Reexamining Ceramic Standardization During Agricultural Transition: A Geometric Morphometric Investigation of Initial – Early Yayoi Earthenware, Japan

1.1 Standardization Theory and Geometric Morphometrics (GMMs)

In recent years, geometric morphometrics (henceforth GMM) investigations of shape have shed a new light on the deep complexities of intra-group idiosyncratic variance and micro-styles (such as in Gandon, Nonaka, Endler, Coyle, & Bootsma, 2020). In particular, outline-based GMM is an exceptionally useful statistical method from which differential degrees of transforming patterns of craft standardization can be accurately assessed and expressed in quantifiable means (a recent example can be found in Wang & Marwick, 2020). In traditional studies of ceramic standardization, the coefficient of variance (CV) test is often employed to quantify the degrees of variance in a percentage form; however, this quantification is often based on linear measurements or size rather than the holistic shape capture seen through outline-based GMM. Ceramics, especially those of the non-wheel thrown vessels of the initial-early Yayoi period, often see variation in size factors as a major hinderance. As such, this study utilizes CV testing based on standardized principal component (PC) scores of the outline-based GMM analysis as a workaround to this issue, while also benefitting from the quantification of variance CV provides.

“Standardization” in material cultural studies is a quantitative measure of ceramic specialization and refers to the process of creating goods that show little heterogeneity (whether this be in form, decorations, etc.), due to a process of production from a “limited range of materials and somewhat formalized or routinized production” (Arnold & Nieves, 1992, p. 93). The study of ceramic standardization has been conducted from the early 1980s (e.g. Rice, 1981) and continues to be a reliable measure of specialization, and how specialization changes over time due to socio-cultural and environmental factors. Standardization in ceramics is most often quantified through the analysis of variability in the form of vessels.

Geometric Morphometrics is often broadly defined as a set of various methods with the explicit goal of extracting and elucidating shape or size (with both combined referred to as “form”). Cooke and Terhune define GMM (referred to as simply “GM” in their example) as “a collection of approaches for the multivariate statistical analysis and visualization of Cartesian coordinate data”, but also liken GMM to a “toolkit” with each “tool” being different data extraction and statistical analysis (Cooke & Terhune, 2015). In this way, GMM is a range of analyses which is impossible to say which toolkit is “correct” and instead should be seen as an idiosyncratic tool itself, of which each scholar can choose how to transform to the needs of any given study. Despite this broad nature of GMM, most modern studies utilizing GMM can be categorized into three major groups: (1) traditional/conventional, (2) landmark-based GMMs, and (3) outline-based GMMs. These categories are based on the means in which data are extracted, by the use of distance measurements, landmarks, or outlines, respectively. Due to the varied nature of “size values,” and the low degree of precise landmarks available, this study utilizes the increasing popular method of 2D outline-based GMMs. In particular, elliptical Fourier analysis (EFA) for outline extraction (Caple, Byrd, & Stephan, 2017; Carlo, Barbeitos, & Lasker, 2011; Kuhl & Giardina, 1982), principal component analysis (PCA) for ordination, dimensionality reduction, and morphospace distribution (Cooke & Terhune, 2015), multivariate analysis of variance (MANOVA) for group-based significance testing, and thin plate spline for transformation visualization purposes, all of which are incorporated in this study. Furthermore, this study utilizes tests of CV from normalized PC scores extracted from 2D outline-based EFA. CV has already been a well-documented means of testing ceramic standardization within archaeological samples (see Roux, 2003; Wang & Marwick, 2020).

1.2 R and the MOMOCS Package

While many open source, standalone programs designed with the distinct purpose of extrapolating and conducting all 3 of the morphometric analysis exist within distinct spheres of academic study, these typically fall short in the number of available analysis useful in the study of craft goods. With the increased utilization of customizable, script-based programs such as R (R Core Team, 2020; v3.6.3) packages dedicated to all spheres of morphometric study have evolved. In the case of the outline-based methods, the MOMOCS package is the most holistic package available as of the time of this study. The MOMOCS package workflow begins with data extrapolation of outlines using EFA, coefficient normalization, and the application of these outline data to popular statistical analysis, such as PCA and cluster analysis (Bonhomme, Picq, Gaucherel, & Claude, 2014). In this study, coefficient normalization is conducted via the default ‘efourier’ MOMOCS method of numerical normalization of the coefficient matrix.

1.3 The Initial-Early Yayoi Period

The Yayoi agricultural transitional period (ca. 900/800 BC∼) is understood in traditional literature as the period of wide-scale socio-cultural changes brought on by the shift from hunter-gatherer to primarily rice agricultural societies in Japan (Yane, 1984); this shift was introduced through various waves of migration from the Korean peninsula to the Japanese archipelago, primarily into the Northern Kyushu region [北部九州地方], and into the Hakata Bay region (Sahara, 1975, 1981; Sugihara, 1961; Yane, 1984). However, in recent investigations into the role of wet-rice agriculture, evidence suggests that not only were Jomon hunter-gatherer groups practicing various forms of millet-based and low-yield rice cultivation (Nakayama, 2010), but also the differential methods of regional introduction of rice agriculture were extremely varied and did not always have stable retention or intensification (Fujio, 2021). Thus, while many texts may refer to the early Yayoi period as simply the beginning of rice agriculture, the complex nature of the acceptance and transmission of “agricultural society” during this transitional period should not be understated.

The preceding Jomon period [縄文時代] (ca. 14000–900/800 BC) is characterized by over 10,000 years of hunter-gatherer societal structures, complex ritualistic ceremonies, and a wide range of regionally and temporally varied material culture, including pottery (Hojo, 2019), with the initial/early Yayoi period seen as not only the roots of agricultural society but also the beginning of the mixture of Jomon and migrant cultural traditions (Hamada, 2018; Mizoguchi, 2013; Okamoto, 1998; Tanaka, 2011). As such, the start of the Yayoi period is delineated from the Jomon period not only through microscopic residue analysis of burnt rice grains on the surfaces of pots (Barnes, 2019; Miyamoto, 2018; Shitara, 2014) but also through the separate material cultural package that was introduced alongside rice cultivation (Tanaka, 2011). Pottery styles of the migrant population find its roots in the Korean peninsula Mumun [無文] tradition of pottery, characterized by undecorated, simple shapes, and a high degree of smoothing techniques utilizing a variety of wooden tools (Imamura, 2011; Kataoka, 1999). The stark differences between these two material cultural systems allow for interesting possibilities in tracing patterns of learning behavior, agent action, and communication systems (a case study on differential micro-styles resulting from these notions can be found in Loftus (2021)). As a result, how to differentiate between the material culture of migrants and the Jomon peoples has continued to be a heated point of contention among archaeological scholars for decades in previous literature (Miyamoto, 2018; Mizoguchi, 2019).

Miyamoto’s (2016, 2017) model of “dual cultural diffusion” theory is a recent mainstay in archaeological discourse within the initial/early Yayoi period. This model presumes, through a macroregional, ceramic chronological analysis that rice agriculture was initially introduced to the Japanese archipelago from approximately the 9th–8thc BC from the Namgang river basin to the Karatsu and Itoshima subregions within the Northern Kyushu region. Yuusu pottery, different in form and production from that of the previous Jomon tradition developed during this period. Following this initial wave, during what is known as the Yuusu 2 phase (7th–6thc BC), a secondary wave of migrants from a separate region in the Korean peninsula further influenced the pottery of the region, resulting in the development of the “Itazuke” style (5thc BC), which was centralized in the Hakata Bay region. Following the development and standardization of a recognizable “Yayoi culture,” of which the Itazuke style of pottery plays a central role, this culture, developed in the Hakata Bay region alone, then dispersed across western Japan. However, no statistical representations of standardization utilizing reproducible, statistical means have been put forth, and it is unclear whether pottery shape indeed decreased in variance during this period, especially within the important Hakata Bay region and what factors played a role in such a switch to standardized shape. Quantifying standardization of ceramics in this period would shed light on the correlation between increases in production intensity brought on by increased population density, and the means in which local populations standardize production to a common goal.

2.1 Materials

The chosen ceramic type is that of the “Tsubo” globular pot [壺形土器] (henceforth “tsubo”) (Figure 1). The tsubo shape was first introduced from the Korean peninsula late Mumun style of pottery (Yane, 1984). Approximately 85% of recovered vessels utilized in this study are found within mortuary contexts and are often the only remaining burial goods from this period in this region of initial contact, with few whole vessels being recovered outside of mortuary contexts, but a number of fragmented pieces found in domestic settings such as pit dwellings. As such, several scholars have attempted to use the tsubo pot as a primary vessel of study when attempting to understand socio-cultural changes from the indigenous Jomon to the “Yayoi” style of burials (Hashino, 2016, 2018; Nakazono, 2004).

Figure 1

An example group of early Yayoi ‘tsubo’ globular pots from the Shimotsukiguma-tenjinmori site. The extended neck with slight outward curve is the characteristic of this period.

Despite the tsubo being one of the smallest vessels in stature in the early Yayoi assemblage, with the average vessel height falling between 15 and 20 cm (Loftus, 2021); this small earthenware vessel holds an important position in the socio-cultural context of the period. This vessel type was one of the only “finer wares” of this period’s assemblage, having examples of delicately burnished and pigmented vessels as well. Due to specimens also being uncovered in domestic settings, scholars have theorized that the original purpose of the tsubo was that of a storage vessel for “non-liquid substances such as cereal grains” (Mizoguchi, 2013) and was transformed into the primary burial good of this period within the Hakata Bay region. As these vessels seemed to be more delicately produced than say, cooking vessels, it is assumed that while used in domestic contexts, the original final purpose of the vessel could be that of being a primary burial good. This burial practice was solidified in this region and was spread across western Japan in later stages, becoming common practice among other rice-cultivating groups across the peninsula. Mizoguchi (2013) argues that this pottery type was of particular central importance to the peoples of this period, representing both life and death as seen through the use of the pot, holding rice grains which guide the calendar and daily lives of agents, and yet being repurposed into the only grave good seen; in this way, it is possible to see how the cultivation of rice came to influence both life and death for the early Yayoi peoples. Being of such central important in this period and region, this pottery type is seemingly a useful candidate when considering the process of standardization during this period, being so directly and significantly connected to rice agriculture in symbolic and functional manners. It is hypothesized that such a central pot may have a more subdued and gradual standardization transition, if previous literature suggesting a non-violent and non-“invasive” form of mutual admixture between migrant and indigenous groups is indeed such (Nakamura, 2011).

The shape of these small vessels changes dramatically over the early Yayoi period, and within this temporal change are also apparent spatial differences eluding to changes in learning strategies (Loftus, 2021). These pots were created without the use of a pottery wheel, with certain examples showing remnants of woven mat and leaf imprints on the bottom of vessels, eluding to a practice of turning the vessel as it is formed and smoothed. According to Miyamoto (2016, p. 62), Yayoi vessels were created following four main attributes, which find origins in the late Mumun culture: “(1) [clay] slab-built using relatively wide clay slabs, (2) attaching the clay slabs onto the exterior rather than interior surface, (3) smoothing with a wooden edge, and (4) firing the pottery in a primitive clay kiln on the ground.” Finally, despite being a descendant of the late Mumun tradition, of which vessels are often left undecorated, a large number of excavated examples in this region are decorated with a variety of incised and painted patterns along the connecting region between the neck and body (Nishimoto, 2007). The relatively gradual change in morphology of this shape type during this period allows for a more accurately representation of potential standardization practices, as compared to other types that transformed into more complex shapes at a rapid pace.

2.2 Periodization

Initial/Early Yayoi period: ∼900/800 BC–300 BC (Hashino, 2018; Miyamoto, 2016). Unlike in many other parts of the world, the process of modern periodization in the archaeological record in Japan is done primarily through complex systems of typo-chronological dating of artifacts. As such, AMS dating is not a widely adopted practice, especially in smaller regional archaeological centers, this being despite this technology being available for over 70 years in Japan (Libby, 1951, 1954). Due to this, what constitutes the ‘Yayoi period’ is still a point of debate among many scholars (Hashino, 2018; Miyamoto, 2018; Yane, 1997). Traditionally, the ‘Yuusu’ type of pottery found in the Northern Kyushu region is seen as a general ‘starting point’ to the study of Yayoi pottery (Okazaki, 1971), and while few AMS dates have been provided by the National Museum of Japanese History (Harunari, Fujio, Imamura, & Sakamoto, 2003), initial results have been marred by much criticism over the years due to a lack of clarity in the charred material tested (Yoshida, 2005), as well as questions regarding archaeological context of tested samples (Fujio, Imamura, & Nishimoto, 2005). As such, while AMS dating is still expanding both in Japan and greater East Asia, typo-chronological control is still of the upmost importance for periodization in Japanese archaeology (Shoda, 2007). Despite these hurdles, many archaeologists have slowly agreed that a push-back starting date of the beginning of the Yayoi period is necessary, with a current general consensus being somewhere around the 9th–8thc BC (Shoda, 2010).

In terms of this study, due to the inclusion of the earliest examples of what is considered “Yayoi” pottery, the Yuusu I type [夜臼1式], the “starting point” of the utilized samples, is within this 9th–8thc BC range, as understood through direct AMS dating on pottery, as well as recent advances in cross-comparative typo-chronological investigations (Miyamoto, 2016, 2018; Nishimoto, 2006; Shoda, 2010). Due to a lack of a sizable amount of reliable AMS dates available for a holistic picture of changes in pottery shape during the transitional period, this study relies on semi-traditional shapes of chronology building to gain temporal control. This was done through a detailed comparison of pottery ‘attributes’ and results of a case study of the same samples, utilizing length-based morphometrics (Loftus, 2021).

A nine-stage chronology (I–IX) of pottery shape was extrapolated from these details, encompassing the entirety of previous chronologies (Yuusu 1–Itazuke 2b) (Figure 2) (I: c = 1, II: c = 5, III: c = 5, IV: c = 3, V: c = 25, VI: c = 30, VII: c = 10, VIII: c = 27, IX: c = 4). This 9-stage system varies slightly from previous works utilizing tsubo pots (such as Hashino, 2018), as the results of the aforementioned length-based morphometric analysis revealed certain previously unseen complexities in shape. To clarify the aforementioned hypothesis in a concise yet holistic manner, these nine stages have been consolidated into the three main morphological stages of the tsubo (as seen in the left column of Figure 2), as noted in previous literature, these being the Yuusu (stage 1; c = 14), Itazuke 1 (stage 2; c = 65), and Itazuke 2 (stage 3; c = 31) phases, respectively. Carbon dating of samples in this region are still few, but the results of Miyamoto (2018) are utilized here for clarity [stage 1: Yuusu (BC 900/800–500), stage 2: Itazuke 1 (BC 500–400), and stage 3: Itazuke 2 (BC 400–300)].

Figure 2

9 Stage chronology of initial/early Yayoi period tsubo pots. Nine stages are divided into three major morphological stages (modified from Loftus, 2021).

2.3 Region

The northern region of Kyushu island [九州地方] (henceforth “Northern Kyushu” [北部九州]), located in the south-west of the Japanese archipelago (Figure 3), is commonly understood as being the first point of interaction between the indigenous hunter-gatherers and the incoming migrant culture (Misaka, 2014). Within this region, the Hakata Bay [博多湾] subregion is understood as consisting of a larger number of settlements of which are often also larger in size as compared to other plains regions, including the moated settlements of Itazuke [板付] and Sasai [雀居].

Figure 3

Left: map illustrating the research area within greater East Asia. Right: the greater Kyushu island region and chosen region of study, the Hakata Bay subregion.

The Hakata Bay region is also the region that is believed to be the first emergence of the “Yayoi culture” (Miyamoto, 2016), consisting of a new package of material cultural goods and a stratifying society increasingly based around the cultivation of wet rice. This region is particularly suitable for wet-rice cultivation due in part to an abundance of tributaries running from the Sangun [三郡山地] and Sefuri [脊振山地] mountain ranges to the east and west, respectfully. These two mountain ranges create a funnel of small streams and larger river basins, which bring nutrient-rich minerals into the Fukuoka and Sawara plains [福岡平野・早良平野], and eventually out to the Hakata bay and Genkai sea [玄界灘] (Figure 4).

Figure 4

The Hakata Bay region, Kyushu Island, Japan, with overlaid sites utilized in this study (1–14).

2.4 Site and Sample Distribution

This study focuses primarily on 2D outlines of full vessels, which is a significantly smaller number than that of sherds, but due to the nature of the tsubo’s depositional context being that of primarily mortuary in nature, the general preservation is much better than that of daily-use styles of pottery during this period; as such, 14 sites yielding full vessels (c = 110) could be incorporated in this study (Figure 4, Table 1).

Temporally, samples tend to increase in number from the beginning of the migration period (Figure 5), similar in trend to population studies (Miyamoto, 2016; Usami, 2020), which assume increases in bay area plain regions as rice agriculture intensified. This aligns with non-traditional methods of population analysis in both Japan and the Korean peninsula, solidifying the likelihood of a migration event during the second stage of this period.

Figure 5

Distribution of tsubo vessels across each temporal stage (stage 1; c = 14, stage 2; c = 65, and stage 3; c = 31); dotted line represents the moving average over the entire period.

2.5 Statistical Methods

This study utilizes silhouetted 2D data of ceramics taken from excavation reports of this period in order to extract outline data using EFA. These outline data then have all size factors removed, so as not to skew results with non-shape data. Elliptical Fourier coefficients are tested utilizing a variety of statistical analysis to extract the variability of shape over time, and in turn, changes in standardization.

3.1 General Prerequisites

3.1.1 Calculation of Harmonic Power

Following the drawing of outlines and coefficient normalization (Figure 6), to holistically complete EFA in a way that captures the highest amount of shape information, it is necessary to calculate harmonic power within the given data set to avoid under/over sampling of harmonics (harmonic coefficients). The current data set of 110 pottery outlines shows 99% of harmonic power captured within the first 10 harmonics (Figure 7). As such, 10 harmonics are used in this study when conducting EFA in this study. A visual representation of reconstructed harmonics and the effect of the number of harmonics on capturing accurate shape data can be seen in Figure 8.

Figure 6

Left: 110 vessel outlines following coefficient normalization. Right: oscilloscope for EFA approach, outlining the curves around outline drawing through the “coo_oscillo” function.

Figure 7

Cumulative harmonic power calculated using the “calibrate_harmonicpower_efourier” function. 99% of harmonic power is captured in the first 10 harmonics (110 outlines) (6 = 90%, 7 = 95%, 10 = 99%).

Figure 8

Visual representation of differential numbers of harmonics (1–10) on captured vessel shape “calibrate_reconstructions_e-fourier” function. Lower harmonics results in less accuracy in capturing vessel morphology, 10 harmonics captures 99% of variance.

3.2 General Morphology

Shape variation along PC axes shows that PC1 and PC2 capture approximately 55% of total variance and combined with PC3 totals ∼73% of variance within the PC space. Each PC correlates with separate morphological features and is visualized in Figure 9. PC1 correlates with the degree of prominence of the elongated neck/lip region, which in previous studies is often referred to as the most important attribute related to temporal shifts in style within this region (Hashino, 2016, 2018; Nakazono, 2004). PC2, on the other hand, correlates highly with the outward flaring of the lip, combined with the width of the neck. PC3 is very obviously correlated with the width of the globular body shape of the vessels. In other words, the majority of vessel variance falls within the upper halves of vessels, especially within the lip/neck region. In terms of sample distribution within the morphospace, there exists a wide cloud of distribution across all quadrants (Figure 10).

Figure 9

Summary of shape variation along PC axes (PC1–3), utilizing the ‘PCcontrib’ function.

Figure 10

Left: PCA of all samples, overlaid with morphospace and 95% confidence ellipsis. Right: PCA of all samples, with overlaid morphology for each sample individually, utilizing the ‘plot_PCA’ function.

3.3 Temporal Transformations

Considering the results from general morphological analysis and MANOVA significance testing; PC scores extracted from the 110 elliptical Fourier outline samples utilized in this study have been understood as representative of unique and useful representations of vessel shape. As such, temporal categories (three stages) have been applied to these results to extrapolate the changes in vessel standardization over the agricultural transitional period. PCA results are able to accurately visualize a wide morphospace distribution in stage 1, which decreases significantly in morphological variance following a wave of migration and population density in the region (stage 2) and onwards (Figure 11, left). Furthermore, the centrality of distribution, as seen through overlaid density contours (Figure 11, right), shows an increase in central density, especially within the post-migration period (stage 3). In summary, the breadth of vessel morphological variability markedly decreased during the migration period and continued to centralize during the post-migration period.

Figure 11

PCA with overlaid morphospace with temporal categories (stages 1–3) overlaid in color. Left: 95% confidence ellipses overlaid; right: density contours overlaid visualize relative centers, utilizing the ‘plot_PCA’ function.

To further examine the degree of variable morphological standardization between the temporal stages beyond the subjective visualization of PCA, each of the three PCs utilized in this study (which are considered representations of unique morphological shape, as seen in Figure 9) was normalized (similar in approach to Wang & Marwick, 2020) for use in the calculation of CV scores (as expressed as a percentage of variance from the mean). CVs of all three PCs show decreases in the variability of vessel shape during the migration period (Figure 12), with an average increase in standardization (decreasing variability) of ∼23% between stages 1 and 2. Post-migration CV scores also show a continued path of standardized shape, with an average increase in standardization of ∼18% from stages 2 to 3. As CVs from all three PCs, which represent distinctly unique areas of vessel morphology, show increasing rates of standardization during the transitional period; it can be understood that the whole of shape of mortuary vessels was made toward a more intensively standardized whole shape following the wave of migration during the Itazuke I phase (stage 2), not just that of certain aspects of vessels. Furthermore, even following the migration period, and despite a decrease in sample size, vessel shape continued to standardize.

Figure 12

Normalized PC1–3 score distribution and overlaid CV scores across three stages.

As CV scores are normalized and tested within different categories than that of the preceding PCA analysis (within each PC rather than in temporal stages), to accurately assess if the outcome of CV analysis is statistically significant or not, this study utilizes the method of testing equality of CVs for all three sets of the normalized PC scores used in CV analysis (Table 3), as outlined in Wang and Marwick (2020) and discussed in detail in Krishnamoorthy and Lee (2014) and Marwick and Krishnamoorthy (2019). P values show that apart from CV comparisons between of PC1 S2–3 and PC3 S2–3, all comparisons (including all three categories as a whole) show statistical significance. With regard to the pre-migration to migration periods, all CV comparisons can be understood as statistically significant when tested using this method. In other words, the majority of intra-categorical comparisons (within PCs) further cement a decrease in vessel morphological variability and an increase in the standardization of production during this period, especially as population density increased during the migration period.

As the three major morphological stage system utilized in this study includes different numbers of minor stages within each (S1 = 4 stages, S2 = 3 stages, S3 = 2 stages), it may be understood that these results may be influenced heavily by this factor alone, instead of the underlying phenomenon presented. However, supplementary evidence when dividing stages evenly into minor stages of three each (S1 = 3 stages, S2 = 3 stages, S3 = 3 stages) (Supplementary Figure 1), overall results show similar degrees of decreasing variability utilizing CV scores (Supplementary Table 1). The only substantial difference in statistical significance is in that of PC3 (17.9% of variance), which shows no statistical significance between all comparisons. This may be due to the increased variability in the body shape of vessels in later stages due to the shifting of morphological stages downward. However, and as PC3 did not show decreased variance from S2–3 in the original data, this region of the pot may need to be separated and further tested in further studies. Despite this, overall results mirror that of those provided in the original chronology, showing an overall decrease in shape variability over the period.