7 13.9 24.2 7.46 22.7Ce 95.4 8.2 90.2 13.9 79.9 8.8 70.6 10.7 55.5 7.9 64.2 7.2 49.9 7.7 58.5 7.7Co 26 9 22.1 18.3 20.5 18.3 15.2 11.8 19.4 12.1 20.6 5.5 22.3 16.3 27.9 8.6Cr 266 15.7 148 11.6 128 15.6 146 10.4 155 17.8 200 8.2 272 17.5 429 26.1Cs 9.91 22 6.69 17 9.71 12.8 7.33 14.1 7.12 9.5 5.57 6.3 11.6 21.8 7.38 11.3Eu 1.43 8.6 1.42 11.5 1.31 12.9 1.23 8.1 1.07 7.1 1.13 6 0.97 10.3 1.11 7.5Fe% 5.24 5.5 5.16 8.7 4.34 10.8 4.4 7.8 4.17 5.7 4.19 4.8 4.03 16.3 4.89 7.3Hf 6.17 7.4 6.7 12.7 5.28 9.8 5.07 8.6 4.32 16.2 4.05 19.3 3.67 12.2 3.99 9.1K% 2.5 9.5 2.16 9 2.6 14 2.51 12.4 2.57 12.4 2.21 12.6 2.2 17.7 2.36 10.1La 51.1 6.6 49.8 13.5 45.4 8.7 39.3 9 31.5 5.8 38.4 9.3 28.5 7.2 33.8 8.2Lu 0.46 12.2 0.51 24.3 0.41 11.4 0.32 13.2 0.26 24.1 0.26 16.6 0.28 15 0.3 16.4Na% 1.11 19.3 0.69 47.4 0.83 27.3 0.93 33.9 1.33 15.9 1.52 14.6 0.84 28.4 0.97 12.5Nd 36.3 8.9 36.8 17.2 34.3 9.4 29.6 14.6 24 10.5 26.1 7.5 20.8 17.1 24.3 12.4Rb 156 11.5 122 22 145 10.5 125 10.4 107 12.7 91.7 4.1 98 13.4 114 11.6Sc 18.5 6.7 17.6 9.3 16.6 14.9 14.8 11.9 12.8 7.9 13.7 5.2 14.2 18 17.3 8.6Sm 7.22 6.8 7.21 13.1 6.39 11.4 5.66 10 4.35 8.9 4.79 5.2 4.19 8.7 4.73 8.5Ta 1.36 10.7 1.46 10.3 1.13 6.5 1.29 8.9 0.86 12.3 1.03 14 0.94 10.4 1.07 13Tb 0.95 10.5 0.99 22.9 0.91 13.8 0.75 14.3 0.64 12 0.6 15.5 0.61 14.5 0.69 13.8Th 14.9 11.2 16.7 11.8 14.6 6.7 10.4 9.4 9.41 9 11.7 4.9 8.27 11.9 11.3 10.8Yb 3.4 10.4 3.49 26.5 2.82 8.1 2.3 14 1.76 14.6 1.92 15 1.95 10.1 2.13 10.6Zn 101 10.2 90.9 16.7 103 10.2 78.9 19.7 91.3 7.5 94.5 12 83.7 19.2 101 10.7P

1282 1121 1229 1026 1076 1191 1010 1225

HB B-5 (n¼ 11) HB B-6 (n¼ 6) HB C-1 (n¼ 7) HB D-1 (n¼ 13) HB E-1 (n¼ 3) HB Clay 1 (n¼ 25) HB Clay 2 (n¼ 13)

Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V. Avg. C.V.

Ba 371 26.4 454 25.2 380 17.6 698 24.7 – – 411 16.2 370 15.2Ca% 9.34 13.4 9.67 18.5 15.8 16.7 5.75 34.4 2.22 42 8.59 20.6 21.2 34.5Ce 46.7 5.2 58.4 5.6 39.1 8 83.5 5.7 64 4.1 59.2 5.1 43.6 9Co 24.6 11.6 26.6 6.3 19.3 11.9 9.6 8.4 47.9 6.2 27 6 17.5 13.7Cr 357 27.4 322 12 281 25.5 62 8.9 324 10.7 192 9.3 230 22.8Cs 5.88 17.8 3.36 7.7 8.91 12.2 6.64 7.4 3.25 5.8 8.36 15.7 18.3 13.3Eu 0.95 5.8 1.03 7.5 0.74 7.6 1.26 7.3 1.89 3.2 1.26 5.2 0.81 9.5Fe% 4.36 10.9 4.14 5.7 3.3 11 2.89 4.2 8.8 3.3 5.12 6.2 3.11 11.5Hf 3.3 5 3.52 4.5 2.67 11.4 6.04 7.2 5.58 2.6 3.89 6.3 3.24 16.7K% 2.32 8.4 1.54 6.7 1.93 16.9 2.38 9.2 1.76 4.7 2.45 21.4 1.81 22.5La 26.5 5.3 31.8 8.2 22.3 7.6 49.4 6 33.1 1.5 33.5 4.7 25.1 9.4Lu 0.24 10.4 0.23 9.1 0.24 28.7 0.29 13.8 0.27 11.4 0.32 13.7 0.27 10.6Na% 0.89 18.1 1.28 11.6 0.64 20.7 2.01 10.1 2.01 8.2 1.14 33 0.74 30.3Nd 20 7.7 23.1 17.3 15.9 15.9 33.3 7.6 27.8 8.5 24.8 16.1 18.5 17.3Rb 97.1 11.2 64.6 19.4 81.2 121 118 9.1 66.8 5.5 93.5 15 99.3 13.6Sc 15 9.9 13.1 6.9 11.6 11.2 8.5 3.3 28.5 4.7 16.5 6.8 11.1 10.8Sm 3.94 4.7 4.31 6.1 3.27 5 5.89 5.1 6.46 4.9 5.05 5.1 3.71 11.4Ta 0.96 8.9 0.88 8.1 0.74 11.3 0.89 7.4 2.39 4.4 1.1 7.9 0.87 12.7Tb 0.62 21.6 0.62 11.9 0.49 15.9 0.7 8.1 0.85 7.1 0.72 13 0.52 20.3Th 8.13 6.1 7.69 6 7.25 7.9 15.8 6.4 5.55 4.8 9.1 8.3 8.8 12Yb 1.75 3.7 1.83 7.1 1.71 20.5 2.23 13.8 2.06 10.8 2.25 5 1.74 13.2Zn 88.3 6.8 73.4 17.4 75.5 10.3 91 39.7 135 17.1 85.1 10.2 67.1 15.9P

1089 1107 973.6 1206 770.2 992 947.3

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–2176 2175

of non-local sources – with the future potential of identifyingsources – will be essential for better understanding the dynamics ofthe political economy of Phrygia and Gordion.

Acknowledgments

This research was funded by the Australian Research Council(DP0558992) and National Science Foundation (0410220). Wethank anonymous reviewers for their comments and suggestions.Sadly, co-author Keith DeVries passed away before the comple-tion of this manuscript. His insights and acumen will be sorelymissed.

References

Erentoz, C., 2002. Jeoloji Haritalari Ankara quad 1:500,000 scale (map) Ankara,Maden Tetkik ve Arama Genel Mudurlugu.

Grave, P., 2009. Melting moments: modelling archaeological high temperatureceramic data. In: Fairbairn, A., O’Connor, S., Marwick, B. (Eds.), New Directionsin Archaeological Science. Terra Australis, vol. 28. Australian National Univer-sity, Canberra, pp. 215–234. Chapter 15.

Grave, P., Kealhofer, L., et al., 2008. Using NAA to identify scales of interactionat Kinet Hoyuk, Turkey. Journal of Archaeological Science 35 (7),1974–1992.

Grave, P., Lisle, L., et al., 2005. Multivariate comparison of ICP-OES and PIXE-PIGE of East Asian Storage Jars. Journal of Archaeological Science 32 (6),885–896.

Henrickson, R.C., 1993. Politics, economics and ceramic continuity at Gordion in thelate second and first millennia B.C. In: Kingery, W.D. (Ed.), Social and CulturalContexts of New Ceramic Technologies, Ceramics and Civilization VI. AmericanCeramic Society, Westerville, OH, pp. 86–176.

Henrickson, R., 1994. Continuity and discontinuity in the ceramic tradition ofGordion during the iron age. In: French, D., Çilingiroglu, A.A. (Eds.), Proceedingsof the Third International Anatolian Iron Age Symposium. Oxbow Press, Oxford,pp. 95–129.

Henrickson, R., 1995. Hittite pottery and potters: the view from late bronze ageGordion. Biblical Archaeologist 58 (2), 82–90.

Henrickson, R.C., Blackman, M.J., 1996. Large-scale production of pottery at Gordion:a comparison of the late bronze and early phrygian industries. PaleoOrient 22,67–87.

P. Grave et al. / Journal of Archaeological Science 36 (2009) 2162–21762176

Henrickson, R.C., Vandiver, P.B., et al., 2002. Lustrous black fine ware at Gordion,Turkey: a distinctive sintered slip technology. In: Vandiver, P.B., Goodway, M.,Mass, J.L. (Eds.), Materials Issues in Art and Archaeology VI. Materials ResearchSociety Symposium Proceedings, vol. 712. Materials Research Society, Pitts-burgh, PA, pp. 391–400.

Kealhofer, L., 2005. Settlement and land use: the Gordion regional survey. In:Kealhofer, L. (Ed.), The Archaeology of Midas and the Phrygians. University ofPennsylvania Museum Press, Philadelphia, pp. 137–148.

Kealhofer, L., Grave, P., et al. Post-collapse: the Re-emergence of polity in iron ageBogazkoy. Oxford Journal of Archaeology, in press.

Korte, G., Korte, A., 1904. Gordion: Ergebnisse der Ausgrabung im Jahre 1900.G. Reimer, Berlin.

Marsh, B., 1999. Alluvial burial of Gordion, an iron age city in Anatolia. Journal ofField Archaeology 26 (2), 163–175.

Marsh, D., 2005. Physical geography, land use, and human impact at Gordion. In:Kealhofer, L. (Ed.), The Archaeology of Midas and the Phrygians. University ofPennsylvania Museum Press, Philadelphia, pp. 161–171.

Rhode, D., 1988. Measurement of archaeological diversity and the sample-sizeeffect. American Antiquity 53, 708–716.

Sams, G.K., 1995. Midas of Gordion and the Anatolian Kingdom of Phrygia. In:Sasson, J. (Ed.), Civilizations of the Ancient Near East. Charles Scribner and Sons,New York, pp. 1147–1159.

Voigt, M.M., 1994. Excavations at Gordion 1988–89: the Yassıhoyuk stratigraphicsequence. In: Çilingiroglu, A., French, D.H. (Eds.), Anatolian Iron Ages, vol. 3. TheBritish Institute of Archaeology at Ankara, Ankara, pp. 265–293. Proceedings ofthe Third Anatolian Iron Ages Colloquium, 6–12 August, 1990.

Voigt, M.M., DeVries, K., et al., 1997. Fieldwork at Gordion 1993–1995. Anatolica 23,1–59.

Voigt, M.M., Henrickson, R.C., 2000. The early iron age at Gordion: theevidence from the Yassıhoyuk stratigraphic sequence. In: Oren, E.D. (Ed.),The Sea Peoples and their World: A Reassessment. University MuseumMonograph, vol. XX. The University of Pennsylvania Museum, Phila-delphia, pp. 327–360.

Young, R.S., 1951. Gordion 1950. University Museum Bulletin 16 (1), 3–20.