Introduction

The comparative study of stoneware found in Southeast Asia has the potential to reveal new insights on their reach of the Southeast Asian Maritime Interaction Sphere (SAMIS), with the choice of sites affecting which aspects of these ceramic trading spheres can be evaluated through quantifiable data. In this chapter, samples studied comprised stoneware found at Kota Cina, northeast Sumatra, a 12th–13th century trading post formed from the coalescence of foreign merchant interest and local systems of social organisation, and Singapore, a 14th–16th century port-settlement which prospered from its geographical location linking the Indian Ocean and South China Sea.

These high-fired stoneware vessels, which remain less well studied and understood relative to contemporary porcelain finds, merely comprise one aspect of the network of goods and ideas traded in Southeast Asia and the SAMIS during three centuries when attitudes concerning trade and exchange swiftly evolved. Regional and mainland settlements ranging from kilns to ports and encompassing intermediaries on land and at sea played various roles in the production, distribution, and use of ceramics, among other materials such as glass and metal, along with food and drink. The South China Sea acted as a conduit for all these entanglements with mainland China along the SAMIS at its most extensive, and in smaller spheres within the SAMIS as well (Miksic 2013, 93–109; Miksic & Goh 2017, 814–26).

The significance of the production and exchange of these high-fired Chinese ceramics is highlighted by their frequent discovery in Southeast Asian archaeological sites, to the extent that many sites can be dated by their correlations with Chinese typology sequences. Ceramics are the main material used to study entanglements between Southeast Asia and China due to the preservation biases resulting from the environment, which degrades many other types of materials but not ceramics (Miksic 2013, 289–90).

The period encompassed by this study coincides with shifts in trade policy both within and between Chinese dynasties, which must be elaborated upon to contextualise the sites, their broader region, and their overall material culture.

The Song Dynasty, a contemporary of Kota Cina, was in crisis by the turn of the 12th century; the north of the country had fallen to Jurchen invaders, and soon all of China would be conquered by the Mongol Khans (Miksic 2013, 99–103). The general reticence of China during previous eras toward commercial engagement with foreign powers via merchants, limited trade to a tributary system where gifts were received from polities in Southeast Asia. This was overturned by the loss of the land Silk Road trade, which made overseas trade necessary for survival. Soon, restrictions concerning private trade were loosened, accelerating engagement with Southeast Asia’s regional powers (Wolters 1970, 35–9).

This policy and the gains it yielded, however, were not enough to reverse the fortunes of the Song Dynasty, which was eventually conquered by the Mongols. The Mongol rulers pursued political as opposed to economic goals in their re-establishment of diplomatic ties with the Malay polities, a state of affairs which persisted until they were deposed by the Han population (Heng 2009, 109–10). The ascendancy of the Ming Dynasty in 1368 and the reimposition of Confucian values led to an overall withdrawal of China from the region, in what has been termed the “Ming gap” in Southeast Asian trade (Wolters 1970, 49–52).

The two sites discussed in this chapter represent two temporal ends of the region’s evolution from the contraction of Srivijaya’s previous hegemony to the establishment of new spheres of influence centered upon Thailand and Java, while trade along the SAMIS became increasingly dominated by private trade until the SAMIS itself was radically altered by Ming dynasty policy curtailing Chinese engagement with Southeast Asia (Miksic 2013, 103–7). Kota Cina’s occupation period highlights this earlier phase at the twilight of the earlier associations and the role of Chinese and Tamil merchants in shaping settlements and patterns of trade and exchange (Miksic 2013, 121–125), while Singapore was contemporary with the later multipolar period (Miksic 2013, 20). Its inhabitants benefited from increased trade volume into and within Southeast Asia and eventually entered literature as a pseudo-legendary kingdom (Miksic 2013, 153–4).

These trends were contemporary with shifts from Chinese re-engagement in trade with the region to the Ming Dynasty’s isolationist policies, alongside the multipolar status quo giving way to the re-emergence of hegemonic polities in the Malay Archipelago, such as Majapahit and Singapore’s successor, Melaka. It is, however, necessary to entertain the hypothesis that variables other than Chinese policies were the main factors in the evolution of new maritime networks in Southeast Asia.

Kota Cina

Kota Cina (lit. “Chinese Fort”) is an archaeological site in northern Sumatra, notable for its vast array of Chinese ceramics. It is not mentioned in ancient sources; Marco Polo’s later account of the region states that there was little sign of settlement in North Sumatra when he arrived there with some Chinese sojourners, necessitating the construction of a walled encampment by his entourage (Miksic 2013, 121–123). Coinage and porcelain from Kota Cina have been dated stylistically to the Northern and Southern Song Dynasties, which overlap with period when Srivijayan hegemony was broken (Miksic 2013, 121–5).

This site does not enter recorded history until the 19th century, when an officer of the East India Company encountered it during a survey of Northern Sumatra; the name appears to have derived from a folktale of Chinese settlers expelling its Indian inhabitants, only to receive divine retribution in the form of a plague of shellfish (Edwards McKinnon 1984, 6–8).

Putting aside these fantastical elements, Chinese and Indian cohabitation was proven to some extent through the 1969–1977 excavations of John Miksic (1979) and E. Edwards McKinnon (1984). The pre-modern settlement appears to have occupied one long episode from the 12th-14th centuries, when settlers associated with massive volumes of Chinese ceramics interacted with others possessing Hindu imagery of the Chola dynasty and Buddhist statuary of South India and Sri Lanka. This interaction of cultures is also evidenced through the wealth of Chinese and Sri Lankan coinage at the site, confirming the 12th-century dates (Edwards McKinnon 1984, vi; 79–83; 106–12; 313–17).

The site’s terminus ante quem was determined from its porcelain. The lack of Yuan or Ming blue-white porcelain at the site indicates its abandonment prior to the trade in this ware, with the site’s abandonment probably dating to the late 13th century (Miksic 2013, 128). Taken together, these finds date Kota Cina’s occupation to the period of Southeast Asian history between the Chola assaults on Srivijaya and the rise of Majapahit (Miksic 2013, 106–11;184–7), and may suggest trends concerning these interactions between Indian, Chinese, and local traders in the economy of Southeast Asia.

Miksic and Yap (1992) aimed to resolve archaeological questions concerning the fine-paste earthenware of Kota Cina, a ware which is unlikely to originate from the immediate environs of Northern Sumatra due to the absence of suitable clay in that region. Using EDXRF (Energy-dispersive X-Ray Fluorescence) data derived from earthenware in southern Thailand, central and eastern Java, along with modern ceramics in Palembang, South Sumatra (Srivijaya’s former capital), Miksic and Yap (1992) measured the proportions of 12 elements using EDXRF to sort them using principal component analysis (PCA) into elemental groups which ought to be correlated with their place of production. Two iterations of PCA (one with all 12 elements, one with 5) were used to demonstrate that the Kota Cina wares (and modern ceramics from Palembang) grouped closer to eastern Java material than southern Thailand’s earthenware (Miksic and Yap 1992, 70–4).

These results overturned their initial hypothesis that the material originated from southern Thailand, despite its geographical proximity relative to eastern Java. Miksic and Yap (1992) suggest that Kota Cina may in fact have been in competition with port-cities in southern Thailand and had deeper connections with the Malay Archipelago preceding its known participation in pre-existing spice trade networks.

Singapore

The antiquity of Singapore, also known as “Temasek”, is primarily suggested by the 17th-century Malay Annals, a text which evolved over time and comprises king-lists and more involved narratives. The history of Singapore presented therein (especially its collapse under Iskandar Shah, its fifth and final king) is chiefly used as polemic, with its more fantastical elements also casting doubt on the historicity of the Kingdom of Singapura as reflected in the Malay Annals (Miksic 2013, 146–54; 229).

However, archaeological proof of Singapore’s material wealth and social complexity has emerged since the mid-1980s following punctuated explorations during British colonisation, and it is now the most extensively studied city of 14th-century Southeast Asia (Miksic 2013, 146–54; 440). Initial archaeological investigations followed the British settlement of Fort Canning, built over an alleged ancient palace complex known as “Forbidden Hill”. Thai and Yuan Dynasty ceramics found in a stratigraphic layer comprising a dark sediment now known as the “Temasek layer” confirm the antiquity of its occupation; numerous excavations of Fort Canning contributed further pieces to this archaeological picture.

Further investigations followed along the Singapore River, mostly spurred by the redevelopment of colonial-era structures, which now overlie the older sites. These have helped in determining the extent of the 14th–15th-century settlement, and in examining variation in concentrations of artefacts, with implications for social stratification in the island. The site of St. Andrew’s Cathedral, erected by the British in 1863, was one such site (Miksic 2013, 240–63).

The vast majority of artefacts found in the “Temasek layer” were ceramics. Rare or unique varieties of porcelain and gold found near the Keramat Iskandar Shah, alongside the presence of earthworks comprising the “Old Lines”—or defensive walls to which Malay and Chinese sources allude—of Singapore, have been used to surmise that the residents of the “Forbidden Hill” enjoyed a privileged status relative to the lowland settlers, which is a pattern found at other ancient Malay capitals (Miksic 2017, 101–22).

Green porcelain forms the majority of porcelain sherds in Singapore; this is of interest in the context of the eclipse of the port of Guangzhou and the Guangdong kilns by Quanzhou and the Fujian kilns. However, the continued presence of white porcelain indicates that Guangdong kilns had not been totally superseded, and it is reasonable to consider that the greenware may have been produced at multiple kilns during this period.

Stulemeijer (2011) aimed to examine the variation within Singaporean greenware, with the explicit goal of characterising this variation only relative to Singapore, as equivalent data from the Chinese kiln sites was not available. Their provenance was instead determined through stylistic analysis by four specialists and comparing the chemical data with the stylistic groups to see if they agreed (Stulemeijer 2011, 93). Instead of principal component analysis, ternary graphs were employed to highlight the variation of key elements. Rubidium, Strontium and Zircon were determined as the most useful in making quantitative distinctions, and scatterplots were used to further explore these (Stulemeijer 2011, 48–71). Most of the sherds were stylistically assigned to Zhejiang, which correlated with their grouping into a fairly coherent Rb-Zr-Sr cluster (Stulemeijer 2011, 98–9). Stulemeijer (2011) concludes that the Singaporean corpus suggests no fewer than three different origins for its greenware, one of which is probably Zhejiang.

With regard to the terminus post quem and terminus ante quem of the Singaporean emporium, the former may be inferred from the presence of Song Dynasty coins, with the caveat that coins remained in circulation long after their minting. Sources from the Song Dynasty contain complaints from court officials about the flow of copper cash out of China, suggesting that Chinese currency was in common use in the societies of some ports by the 12th century. The end of Temasek’s activity as an emporium is difficult to precisely pinpoint, but porcelains of the middle and late Ming Dynasty as well as Vietnamese and Thai ceramics of the 15th century have been found at Empress Place indicating that Singapore received some foreign imports, albeit on a considerably reduced scale, long after the conquest recorded in the Malay Annals which can be linked to historical events of the late 14th century (Miksic 2013, 310; 330–6).

Singapore hence formed a port-settlement of archipelagic Southeast Asia which appears to have been largely autonomous and not hegemonic like Majapahit or Melaka, given that the Singaporean rulers never claimed suzerainty over any external territories despite connections with Johor and Bintan (Xin, 2015, 352–55); Singapore existed contemporaneously with numerous mainland kingdoms which had arisen following the decline of the Khmer Empire. Foremost among these was Ayutthaya, which might be identified with the “Siamese” held responsible for Singapore’s downfall (Miksic 2013, 162–3; 356–7); Siam also included kiln complexes at Sisatchanalai and Sawankhalok, famous for stoneware production (Grave al. 2000, 170–2). Singapore probably paid tribute to both Majapahit and Ayutthaya.

The city-state capitalised on its geographic location to prosper through entrepôt trade; no evidence has been found for great religious or political significance for the site, unlike the capitals of its successor-states, the Melaka and Johor Sultanates (Xin 2015, 352–5). The St. Andrew’s Cathedral site may be interpreted as a non-elite settlement within the autonomous kingdom of Singapura, containing material typical of the realm, and it may be hoped that studying its stoneware might reveal facets of the socioeconomic dynamics concerning the trade in which this mercantile island-kingdom was deeply engaged. Singapore no doubt had a significant relationship with the residents of the Riau islands to its south; that area was a major source of local commodities, which would have been of interest to Chinese traders. The Chinese, however, did not visit those islands, which were associated with pirates and treacherous navigational hazards.

Aims, Materials and Methodology

The following section describes the potential contribution of laboratory analysis to the comparison between the Kota Cina and Temasek sites. The author has decided to focus on an understudied type of Chinese utilitarian ware to see whether this type of study can yield insights into other aspects of Southeast Asian cultural evolution beyond technology. Stoneware has long been at a disadvantage to porcelain in ceramic studies due to its form and function. As stoneware was rarely traded as a commodity itself, its form does not carry much social value relative to porcelain; therefore much less care was put into vessels’ appearances and the corresponding result of this lack of attention to form means that it is difficult to distinguish stoneware by kiln or style anywhere to the same extent as porcelain.

If one is to gain information concerning the more utilitarian aspects of life which stoneware artefacts represent beyond comparing the presence and volume of stoneware as divided into the categories of buff ware, brittle ware and mercury jars, data intrinsic to the sherds and vessels, such as their chemical makeup, is required. The relative proportions and variances of characteristic oxides as listed below will give clues to both the formulae used to make their body paste clays and glazes, as well as the consistency of quality control among the vessels received and discarded at these archaeological sites.

With regard to the two sites sampled in this study, Kota Cina (KTC) represents a sample of settlement in Southeast Asia–specifically at the northern end of the Straits of Melaka—where Chinese and South Asian presence is strongly suggested, and was established during the Song Dynasty. Temasek was coterminous with the Yuan and early Ming Dynasties, and the southern end of the Straits of Melaka, where proximity to China no doubt played a role in shaping local society and economy. St. Andrew’s Cathedral (STA) is unique in Singapore for its location, being situated neither in the presumed elite areas upon Fort Canning Hill nor by the Singapore River, which would have been the entrance for trading ships and boats in the settlement. Buff ware, brittle ware and mercury jars are present in the archaeology of both sites, with a fourth category—tempered jars, typified by thick walls and green glazes—present at Kota Cina but not Singapore. The general comparisons to be made are between categories at each site to determine if typological differences tally with formulaic ones, then between sites per category to see if it is possible that typological similarities belie shifts in body paste and glaze composition over time, or between the trade witnessed at the various points along the SAMIS.

Specifically, the following questions will be raised:

• To what extent is the visual variation within and between the “Buff Ware”, “Brittle Ware” and “Mercury Jars” at both sites correlated with their microstructure and elemental composition?
  • What differences in production (raw material, quality control, techniques) can be inferred from these variations?
  • What can these production differences inform us about the market preferences at these two settlements?
• To what extent can we observe changes in the overall paste and body recipes as well as the consistency of these formulas between 12th–13th century Kota Cina and 14th–16th century Singapore, and infer changes over time in stoneware production sequences, vis-à-vis parallel developments in the Chinese ceramic industries?
• Can traces of any substances which may have been contained within these vessels be detected on their interior surfaces, such as mercury in mercury jars?

Previous chemical studies on Chinese ceramics in China and outside the mainland in settlements connected directly to the maritime Silk Road, or ones which traded with those ports, can be divided into two categories: those which use techniques other than SEM-EDS to detect trace elements (in quantities of less than 0.1%) taken as markers of their geological origins and those which measure major and minor elements (exceeding 1% and 0.1% respectively) to identify the hand of human intervention in elements associated with ceramic recipes and technologies.

The former, which includes the only scientific analyses of ceramics excavated from Kota Cina and Singapore—earthenware at both and porcelain in Singapore—have been successful at identifying statistically coherent clusters of similar ceramics but not necessarily connecting them to specific kilns or clay deposits to prove their provenance. The latter have generally focused on Chinese porcelain with rare explorations of Southeast Asian products like Vietnamese ceramics (Simsek et al. 2015; Polkinghorne et al. 2019), with stoneware largely excluded from these studies. Some have been able to identify body-paste and glaze recipes and changes of these over time (Wen et al. 2007), but usually do not conduct much statistical analysis to identify the distinctiveness of these formulas between kilns (Zhou et al. 2015) to eliminate apparent divergences instead being the products of statistical uncertainty.

These trends in chemical analyses of archaeological ceramics suggest a need to develop a framework to enable further research concerning ceramic production and distribution on this overlooked body of material. This may prove useful not merely for the wealth of stoneware currently stored within the NTU Archaeological Laboratory, but for numerous other contemporary vessels and sherds—both those broadly classified as “trade wares”, with little further analysis, and those which have been stylistically connected to kilns and traditions in order to supply evidence confirming or challenging their identifications.

The material for this project comprised the following:

Site	Ceramic Types	No.	SEM samples
St. Andrew’s Cathedral (STA)	Buff ware	21	19
	Brittle ware	25	19
	Mercury jars	22	10
Kota Cina (KTC)	Buff ware	10	10
	Brittle jarlets	10	10
	Tempered jars	5	5
	Mercury jars	18	12

Table 1: Types and numbers of ceramic sherds in St. Andrew’s Cathedral and Kota Cina obtained from the NTU Archaeological Laboratory in Singapore.

The STA sherds were salvaged from trenches dug for the purpose of laying pipes and cables. The sherds correspond to types well-known from the controlled excavations at the STA site and elsewhere in Singapore. There is little information concerning their archaeological context besides the general 14th-16th century “Temasek layer” in the churchyard. These sherds were chosen instead of sherds with precise provenance information from the archaeological site because the analysis involved some damage to the artefacts, and this was essentially a pilot study. I opted to take samples at random from the bags in which they were stored, with the objective of testing as wide a variety as I could visibly discern of sherds and glazes in terms of size, thickness, and colour.

The KTC sherds were a relatively narrow sample from the site as they were obtained from archaeological investigations of the site in the 1960s–1970s, now stored in the NTU Archaeological Laboratory; there is no context information for these, making them as undifferentiated as STA’s salvage. This sample represents a majority of the laboratory’s collection of KTC stoneware. The KTC sherds are mostly smaller than the STA ones; this might indicate a sampling bias concerning the range of forms and vessels.

The first step in this project was to sort the sherds into the various fabric groups:

Figure 1: The workflow of the visual examination employed in this project.

Groups were created first by similarity in glazes and then body fabrics, such that the glazes could all be studied together. This may lead to greater heterogeneity within the glazed fabric groups than the unglazed groups, as the same glazes may have been applied on multiple body compositions; compositional analysis may hence reveal if this is the case, or if sherds intended for the same glaze were all made together. All glazed sherds were sampled, and (usually) a maximum of three sherds from unglazed fabric groups were selected.

In order to logically organise these sherds and permit further sorting and characterisation, I consulted Mr. Antonis Dimopoulos, a DPhil candidate pursuing ancient DNA research at RLAHA, and enquired as to the suitability of the MySQL database, which he was employing to organise his ancient DNA samples. Finding it to be appropriate, I adopted his system with help from Mr. Devesh Batra, a DPhil candidate at the Oxford Big Data Institute, utilising the MySQL graphics user interface Sequel Pro.

Fabric Groups

The resultant fabric groups are as follows:

Fabric group	Description	No.	SEM
STA-BUF-AA	Brown glaze, exterior	4	4
STA-BUF-AB	Brown glaze, interior	2	2
STA-BUF-BA	Red slip and brown glaze, exterior	1	1
STA-BUF-BB	Red slip and brown glaze, exterior	1	1
STA-BUF-CA	Grey slip/glaze, exterior	1	1
STA-BUF-CB	Grey slip/glaze, both sides	2	1
STA-BUF-D	Unglazed yellowish fabric	3	3
STA-BUF-E	Unglazed grey fabric	5	4
STA-BUF-F	Unglazed fabric, pores on both sides	1	1
STA-BUF-G	“Earthenware”-like fabric	1	1
KTC-BUF-A	Glazed, red-pink fabric	2	2
KTC-BUF-B	Glazed exterior, grey fabric	1	1
KTC-BUF-C	Glazed on both sides, grey fabric	4	4
KTC-BUF-D	Glazed with multiple patterns	3	3
STA-BRI-A	Unglazed grey fabric	7	3
STA-BRI-B	Dark exterior grey fabric	2	2
STA-BRI-C	Pores on both sides	5	3
STA-BRI-D	Yellowish-green glaze exterior	6	6
STA-BRI-E	Striations similar to mercury jars	1	1
STA-BRI-F	Extremely eroded interior	1	1
STA-BRI-G	Reddish-pink fabric	2	2
STA-BRI-H	Two fabric colours glazed interior	1	1
KTC-BRI-A	Green glaze exterior	4	4
KTC-BRI-B	Deformations due to firing	2	2
KTC-BRI-C	Brown glaze exterior	1	1
KTC-BRI-D	Mottled surface	1	1
KTC-BRI-E	Two fabric colours	1	1
KTC-BRI-F	Black glaze interior and exterior	1	1
KTC-TMP-A	Green glaze exterior	1	1
KTC-TMP-B	Brown-green glaze exterior; speckled interior	1	1
KTC-TMP-C	Brown glaze exterior	3	3
STA-MER-A	Dark grey surface; exterior	11	4
STA-MER-B	Brown-grey surface; both sides	6	3
STA-MER-C	Grey body; black pores on interior	5	3
KTC-MER-A	Grey fabric with yellow surface colouration	6	4
KTC-MER-B	Grey fabric; unglazed	6	3
KTC-MER-C	Grey-yellow fabric; unglazed	4	3
KTC-MER-D	Dark fabric; unglazed	1	1
KTC-MER-E	Light grey fabric; unglazed	1	1

Table 2: Fabric groups, their descriptions, and numbers of ceramic sherds present and sampled from St. Andrew’s Cathedral and Kota Cina.

A Munsell colour chart was employed to describe body fabrics, as is common in archaeological ceramic studies (Whitbread 2017, 203); Stulemeijer (2011) employs this methodology in his characterisation of Singaporean greenwares. All the sherds were measured in this way, then those which were cut with coarse and fine diamond-bladed saws to produce smaller ‘sub-samples’ which could be fit into the SEM-EDS’s sample chamber, with new colour measurements taken from freshly-sawn surfaces.

Figures 2a and 2b: Some of the buff ware fabric groups from a) STA and b) KTC.

Figures 3a and 3b: Some of the brittle ware fabric groups from a) STA and b) KTC.

Figures 4a and 4b: Some of the mercury and tempered jar fabric groups from a) STA and b) KTC; tempered jars were found at only KTC.

Chemical Analysis

XRF was selected as a means of investigating the contents of mercury jars. Although the mechanical properties of elemental mercury such as its non-wetting nature at room temperature (Aligizaki 2005, 61) probably dictate otherwise, prolonged containerisation of mercury in these relatively porous wares might have resulted in its retention as mercury vapor; Raman spectroscopy, which sadly was unavailable for this study, is able to detect cinnabar residues on archaeological pottery (Mioč et al. 2004), and XRF analysis is able to detect mercury in Mexican gall-ink (García et al. 2014), as well as mercury-vapor in modern contaminated soils (McComb et al. 2014). Despite these exceptions, the probability of finding any retained mercury remained remote, although it was still decided to employ XRF to test its ability to discern any such substances. Benefits of XRF in this context were the relative ease of its use and availability within RLAHA, and the lack of a requirement for destructive cutting for sub-sampling. Both the interior and exterior surfaces of the mercury jars were analysed (for a standardised period of at least 360 seconds) to reduce the possibility of false positives; if unambiguous mercury peaks could be identified, this methodology would be extended to the rest of the corpus.

Sherd fragments were then ground down to ensure flatness and then embedded into epoxy resin to permit their study within the SEM, a technique the chief strength of which is its ability to collect high-resolution data concerning the micro-texture and elemental composition of properly prepared ceramic sherds (Tite et al. 1982). A vacuum chamber was used to extract the bubbles from the epoxy, after which the pieces of resin containing the sherds were ground and polished for further flatness before being carbon-coated to avoid surfaces being charged by exposure to the electron beam and thus becoming unable to reflect a coherent signal (Reed 2005, 156–7). Once inside the JSM-5910 SEM, the electron beam was used to generate images of the microstructure of the sherd fragments in secondary emissions mode; as the brightness of backscatter-emissions is a function of the atomic number of the area being scanned, it was used as a proxy for the variations in atomic masses across the study area (Reed 2005, 54–7). Looking at the generated backscatter image, I utilised the analysis software INCA 5.05, to generate target areas which appeared chemically consistent (brightness correlating with atomic density), and hence measure the electron signal returned from the sherd surface of these areas in semi-quantification mode, following the example of Tite et al. (1982); due to the roughness of the surfaces being studied, full quantification was neither desirable or feasible (Loehman 2010, 271).

Results – XRF analysis

The analysis was undertaken with the specific intention of finding mercury or mercury vapour which might have leached onto the coarse surfaces of the mercury jars. Of the 40 mercury jar sherds analysed with the XRF, none of them returned any mercury
peaks. This leads to one of two conclusions:

1. There is no mercury to be found on the interior surfaces of the mercury jars, indicating there is no physical evidence for their use for containerising mercury.
2. If there is mercury retained on the interior surfaces of the mercury jars, it is in concentrations which are not easily detectable by XRF.

These results demonstrate at the very least that XRF cannot be used in this fashion to prove that these vessels contained mercury, perhaps necessitating its identification via SEM-microprobe analysis, which would be able to identify mercury-rich phases via backscatter imaging, as mercury’s high atomic number would be reflected as a bright tincture on the rendered image (Reed 2005, 54–7). As SEM-microprobe analysis requires more preparation than XRF, this is unfortunate news for future researchers who may wish to resolve this question but are short on time or resources for dedicated microprobe analysis.

Otherwise, we may potentially turn to other forms of higher-resolution elemental analysis, such as mass spectrometry, which are able to discern much lower elemental concentrations (Pollard and Heron 2008, 27–31). Other potential contents, such as wine and organic residues cannot be discerned by XRF or SEM-microprobe analyses; liquid, gas or supercritical fluid chromatography would be more suited to this task (Pollard and Heron 2008, 66–74).

Results – Body compositions

For the purposes of this comparison, I have chosen to focus on aluminium, iron and potassium; all other chosen elements had mean values of ~1% and are less likely to have had a significant impact on the firing or quality of the ceramic body. Aluminium, even outside of visible formations, contributes refractory properties to ceramics, and iron and potassium conversely act as fluxing agents. As silicon is the most abundant element in clays and ceramic fabrics, it normally comprises the balance of the rest of the elements; hence, its compositional proportions are not focused upon in these analyses.

Oxide Significance

Al2O3	Refractory properties; addition above naturally occurring levels indicative of deliberate clay mixing
Fe2O3	Naturally occurring fluxing agent; usually selected against as part of the high-fired clay production sequence
K2O	Naturally occurring fluxing agent desired in the forms of feldspars and “china stone” to produce clay that vitrify over long periods of time in the kiln to produce hard body fabrics
SiO2	Majority oxide/element in sediments used in ceramics; forms balance of elemental composition; may also be present in the form of quartz-grain inclusions (not statistically analysed)

Table 3: Important oxides used in this study as proxies for quality control and variability in production sequences. Oxide details from Hamer (1975) and Hamilton (1982).

Aluminium oxide

Figure 2: Al2O3 proportions from the STA and KTC sherds, divided into broad categories with outliers separated.

All of the categories at STA and their outliers are statistically distinct from each other. Brittle ware contains more Al2O3 than mercury jars, and buff ware contains more Al2O3% than brittle ware on average. STA-BUF-A has an appreciably high alumina concentration relative to the other buff wares. The same trend holds for STA-BRI-A. Overall, the distinction of buff wares from brittle and mercury jars is vindicated (and perhaps, surprisingly, mercury jars also prove to be distinct from buff ware); however, we must resist over-interpretation of these statistical differences, given that the observed variances may not be particularly useful in terms of ceramic performance.

With the exceptions of STA-BUF-06© and STA-BRI-02(B), all of these body-fabrics are higher in Al2O3% content than naturally occurring stoneware clays in southern China (Wood 1999, 29), possibly indicating some level of artificial addition of kaolin-rich clay to meet the requisite compositions, or simply different clay sources and production centers, even outside of China; it will be necessary to compare Chinese ceramics with clays to see if they directly correspond to them, or also demonstrate such an addition.

All four categories of stoneware at KTC seem to have at least two different peaks as concerns mean Al2O3, most obviously seen in the mercury jars and tempered jars, but also more subtly in the buff and brittle wares. While the consistency of this pattern may suggest the existence of at least two discrete sets of clay sources, it is evident that they do not strongly correspond to differences seen in the other oxides. It is, of course, entirely possible that the clay sources differ only by alumina content.

Buff wares at KTC seem to be divided into the higher-Al2O3 KTC-BUF-A and B, and the lower-Al2O3 KTC-BUF-C and D. It is the latter two fabric groups which display the dazzling multicoloured glazes, which raises questions concerning the intentionality of glazes and surface treatment vis-à-vis the perennial question of “quality”. KTC-BRI-D is significantly lower in alumina content than the rest of the brittle ware. Within mercury jars, KTC-MER-A, E, and perhaps KTC-MER-18(B) might be representative of a low-alumina clay. If the yellow colouration on the exteriors of the KTC-MER-A sherds is indicative of a degraded glaze, there may be parallels with the KTC buff wares.

Iron oxide

Figure 3: Fe2O3 proportions from the STA sherds, divided into broad categories with outliers separated.

We can see that both major categories of buff and brittle, distinguishable by more iron oxide in the latter, have significant outliers, with STA-BRI-G’s case vindicating the hypothesis that its clay matrix was derived from laterite soils.

Among buff wares, STA-BUF-02(B) and STA-BUF-20(E) have high Fe2O3%, hewing much closer to brittle wares; the rest of STA-BUF-E has uncharacteristically low Fe2O3%, even for buff ware. This can also be seen among the brittle wares, where STA-BRI-A and STA-BRI-17(D) conversely have significantly low Fe2O3%. Although mercury jars have a lower Fe2O3% than either buff or brittle ware, it is not statistically distinct from the former. However, as with alumina, this commonality alone may be insufficient to declare mercury jars as sharing some or all of the same chaîne opératoire as buff ware.

Figure 3: Fe2O3 proportions from the KTC sherds, divided into main categories with outliers separated.

Fe2O3% appears to be a more promising discriminant among the KTC categories. As with STA, the brittle wares have a greater mean Fe2O3% than the buff wares; unlike STA where both are enriched in iron oxide relative to mercury jars, the mercury jars have Fe2O3% values between these two categories, as do tempered jars.

KTC-BUF-A and two sherds within C (hence “CA”) appear to have greater affinity with brittle wares than buff wares due to their higher ferrous component, as do KTC-MER-05(A) and KTC-MER-D relative to the other mercury jars. The pinkish fabrics in KTC apparently do not denote the use of laterite clays as they did for STA-BRI-G. Unlike STA, mercury jars do not cluster strongly and separately from the other body fabrics.

Despite the enrichment of KTC-BUF-A and CA relative to the other buff wares, they notably do not cluster together, forming grounds for the suggestion that the corpus represents not just two but possibly three different production centres, albeit based on the small sample of four out of ten sherds. These risks are highlighted with the issues encountered with the brittle wares, where almost no fabric groups cluster with each other.

Unlike the buff ware outliers, KTC-MER-A and D do group together. KTC-TMP-C’s differences from the other two tempered-jar sherds are further emphasised by their statistical separation concerning Fe2O3%. The ferrous content of both buff and brittle wares increased by statistically significant amounts from KTC and STA, which is unusual considering that alumina also increased for both materials between sites. Mercury jars decreased in ferrous content while also decreasing in alumina, suggesting the opposite trend to both other categories. The only category at STA with which KTC tempered jars correspond is its buff wares; it does not match any other STA categories or outliers.

None of the buff ware outliers at either site correspond to the bulk of buff wares at the other, besides STA-BUF-E with KTC buff wares. Similarly, it appears that the brittleware and mercury jar outliers are exceptions at both sites, indicating that their grouping outside the main categories must be explained by means other than those of ingression of new production centres to the markets of KTC, or the resilience of old kilns at STA.

Potassium oxide

Figure 4: K2O proportions from the STA sherds, divided into main categories with outliers separated.

Within the buff wares, there appear to be two sub-groups: the low-K2O STA-BUF-D, E and F, and the high-K2O STA-BUF-A and C; STA-BUF-B appears to be an unusually high-K2O group, even outside of the latter, and STA-BUF-G is the sherd with the lowest K2O% out of the entire corpus, perhaps reflecting its more earthenware-like characteristics.

There appears to be a similar bifurcation among the brittle wares, with STA-BRI- B and E forming a low-K2O group and the rest forming a high-K2O group, besides STA-BRI-G. STA-BRI-H has a potassic component much more similar to buff-wares than those of the brittle-wares, furthering the hypothesis that it is probably misclassified. Mercury jar fabric groups are statistically indistinguishable from each other in terms of K2O%.

Figure 5: K2O proportions from the KTC sherds, divided into main categories with outliers separated.

Among the buff wares, only the high-K2O KTC-BUF-A and B group together; KTC-BUF-C and D, while both lower in potassic content than the former two fabric groups, do not cluster with each other, furthering the suggestion that at the very least the buff wares represent at least three different production sites.

Although there appear to be similar low-K2O outliers among the tempered jars, these are not statistically distinct from the rest of their respective categories. The first two analyses in KTC-MER-06© are abnormally high (6.11%; 6.24%). As the other two analyses fit neatly within the range of the rest of the mercury jars, their significance is unclear. It may simply be that the centre of the sherd was anomalously high in potassic content.

KTC-BRI-B is a low-K2O outlier, and the case may either be that the rest cluster into a bimodal distribution, or that KTC-BRI-F is the opposite in that it represents a high-K2O ceramic. At any rate, the brittle wares, like the buff wares but through a different pattern, appear to represent at least three different clays and/or production centres.

Results – Glazes

SEM analysis of glazes necessitates measurement of a different set of elements from body fabrics due to the different mechanisms involved in firing glazes. As the intended goal is the achievement of a totally vitrified outer layer, it is fluxing agents to which we must turn to discern variation in glaze production sequences.

I have chosen to focus primarily on calcium, potassium, and lead due to their relative abundance within this corpus and their known uses in Chinese kilns and other production sites. Iron will also be discussed as its relative abundance suggests that it is the primary colourant used in this corpus. Other oxides found in similar proportions to these were those of magnesium, phosphorus, and manganese, all of which have various useful properties in glazes (Hamer 1975, 42; 56; 161–4; 176; 188; 193–5; 221–2; 231–2).

Table 4: Elemental proportions in ceramic glazes observed in STA and KTC, with assumed chief fluxing agents bolded.

Fluxing agents

Most of the STA glazes and all KTC glazes have calcium as their primary fluxing agent. This is entirely consistent with the extensive use of lime in Chinese glazes (Wood, 1999, 30–3), and may indicate that the kilns making these glazes borrowed from the same pottery traditions. The yellowish colouration on surfaces of KTC-MER-A sherds contains CaO% values very similar to the KTC glazes, with no other factors in this consideration excluding the possibility that these are the retained interfaces of degraded glazes. This aligns with the known preponderance of glazed mercury jars in earlier periods (Wong 2016: 14).

KTC-BUF-C and D were noted during visual examination for the use of visually different glazes on their interiors and exteriors, with the latter even expressing multicoloured patterns in some cases. Under the SEM, only the latter groups’ glazes were discernible in backscatter mode. It is evident that there is about a 7–9% difference in CaO concentrations between the surfaces of KTC-BUF-D, suggesting that there is a substantial difference in the lime content of their glazes. KTC-BUF-07 and -09 have more lime on their exteriors, while KTC-BUF-10 has more on its interior. These trends raise questions concerning the motivations of potters formulating entirely different compositions of glazes for the same vessel.

Minus the one lead-glazed sherd, all sherds not utilising lime glazes instead used potash as their chief fluxing agent. These belong exclusively to STA-BUF-02(B), 06©, and 12©, all of which are relatively matte. Although STA-BUF-13© was grouped with the latter two sherds in the same group, its matte “glaze” is calcium-rich, in keeping with the majority of the corpus. The use of potash in glazes was not unknown to Chinese potters, but was not their sole domain either (Wood 1999, 215–16). Evidently STA demonstrates a somewhat greater variability than KTC in overall terms concerning main fluxes.

Only STA-BRI-21(D) appears to have a concentration of MgO comparable to its CaO content. At 9.0%, this exceeds all other MgO proportions among both bodies and glazes, indicating the deliberate addition of magnesia far above naturally occurring levels and potentially signifies a production sequence divergent from the rest of the brittle wares. Among the entire corpus, the sole lead-fluxed glaze was that of STA-BRI-20(G). This, combined with the sherd’s uniquely high ferrous proportions in its body fabric, suggests that it represents a unique ceramic tradition exhibited in this one sherd at STA.

Colourants

The versatility of both iron and manganese oxides as colourants, as explained above, was well-known to potters of the Song-Yuan period. Almost all glazes contained at least 2.0% Fe2O3 and at most 2.0% MnO, indicating that the former and not the latter is serving as the primary colourant for practically all sherds.

Comparison with Chinese ceramics (from Wood 1999)

Figures 6a and 6b: Fe2O3% vs. Al2O3/SiO2 ratios of a) STA and KTC means per category and outliers, b) Chinese ceramics sampled in Wood (1999), represented as grey squares in 6a.

KTC	STA	Category
♦	●	Buff ware
♦	●	”	outliers
♦	●	Brittle ware
♦	●	”	outliers
♦	●	Mercury jars
♦		”	outliers
♦		Tempered jars

Porcelain		Stoneware		Clay
●	Jingdezhen (& celadons)	♦	Jian ware	■	Fujian Neolithic Clay
●	Jingdezhen Yingqing ware	♦	Yixing red stoneware	■	Southern China
●	Longquan (& celadons)	♦	Ding glazes	■	stoneware clays
●	Longquan Guan ware
●	Hangzhou Guan ware
●	Dayao ware	♦	Northern Cizhou
●	Jizhou ware		glazes

Table 5: Legend for Figures 6 and 7. Values are derived from Wood (1999).

With regard to ferrous content vis-à-vis the alumina/silica ratio, we can see that kilns overall tend to be differentiated by their Fe2O3 concentrations while variability in both contemporary firings and trends over time is more discernible by the change in Al2O3/SiO2. We can see that the group with which the KTC and STA stonewares share the most affinity in both dimensions is that of the Hangzhou Guan wares. Totally equating them with the production sequences represented by the Hangzhou kilns would, however, be fallacious as Guan wares are a precise formulation derived from the need for the retention of exterior quality despite an artificial dearth of clay materials for the bodies, hence the infamously thick glazes (Wood and Li 2015, 628–9).

The other kilns with which the KTC and STA wares broadly correlate are Longquan, with their distribution essentially being neatly sandwiched between the Longquan porcelains, their Dayao celadons and Longquan Guan wares. Hence it is possible that the Chinese ceramics at both sites derived from the same technological suite available to the Longquan potters, a conclusion which dovetails neatly with the known trends concerning the rise of Zhejiang kilns channelled through Quanzhou over the period represented by both sites. The ferrous proportions of Jingdezhen wares suggest a more concerted effort in selecting against these than at the other kilns.

KTC and STA wares are notably strongly differentiated from the high-Fe2O3 stonewares such as the Fujian-made Jian and Jiangsu-made Yixing wares; even the lateritic stonewares of STA-BRI-G do not demonstrate either the ferrous component or alumina/silicon ratio seen in those technological traditions. It is evident that STA-BUF-12© is truly exceptional in terms of its alumina content and probably represents a clay source which cannot be correlated with any of the kilns in this study. STA-BRI-02(B) represents the opposite scenario: it has a smaller alumina content than the lower boundary of the Southern Chinese stoneware clays surveyed by Wood (1999).

However, before concluding that the Longquan kilns are the most likely candidate for the production sequences evidenced by the KTC and STA sherds, let us look at the potassic content for further clues as to the chaînes opératoires concerning these stonewares.


Figures 7a and 7b: K2O% vs. Al2O3/SiO2 ratios of a) STA and KTC means per category and outliers, b) Chinese ceramics sampled in Wood (1999), represented as grey squares in 7a.

The K2O content of the various body ceramics reveals a somewhat more nuanced picture than the Fe2O3 plot alone. Potassic content appears to be more variable over time than ferrous content within the same kiln. It is important to recall that potash generally needs to be added to clays as a flux, necessitating an externally derived source beyond certain levels.

The range with which the KTC and STA corpus groups the best is the Hangzhou Guan wares, but its doubtful candidature has already been outlined above. Outside of this particular “match”, the situation diametrically flips compared with the Fe2O3% vs. Al2O3/SiO2 plot. Now, it is Jingdezhen with which the KTC and STA ceramics correspond best, as both have less potassic content than the Longquan wares. However, this is insufficient ground to reject the premise outlined above.

Differences in potassic content are more suggestive of technological as opposed to geological differences; the Jingdezhen and the Southeast Asian K2O% values notably are well within the natural ranges expressed by the surveyed clays. It may be that the KTC and STA wares were made in a similar fashion to the Longquan porcelains, with the latter containing additional potash-rich materials relative to the stonewares to permit porcelain firing. Naturally, the opposite conclusion may be true, namely that the KTC and STA wares are derived from the same source as the Jingdezhen porcelains but less well-selected against ferrous impurities; this theory, however, raises questions as to why their alumina content is relatively higher than most of the Jingdezhen porcelains. The sole potassic outlier, STA-BUF-06©, is not very far off from the Longquan kilns’ K2O% vs. Al2O3/SiO2 distributions.

It may be hoped that with future studies concerning these kinds of materials in both China and Southeast Asia, the compositions of these neglected stonewares may one day be plotted as extensively as porcelain measurements and possibly be used to distinguish not only between source recipes but their distributions within their target markets, permitting the direct connection of ceramic sources with their ultimate destinations. Undoubtedly, the addition of this independent line of evidence will allow future researchers to confirm or develop theories and trends concerning differences between the preferences for different Chinese ceramics in various markets of Southeast Asia. If the existence of different preferences in different markets is confirmed, we can proceed to try to discern how much of this difference is due to trade routes, how much to the positions of sites at different levels of the socio-economic hierarchy, etc. We could also measure the changes in the flow of various types of Chinese ceramics to Southeast Asia (and further west) at different periods.

The glazes seen at STA and KTC exhibit K2O and CaO proportions largely similar to those seen within the remit of contemporary Chinese pottery; even the low-calcia high-potash buff wares at STA share antecedents with some of the alkali glazes seen in China. The excavated sherds’ chemical compositions are broadly compatible with the technologies witnessed in Chinese ceramics; however, they exhibit somewhat greater variability than the ranges measured by Wood (1999), with the KTC buff wares’ differences between sherd sizes marking an extreme of this trend. It is due to this variability that difficulties emerge concerning the ability to pinpoint any one kiln as being the majority supplier (or even being able to discern clusters which might correspond to specific kilns) for Singapore and Sumatra, even given evidence from the body-fabric affinities.

Unlike the K2O vs. CaO ratios, the Southeast Asian material stands in stark contrast with the surveyed Chinese ceramics due to their relatively elevated MgO and P2O5 content. The only wares with similar glazes differ in body composition; these were the high-iron Jian ware and the high-alumina, high-potash Jizhou ware. It is necessary to temper previous deductions with the caveats that while the Jingdezhen and Longquan kilns share body-ceramic technology with the excavated sherds, other formulae must be found for their glazes.

Curiously enough, MgO and P2O5 both appear to exhibit statistically significant linear correlations with CaO, suggesting that plant or perhaps bone ash was being used for lime content in the glazes of the sherds excavated from KTC and STA compared to the corpus surveyed by Wood (1999). It may be that the glazes seen in Southeast Asia represent a different market with different demands from Chinese customers, influencing technologies employed by kilns. Conversely, these may be glaze recipes not from China, but of kilns in Southeast Asia—one candidate is Sisatchanalai in Thailand, a prolific contemporary stoneware producer (Grave et al., 2000)—and the body fabrics in fact represent their products, which have successfully copied Chinese body-ceramics but not their glazes, either by choice or due to geological limitations.

Either way, these data, in conjunction with the other information from the glazes, have posed numerous questions regarding the technology contributing to the corpus in Singapore and Sumatra and their potential origins, and it may be hoped that future work on these under-appreciated vessels may open up new lines of evidence.

Discussion

Kota Cina and St. Andrew’s Cathedral, Singapore

Overall, what can these compositional data and scatterplots truly inform us about the nature of, and changes in, stoneware production as manifested along the SAMIS between 12th-13th century Sumatra and 14th-16th century Singapore? Having established that the suite of body ceramics available at both sites largely falls within the spectrum of contemporary Chinese kilns, what further details can be teased out of the variations in measured oxides in both the body fabrics and glazes?

Comparing the mean proportions of the various metrics (Al2O3, Fe2O3, K2O) chosen to explore the “quality” and production of these stoneware vessels is perhaps most instructive in the identification of outliers, which illuminate the relative diversity of products received at both ports. Here, STA contains slightly more exceptions to the general ceramic ranges; its outliers are also more dramatic, such as the extremely Al2O3-rich BUF-12©, Al2O3-poor BRI-02(B), Fe2O3-rich BRI-G and K2O-rich BUF-06©. Compounded by the presence of potassic and lead alkali glazes, these trends suggest a more diverse clientèle being served in Singapore than Kota Cina. It is also possible that more Chinese kilns were competing with each other for the export market in the 14th century than in earlier periods. The proportion of Chinese to local ceramics is also greater at STA (50:50) than at KTC (30:70).

In practically all categories and metrics, the STA material demonstrates greater variability from KTC, but there is a risk of over-interpretation as STA buff and brittleware samples outnumber KTC ones two-to-one. Given the occupation timespans of the two settlements (about two centuries for Kota Cina and three for Temasek) this may suggest that the greater volume of trade seen at STA may also be correlated with a broader range of sources either concurrently or over a longer period, implying diversity not only of recipients but of producers during the period of Southeast Asian political history represented by ancient Singapore. Of course, there remain major exceptions such as mercury jars, which have proven to be a surprising contrast to the overall trend. Despite similar sample sizes, the KTC mercury jars show considerably more variability and outliers than the STA ones. Combined with the knowledge that the KTC-MER-A sherds demonstrate the retention of glazed jars into its period of habitation, this suggests the possibly that by the time of STA’s occupation, mercury jar production might have become monopolised by just one kiln complex, and further implies that whichever industry these jars facilitated had also similarly coalesced.

Southeast Asia and China

With regard to the exact relationship between the Southeast Asian finds and the known products of Chinese kilns as measured by Wood (1999) as plotted in the above graphs, the results of their comparison seems more nuanced than the initial apparent similarity of their technology; this is with the caveat that relatively few samples of each type of porcelain were taken, and barely any stoneware—all of the ceramics in Wood (199) were all measured with XRF, which is markedly less accurate than SEM-EDS. For the most part, the body ceramics seen in KTC and STA, while divergent between each other to some extent, are conservative in their formulation vis-à-vis known Chinese ceramic formulae, minus the occasional outlier, some of which also fall within the range established for some of the Chinese stoneware.

This contrasts considerably with the patterns seen in the glazes. Despite sharing affinities in overall lime / lime-alkali flux formulation, the sherds found at KTC and STA differ massively from Chinese norms with respect to magnesia and phosphoric content which are themselves indicators of plant ash as the source for calcium, not to mention the divergent proportions of iron-oxide colourants employed.

Several causes for these dramatic divergences may be postulated:

• These are Chinese wares which have been glazed with formulae which have yet to be scientifically recorded.
• These are ceramics which have managed to replicate Chinese ware, but use glazes of different flux formulation.
  – Plant/bone-ash flux formulae were chosen due to a lack of limestone.
  – Plant/bone-ash flux formulae were chosen due to prevailing ceramic technological traditions.

The second hypothesis may be dismissed due to the vast abundance of limestone seams and karsts in the region (Gillieson 2005, 174). Between the other two hypotheses, the first is essentially an argument from lack of evidence; conversely, it is known that wood-ash glazes have been employed extensively in Southeast Asia from at least the Angkorian period, lending weight to this theory (Miksic 2009, 51). Ideally, a broader investigation utilising element quantification aimed at Southeast Asian stoneware bodies, along with wood-ash and limestone fluxes, would be able to provide a suitable corpus to compare these technologies. This would help to establish that the KTC and STA sherds could possibly be accounted for by less distant kilns, in which case, more questions are raised concerning the probable nodes within the SAMIS, to which these two sites were connected via trade lay in these critical centuries.

More research on sherds from Chinese kiln sites and port cities should also be undertaken. It is possible that the stoneware vessels intended for export were made at different kilns from those destined for Chinese domestic markets. There are, however, very few publications devoted to the habitation areas of Chinese ports which were in contact with Southeast Asia within the SAMIS. Not all is lost, however; while it previously was difficult if not impossible to make meaningful comparisons between STA and KTC wares and those exported by the ports of Guangdong, Fujian, and Zhejiang, chemical analyses such as those recently performed upon the shipwreck of the Nanhai One, which sank near Guangzhou during the Southern Song era (Zhou 2019, 12880–1), have the ability to provide data for these very comparisons, given the positive identifications of Longquan, Jingdezhen and Dehua material on board which were measured in that study. Regarding shipwrecks which may help with understanding ancient Singapore’s ceramic trade, the Temasek shipwreck off Pedra Branca (the cargo or which survived Singapore’s coastal conditions, but not the ship itself) has been identified as one such contemporary based upon comparisons with Singapore’s terrestrial archaeology including STA, with further study pending (Flecker 2022, 23–39).

Although neither vessel contained evidence that their destinations were Kota Cina or Temasek respectively, their cargoes are still of great interest for representing the possible range of materials exported and received along the SAMIS in relation to these two settlements. Future studies ought to compare their porcelain compositions as well as their stoneware ones to test if the porcelain on board tallies formulaically with those deposited at these sites. Should that prove to be the case, this would provide circumstantial evidence for their manifest possibly including trade at these settlements, with similarities and differences in their recovered stoneware necessitating interpretation in this context, i.e. if the cargo theoretically could have been imported at Kota Cina and Singapore or if its destination was elsewhere within the SAMIS, such as Chinese domestic consumption. To this date, only statistical aggregates have been released concerning body-paste formulas aboard the Nanhai One, and the Temasek shipwreck is still in early stages of study; it is hoped that both may be measured in sufficient numbers for meaningful statistical tests to be conducted upon their cargoes to compare them with existing data from these two settlements.

Conclusions

The routes taken by merchants, envoys, and other sojourners within the SAMIS were more than just means of conveying stoneware (or goods contained therein) from one point to another. It would be fallacious to assume that every site was a scale-model microcosm of overall trends, even when attempting to differentiate Kota Cina from Singapore both geographically and temporally. Stoneware hence only tells us part—albeit a hitherto overlooked part—of the dynamics which governed people’s lives in these times and places.

Even within ceramics, earthenware, stoneware, and porcelain fulfilled very different roles in society. These artefacts also undoubtedly differed by social stratum between and within elites and non-elites. Although I have assumed that both Kota Cina and St. Andrew’s Cathedral are non-elite settlements, STA is possibly connected to the elite residents on “Forbidden Hill”, which might account for its greater variability. No information exists regarding the possible location of structures associated with political leaders at Kota Cina.

Having demonstrated that we may have to look beyond surveyed Chinese kilns for the source of the glazed sherds—not to mention the compositional outliers—further questions may be raised concerning the means (i.e. ships) which brought them from kilns to ports, and from there dispersed to other settlements. Although shipwrecks can tell us much about the ratios of ceramic vessels upon them, it would be beneficial to have a larger sample in order to gauge the degree to which they represent all shipping involved in trade in the region.

This brings us to a more contentious issue: namely, the degree to which pottery is representative of trade and exchange as a whole. Inasmuch as ceramics are highly versatile in use and highly taphonomically resistant, permitting study of their essentially unaltered body and glaze fabrics centuries from their final discard or archival, they are neither foodstuffs nor dedicated tools. As Fahy (2014) surmises concerning contemporary maritime shipwrecks, ceramics benefit from the wealth of existing literature, whilst metal artefacts and ingots, and especially organic remains, are generally understudied.

A comprehensive picture can only truly emerge through the analysis of these multiple dimensions, perhaps first through simple comparison of terrestrial and maritime deposits and seeing if settlements, or combinations of settlements, can account for the volumes of trade on the seas; following this, real effort will have to made concerning the bio-archaeology and material science of the more perishable goods to gain understandings as to their norms and variability concerning both producers and recipients.

Perhaps through harmonising all of these, it may one day be possible to gain a fleeting glance of the most ephemeral exchanged commodities—the ideologies and philosophies which were espoused, shared, and debated by the inhabitants of these polities, through using these more tangible lines of evidence as proxies. In order to draw truly instructive conclusions from just ceramic data, a whole suite of material from a variety of contexts—Chinese and Southeast Asian, terrestrial and maritime—will have to be sampled for body and glaze compositions, perhaps finally resolving mysteries concerning ceramic traditions and transmission thereof in both knowledge and products; the addition of data from Southeast Asian high-firing kilns will make considerable progress in adding nuance to the network of ceramic producers of the region and the extent to which Chinese exports dominated the SAMIS.

In conclusion, this study opens up new avenues for high-resolution studies of ceramics. This procedure demonstrates how the wealth of material culture available to researchers can be exploited to investigate this region and its history, not merely in terms of stoneware, but to form a picture of life in and beyond the port-settlements of Kota Cina and Singapore as they stood at the junction of land and sea.

Bibliography