Lapita on Wari Island: What's the Problem?

Abstract

The 2007 excavation of Kasasinabwana Shell Midden opened a new chapter on Lapita on the south coast of Papua New Guinea. We look to establish the degree to which the Kasasinabwana assemblage fits into the current understanding of Lapita colonisation by investigating modes of pottery production utilising physico-chemical analysis of the ceramics and patterns of obsidian exploitation using portable X-ray fluorescence (pXRF). Three separate ceramic Chemical Paste Compositional Reference Units (CPCRUs) are identified along with the presence of calcareous non-plastic inclusions in layers associated with possible colonisation phases. Obsidian is present from around 2000 b.p. and its appearance seems to correspond with the emergence of Early Papuan Pottery (EPP). The Lapita ceramic production model fits well with Late Lapita production.

Introduction

In 2007, a site with plainware pottery dating TO between 2800 and 2300 years ago was identified and excavated on Wari Island, Southeast Papua by archaeologist Yo Negishi (Negishi and Ono 2009). The site is called Kasasinabwana Shell Midden (PNG site code BALU). This was the first in situ finding of pottery dated to the third millennium b.p. along the southern coast of Papua New Guinea. Prior to this, the only securely dated pottery of comparable age was found in the Bismarck Archipelago and belonged to the Lapita cultural complex. The ceramics found in the lower layers of Kasasinabwana were named Kasasinabwana Plain Pottery (KPP) and noted as appearing exceedingly similar to the plainware Lapita pottery found further to the northeast in the Bismarck Archipelago.

At the time of the discovery of Kasasinabwana, the south coast of Papua was thought to have been colonised by pottery-producing people around 2200–2000 years ago (Summerhayes and Allen 2007:99). Kasasinabwana therefore extended the known sequence by between 300 to 800 years and provided the first evidence for Lapita colonisation of the southern parts of Papua New Guinea, either along the mainland coast or in the Massim region. The subsequent discovery of the slightly older Caution Bay sites added to the significance of the Lapita presence in southern Papua New Guinea, providing a new dimension to the story of Lapita colonisation of the Pacific (McNiven et al. 2011).

Although widely accepted by most, the finding by Negishi on Wari Island has been criticised. Irwin (2012:9) questions the integrity of Kasasinabwana, calling it "problematical." On the basis of dating the early assemblage and querying the shape of vessels, Irwin (2012:9) notes in reference to the term "Kasasinabwana Plain Pottery" that:

There is a magic in names. Once let a hatful of miserable fragments of fourth-rate pottery be dignified by a 'Name,' and there will follow inevitably the tendency of the name to become an entity, particularly in the mind of him who gives it. Go a step further and publish a description and the type embarks on an independent existence of its own. At that point the classification ceases to be a 'tool,' and the archaeologist becomes one.

The aim of this article is to look again at the Wari assemblage and present the results of a physico-chemical characterisation of ceramic samples in order to demonstrate that major changes in ceramic procurement and production occurred over time. In addition, a technological and chemical analysis of the obsidian artifacts are undertaken in order to ascertain the nature of their procurement and distribution. We will then address how "problematical" this site is.

Archaeology of wari island

Wari Island, also known as "Ware" or "Teste," is a solitary island located in the southwest of Milne Bay Province (Negishi and Ono 2009:26) (Fig. 1). The island has

Fig 1.

Location of Wari (Ware) Island.

an elongated shape, approximately 4 km long and 1 km wide (May and Tuckson 2000:84). It is a "basic volcanic" island, as are many of the small islands in its vicinity (Smith and Davies 1972:6). However, a lack of geological work completed on the island inhibits further clarification as to the specific identification of rock types.

Wari has a history of pottery production and the ceramics produced there are historically and regionally famous for being the most superior ware of Milne Bay (Negishi and Ono 2009:26). This is evidenced by the presence of several Wari vessels in every household, even on surrounding islands where other pottery was also produced (May and Tuckson 2000:85). Pottery is still being produced on Wari, mainly for trade, and the distribution of Wari vessels reaches as far north as Kiriwina in the Trobriands and Dobu Island in the D'Entrecasteaux Islands (Negishi and Ono 2009:44). Wari was also frequented by Mailu traders as recently as the 1980s (Negishi and Ono 2009:27).

Fig 2.

Location of Kasasinabwana excavation (after Negishi and Ono 2009:28, fig. 4).

Kasasinabwana Shell Midden

The Kasasinabwana Shell Midden site is located on the south coast of Wari Island. It is one of a complex of middens referred to as "Kasasinabwana village" by local residents (Negishi and Ono 2009:28) (Fig. 2). A sandy beach is only 12 m distant from the site. Information obtained from oral histories gives the impression that the village was inhabited until the recent historic period (Negishi and Ono 2009:28). Indeed, a clan leader explained that an old, weathered post in the locale had belonged to a chief's house that was sketched by Otto Finsch (1888:280) during his visit to Wari Island in the late nineteenth century. It is therefore unsurprising that this area of the south coast is described as being the oldest inhabited place on the island (Negishi and Ono 2009:28).

This area was selected for a test excavation due to the presence of shell, ceramic sherds, and obsidian flakes on the surface. The purpose of the investigation was to establish a ceramic chronology for the Massim (Negishi and Ono 2009). The excavation consisted of a 1 2 m test trench that spanned seven stratigraphic layers (Layers I, IIa, IIb, IIc, III, IV, V) (Negishi and Ono 2009:30, fig. 6). Negishi and Ono divided Layer II into three layers (IIa, IIb, IIc) based on colour. A sub trench of 0.5 2 m was opened in Layer III, but excavation ended at Layer Vas water began to well up halfway through the excavation (Negishi and Ono 2009:29).

Shellfish was the main component of the midden, with a total of 277 kg extracted from the trench (Negishi and Ono 2009:39). The vast majority of the shellfish were found in the most recent layer (Layer I), which was also the thickest layer and indicated intensive exploitation. Change over time in the sediments was noted by Negishi, with Layer IIa being dark soil that is markedly distinct from the sand-based sediments found beneath. Negishi also noted the stratigraphic integrity of the strata due to the presence of thin horizontal coral plates between Layers IIc to V.

Dating

Accelerator Mass Spectrometry (AMS) radiocarbon dating of three samples of Tridacna shell yielded dates of 2300–2600 cal. b.p. (Wk 25605) and 2600–2800 cal. b.p. (Wk 26604) for Layer V, and 2000–2300 cal. b.p. (Wk 25603) for Layer IIb (samples calibrated using OxCal) (see Negishi and Ono 2009:30, table 1 for further details). Chronologically, the earliest levels of the site (III, IV, V) can thus be assigned to the Late Lapita period (Summerhayes 2000a:129). Lapita settlements first appear at a number of sites in the Bismarck Archipelago dated to 3300 years ago (Summerhayes 2010). These Early Lapita sites were beach stilt-house settlements, containing locally made intricate dentate stamped pottery making up a third of all vessels (50 percent were plain water and cooking vessels, the rest have incisions and other designs), sharing similar stamped designs and vessel forms from sites separated by hundreds of kilometres, suggestive of high levels of interaction between settlements. Pot production patterns based on physico-chemical analysis suggest a highly mobile society (Summerhayes 2000b). Over time these settlements increased in number and distribution. By 3000–2900 years ago, settlements entered Remote Oceania and the south coast of Papua for the first time. Towards the middle of the third millennium b.p., the later Lapita settlements were more sedentary in nature and the dentate designs became more simplified and crude, eventually dropping out of the pot assemblage. The beginning of the Wari Island assemblage dates to this later Lapita period of time.

The Upper Layers (I, IIa) are estimated to date from the past 500 to 1000 years to recent levels (as evidenced by the presence of historic glass in the assemblage), while the Middle Layers (IIb, IIc) are thought to be from 1600 to 2300 cal. b.p.(Negishi and Ono 2009:48), which is the period after the KPP (later Lapita) and similar to the period characterised by Early Papuan Pottery (EPP). This placement of the Middle Layers is based on being sandwiched between two periods dated by radiocarbon and pot style similarities with what is termed EPP. "Early Papuan Pottery" was defined by Summerhayes and Allen (2007) to account for a regional ceramic sequence along the south Papuan Coast from ca. 2200 to about 1200 years ago. Although EPP developed out of the preceding Lapita period, its abrupt appearance along the south Papuan coast at a number of sites spread over 400 km has suggested to some that it could represent a colonisation event (Summerhayes and Allen 2007).

Ceramic and Lithic Assemblage

A total of 165 pottery sherds were excavated: 40 rims, 6 carinations, and 119 body sherds. The earliest levels of the site (V, IV, III) contained 49 sherds, the middle layers (IIc, IIb) 81 sherds, and the upper layers (IIa, I) 35 sherds (Table 1).

A total of 79 stone artifacts were recovered from the excavation (Negishi and Ono 2009:31, table 2). Of these, 63 (80 percent of the assemblage) were obsidian and three were historic glass. Other raw materials included schist, serpentine, shale, sandstone, chert, and chalcedony, but in very small quantities (Negishi and Ono 2009:38, table 3). It is important to note that the majority of obsidian was found in the top layer (Layer I), with only four obsidian artifacts found in the 1600–2300 cal. b.p. levels.

Table 1.

Rims and Body Sherds by Layer

Methods

Pottery

Fabric Analysis—Macro-fabric analysis was performed with each sherd examined under a binocular microscope at low magnification (20) to determine preliminary fabric groups. The identification of macro-fabrics was difficult due to the nature of the pottery surface, and indeed the macro groupings were completely superseded by the

Table 2.

Ceramic Samples for FEG SEM Analysis

results of the physico-chemical analysis. The physico-chemical analyses provided a more accurate representation of fabric groupings based on non-plastic inclusions. A representative sample based on initial visual identifications of fabric types resulted in just over a quarter of the ceramic assemblage being subjected to physico-chemical analysis using a field emission gun scanning electron microscope (FEG SEM) (Table 2).

Samples selected from each of the fabric groupings were mounted in epoxy briquettes and polished to 1 mm. Physico-chemical analysis utilised the Zeiss Sigma VP FEG SEM located at the Otago Centre for Electron Microscopy (OCEM) using an XMax20 silicon drift energy dispersive X-ray (EDX) detector and AZTEC acquisition and processing software. A single micrograph was taken at 100 magnification to record the fabric and another at 2000 magnification to examine the clay matrix. To record the non-plastic inclusions within the fabric, an area of the sample was map-scanned using the "TruMap" scanning function for 7 min with a rendering rate of 100. "Point and ID" analysis of the clay matrix was carried out at 2000. Five points were shot for 15 s each. The points were selected manually and the utmost effort was made to select points that contained only clay.

The geochemical data, consisting of levels of sodium (Na), magnesium (Mg), aluminium (Al), silicon (Si), phosphorous (P), potassium (K), calcium (Ca), titanium (Ti), and iron (Fe) collected from "TruMap" and "Point and ID," was transformed into oxide percentages (O%) prior to further analysis. As the TruMap function creates a layer-cake of the various components by grouping them by chemical similarity, only the layers containing temper were utilised for analysis as the clay data had already been collected. The grouping constraints were changed manually until it was deemed that all the non-plastic inclusions present in the image was accounted for in the data seen visually. (Elemental data for each separate temper type was identified with the aid of the tables and descriptions in Deer et al. 1985.)

Chemical Paste Compositional Reference Units (CPCRUs) were identified using principal component analysis (PCA) using MVARCH (Wright 1991). PCA is a multivariate statistical technique used to help identify groupings of chemically similar objects. It is basically an orthogonal transformation of the original variables into a new set of uncorrelated variables, constituting the principal components (Chatfield and Collins 1980:57). The major aim of PCA is to reduce the number of attributes to a few dimensions so that the data can be plotted and clusters identified visually. As noted in Summerhayes (2000b:39), the first component should account for the greatest amount of variation in the data, the second component the second greatest, and so on, with the components decreasing in order (Chatfield and Collins 1980:57).

Petrographic Analysis

Due to time constraints, the eight samples from Layer III were selected for thin section petrographic microscopy to be undertaken by James Scott at the Geology Department of the University of Otago. The samples were thin-sectioned and polished to a final thickness of 30 mm (0.03 mm) and analysed using a polarised light microscope.

Obsidian

Technological Analysis

An attribute analysis was carried out on all excavated obsidian and glass artifacts in order to provide insights into the nature of obsidian access and use at Kasasinabwana (Table 3). The maximum length, width, thickness, and weight of each artifact was measured in order to determine the overall quantity of obsidian present. Flake scars were counted for each artifact and the presence of retouch was noted. Intensity of retouch was measured by estimating the percentage of retouch compared to the total circumference of the artifact.

Chemical Analysis

Portable X-ray fluorescence (pXRF) analysis was carried out on all 63 obsidian artifacts in the XRF laboratory in the Department of Anthropology and Archaeology at the University of Otago using a Bruker Tracer III-V Portable EDXRF analyser. This instrument contains a rhodium tube from which X-rays are emitted and a Peltier-cooled, silicon PIN diode detector (Chen 2012:94). The X-ray tube was operated at 40 kV 30 mm using Bruker's green filter for 300 s. Elements analysed included Mn, Fe, Rb, Sr, Y, Zr, and Nb. Each sample was placed with the flattest surface over the machine window in order to optimise the data collected. A United States Geological Survey (USGS) standard BHVO2-Hawaiian Basalt was shot at the start and end of each session, and was used as every twelfth sample during a run to determine the precision and accuracy of the pXRF. The results of this, presented in Table 4, show a generally high level of precision overall for analysed elements. Results were calibrated to parts per million (ppm) in S1CalProcess using Bruker's OB40 obsidian factory calibration. The resulting data was compared to 37 reference samples from obsidian sources in Papua New Guinea (West New Britain, Admiralties, West Fergusson, and East Fergusson), which were all shot using the same settings on the same machine as described above.

Table 3.

Obsidian and Glass Artefact Assemblage

Table 4.

Error Ranges of Bruker Tracer III-SD during Analysis

Results: pottery

Vessel Forms Identified by Negishi and Ono

All but one of the sherds from the earliest layers (Layers V, IV, III, dated 2800–2300 cal. b.p.) are non-red slipped plainware bowls with everted rims, which are similar to Late Lapita pottery from the Arawe Islands, Southwest New Britain (Negishi and Ono 2009; Summerhayes 2000a, 2000b). These are the ceramics named "Kasasinabwana Plain Pottery" by Negishi and Ono. One sherd from Layer V is red slipped. Plainware of various forms with occasional red-slipping are found in the layers corresponding to EPP (Negishi and Ono 2009:34). The sherds from the upper recent layers (Layers IIa, I) have incised and applique decorations.

Results of FEG SEM Chemical Analysis of Clays

Three separate CPCRUs were identified on the first and second components and the first and third components (Fig. 3, Fig. 4). These correspond to three separate clay sources. Figure 3 shows the results on the first and second components, which made up 70 percent of variation. The first three components together make up 81 percent of variation. Here, Mg and P load heavily on the first component, while K and Ti load heavily on the second component; Mg and Ti also load heavily on the third component. CPCRU 1 is found throughout all layers of the site, while CPCRU 2 appears in all layers with the exception of the most recent. CPCRU 3 is represented by a single sherd from Layer V, the oldest layer.

Fig 3.

Principal components 1 and 2.

Fig 4.

Principal components 1 and 3.

FEG SEM Non-Plastic Inclusion Analysis

A range of tempers are used throughout the sequence at Kasasinabwana, including calcareous, Pyroxene, Light minerals (Quartz and Feldspar), Pyroxene/Light, Quartz/Amphibole, and Temper-free (Table 5). The most common temper type is Pyroxene, which dominates the assemblage from Layer III onwards. Other common fabrics include Light (Quartz/Feldspar) and a mixture of Pyroxenes and Light minerals. Other temper types are represented by a single sherd each: Quartz/Amphibole in Layer I, Ilmenite in Layer IIa, and no discernible inclusions (Temper-free) in Layer III.

It is important to note calcareous non-plastic inclusions are only used in the early layers of possible colonisation phases. It is used as the sole temper in one sherd from Layer V (earliest occupation) and in mixed Calcareous/Light fabrics in Layers IV (earliest occupation) and IIc (EPP) (Fig. 5).

Table 5.

Dominant Fabrics as Identified Through FEG SEM and Petrographic Analysis

Fig 5.

Distribution of non-plastic inclusions over time.

Results: obsidian

Sourcing Using pXRF

It is clear from Figure 6 that the obsidian is split into two distinct sources: West Fergusson (15 samples) and East Fergusson (48 samples). The East Fergusson pieces

Fig 6.

Bivariate analysis of obsidian artefacts and source material (× = artefacts).

cluster much tighter than those from West Fergusson, which shows a slightly larger amount of variation. This has been noted by previous authors as there appears to be different flows within this region (Ambrose 1976; Reepmeyer et al. 2016; White et al. 2006).

Of the pieces from the earlier layers of the site, the artifacts from Layers IIc (n =1) and IIb (n = 2) are sourced to West Fergusson, while the piece from Layer IIa is from East Fergusson. Of the 59 obsidian pieces from Layer 1, 46 were sourced to East Fergusson and 13 to West Fergusson.

Technology

Obsidian is not introduced at Kasasinabwana until Layer IIc, and even then remains in very small quantities until the most recent layer, Layer I (Table 3). The rapid increase of obsidian artifacts in Layer I can be attributed to Wari's interaction in the Kula ring. At the time of European contact, Wari Island and neighbouring Tubetube Island dominated trade in the region (Macintyre 1983:11).

The vast majority of obsidian artifacts are cores or core tools (cores exhibiting use or retouch) (Table 6). There is a lack of cortex, with only three artifacts, or 5 percent of the overall assemblage, still exhibiting cortex. These came from three different layers: Layers I, IIa, and IIb. Many of the artifacts have been heavily flaked or show signs of being repeatedly bashed prior to discard. A large portion of the cores are themselves made on flakes. While the majority of the assemblage has retouch, low levels appear to be more common, with 10–20 percent of the overall circumference being the most frequent amount of retouch. Altogether the technological evidence shows a high degree of reduction and use.

There is a good deal of variation in the size of the obsidian artifacts, although this is problematic due to small sample sizes in all layers except for Layer I (Fig. 7). The earliest obsidian in Layer IIc weighs 5.19 g. The average weight then falls to 1.55 g in Layer IIb, to 0.96 g in Layer IIa, and finally increases to 1.93 g in Layer I. In terms of size, it is interesting to contrast the Kasasinabwana assemblage to the EPP site of Mailu, where the initial colonisation phase is represented by a mean obsidian weight of 2 g (Irwin 1991:506; Irwin and Holdaway 1996:227), before dropping to a mean weight of approximately half a gram and remaining at this level for 2000 years. This indicates that

Table 6.

Technology of Obsidian/Glass Artefacts by Layer

Fig 7.

Dimensions of obsidian assemblage.

on the whole, the Kasasinabwana material is larger, which may reflect its closer proximity to Fergusson obsidian sources than Mailu.

Discussion

Late Lapita

The Early Lapita colonisation period is usually characterised by highly mobile communities, as shown by the complexity in ceramic production, where a wide variety of local and non-local clay sources and tempers were used to produce identical wares (Summerhayes 2000b:225–226; Summerhayes and Allen 2007:107). Later occupation is characterised by sedentary village life, as indicated by reduced production and local ceramic production (Summerhayes 2000a, 2000b). The KPP (Layers III, IV, V) ceramic assemblage is generally consistent with the Summerhayes (2000a) model for Late Lapita pottery production, expressing a sedentary, less mobile community.

The complexity of pot production strategies seen in Early Lapita is not witnessed at KPP. Only three CPCRUs are found in the earliest layers, while later layers have just two. The CPCRU unique to the Late Lapita layers, however, is made up of a single sample tempered with Pyroxene inclusions. Calcareous non-plastic inclusions are also found in this early Late Lapita layer. The only other layer where they are found is Layer IIC, found with Light inclusions. A decline in the use of this inclusion type has been noted elsewhere in the Pacific where it often drops out of use shortly after colonisation (Shaw 2014:233; Summerhayes 2000a, 2000b). Of particular note is the absence of obsidian in the Late Lapita layers indicating isolation from outside exchange networks. This absence makes Wari unique in the Lapita world, where obsidian is found, albeit in small quantities away from the source, and does suggest some communication with outside Lapita communication networks. It could be that what we are witnessing at Kasasinabwana is the terminal phase Lapita settlement. Earlier Lapita may be elsewhere on the island; future work would resolve this.

EPP

In Layer IIc, at the beginning of the middle layers dated to ca. 2300–1600 cal. b.p. that mark the beginning of the EPP (Negishi and Ono 2009:45), there is an increase in temper variety. Calcareous inclusions reappear in combination with Light inclusions in this layer and another new fabric combination, Pyroxene/Light, appears for the first time. Following this transformation, calcareous temper disappears and the ceramics maintain a relatively stable production model that continues to utilise two clay sources. This change in pottery production in the EPP could be a result of an increase in interaction along the south Papuan coast or indicative of a secondary colonisation event, as argued in Summerhayes and Allen (2007). In addition to this change, obsidian appears at the site for the first time and persists in very small quantities until the upper layers, indicating a break from Late Lapita isolation and the beginning of interactions between communities. More work is needed to cover this crucial time period for this part of Papua.

Recent Layers

Only one CPCRU is present in the uppermost layer (Layer I), indicating specialised pottery production from a single area. However, the number of non-plastic inclusion types does not correspondingly decrease but instead increases again, with the addition of a quartz/amphibole temper. This may be a result of exchange of temper as part of an increase in exchange, as this layer is thought to correspond to the Kula ring era in which Wari Island was involved in its role as a specialised pottery producing centre (Macintyre 1983:11). Wari was known as a producer of high quality pottery for the Kula ring (May and Tuckson 2000:85). Also of importance is its uniqueness compared with other pottery traditions from the Massim. Shaw and Dickinson (2017) argue that pottery from this study region contains no shell tempering, making Wari Island pottery very distinctive.

Absence of Obsidian

Obsidian has been modelled along the south coast as being present in significant quantities and in its largest size during a colonisation phase (Irwin 1991:506). This is seen not only in Early Lapita contexts within the Bismarck Archipelago, but also in the EPP along the south Papuan Coast (Allen et al. 2011; Summerhayes and Allen 2007). This is due to obsidian being viewed as a tangible link to the established homeland (Kirch 1988:113). The overall abundance of materials during a colonisation period is due to the high level of interaction between the colonists in the form of exchange networks (Irwin and Holdaway 1996:228; Kirch 1988:113; Summerhayes and Allen 2007:109). The absence of obsidian in the early layers of Kasasinabwana reinforces its attribution to the Late Lapita period. Its appearance from Layer IIC in the initial EPP fits well with both Irwin's (1991) and Kirch's (1988) models for obsidian distribution which explicitly state that the movement of obsidian is an integral part of the colonisation process (Irwin 1991:503; Irwin and Holdaway 1996:228; Kirch 1988:113–114).

That obsidian is present in the secondary EPP colonisation fits well with the data from Oposisi, which also suggest a colonisation phase (see below), even though it is also present in very limited numbers, consisting of a total of three pieces. Oposisi was excavated by Vanderwal (1973) in 1969; despite the large excavation, only two pieces of obsidian were recovered from the site. Seventeen pieces were subsequently excavated by Summerhayes and his team (Allen et al. 2011:69) in their re-dating of the site. However, a relative lack of obsidian at this period is also noted at the Middle Lapita sites at Caution Bay, where obsidian begins with the first settlement at Bogi 1 (McNiven et al. 2011:4; Mialanes et al. 2016:254). The obsidian from Caution Bay can only be described as minute in size. Thus, despite the significant transport of obsidian from the source areas in Bismarck Archipelago out into the Pacific with the initial Lapita colonisation, the same principle does not appear to apply to the colonisation of the Massim and the south Coast of Papua. It is obvious that something different is happening in this region at this time, but given that there is scant evidence for this time period, it is difficult to speculate about exactly what was taking place. As noted above, it may be that we are just excavating later terminal Lapita without having yet found an earlier Lapita occupation. There remains much more work to be done in this newly discovered period of Papuan history.

Conclusion

On the basis of dating the early assemblage and querying vessel shapes, let us return to Irwin's earlier statement on the integrity of Kasasinabwana. First, the chronology. Irwin argues that "the cultural and/or natural distribution association of the shells samples for dating are not clear" (Irwin 2012:9). We would argue the opposite. The shells came from secure stratigraphic contexts as outlined in Negishi and Ono (2009). Also the pottery production presented in this research sits well with the clear chronology provided by radiocarbon dates (which in themselves form an unambiguous chronology).

Second, the shape of pottery. Irwin notes that, although Negishi and Ono argued that sherds are from "everted rims and probable flat bottoms," he sees no indication of flat bottoms and "several of the sherds illustrated could be from vessels with direct rims, not everted ones" (Irwin 2012:9). Like Irwin, we see no evidence of flat bottoms—although Negishi and Ono (2009:34) did stipulate "probable" flat bottoms. Unlike Irwin, however, we have viewed the actual sherds and can vouch that they do come from everted rims, with the sherds terminating at fractures separating the outward top part of the everted rim from the body. Although acknowledging that what he calls "the so-called KPP sherds" could be found in Lapita assemblages, Irwin (2012:9) argues that many of the KPP sherds could be found in dated EPP sequences along the south Papuan Coast. He notes that similar everted forms were found at Mailu (where he undertook Ph.D. research nearly 50 years ago) and that similar forms were identified by Bulmer (1999) and Vanderwal (1973) "among the inventory of EPP forms" (Irwin 2012:9). This is true: everted plain globular vessels are found in EPP assemblages. Indeed, they are also found in many assemblages at all points in time within New Guinea archaeological assemblages and indeed many contemporary Papua New Guinean pottery assemblages. Yet their similarity to other pottery should not have been used to cast doubt on the age of these forms, since this form is also common in the Lapita assemblages from Adwe, Apalo, and Paligmete from the Arawe Islands (Summerhayes 2000b) and from Kamgot from the Anir Island group.

In summary, the pottery from Kasasinabwana excavated by Negishi demonstrates occupation along the south coast during the later Lapita period and into the EPP and more recent periods. It also draws attention to how archaeologists accept or reject findings based on other than empirical observations. On the basis of this review, we see nothing "problematical" in this site. We do see, however, that a lot more research needs to be undertaken in this important region.