Introduction

Southeast Asia's earliest copper-base consumption and founding assemblage has been dated to c. 2000 BCE at the site of Ban Chiang (White 2008; though this chronology is disputed, e.g. Higham and Higham 2009; Higham 2010). Thereafter, such evidence appears with increasing regularity in mid/late second and first millennia BCE regional contexts, especially in northeast Thailand (e.g. Ban Non Wat, Non Nok Tha; see Fig. 1). Prehistoric extractive metallurgical evidence is much rarer and is currently published at only two locales: Phu Lon and the Khao Wong Prachan Valley (hereafter ‘KWPV’ or ‘the Valley’; see Fig. 1). Phu Lon offers extensive first millennium BCE copper mining evidence (Pigott and Weisgerber 1998), but only the KWPV provides a confirmed copper-smelting assemblage (Pigott et al. 1997).

Fig. 1
figure 1

Location map with the sites and locales of Ban Chiang (BC), Ban Lum Kaeo (BLK), Ban Non Wat (BNW), Khao Sai On (KSO), Khao Sam Kaeo (KSK), the Khao Wong Prachan Valley (KWPV), Non Nok Tha (NNT), Noen U-Loke (NUL), Phu Lon (PL) and the Loei-Petchanbun Volcanic Belt (LPVB)

Full size image

Iron Age (c. 500 BCE to c. 500 CE) Valley copper production coincides with a period of seemingly accelerating social complexity in contemporary Southeast Asian societies, attested by the appearance of site size hierarchies, defensive and/or hydraulic earthworks and the marked ranking of individuals and groups in burial traditions (e.g. Bellina and Silapanth 2008; Higham 2004; O'Reilly 2008; Pautreau 2007; Stark 2006; White 1995). This phase is metallurgically manifested by an increasing deposition of metal, especially copper-base bangles and rings, in predominantly funerary contexts (e.g. Higham 2002, 2010), which we assume, due to the relative paucity of settlement evidence, to represent a notable escalation in general consumption. How this potentially elevated regional demand for metal may have been reflected in Iron Age KWPV supply behaviour is an underlying theme of this paper.

Geology

Mainland Southeast Asia lies on the boundary between the Indian and Eurasian tectonic plates, which accounts for the region's magmatic activity and the relative abundance of zoned metallogenic deposits (Gardner 1972; Lai et al. 2006; Sitthithaworn 1990; Takimoto 1968; Workman 1977). Splitting Thailand asymmetrically into two, the Loei-Petchanbun Volcanic Belt (hereafter ‘LPVB’) runs approximately 400 km SSW from Phu Lon, on the banks of the River Mekong, before turning ESE near the KWPV and continuing for approximately 200 km to the Cambodian border (Fig. 1). It has been assumed that copper supply from the LPVB may have been at least partially fulfilling demand from what appears to be the major locus of prehistoric metal consumption, the Khorat Plateau (see Fig. 1; e.g. Higham 2010; White and Pigott 1996: 158).

The Valley geology consists of Permian calcareous and argillaceous sedimentary rocks with intrusive igneous rocks of acidic or intermediate composition, like andesite, diorite, granite and granodiorite (Cremaschi et al. 1992; Nakornsri 1981; Vernon, unpublished). The contact zone between the igneous and sedimentary rocks is comprised of metamorphic products such as marble and skarn, which are characterised by inclusions of garnet, vesuvianite, wollastonite and quartz (Vernon, unpublished). It is also in this metamorphic contact zone that deposits of the copper minerals chalcopyrite and malachite have been recorded in association with haematite and magnetite at Khao Tab Kwai (100.657° E, 14.983° N) and Khao Phu Kha (100.667° E, 14.950° N; Fig. 2; see, e.g. Bennett 1988b: 128; Natapintu 1988: Tables 1, 2 and 3; Vernon, unpublished). Khao Phu Kha does have extant mining galleries but there is no dating evidence to suggest that these are contemporaneous with Iron Age Valley smelting. In contrast, Khao Tab Kwai has been exploited in recent decades for its iron oxides and any ancient mining evidence that may have existed would have been destroyed.

Fig. 2
figure 2

Quickbird image of the KWPV sites and mineralisations: Khao Phu Kha (KPK), Khao Tab Kwai (KTK), Nil Kham Haeng (NKH), Non Pa Wai (NPW). Width of field is 5,000 m

Full size image

Archaeology

The presence of extensive copper production debris in and around the KWPV was noted in the late 1970s by archaeologists from the Central Thai Archaeology Project (e.g. Natapintu 1984), but it was the Thailand Archaeometallurgy Project (hereafter ‘TAP’; see, e.g. Natapintu 1988, 1991; Pigott and Natapintu 1988; Pigott et al. 1992; Pigott et al. 1997) and the Lopburi Regional Archaeology Project (e.g. Ciarla 1992, 2005; Ciarla 2007, 2008; Cremaschi et al. 1992) that brought the locale to international attention (Fig. 2). Overlaying Neolithic (c. 1800 to c. 1300 BCE) and Bronze Age (c. 1300 to c. 700 BCE) cemeteries (the latter with funerary evidence for copper-base founding [moulds] as well as consumption activities; see Pigott and Ciarla 2007; Pigott et al. 1997), Non Pa Wai (hereafter ‘NPW’—100.678° E, 14.971° N) has 5 ha of early Iron Age (c. 500 to c. 300 BCE) extractive metallurgical debris up to 4 m deep in a loose astratigraphic ashy/powdery matrix with relatively sparse evidence for settlement, though one might assume that metalworkers still lived there (Pryce 2009: Fig. 2.8). Conversely, Nil Kham Haeng (hereafter ‘NKH’—100.656° E, 14.957° N) has more than 3 ha of later Iron Age (c. 300 BCE to c. 500 CE) production material up to 6 m deep, contained within a dense, probably hydrodynamically layered, gravelly matrix of crushed mineral and slag, along with habitation evidence (e.g. post-holes, pottery and faunal remains; Pryce 2009: Fig. 2.10).Footnote 1 There were no clearly discernable metallurgical activity areas at either NPW or NKH, but the many thousands of kilogrammesFootnote 2 of mineral, technical ceramic and slag at these neighbouring and chronologically abutting sites attest to the industrial production of raw copper, presumably far in excess of local requirements (Mudar and Pigott 2003; White and Pigott 1996).

Previous research

From mineralisations to smelting, founding and burial practices, the Valley evidences a prehistoric metallurgical behavioural sequence of unprecedented scale and completeness within Southeast Asia (e.g. Pigott and Ciarla 2007). As such, the KWPV naturally became a locus for regional archaeometallurgical studies during the 1980s and 1990s, with original laboratory-based contributions from Anna Bennett, William Rostoker and Dong Ning Wang (Bennett 1988a, b, 1989, 1990; Rostoker et al. 1989; Wang et al. 1994). Their combined reconstructions depicted Valley metalworkers producing copper from local sources of oxidic and sulphidic copper minerals (predominantly malachite and chalcopyrite, respectively) using ceramic crucibles and perforated chimneys (Fig. 3; see, e.g. Pigott et al. 1997). Analyses of metal artefacts from NKH and other sites suggests that Valley copper had a highly variable (0 wt.% to approximately 20 wt.%) sulphur content as well as occasional trace arsenic, but the dearth of NPW artefacts mean the composition of early Iron Age metal is unknown (Bennett 1989: Table 7). This generalised reconstruction of central Thai metallurgy was well founded and has stood for over 20 years, but lacks technological resolution on the inter-site scale, as well as explanatory mechanisms for technological development and variation (Pryce 2009: 74–78).

Fig. 3
figure 3

Previous TAP reconstruction of KWPV smelting installation (left), courtesy of Ardeth Abrams (Ban Chiang Project); reconstructed perforated ceramic cylinder from NKH (top right), courtesy of TAP; near-complete crucible from NPW (bottom right), photo by Roberto Ciarla, courtesy of TAP

Full size image

Research aim

Although deposit formation processes at NPW and NKH seem to have differed significantly, the sites are united by their both having complex stratigraphy (Ciarla and Natapintu 1992). Although TAP researchers are engaged in resolving these issues (e.g. Rispoli et al. 2009), the current predicament militates against investigations of intra-site technological change, but the chronological and spatial contiguity between NPW and NKH provides an excellent opportunity for inter-site comparison. The Valley metallurgical assemblage consists of minerals, crucibles, pyrotechnological structures, slag, moulds and metal, but the extractive metallurgical focus of the present study dictated a concentration on the first four artefact classes. Detailed microstructural and microanalytical data can be found in Pryce (2009), but in this paper, we primarily concern ourselves with the identification and attempted explanation of changes in copper-smelting behaviour at NPW and NKH from c. 500 BCE to c. 500 CE. Thus, we present different aspects of the analytical results from each site side by side, before using this comparison to address issues related to the origins of metallurgy in Southeast Asia and its subsequent evolution.

Methodology

Given the increasing recognition that archaeometallurgical studies are often far from statistically significant (e.g. Humphris et al. 2009), we attempted to design a KWPV sampling strategy that was, to some degree, representative of ancient industrial activities. Using a stratified sampling frame, artefact bags were randomly picked from selected contexts extending horizontally and vertically across each site (Orton 2000). In total, 38 slag samples (18 NPW + 20 NKH), 19 mineral samples (13 NPW + 6 NKH) and 19 technical ceramic samples (13 NPW + 6 NKH) were chosen from the TAP samples currently held at the University of Pennsylvania Museum for analysis at the UCL Institute of Archaeology's Wolfson Archaeological Science Laboratories (Tables 1, 2 and 3).

Table 1 Names, context numbers and macro-characteristics for KWPV mineral samples
Full size table
Table 2 Names, context numbers and macro-characteristics for KWPV technical ceramics
Full size table
Table 3 Names, context numbers and macro-characteristics for KWPV slag samples
Full size table

Bulk chemical analysis of milled (<50 μm) powder pellets was performed by polarising energy-dispersive X-ray fluorescence ([P]ED-XRF) using a Spectro X-Lab Pro-2000 unit (‘Turboquant’ algorithm for technical ceramic/mineral and ‘Slag_Fun’ for slag; see Veldhuijzen 2003). From NPW, crucible fragments with little or no visible slagging were selected, so analyses predominantly represent the fabric chemistry. This was not possible with NKH crucibles due to uniformly heavy slagging and fabric analyses were performed using scanning electron microscopy with energy-dispersive X-ray fluorescence spectrometry (SEM-EDS) area scans on polished sections. Reference materials (certified and ‘agreed’) were used to check accuracy and precision, and all analyses were repeated three times to ensure data quality (Table 4). Differences from the certified values range from several hundredths or tenths of a weight percent for the minor oxides to a few weight percent for the major oxides—one seeming exception was the detection of iron oxide in ‘Swedish Slag W25:R’, but the consistent variance corresponds to the stoichiometric difference between Fe2O3 (reported) and FeO (the valency expected in slag). Precision was decreased in those oxides with lower concentrations. Trace elements below 10 ppm are reported as ‘<10’.

Table 4 Reference materials [P]ED-XRF bulk chemical analyses; all detected elements reported (data not normalised)
Full size table

The slag bulk chemistry guided sub-sampling for microanalysis by identifying chemical variability and those samples from the two sites with overlapping and outlying values; no mineral samples were examined microanalytically and technical ceramic samples were selected on the basis of the range of phases and interfaces visible in section. Reflected-light microscopy of polished sections under plane- and cross-polarised light was used to identify areas of interest prior to carbon coating and further analysis using an electron microprobe fitted with an energy-dispersive spectrometer (SEM-EDS). The JEOL JXA8600 unit was equiped with an Oxford Instruments INCA X-sight EDS system, controlled by INCA software via a PC. A Co standard was used to calibrate the EDS analyser and was scanned every session to guard against analytical drift, which was found to be practically nil. The samples were analysed with a 10-mm working distance, an accelerating voltage of 15 kV, a current of 1.5 × 10−8 A and a dead time of approximately 30%. The instrument had an accurate detection limit of approximately 0.3 wt.% for most elements; lower values are reported as indicative of presence/absence only. Oxygen was added by stoichiometry where appropriate, taking into account the likely predominant iron valency in slag (Fe2+) and ceramics (Fe3+). All data has been normalised to 100 wt.% but analytical totals are provided. Na2O is not reported for SEM-EDS analyses due to volatilisation-based inaccuracy. For both [P]ED-XRF and SEM-EDS data, ‘nd’ means ‘not detected’ in the tables.

Results

Minerals

The mineral species excavated at NPW and NKH consist of small fragments of copper carbonates and sulphides, as well as siliceous and ferruginous host rock, e.g. malachite, chalcopyrite, quartz, magnetite and pyrite (Tables 1, 2, 3 and 5; see Vernon, unpublished field notes from the TAP geological survey of the Khao Wong Prachan Valley, 18–23 February 1988). The lack of clear stockpiling and/or discard contexts, combined with the unknown mineral suite available to Iron Age metalworkers from Khao Tap Kwai and Khao Phu Kha (Fig. 2), means we do not know the grade and composition of minerals actually selected by Iron Age metalworkers for their copper-smelting charges. The only discernable difference between the NPW and NKH assemblages is the increased proportion of pyritic minerals at the latter (Pryce 2009: 120–123, 172–175). The question of whether this patterning is due to the depletion of oxidic minerals or an increased preference for sulphidic minerals over the Iron Age production sequence can be partially satisfied by the slag evidence (see the “Slag” section), but we do not consider the mineral data to be especially useful in our reconstruction due to our reservations over their representativity.

Table 5 KWPV mineral sample [P]ED-XRF bulk chemical analyses; selected major and minor oxides and trace elements after data normalisation, analytical total presented
Full size table

Technical ceramics

Consideration of the technical ceramics was far more rewarding but space dictates that they are only touched upon here and will be addressed in detail elsewhere (see Pryce 2009). The longstanding TAP copper-smelting reconstruction (Fig. 3) combined the sole near-complete example of a crucible from NPW with the sole reconstructed example of a perforated ceramic furnace from NKH (Pigott et al. 1997). As previously mentioned (see the “Previous research” section), this offered a useful generalised image of KWPV metal production, but ultimately one that blurred important variation between the two assemblages.

Crucibles

Although fragmentary, many of NPW's crucible fragments are identifiable as free-standing vessels made of a tough fabric, despite their being highly vitrified and slagged on the interior surface (Fig. 4a; see Pryce 2009: Chapter 5). Contrastingly, those from NKH are mostly amorphous, extremely friable and often have grit-like inclusions on what appears to be their unslagged exterior surfaces (Pryce 2009: Chapter 6). These characteristics understandably led to suggestions that these ‘crucible slag skins’ represented the clay linings of bowl furnaces (Pigott et al. 1997), but could not account for how NKH metalworkers were able to lift and pour molten materials in the absence of portable crucibles. However, detailed examination of the NKH technical ceramic assemblage held at the Penn Museum revealed a few examples with the smooth curvilinear exteriors of free-standing vessels (Fig. 4b). This suggests that most ‘crucible slag skins’ are probably crucible fragments with highly abraded exterior surfaces, rendering unnecessary the bowl furnace interpretation and explaining how hot liquids were manipulated on site. Given their chemical and microstructural similarities (Table 6; see also Pryce 2009; Pryce et al., forthcoming), we are as yet unable to explain the friability of NKH crucible fabrics compared to those from NPW (Fig. 2). In both cases, the fabrics are relatively coarse, tempered with quartz and organic material and the vitrification gradients make it clear that the heating was from the inside (Pryce 2009).

Fig. 4
figure 4

Curvature of crucible fragments from NPW (left) and NKH (right)

Full size image
Table 6 NPW (top) and NKH (bottom) technical ceramic sample [P]ED-XRF bulk chemical analyses; selected major and minor oxides and trace elements after data normalisation, analytical total presented
Full size table

Tuyères

Neither tuyères, tuyère fragments nor technical ceramic nozzles for organic tuyères have been recovered from NPW and NKH, which begs the question ‘how was air supplied to the smelting reaction’ (see below).

Furnace chimneys

Following a review of TAP databases and field notebooks by Fiorella Rispoli, Roberto Ciarla and Vincent C. Pigott, it was ascertained that perforated ceramic chimneys and readily identifiable fragments thereof were only attributable to NKH. Occasional fragments of technical ceramic with a similar fabric and thickness to the chimneys were recovered at NPW, but the lack of any articulated wall sections or unambiguous perforations means we cannot assume that the two artefact classes are closely related. It follows that the perforated furnace should be removed from the NPW reconstruction.

In fact, Ciarla's 1986 excavation notebook contains a sketch of a possible NPW smelting pit with a squat ceramic rim (reproduced in Pryce 2009: Fig. 5.5), comparable to that noted by Craddock (1995: 180) at Los Millares (Iberia) and by Golden et al. (2001) at Shiqmim (Israel). These low-profile structures might allow a deeper charcoal bath, but they cannot be thought of as furnaces in the sense of encouraging updraft, facilitating very high temperatures and/or very low partial pressures of oxygen (e.g. Merkel 1990; Pryce et al. 2007). Furthermore, 10 full-scale tests conducted by Pryce (2009: Chapter 7) at Fiavè (Trentino, Italy) with reconstructed wind-blown perforated furnaces suggested that these enigmatic cylinders should also be decoupled from the smelting chaînes opératoires at NKH and nearby Khao Sai On (100.730° E, 14.785° N) where they have also been reported (Ciarla 2007, 2008). Whilst the experiments successfully produced copper, in each instance, the chimney interior was vitrified and slagged due to excessive heat exposure and contact with the iron-rich charge. None of the archaeological furnace samples are vitrified or slagged and the presence of intact surfaces indicates that the post-depositional loss of these layers is unlikely. Furthermore, the absence of tuyères from the archaeological record means we cannot speculate that air was directed through the chimney perforations to build temperature in the crucible below as organic blowpipes would soon burn back (also field-tested informally with green bamboo), leaving the chimney to vitrify. When we consider that the crucible fabrics are compositionally very similar (Table 6—and thus with comparable refractory properties, see, e.g. Martinón-Torres and Rehren 2009) but suffer far higher levels of heat damage, we are inclined to suspect that the NKH perforated furnaces have had nothing like the same usage profile. However, one might reasonably speculate that they were involved in crucible-based foundry processes instead—thus being exposed to the fluxing effects of fuel ash but not of iron-rich minerals. The relatively enclosed form of NPW and NKH reaction vessels implies that archaeologically invisible (organic) forced draught systems must have powered both smelting processes, but that the delivery nozzle was frequently withdrawn and/or dampened (no mud allowed—that would leave some evidence). Blowpipes are a possibility, as is the recent historical Southeast Asian tradition for teak or bamboo piston bellows (e.g. Anonymous 1886; Bronson and Charoenwongsa 1994; Wake 1882).

These revisions to the probable role of technical ceramics in smelting operations at NPW and NKH are schematised in Fig. 5 and can be seen to indicate a strong continuity in Iron Age Valley metallurgical behaviour (Pryce 2009; Pryce et al., forthcoming). The remainder of this paper concentrates upon the morphological, chemical and microstructural evidence offered by excavated NPW and NKH slag samples.

Fig. 5
figure 5

Schematic evolution of KWPV technical ceramic interpretation: a previous TAP reconstruction (e.g. Pigott et al. 1997), b taking into account the NKH crucible evidence and c taking into account the archaeologically unattested heat damage seen on experimental furnaces

Full size image

Slag

Thousands of kilogrammes of slag were recovered from NPW and NKH, quantities that must be multiplied many fold when one considers the tiny proportion of the sites excavated (Pigott et al. 1997). Whilst the scale of both slag deposits certainly prompts the adjective ‘industrial’, the aforementioned uncertainties over the mineral (and fuel) inputs dictate that we are unable to estimate by ‘mass balancing’ (e.g. Maldonado and Rehren 2009) the total amount of metal represented by the slag deposits, and the current dearth of intra-site chronology frustrates discussions of production rate. However, the morphological, chemical and microstructural characteristics of excavated slag samples differ significantly and systematically between the two sites, which may be interpreted as partly representing diachronic variation in Iron Age Valley metallurgical behaviour.

Morphology

NPW

Derived from measurements of complete or near-complete examples, the NPW slag morphology appears to be based on plano-convex cakes with a diameter of approximately 15 cm and a mass of up to 1.5 kg (Fig. 6). The form and volume are comparable to the NPW crucibles (Fig. 4), and thus commensurate with the smelting reaction having taken place within these vessels. However, it seems likely that the slag was poured into a depression in the ground whilst still semi-molten, as evidenced by the ‘stiff’ folds and ridges of the cakes' upper surfaces (Fig. 6) and the presence of ceramic sherds and stones embedded in the lower surfaces. There is no sign of an ‘ingot meniscus’ caused by slag floating on freshly smelted molten copper, suggesting that the metal was separated during the pour. Fragments of embedded charcoal are occasionally visible. Most of the other samples appear to be fragments of these cakes, ranging from approximately 5 cm to approximately 10 cm in diameter and several tens to several hundreds of grammes (Fig. 6; Table 3). The level of fragmentation does not correspond with the expected meticulous crushing by metalworkers diligently recovering copper prills or preparing to re-smelt slag (e.g. Bachmann 1982). Given that macroscopic copper prills were not noted, it is, at present, unknown why some of the cakes are complete and why some are fragmentary, but it does appear that early Iron Age NPW metalworkers regarded slag as waste not worthy of further processing. A cut and polished profile (Fig. 6) reveals siliceous minerals dissolving into the melt, as well as large silvery fragments of unreacted mineral iron oxide—features also noticed during microanalysis.

Fig. 6
figure 6

Slag cakes, slag cake fragments and a sectioned slag cake from NPW (top) and slag cakes, slag cake fragments and slag casts from NKH (bottom)

Full size image
Fig. 7
figure 7

Scatter plot showing the lack of correlation between copper oxide and iron oxide levels in [P]ED-XRF bulk chemical analyses of NPW (diamonds) and NKH (crosses) slag—symbols for mean values are enlarged

Full size image

NKH

There are two distinct primary slag morphologies in the NKH assemblage: cakes and casts (Fig. 6). The most complete slag cakes suggest original dimensions of approximately 7 cm to approximately 10 cm in diameter and up to 0.5 kg in mass, i.e. a third of the maximum weight of the NPW slag cakes (Table 3). We do not know the original dimensions of NKH crucibles, but the form and volume of slag cakes are consistent with the expected waste from a single crucible smelt. This is corroborated by the uniform underside of NKH slag cakes (Fig. 6), which, as opposed to NPW, would suggest they solidified within the crucible—though there is no sign of ingot formation. Slag cake fragments range in diameter from approximately 0.5 cm to approximately 5 cm and have a mass of up to several tens of grammes. Especially with respect to the uniformity of NKH's later Iron Age matrix, the relatively standardised size distribution of the slag cake fragments could be regarded as evidence of deliberate crushing to mechanically extract copper prills, but the lack of extant macroscopic copper prills raises the possibility that NKH slag was re-smelted to extract dissolved copper or used as a flux (see Rostoker et al. 1989). The ‘slag casts’ were so named (Pigott et al. 1997) because their consistent dimensions and cylindrical shape strongly suggests they formed in a mould of some sort (Fig. 6). Complete slag cast examples range from c. 5 cm to c. 6 cm in diameter and from c. 50 g to c. 100 g in mass. Slag cast fragments range in diameter between approximately 3 cm and approximately 4 cm and in mass from approximately 20 g to approximately 30 g (Table 3), but appear to be broken rather than crushed. Compared to NPW slag, both the NKH cakes and casts are relatively homogeneous, with unreacted mineral, ceramic and charcoal inclusions being scarcer and largely present towards the surfaces. It is not easy to explain the presence of cake and cast morphologies at NKH, but there is, at present, no more convincing explanation than the reprocessing of cakes. Considering the slightly more homogeneous and occasionally glassy texture of the slag casts, it may be that these were the second-stage waste product and the extant cakes were either awaiting crushing, misplaced or rejected for some reason.

Chemistry and microstructure

In a reasonably uniform geological environment, as exists in the KWPV, chemical patterning in slag can be regarded as a partial proxy for technological behaviour (e.g. Rehren et al. 2007). Therefore, we have endeavoured to emphasise systematic chemical and microstructural variation in slag samples that could relate to differences in the smelting chaînes opératoires at NPW and NKH. [P]ED-XRF bulk chemical data (Table 7) were used to assess inter-sample variability and what this might mean in terms of collective technological choices in Valley copper production. SEM-EDS (Table 8) scans of areas of slag that had been fully molten enabled us to strip out the chemistry of unreacted materials to assess operating parameters and intra-sample variability. Slag microstructures at both NPW and NKH contained skeletal and euhedral olivine crystals, euhedral and dendritic magnetite spinel, interstitial glass and unreacted siliceous and ferrous mineral fragments, indicating non-equilibrium reactions in variable mid-pO2 redox conditions with relatively slow coolingFootnote 3 (e.g. Donaldson 1976; Kongoli and Yazawa 2001: 585). Overall, the largely comparable NPW and NKH slag microstructures are consistent with the similar chemistry of their molten phases; the significant difference lies in the higher abundance of residual minerals in NPW samples.

Table 7 NPW (top) and NKH (bottom) slag sample [P]ED-XRF bulk chemical analyses; selected major and minor oxides, and trace elements after data normalisation, analytical total presented
Full size table
Table 8 SEM-EDS area scans of NPW (top) and NKH (bottom) slag matrices, selected major and minor oxides after data normalisation, analytical total presented
Full size table

Copper content

Plotting P[ED]-XRF copper oxide against iron oxide data (Fig. 7) indicates that copper losses were elevated and variable at NPW and reduced and stabilised at NKH. Microscopic copper prills are common but not abundant in slag samples from both sites (Pryce 2009: 161–162, 210–211) and SEM-EDS area scans of slag matrices suggest that dissolved copper oxide levels are relatively low at NPW and NKH (approximately 1.8 wt.% versus approximately 0.6 wt.%, respectively, see ibid., Tables 7 and 8). The absence of a significant correlation between iron oxide and copper oxide in the P[ED]-XRF slag data from NPW (R 2 = 0.016) or NKH (R 2 = 0.089) suggests that the discrepancy between bulk and phase analyses is best explained by copper losses being primarily caused by unreacted ore minerals, with poor phase separation due to viscosity a long second (e.g. Davenport et al. 2002: 273).

Sulphur content

Microscopic prills of ‘matte’, a copper–iron–sulphide compound, were detected in slag samples from NPW and NKH, indicating that sulphur was entering both smelting systems (Fig. 8). The P[ED]-XRF data indicate that the sulphur content in NPW slag was elevated in some cases, but it was not correlated with copper (R 2 = 0.025), as one might expect with the formation of matte (Fig. 9). Instead, the sulphur content may be explained by the presence of unreacted sulphidic minerals (Pryce 2009: 156–159) and suggests that, whilst sulphur was sporadically incorporated within the smelting charge, it was probably unintentional and could just reflect heterogeneity in the oxidised mineralisation (e.g. Stos-Gale 1989), a form of serendipitous ‘co-smelting’ (Rostoker et al. 1989). However, another possibility is that mineral species were sometimes misidentified by ancient metalworkers, as is consistent with the highest level of sulphur being due to large fragments of sphalerite in slag sample NPWMS7 (Pryce 2009: Fig. 5.41).

Fig. 8
figure 8

Micrograph at ×50 magnification under plane-polarised light of matte prills within a matrix of mid-grey euhedral olivine crystals, light grey magnetite dendrites and a dark grey glass. NKH slag sample NKHMS7. Width of the field is 2 mm

Full size image

In the NKH samples, the sulphur content is much lower, but the stronger correlation with copper (R 2 = 0.359) suggests the more consistent formation of matte, with fewer unreacted minerals (ibid., 206–209). The much increased uniformity could be explained by the systematic inclusion of sulphidic minerals intermixed amongst oxidic minerals within the smelting charge. This could have been the result of oxidic copper ores being gradually exhausted and the supergene or hypogene mineralisation below mined into (Vernon, unpublished field notes from the TAP geological survey of the Khao Wong Prachan Valley, 18–23 February 1988) or deliberate selection on the part of NKH metalworkers. The low level of sulphur in their slag, despite the probable increased input of sulphidic material, could be accounted for by the incorporation of an ore roasting stage in the later Iron Age chaîne opératoire, but there is no direct evidence for this. Given the uncertainties, we prefer to consider the later Iron Age technology one based on deliberate co-smelting rather than the more rigorous process of matte smelting (e.g. Craddock 1995). The variable sulphur content of prills embedded in Valley slag (Pryce 2009: 161–162, 210–211) is consistent with that detected in metal artefacts found locallyFootnote 4 (Bennett 1989; Wang et al. 1994), but the volatilisation of sulphur with each heating cycle in production and recycling means that this element (and, e.g. arsenic) cannot be used as a reliable indicator of Valley copper (Pryce et al. 2010).

Fig. 9
figure 9

Scatter plot showing the lack of correlation in [P]ED-XRF bulk chemical data between copper and sulphur compounds in NPW (diamonds) slag, compared with a weak relationship in NKH (crosses) samples—symbols for mean values are enlarged

Full size image

Tin content

NPW slag contains almost no trace of tin, but NKH samples irregularly contain tin at trace levels (Table 7). Due to the deficiency of tin in the local geological system (Table 5; see also Vernon, unpublished field notes from the TAP geological survey of the Khao Wong Prachan Valley, 18–23 February 1988), the tin patterning suggests that non-Valley minerals and/or artefacts could have entered the NKH production system, presumably during the refining and/or recycling of bronze. The lack of correspondence between tin content and cake/cast slag morphologies indicates that, if the NKH smelt was a two-stage process, the tin-bearing materials were charged from the off. From NPW's Bronze Age funerary evidence, we know that copper-base metal was present in central Thailand by the late second millennium BCE, but the presence of tin at NKH could reflect an increased circulation of copper-base artefacts in the late Iron Age.

Ceramic/fuel ash contribution and slag liquidus

A further significant difference between the NPW and NKH slag bulk chemistries concerns the relative abundance of alkali and earth elements and oxides associated with ceramic and fuel ash (e.g. calcia, magnesia, potash and titania; see Jackson et al. 2005; Merkel 1990: 110), with higher average values for the latter site in all cases. The moderate positive correlation between these oxides [as illustrated in Fig. 10 for potash and titania (NPW R 2 = 0.527 and NKH R 2 = 0.613) but also identified for MgO, CaO, Al2O3, Sr and Zr] indicates a trend towards increasing levels of ceramic degradation and/or fuel ash dissolving into the melt at NKH (Fig. 10). This suggests the later Iron Age smelting process was hotter, longer and/or had a higher fuel/ore ratio than its NPW predecessor—the first option implying increased forced blast and all potentially impacting local fuel reserves.

Fig. 10
figure 10

Scatter plot showing the positive correlation of titania and potash in [P]ED-XRF bulk chemical analyses of NPW (diamonds) and NKH (crosses) slag samples, with increased concentrations in the latter—symbols for mean values are enlarged

Full size image

Of those technical ceramic/fuel ash/non-ferrous gangue minor oxides, MgO and CaO would typically be regarded as fluxes, producing a more fluid melt with less impedance to the agglomeration of copper prills, whereas Al2O3 is normally considered a network former (Davenport et al. 2002; e.g. Eisenhüttenleute 1995; Gilchrist 1989) increasing viscosity and thus negatively impacting copper separation. However, the behaviour of these components can vary considerably and even inverse depending on their abundance and the prevailing partial pressure of oxygen during slag formation (e.g. Kongoli and Yazawa 2001). Liquidus temperatures for NPW and NKH slag matrices (Table 8) were calculated using intermediate oxygen partial pressure thermodynamic models specifically designed for modern copper-smelting systems (Kongoli and Yazawa 2001; Pryce 2009: Chapters 5 and 6).Footnote 5 We suggest that the figures, approximately 1,240°C for both sites, should not be taken too literally considering the many assumptions and error margins in their production, but it could be interpreted that process temperatures at NPW and NKH were roughly equal. However, one must recall the fundamental difference between a liquidus (minimum) temperature and the actual reaction temperature. Experimental archaeometallurgical research (e.g. Catapotis et al. 2008; Merkel 1990) has demonstrated that ancient furnace designs can generate heat several hundred degrees in excess of liquidus. Indeed, actual temperature must exceed liquidus (the point at which the slag is just molten) if any significant density separation of metal and slag is to occur. Considering the comparable degree of slag/crucible contact in both Valley reconstructions (Fig. 5), it is conceivable that the concentration of alkali and alkaline earth components in NKH slag was indeed caused by actual process temperatures substantially in excess of the liquidus, the NKH smelt running for a longer time or with a higher fuel to ore ratio than the average NPW smelt. If the NKH slag liquidus represents minimum operating temperatures, then the enormous heterogeneity of NPW slag, containing abundant residual minerals, suggests its calculated liquidus reflects actual or maximum temperatures achieved.Footnote 6 A more effective generation and distribution of heat (perhaps through refinements in forced draught delivery) may have been an important factor in the improving trend of copper production in the late prehistoric Valley.

Residual magnetite presence

We have made the case for an increased thoroughness and expertise in smelting charge preparation being an important factor in the development in Iron Age Valley copper production. It is then deeply intriguing that a contradictory vein can be adduced by the presence of residual magnetite fragments in NPW and NKH slag (Fig. 11). These almost ubiquitous mineral inclusions cannot realistically be regarded as a flux (‘Any substance that is mixed with a metal etc. to facilitate its fusion…’—OED) considering their vast excess and frequently sharp unreacted boundaries, i.e. if ‘fluxing’ took place, it was massively and systematically ineffective (cf. Bennett 1989: 332). Neither can the magnetite be regarded a source of solid oxygen to drive sulphur out of the smelting system, as could have been argued for the addition of haematite (Kaiura and Tohuri 1979), and has been argued for malachite (Burger et al. 2007). The theoretically sound idea of Rostoker et al. (1989) of the magnetite (or slag as was originally suggested) functioning as a gas-trapping mineral blanket does not satisfy as there is no archaeological basis for it in the prehistoric Valley slag evidence.

Fig. 11
figure 11

Micrographs under plane-polarised light of Valley slag samples: residual magnetite in NPWMS6 (top left, ×50, width of the field is 2 mm) and NKHMS18 (top right, ×50, width of the field is 2 mm), trigonal intergrowths of haematite in residual martitised magnetite in NPWMS7 (bottom left, ×100, width of the field is 1 mm) and sulphidic phases in residual magnetite in NKHMS13 (bottom right, ×50, width of the field is 2 mm)

Full size image

Considering the proximity of magnetite and sulphidic copper minerals at Khao Tab Kwai and their shared characteristics of high density and lustre, residual inclusions in NPW slag could be interpreted as mineralogical misidentification, which would be entirely coherent with the generally poor charge composition in NPW production and the occasional presence of sulphide inclusions within the magnetite matrix (Fig. 11). How then does this correlate with residual magnetite presence in NKH slag, when all the other evidence points towards a substantial improvement in smelting charge formulation? Though its technical purpose remains unexplained and is perhaps inexplicable, it would appear that the addition of crushed mineral magnetite is a unifying behavioural characteristic of NPW and NKH copper smelting—complementary rather than contradictory evidence for the general trend of technological continuity in the Iron Age Valley.

Deposit texture

Finally, important evidence for technological change in Valley copper production can be deduced from the nature of the NPW and NKH archaeological deposits themselves. The 3 to 4 m of loose ashy/powdery NPW deposit certainly manifests approximately two centuries of industrial metallurgical activity, but the lack of conscientiousness and efficiency is encapsulated by the frequent presence of whole slag cakes whose remnant copper content is sufficient for them to be considered ‘ores’ (Table 7). The texture and composition of the NKH matrix stands in monumental contrast. In excess of 3 ha and up to 6 m of pulverised mineral and slag is a poignant testimony to approximately 800 years of massive labour expenditure and the sheer determination of metalworkers smelting and quite possibly re-smelting in the pursuit of copper (Pigott et al. 1997; White and Pigott 1996). The textural similarity of the contemporaneous nearby smelting deposit at Khao Sai On (Ciarla 2007, 2008; see Fig. 1) perhaps indicates that this extractive metallurgical fervour was not a phenomenon exclusive to NKH. Although Khao Sai On is not on the same scale as the Valley sites, it and neighbouring surface scatters suggest an increased prospection and exploitation of small-scale copper deposits in the local area, perhaps corresponding to an Iron Age ‘Copper Rush’ (Pryce et al., forthcoming).

Discussion

Intensification in KWPV copper production

We are happy to concede that the explanation of metallurgical behavioural change is more difficult and tentative than its identification. However, the majority of the evidence presented substantiates the general premise that the improved temperature generation and formulation of smelting charges were critical technological changes over the NPW/NKH transition, c. 300 BCE. It is probably safe to regard the NKH smelting technique as a largely autochthonous development of NPW practice (Fig. 5), but it is improbable that local needs alone could be responsible for the unquantifiable but certainly enormous output of Valley metalworkers. We consider that the apparent trajectory of increasing intensity, standardisation and efficiency of KWPV extractive metallurgical behaviour may have been in response to stimuli from the wider regional arena. This is not a new idea (e.g. Mudar and Pigott 2003), but the revised Valley chronology (Rispoli et al. 2009) is now wholly contemporaneous with the apparently much increased incidence of copper-base grave goods during the Thai Iron Age, c. 500 BCE to c. 500 CE (e.g. Higham 2004; White 1988: 177). Within the limitations of a predominantly funerary dataset (see Nakou 1995 for the EBA Aegean metallschock), if we assume that metal deposition in regional burial contexts is even partly correlated to general Iron Age consumption behaviour, then this ‘starburst’ of competitive social display (from c. 1000 BCE in Higham and Higham 2009: 138) might well constitute the motivation for the industrial-scale production evidenced in the Valley. This could be seen as especially well evidenced at settlements in the upper Mun River catchment just a few hundred kilometres to the east over the LPVB (Fig. 1), some of which (e.g. Ban Lum Khao and Ban Non Wat) have evidence for the alloying and casting of copper-base artefacts (e.g. Higham and Kijingam 2009; Higham et al. 2007; Higham and Thosarat 2004).

If Valley copper was largely destined to fulfil regional demand, but there is no evidence of exogenous control over those metalworkers (White and Pigott 1996), what goods or services were supplied in return for metal remains a frustrating unknown. Mudar and Pigott (2003) have suggested that foodstuffs could have been imported to counteract claimed agricultural marginalism, but recently published archaeobotanical studies suggest that the Valley had a wide variety of crops and could have sustained itself (Weber et al. 2010). Although the social standing of local metalworkers relative to their non-metal-producing neighbours cannot, at present, be reliably estimated, Shennan's (1999) economic model for Bronze Age copper smelting in the Austrian Alps has possibly interesting analogies for the Iron Age Valley. The Mitterberg example described an exchange system whereby small autonomous communities living in areas unsuited for farming mined and smelted copper to gain access to agricultural products and exotic goods (ibid., 360–362). The low labour efficiency of Mitterberg copper production meant that metalworkers were consistently at an economic disadvantage when participating in regional exchange networks, a factor partially responsible for generating and proliferating social stratification in the area. Up to now, we have made a case for low-efficiency NPW and higher-efficiency NKH copper production, which would superficially suggest that NKH metalworkers may have been able to exchange copper on improved terms relative to their NPW antecedents. However, the definition of ‘efficiency’ we have previously employed has related to the loss of copper product and not necessarily to labour efficiency as per Shennan's (1999) hypothesis. Whilst NKH copper production was certainly more effective in reducing metal loss, this ‘efficiency’ probably came at the price of a much increased labour input, e.g. as required by the higher temperatures and the potential crushing and re-smelting of slag. Therefore, we cannot currently estimate whether the c. 300 BCE shift in Valley metallurgical ethos coincided with a modification of metalworkers' participation in regional exchange systems or whether this had an effect on their relative social status.

Though most likely an incomplete account, the consumption of bronze in the Valley area suggests that one of the mediums of regional exchange could have been the provision of tin, whether in the form of tin/bronze ingots and/or bronze artefacts. For the direction of this interaction, we might automatically turn eastwards towards the seemingly well-connected (to Laotian tin fields?) upper Mun River sites (Fig. 1), but we should not ignore the possibility of an about-face to the extensive tin deposits of west-central Thailand, also a couple of hundred kilometres distant (e.g. Bennett and Glover 1992; Coote 1990; Kanjanajuntorn 2006). With even wider repercussions, Pigott et al. (1997) noted that, in late prehistory, higher sea levels would have meant the Gulf of Siam was significantly closer to the KWPV than today (e.g. Sinsakul 2000), and we should not disregard the possibility of Valley metal having been exchanged over long ranges. Valley copper smelting is wholly contemporaneous with the high tin bronze cassiterite cementation process tentatively identified at the peninsular urban settlement, industrial centre and entrepôt of Khao Sam Kaeo (Bellina 2008; Murillo-Barroso et al. 2010; Pryce et al. 2008; see Fig. 1). The lack of known peninsular copper sources suggests that KWPV copper alloyed with tin at Khao Sam Kaeo may have played a role in the fluorescence of trans-Asiatic maritime exchange networks during the Iron Age (e.g. Bellina 2007; Bellina and Glover 2004; Dussubieux and Gratuze 2003; Hung et al. 2007). These possibilities are currently being investigated by the Southeast Asian Lead Isotope Project (e.g. Pryce et al. 2010).

Experimentation and the ‘origins’ of Valley metallurgy

Because the evidence for the earliest bronze metallurgy in the region indicates that it appeared fully-developed, and no signs of an experimental period have been found, the scholarly consensus, despite occasional flashes of discussion (e.g. Higham 1996, 2002, pp. 166, 353, 2006, p. 19; vide Sherratt 2006, pp. 43–44) is that metallurgy—the system of manufacturing, distributing, and using metals and metal objects—was derived from elsewhere (White and Hamilton 2009: 358 [our emphasis]).

Questions on archaeological origins, technological or otherwise, frequently linger on the boundaries of speculation due to the unlikelihood of the material record preserving ‘firsts’ (Killick 2008: 3045), but with regard to the Southeast Asian ‘origins’ debate, there has been no specific discussion of what ‘an experimental period’ might actually look like archaeologically. Although chronologically dislocated from Southeast Asia's earliest metal consumption and founding activities by between 500 and 1,500 years (depending on your chronology; see Higham 2010; White and Hamilton 2009), the Valley smelting evidence may offer some insights into the social context of production and regional variation in the adoption and adaptation of metal technologies. However, lacking well-defined workshops (see, e.g. Qantir Pi-Ramesses, Pusch 1995), archaeometallurgists often struggle to identify physical evidence for the social context of production as per Costin's (1991, 2001) interpretive framework. Though laudable efforts have been made with regard to Valley and Thai metal production (White and Pigott 1996), the data are necessarily circumstantial and the arguments largely ‘absence of evidence’-based. An important aspect of the present study was to make interpretive headway with the evidence we have in abundance: slag and, in particular, variation in slag chemistry.

Archaeometallurgical application of the Weber fraction

Akin to ‘mutation’ in evolutionary biological models, it has been argued that a cultural equivalent, ‘copying error’, can introduce new behavioural and/or artefactual variation due to the inability of humans to exactly replicate (e.g. Collard et al. 2008; Eerkens and Lipo 2007; Henrich 2001; Shennan 2008). A quantitative measure of ‘copying error’, the ‘Weber fraction’, has been derived from cognitive psychological studies of human sensory perception and motor skills in a wide range of skilled and unskilled social learning environments, from various cultural backgrounds (Eerkens 2000; Eerkens and Bettinger 2001; Eerkens and Lipo 2005, 2007). The Weber fraction dictates that, without aids (i.e. a ruler or mould), people will, on average, introduce a replication error of approximately 5% for each successive generation, presupposing they only have reference to the preceding generation. Ignoring any other cultural transmission mechanisms that may be operating on variation, the cumulative effect of up to 5% ‘copying error’ per generation can rapidly produce substantial behavioural and/or artefactual divergence. If the ‘coefficient of variation’ (CV) per generation within an assemblage is approximately 5%, then the null hypothesis is that all material culture change could be accounted for by ‘copying error’ and there is no need to invoke more complex transmissions of cultural information (Bentley et al. 2004: 1449). If the observed CV is significantly less than 5%, then some sort of biassed transmission may be responsible for constraining peoples' choices. Likewise, if the observed CV is significantly in excess of 5%, then we may be seeing the introduction of new behavioural variants through experimentation and innovation. A number of studies on modern and ancient data have demonstrated that much material culture variation can be accounted for by cumulative copying error without recourse to more complex social interactions (e.g. Eerkens and Lipo 2005). However, this does not mean we can blithely apply the Weber fraction to metal production assemblages as there are at least two major problems to be addressed:

  • The 5% CV figure relates to the variability of final artefacts, the entities upon which people can presumably exercise their full sensory judgement (e.g. hearing, sight, smell, taste and touch) of replication fidelity. Slag is a by-product and, whilst metalworkers may have sought to constrain or exaggerate variation in their products, we do not know what quantitative effect this would have on by-product variability

  • Ancient metalworkers would not be able to control variation in slag chemistry to the same degree as artefact variability (±5%) by human senses alone. However, the effects of slag variation would have been detectable in terms of the performance and efficacy of the smelt, slag formation and behaviour during the smelt and its physical characteristics when cooled.

Therefore, it is not unreasonable to expect metalworkers to have correlated, to some degree, the quality of their product and by-product with the composition of their smelting charge. Exciting preliminary results have been provided by Humphris et al. (2009) with traditional nineteenth century iron smelting in Uganda, but we require further detailed case studies correlating the social context of metal production with product/by-product variability to build a reliable cross-cultural interpretive framework (vide Roux 2007), and this would preferably include non-ferrous metals (e.g. Anfinset 2000 in Nepal). In lieu of this, the Weber fraction in particular remains a qualitative though useful means of interpreting slag variability, whereby high relative CV might equate to experimentation and low relative CV possibly indicating standardisation (Pryce 2009: Chapter 8; see Charlton et al. 2010 for an historically insightful and methodologically advanced fully quantitative evolutionary approach; albeit with superior technological and chronological resolution than that available in the KWPV).

‘Origins’

In light of the revised Valley chronology (Pryce et al. 2010; Rispoli et al. 2009), we must now make a sharp differentiation between the earliest currently known evidence for local copper-base consumption and founding (Bronze Age burials, c. 1300 BCE) and the first definite evidence for copper smelting at NPW (Iron Age industrial deposit, c. 500 BCE). Thus, by c. 500 BCE, copper-base metal may already have been used in the Valley for up to 800 years. We do not know how evenly distributed this metal was amongst the Valley populace, but surely the degree of late second millennium/early first millennium BCE bronze exposure rules out any suggestion the NPW technology might be an independent invention of copper-base extractive metallurgy. As for the introduction of smelting technology by foreigners, there is no evidence to suggest the Valley was even partially re-peopled during the early first millennium BCE (Pigott et al. 1997; Rispoli et al. 2009). This leaves the processes of adoption of foreign technology and local innovation of technology as the most plausible explanations for the appearance of the NPW smelting process. However, we suggest any adoption process was far from a complete transmission of extractive metallurgical knowledge and know-how. Mining and smelting require different skills than alloying and casting, and there is likewise a distinct dissimilarity between the rudimentary copper production of early Iron Age NPW described here and the relatively sophisticated founding techniques attested by copper-base artefacts and ceramic moulds in underlying Bronze Age strata (Pigott and Ciarla 2007). Laying ourselves at the mercy of future research, we are inclined to regard the inefficient and non-standardised nature of NPW smelting in the Valley as representing, at the most, a very imperfect and/or selective adoption of an as yet unknown technology or, more probably, a local innovation derived from familiarity with metals rather than direct learning from specialists or migration of metallurgists.

Considering that Valley metalworkers appear to have been familiar with high-temperature founding processes and technical ceramics from the late second millennium BCE, the exploitation of local minerals using the simple NPW crucible-based reaction is not an unreasonable step in an environment of growing regional metal demand. Indeed, in almost every measure, the NPW slag gives the resounding impression of a metallurgical process not yet perfected and it is tempting to label the early Iron Age technique an ‘experimental’ mode of production. This proposition is supported by the consistently high CVs for inter-sample (smelting charge) and intra-sample (process parameter) variation (Tables 7 and 8), relative to later Iron Age NKH (e.g. Eerkens 2000; Eerkens and Lipo 2007; Humphris et al. 2009). There is no pejorative sense to describing the NPW process as rudimentary or unsophisticated—the metalworkers appear to have employed a technology suited to their needs and environment and probably had no need to avoid wastefulness. Indeed, if our interpretation reflects historical reality, then NPW metalworkers achieved an admirable feat in developing their knowledge of bronze founding principles into an effective means of primary copper production, which then followed a trajectory of increasing proficiency and intensity into the relative efficiency and uniformity of the later Iron Age NKH smelting process. Similarly, there is no suggestion of any inevitability in this development, which may largely reflect changing social contexts of Valley metal production in relation to varying regional copper demand. Our interpretation is acknowledged to be tentative and we are by no means suggesting a return to Southeast Asian metallurgy being seen as an independent invention (e.g. Bayard 1980; Solheim 1968), but it does seem that the region's metallurgical history may be far more complex than has been hitherto assumed (Pryce et al. 2010).

The KWPV and the ‘southern metallurgical tradition’

The implications of our KWPV findings can be discussed on the regional stage. White and Hamilton's (2009) term, ‘common Southeast Asian crucible production’, refers to the relatively small (approximately 10 cm diameter) spouted crucibles known for some time from Ban Chiang (Vernon 1997), Ban Na Di (Higham 1988), Ban Non Wat (Higham and Kijingam 2009), Non Nok Tha (Bayard and Solheim 1991 in White and Hamilton 2009) and Phu Lon (Vernon 1996–1997). These crucibles may appear from the early second millennium BCE onwards and for White and Hamilton (2009) are characteristic of a mobile, small-scale, low-capital investment, though not unsophisticated (refractory quartz slurry lagging), technology practiced by widely distributed craft production communities satisfying relatively local demand in a decentralised prehistoric economy (White and Pigott 1996). These characteristics constitute the ‘southern metallurgical tradition’ (White 1988; White and Hamilton 2009) and in this their interpretation seems perfectly reasonable. White and Hamilton (2009) go on to contrast the ‘common’ crucible production with a geographically distinct variant known as ‘KWPV crucible production’, referring to the technology that has been the focus of this paper. Whilst recognising that prehistoric Valley metallurgy is also relatively simple, low-capital investment and commensurate with a ‘community craft specialisation’ organisation of production, White and Hamilton (2009) crucially do not distinguish between the stages of production represented by the archaeometallurgical remains. The significant morphological and volumetric differences between the ‘common’ versus KWPV crucibles could well reflect their variant primary functions. The ‘common’ crucible, absent from Valley assemblages, seems to represent only founding activities, and even at Phu Lon where copper smelting is very likely, the crucible evidence has not yet demonstrated it (Vernon 1996–1997). In contrast, though Valley crucibles are likely to have been used for melting and casting metals too, it seems their principal use was smelting. Therefore, White and Hamilton (2009) are not comparing like with like.

The relative proliferation of the ‘common’ crucible may, to some extent, reflect the relative intensity of archaeological site prospection in northeast Thailand. Here, secondary production centres appear to have been widely distributed amongst potential consumers and metalworkers perhaps had better access to alloying materials like tin and lead via riverine exchange networks. The larger Valley crucibles represent primary production sites whose location is to some extent constrained by the availability of minerals and fuel and are subsequently fewer in number. Thus, whilst Valley crucibles appear and may well be geographically distinct, it is also possible that this variant may one day be reported from as yet undiscovered production sites located in the comparatively under-explored uplands of the LPVB (Fig. 1). Given the current level of evidence, it may be equally valid to state that, at the regional metallurgy scale, the ‘common’ and ‘Valley’ types of crucible production should perhaps be considered as necessarily complementary rather than awkwardly opposing characteristics of the ‘southern metallurgical tradition’.

Conclusion

The original research aim was to produce a diachronic account of change and continuity in Iron Age Valley extractive metallurgical behaviour. Improved reconstructions of copper-smelting activities at NPW and NKH have been developed, providing significant revisions to previous efforts (e.g. Bennett 1989; Pigott et al. 1997). Nevertheless, it is acknowledged that gaps and conflations probably continue to exist in the Valley chaînes opératoires offered here. In our reconstruction, the early Iron Age NPW process may be summarised as the inefficient and non-standardised crucible-based reduction of oxidic and some sulphidic copper ores by ‘serendipitous co-smelting’, whilst the NKH process was the more efficient and standardised crucible-based reduction of sulphidic and some oxidic copper ores by ‘deliberate co-smelting’, both requiring archaeologically invisible forced draught delivery mechanisms. The reconstructed chaînes opératoires were used to identify those characteristic technological choices, or artefactual representations of them, we believe were most definitive of the two technological styles currently discernable. We argue that decreasing chemical variability and copper loss in the NKH slag samples when compared to those from NPW represents the increasing standardisation and intensification of Valley copper-smelting processes during the course of the Iron Age. It is also proposed that this shift in metallurgical ethos may be interpreted as the result of copper-producing communities responding to rising demand for copper/bronze in increasingly ranked late prehistoric Thai communities. Furthermore, due to the relatively rudimentary early Iron Age copper-smelting process attested at NPW, we propose that prehistoric extractive metallurgy in the KWPV may have been a local innovation derived from experimentation and up to 800 years of familiarity with copper-base founding techniques.