CHAPTER 4

THREE LAPITA VILLAGES:

EXCAVATIONS AT TALEPAKEMALAI (ECA), ETAKOSARAI (ECB),

AND ETAPAKENGAROASA (EHB), ELOAUA AND EMANANUS ISLANDS
PATRICK V. KIRCH

From - 1500 sc until sometime in the
; early-first millennium Bc, three villages

stood on the shores of Eloaua and
?1     i    { 1 Emananus islands, each visible to the

k\  others. One, to which we have given
/X,   A i  the contemporary Mussau place-name

of Talepakemalai (and the formal PNG
site designation ECA), at times covered as
much as 82 ha or more and was made up of
numerous stilt or pole-houses built out over a shallow
reef flat. The other two, which we call Etakosarai (ECB)
and Etapakengaroasa (EHB), were much smaller-mere
hamlets really. These three settlements are among the
earliest and the most westerly of any Lapita sites known
in Near Oceania, and thus are of critical importance in
our emerging knowledge of the Lapita cultural com-
plex and its significance for Oceanic prehistory.

This chapter presents the detailed results of exca-
vations conducted at these sites from 1985 to 1988. My
emphasis here is on aspects of stratigraphy, depositional
sequence, spatial and stratigraphic distributions of ma-
terial culture, and chronology. Analyses of faunal and
floral remains recovered from these sites are reported

on in Volume II of this monograph series, while the
ceramics and portable artifacts-as well as an interpre-
tive overview of the sites-will be the subject of Vol-
ume III.

THE TALEPAKEMALAI SITE (ECA)

Talepakemalai has a distinctive history in the annals of
Lapita archaeology. When the LHP was conceived in
1983-84, ECA was the earliest and the most westerly
Lapita site known, making it of obvious importance to
the Project's objectives. The discovery-early during the
1985 field season-of extensive waterlogged deposits
with preserved stilt-structure posts was the first indica-
tion that Lapita settlements had included pole dwellings
situated over reef flats, raising new issues of survey
design and sampling. The 1985 excavations yielded an
unprecedented array of ceramics and non-ceramic ob-
jects, and the 1986 season added numerous well-pre-
served macrobotanical remains indicative of extensive
arboriculture (Kirch 1989). Talepakemalai, with an area
in excess of 82,000 m2, remains the largest Lapita site
on record (Kirch 1997:167), and its various deposits

THREE LAPITA VILLAGES

span a time range from the mid-second millennium until
the early-first millennium BC. Over this time span Lapita
pottery changed radically, so that the early assemblages,
rich in finely executed dentate-stamped vessels, might
hardly be seen as a part of the same tradition as the
later pottery, characterized by large, globular jars with
incised decorations and notched rims. Yet the ceramic
progression is particularly well documented at
Talepakemalai, making this site a key to the interpreta-
tion of the Lapita ceramic sequence in the Bismarck
Archipelago.

THE SETTING

Talepakemalai occupies the larger part of a coastal ter-
race of unconsolidated calcareous sands forming an
isthmus between the N and S upraised limestone blocks
of Eloaua Island (Fig. 4.1). This sandy terrain lies - 1-
2.36 m asl, with the higher ground situated away from
the present shoreline and closer to the edge of the S
upraised limestone block. The history of the site's dis-
covery is described below, and resulted from the con-
struction of a small-plane landing strip across the wid-
est part of this coastal terrace (see also Chapter 1). An
aerial view of this part of Eloaua Island is shown in
Figure 4.2.

The geomorphology of the unconsolidated sand
flats within which Talepakemalai is situated offers sig-
nificant clues to the depositional history and formation
of the site. Fortunately, the airstrip running across this
part of Eloaua Island provided an ideal transect from
lagoon shore to seaward shore, as shown in Figure 4.1.
Walking this transect in 1985, I recognized that a series
of alternating depressions and ridges along both shore-
lines were probable indicators of former beaches, evi-
dence for coastal progradation. lt also became appar-
ent that the central part of this transect, on which the
aircraft landing strip itself had been situated, was some-
what higher than the modern beach terraces.1 In order
to obtain precise elevation data on these topographic
features, I surveyed several elevation transects across
the coastal plain using a Leitz telescopic level and stadia
rod; selected transect profiles are shown in Figure 4.3.
As can be seen in the diagram, the modern beach ridges
on the seaward and lagoon shores stand at - 1 m asl.2
Inland of the modern beach terrace, however, one rises
up onto a slightly higher terrace, which stands between
1.8-2.35 m asl.

I realized early on during the 1985 season that this
somewhat higher terrace most likely represented a
paleobeach ridge that had formed in relation to a some-
what higher sea level than at present. My reasoning was
based on the assumption that unless there had been a
major change in local reef topography and/or coastal
wave energy regimes, the - 1 m differential between
high water and beach ridge should have pertained in the
past as well as at the present. Thus the - 2 m asl terrace
was likely to have formed at a time when the local sea
level was - + 1 m higher than at present. This hypothesis
fit well with what we knew about a + 1-1.5 m higher
sea level in the mid-Holocene, as described in Chapter
1. Moreover, as we began to plot the distribution of
surface pottery sherds over the cleared transect of the
airstrip, we discovered that these archaeological remains
were found almost exclusively on the higher, paleobeach
ridge and terrace.

This slightly higher, central part of the Talepakemalai
Site, through which the airplane landing strip was cut
and which corresponds to the area of surface pottery
and shell midden distribution, is covered with a degraded,
second-growth vegetation dominated by Pandanus
tectonius, Moinda citnfolia, andMacaranga sp., along with
low shrubs, ferns, and grasses. This degraded scrub con-
trasts markedly with the dense climax forest bordering
the site on the N and covering part of the waterlogged
zone (e.g., the N end of the W250 transect). World War
II aerial photos of Eloaua show the distinctly degraded
Pandanus scrub occupying the same area as today.3 This
second-growth vegetation pattern is the result of short-
fallow cultivation by the Eloaua villagers, who regularly
cut and burn root-crop gardens here (particularly of
Dioscorea yam and sweet potato, with some taro). They
regard the dark loamy soil found in this area as espe-
cially fertile, prized for gardening.4 In fact, the dark loam
is an anthropogenic soil, consisting of an extensively
reworked Lapita midden deposit, heavily enriched with
finely dispersed charcoal, bone, shell, and other organic
matter (see Chapter 2). How long the ECA Site has
been gardened in this manner cannot be determined,
although thorough chumning of the cultural deposit over
the higher portions of the site, and the small, fragmented
and worn potsherds suggest that such gardening on the
site has a long antiquity. Utilization of anthropogenic
midden soils for root-crop cultivation is not unique to
Eloaua, and is a regular occurrence throughout the SW
Pacific.5 The ECA situation is paralleled, for example,

69

THREE LAPITA VILLAGES

8

C,

u)

0

(0

E

D3
?)
0(0U

(n 0

?"-Z

1'

0     *

5<- -n

4L).

~)t 0

c:
00

c c c
0.0

<00-

7rc

70

U)
3

c -O

m r-
c: (U
m -rA

E
w

Z,

"qnq

THREE LAPITA VILLAGES

FIGURE 4.2 Aerial view of part of Eloaua Island, showing the small plane runway which cuts across the ECA Site,
and was responsible for its initial discovery in 1973.

by the Lapita occupation zone on Niuatoputapu Island
in Tonga (Kirch 1988a:38-41) where an identical an-
thropogenic midden soil is regarded by the island's in-
habitants as the best edaphic environment for the culti-
vation of Dioscorea esculenta yams.

Being prime agricultural land, the ECA Site area is
subdivided into a number of discrete, named land sec-
tions, held by different clans. Five land section names
were given to me by Ave Male: Talevungateo (7ok Pisin
translation, 'placeem i gatwada'), Etanamakapa ('olgeta
kaikaipinis'), Talepongakosa (?liklik hill longgraun'),
Taleliubua (no translation obtained), and Talepakemalai
(?olestapundadispeladiwainam belong em mala?; the malai
tree is the vi-apple, Spondias dulcis). The Talepakemalai
tract, belonging to Male's clan, includes that part of the
ECA Site where we focused our excavations (Areas A,
B, and C). We have therefore appropriated this garden
toponym as the name for the site, although strictly speak-

ing it applies only to the N-central portion of the entire
ECA Site area.

SITE DISCOVERYAND PRIOR EXCAVATIONS

As is sometimes the case in archaeology, the discovery
of Talepakemalai resulted from a succession of seren-
dipitous incidents. In 1973, the S.D.A. mission com-
menced clearing of a small aircraft-landing strip on
Eloaua Island to allow-for the first time-access to
Mussau by air, in this case by the mission's single-engine
plane. A small gas-powered tractor equipped with a
rear-mounted grader blade was shipped to the island
to assist the islanders in leveling the ground for the air-
strip. After clearing the scrub brush along a 3040 m
wide strip across the island, and following minor grad-
ing and leveling with the tractor, the airfield surface was
paved with a layer of crushed coral limestone obtained

71

8

w

D2
Z0

8     8     8

w     w     w     u

E

0

O W) O                    E S               /              L O                        7DO  0  z
cr X /)L u)                                f) o c oofc          U0

Mt g . C > ~~~~~~~~Mid-Holocene calcareous sands'"\i: \

r     La~te Holocen                                                           HoloceoS<S0a U?0 t 00<t,

prograded sands           a<z<)     a      nf                        rgae     ad

J--ZL~~~~~~~~~~~~e- .7' ,GhA, -,t f~;?" +Z .                          -Ryben-Herzberg Aqui f r_ *  -E A,
(b)                 ~~~~~~Mean sea level

s

2 m.

1 -     Vertical scale x 100

0-        1

0       10b0    200     300m

:2
8   D

Uf)0

8     8      8

z     z      z

N

O

?)  E o

L C 0

O)

;  ELL  U

?:     Late Holocene

~~prograded sands
ben-Herzberg Aquifer .. ' .

FIGURE 4.3 Elevation profiles across the ECA Site, showing geomorphic features associated with the mid-Holocene
higher sea level. (A) E-W profile across Eloaua Island; (B) N-S profile through the central part of the ECA Site.

from a small quarry at the foot of the first uplifted
terrace escarpment.6 During the clearing and leveling
operations, "broken pots and some bones were un-
covered and noticed by villagers working on the
project" (Egloff 1975:15).

At the same time that the SDA airstrip was under
construction, Brian Egloff of the PNG National Mu-
seum and Art Gallery made a trip to Mussau "as an
ethnographer to collect weaving, but on the first day
[Site] ECA was located and I tried to define the site
before the medical boat arrived to pick me up" (Egloff,
pers. comm., 1985). According to the published report
Deku Malua, a teacher at the Malakut Primary School,
brought Egloff a large pot sherd decorated in fine den-
tate-stamping of classic Lapita style, tipping him off to
the presence of the site (the sherd is illustrated in Egloff
1975, fig. 14a, and also in Bafmatuk et al. 1980:80).
Egloff immediately recognized the significance of a
primary Lapita site in the Bismarck Archipelago, at a
time when only the Watom and Ambitle sites were yet
known, and he conducted several trial excavations in

three areas adjacent to the air strip (see Egloff [1975,
map 4] for excavation locations). These tests yielded
pottery (much of it with Lapita-style decoration), some
obsidian flakes, and a small sample of vertebrate faunal
material. These finds were soon reported in a short
monograph (Egloff 1975). In this report, the ECA Site
was summarized as follows:

Site ECA proved to be an extensive midden with

considerable potential. Six one by two metre units were
excavated with cultural debris being generally encoun-
tered no deeper than circa 30 cm below the surface of

the ground. However, one specific area of the site, that
being where units AA, AB, and B are located is

particularly rich in cultural debris and appears to have

undisturbed midden lying in a shallow depression in the
surface of the coral sand subsoil (Egloff 1975:15).

The ceramics were reported in detail, and Egloff
noted the presence of "pedestalled vessels," remarking
that similar pedestalled pottery had recently been re-
covered from archaeological sites at Collingwood Bay
in eastern Papua (1975:29-30). Unfortunately, "material

suitable for radiocarbon dating was not recovered,"7

THREE LAPITA VILLAGES

72

THREE LAPITA VILLAGES

and therefore no direct assessment of the site's age was
possible.

In 1978, Egloff organized a second trip to Eloaua
accompanied by two staff members of the PNG Na-
tional Museum (Francis Bafmatuk and Resonga Kaiku).
Kaiku was enrolled in an undergraduate degree pro-
gram in archaeology at the University of PNG, and
?the second trip was basically designed to provide
Resonga with sufficient data upon which to construct a
thesis" (Egloff, pers. comm., 1985).8 The team ?con-
ducted a test excavation" of unspecified dimensions on
the N side of the landing strip in a garden belonging to
Ave Male (re-examination of this area in 1986 revealed
that the 1978 excavation extended - 9 x 2 m9). The only
published report of this National Museum excavation
is a popular article by Bafmatuk et al. (1980). This article
reports the first two radiocarbon dates for the ECA
Site, both run on samples from a single "coral oven"'0:
3030 + 180 BP (GX-5498) and 3900 + 260 BP (GX-
5499). The discrepancy between these two dates did
not go unremarked by the excavators, who opined that
the uppermost sample could have been contaminated
by charcoal from recent gardening (Bafmatuk et al.
1980:80). They believed that the lower, and older, sample
had produced a correct age determination: "The low-
est sample, dated at close to four thousand years ago, is
considerably older than the dates obtained from Lapita
sites in Eastern Melanesia and certainly ranks as one of
the oldest Lapita-associated dates in Melanesia"
(1980:80). The early GX-5499 date continued to be cited
for the next several years (e.g., Allen 1984:188; Anson
1986:162) as evidence for initial Lapita occupation in
the Bismarck Archipelago - 4,000 BP. Anson (1986) em-
phasized the older ECA date of 3900 BP in his argu-
ment for a distinctive ?Far Western" Lapita style ante-
dating the earliest Lapita to the east which was ?present
in the Bismarck Archipelago for several centuries be-
fore moving eastward" (1986:162).1i However, as our
excavations and extensive radiocarbon dating of mate-
rials from ECA were to demonstrate, the 3900 BP date
is unquestionably too old to be an accurate age for cul-
tural occupation at Talepakemalai and must be rejected.

In sum, at the commencement of our 1985 field
season, two prior test excavations had been carried out
at ECA, with only the initial 1974 results available in
published form. While the ECA midden deposits were
known to be extensive, lying on both sides of the land-
ing strip, the boundaries of the cultural deposits and

therefore the areal extent of the site had not been de-
termined. The range of ceramic variation could be partly
ascertained from Egloff's 1975 monograph, but the
faunal sample (especially invertebrate remains) was in-
adequate to interpret the site's economic prehistory, and
non-ceramic artifacts were extremely limited. The age
of the site was suggested by the two radiocarbon dates
cited above, which as they derived from the same puta-
tive "coral oven" yet had ages 870 years apart, hardly
inspired confidence in an interpretation of a 4,000-year
old occupation. Four problems thus defined the initial
research objectives for our 1985 season at ECA: (1) to
determine the areal limits of the site; (2) to refine the
stratigraphic interpretation; (3) to ascertain the chronol-
ogy of occupation; and (4) to obtain and analyze an
expanded sample of ceramics, material culture, and fau-
nal-floral materials.

EXCAVATION OB/ECTIVES AND STPATIGPAPHY

As the work at Talepakemalai proceeded over the course
of three field seasons, our research objectives evolved
considerably, requiring changes in excavation strategy
and in specific procedures. At the start of the 1985 sea-
son, we had no idea that the site would prove to be so
extensive or would contain a waterlogged component
yielding abundant organic materials as well as ceramic,
lithic, and shell artifacts, all necessitating special excava-
tion and recovery techniques. During the 1986 and 1988
seasons special efforts were made to address these con-
ditions, and particular research questions were also out-
lined (see Chapter 1). The following summaries of the
three seasons of fieldwork at Talepakemalai provide an
overview of the changing research objectives and exca-
vation strategies applied during that period.

THE 1985 SEASON

Initial objectives at ECA were defined on the basis of
the prior PNG National Museum test excavations. Ar-
riving on Eloaua in early August 1985, and following a
reconnaissance of the ECA Site with Ave Male as my
guide, 1 decided that the highest priority was to define
systematically the boundaries of the site's cultural de-
posits, including the construction of an accurate site map.
For this purpose, a series of intersecting ytenatic transa

with test units positioned at regular intervals was demed
to be the most appropriate excavation strategy. I hW4

73

THREE LAPITA VILLAGES

previously applied a transect approach in defining large
open midden sites in both Tikopia and Niuatoputapu
(Kirch and Yen 1982; Kirch 1988a), and these provided
a methodological precedent. The systematic transect
units were expected to yield a representative sample of
ceramics and other cultural remains, thus meeting our
objectives of defining the site's cultural content.

A datum was established - 20 m S of the landing
strip near its E end (see Fig. 4.1), on the highest part of
the paleobeach terrace, and a coordinate grid system
was staked out running out from this datum in four
cardinal directions. This main site datum has an eleva-
tion of 2.36 m asl. A primary E-W transect was laid out
along the NO line, with test excavation units positioned
every 100 m parallel to the landing strip, beginning at

E1O0N0 and extending to W500N0.12 Excavation of

these units rapidly revealed that the cultural deposits
through this central part of the site were uniformly shal-
low, generally consisting of between 15-25 cm of dark
loam containing highly fragmented and worn sherds
along with some bone and shell midden. These depos-
its had been extensively reworked through intensive gar-
dening; only in one unit (W1O0NO) did two pits or pos-
sible postmolds extend below the reworked garden soil
into the basal calcareous sands. The initial NO transect
units also served to define the E and W boundaries of
the site, and demonstrated that these coincided with low,
swampy depressions marking former shorelines.

The second phase of transect testing commenced
along perpendicular N-S transect lines, at W100, W200,
and W400. At the same time, intensive systematic sur-
face walking and recording the presence of sherds, shell
midden, fire-cracked rock, and other cultural debris also
helped to define precisely the areal extent of the site.
Rather than several concentrations of midden as de-
picted in Egloff's map (1975, map 4), we determined
that cultural materials were distributed more-or-less

continuously over an area of at least 72,500 m2 (the area

within the dashed line labeled ?limits of surface pot-
tery" on Fig. 4.1). lt was also evident that this area of
surface distribution of cultural materials corresponded
to a former beach terrace, - 0.8-1 m higher than those
portions of the coastal flat bounding the site on the E,
N, andW.

Just as these surface survey and transect operations
were beginning to suggest that defining the site's bound-
aries would be a straight-forward matter-and that the

cultural deposits might prove to be uniformly thin and

heavily disturbed by gardening activities-the situation
became complicated in Unit W200N100 (originally des-
ignated TP-13). This square was positioned just 7 m
east of the trench excavated in 1978 by Bafmatuk,
Egloff, and Kaiku (1980), and we therefore fully ex-
pected our new excavation to reveal the same shallow,
single-component stratigraphy reported by them. 13 When
proceeding to excavate below the dark loamy garden
soil, we were therefore surprised to discover pottery,
obsidian, bone, and shell midden extending well into
the white, calcareous beach sands down to 100 cm, where
the deposit became strongly cemented, prohibiting fur-
ther hand excavation. The large sherds in the lower white
sand layer of Unit W200N100 lacked the broken and
worn sherd edges typical of those from the upper gar-
den soil. These fmds contradicted the simple stratigraphic
interpretation of the site suggested by Egloff (1975),
and by Bafmatuk et al. (1980), at least for this N por-
tion of the site, and made further testing essential.

Another transect unit was opened up 50 m farther
N along the W200 line (W200N150, originally desig-
nated Test Unit 14), even though this location lay be-
yond the putative former shoreline, and despite the scar-
city of cultural materials on the ground surface. (Based
on the surface survey and geomorphology, the
W200N150 location was initially judged to lie beyond
the site's periphery.) While the upper 65-70 cm of this
unit yielded only limited cultural material, at - 70 cm bs
we encountered a concentrated deposit of large, un-
weathered sherds, nearly half of them carrying classic
dentate-stamped Lapita decoration. At this depth the
sandy deposit became waterlogged, and it was clear
that we were at the tidally-fluctuating contact with the
brackish-water Ghyben-Herzberg aquifer. Directly as-
sociated with the thick deposit of sherds were abun-
dant large obsidian flakes, two whole Conus-shell rings,
a small Conus-shell bead, and a perforated pig-tusk pen-
dant. Our collective fieldnotes and my personal journal
for this day reveal the dual sense of exhilaration and
confusion created by the emerging finds in the
W200N150 pit. Testing what we thought would be the
N edge of the site, we had unexpectedly encountered
an extremely rich, undisturbed Lapita deposit, inundated
by the island's freshwater lens.

At this point I temporarily halted the systematic
transect excavation strategy, and chose two locations
for expanded areal excavations. Area A consisted of a
2 by 3 m block defined by Units W228-229/N100-

74

THREE LAPITA VILLAGES

102, 20 m W of the 1978 PNG National Museum
excavation (and 28 m W of our TP-13). Situated on the
higher beach terrace (- 1.90 m asl) just above what we
interpreted as the former shoreline (represented by a
N-dropping slope of - 0.8 m), Area A was anticipated
to reveal the stratigraphy of this elevated terrace, and to
confirm whether cultural materials extended apprecia-
bly below the disturbed dark loamy garden soil (Fig.
4.4). Indeed, the lower white sands yielded consider-
able plain pottery, dominated by large jars with restricted
orifices and flaring rims. However, only three sherds
had dentate-stamped designs.

A second 6 m2 areal block was laid out to incorpo-
rate the W200N150 (rest Unit 14) pit which had pro-
duced the striking Lapita finds in waterlogged deposits
below 70 cm depth. This block-designatedArea B-
was defined by Units W200-201 and N149-151.14 Strip-
ping off the dry, upper deposits (the dark garden soil,
followed by a gray sandy deposit), we found that cul-
tural materials were infrequent, especially in the grey
sandy deposit as one approached the water table at  60-
70 cm bs. Just as the deposit turned white and became
increasingly damp and sodden, a thick concentration of
Lapita sherds appeared along with obsidian, bone, and
shell midden. Unequipped for a proper ?wet site" ex-
cavation, we faced several technical problems. First was
the matter of removing the brackish water which seeped
continually into the pits. Lacking pumps, we resolved
this problem by always keeping one unit lower than the
others, and using buckets to bail out the sump. As we
proceeded deeper, the Area B excavation would flood
completely overnight, requiring an hour or more of
heavy bailing each morning. The water table fluctuated
tidally, and we began to use a tide table to plan work
around predictable periods of low tides and hence lower
water tables in the excavation. Gradually, we were able
to extend the Area B excavation well into the water-
logged deposit.

A second problem was that the thoroughly satu-
rated sherds were extremely soft and fragile when first
exposed, although they became relatively hard after dry-
ing. Experiments with different drying conditions re-
vealed that rapid sun-drying was detrimental, sometimes
resulting in severe shrinkage, warping, and cracking.
Therefore, as sherds were removed from the water-
logged sands, they had to be carefully laid out on trays
(we used damaged sieving screens for this purpose) and
kept in cool, shady conditions to dry out slowly over

several days. The floor of our field house became an
obstacle course of hundreds of sherds drying in the
most protected situation we could afford them.15

We were now excavating in a completely undis-
turbed deposit, for not only were the sherds, artifacts,
and faunal materials in an excellent state of preserva-
tion, but we encountered matching sherds making up
large portions of vessels lying intact just as they had
been discarded and broken (ig. 4.5). Taking advantage
of several days of low tides, we approached the base
of the thick concentration of Lapita materials, and ex-
posed a deposit of coral reef rubble which appeared
to represent the pre-cultural surface. When we had first
broached the waterlogged deposit in the W200N150
(TP-14) test, the highly decomposed top of a vertical
wooden shaft or post had appeared. At first we thought
this might be the base of a rotted tree root, but expos-
ing more of this object as we proceeded into the wa-
terlogged zone, the wood preservation became increas-
ingly good. On August 28, taking advantage of a par-
ticularly low (+ 0.7 m) tide, we brought the Area B ex-
cavation to its base, and removed the object, which
proved to be the worked base of a wooden post which
had been set upright into the coral reef rubble. Adz
marks were visible on the tapered point of the Post,
which was - 65 cm long. There was little doubt that this
post was directly associated with the Lapita materials
(as radiocarbon dating would subsequently confirm),
and the realization that we had a site with intact Lapita
wooden architecture occasioned excitement both among
the archaeologists and our team of Eloaua workers.16
Several other smaller wooden stakes appeared in Units
W200N151 and W201N151 as we cleared the deposits
down to the coral rubble layer.

At this juncture-despite the exciting nature of the
Area B finds-it was essential to call an intermission in
the ECA Site work, in order to test two other Lapita
sites: ECB on Eloaua and EHB on Emananus Island.
Neither of these sites proved to contain a waterlogged
component as at ECA, and after sufficient transect tests
were completed, I decided to conclude the 1985 sea-
son with a further expansion of the Area B excavation
at ECA. We therefore gridded out another 6 m2 area
adjacent to the existing block, comprising Units W198-
199/N149-151. This excavation expanded and clari-
fied our understanding of the Area B stratigraphy, and
yielded additional well-preserved wooden posts, form-
ing the corner of a structure. Since we had determined

75

THREE LAPITA VILLAGES

FIGURE 4.4 The completed Area A excavation, during the 1985 field season.

that the Lapita materials had been deposited sub-tidally
and had not been disturbed, it appeared that the posts
were the supports for a stilt- or pole-house that had
been elevated above the reef flat at a time when the
island's shoreline was situated - 30-40 m S of Area B.
After completing this 6 m2 expansion to Area B, the
base of the excavation-with wooden posts left in
situ-was protected by laying down a sheet of thick
plastic, followed by back-filling.

THE 1986 SEASON

On 1 September 1986 we recommenced the
Talepakemalai Site excavations. I intended to expand
the Area B areal excavation block by at least double to
expose more of the stilt-house architecture and to in-
crease the sample of associated cultural materials. Also,
it was critical to refine our model of the site's geomor-
phology and depositional history, in regard to former

shorelines and to the progradation of the coastal flat
that buried and preserved the Area B wooden architec-
ture. The systematic transect approach was again the
most appropriate, this time utilizing close-spaced
transect units extending outwards from the Area B block
in order to fit Area B into a micro-stratigraphic and
geomorphological context.

Having learned from the difficult experience of
excavating the waterlogged Area B deposits without
adequate equipment in 1985, we arrived on Eloaua in
1986 armed with the tools to allow us to tackle the
site's unique conditions. I had purchased in the United
States two sizes of manually-operated diaphragm-type
pumps intended for use as bilge-pumps on small ma-
rine craft.'7 These not only allowed us to continually
remove water from the pits, but the outflow could be
used to wet-sieve the excavated sediment, greatly en-
hancing the recovery of cultural materials. Without wet-
sieving, the water-saturated sand could be passed

76

THREE LAPITA VILLAGES

through 5 mm mesh only with great difficulty; the use
of the water-outflow from the pumps allowed us to
screen materials through 3 mm mesh, and to recover
fme bone and floral materials without significant break-
age. We were also equipped with customized, preprinted
field recording forms (see Chapter 3), better surveying
and photographic gear, and-very importantly-an
abundant supply of waterproof labels, plastic bags, and
cotton wool for packing and protecting the fragile wa-
ter-logged pottery.

With Terry Hunt and Marshall Weisler as excava-
tion supervisors, I engaged a team of 13 to 15 Eloaua
islanders to assist in the excavations.'8 The back-filled
12 m2 Area B block from 1985 was re-opened, and the
plastic sheeting which had protected the floor of the
excavation was removed. A new block of six squares
adjoining the 1985 excavation on the E side was gridded
out (Fig. 4.6), comprising Units W195-197/N150-151.
This was matched several days later by a S expansion
(Fig. 4.7), comprising Units W198-200/N147-148. In

total, we exposed 24 m2 at Area B, within which we
defined two alignments of anaerobically preserved
house posts meeting at what appeared to have been a
reinforced corner. Initially, I had hoped to trace these
alignments to additional comers, and eventually expose
an entire stilt-house outline. This goal was thwarted by
the shear volume of cultural materials recovered, and
by the massive amounts of time and labor necessary to
wet sieve, recover, and process them. I decided that it
was a higher priority to conscientiously recover all ver-
tebrate and invertebrate faunal materials retained in the
5 and 3 mm mesh sieves, as well as 100 percent of the
ceramic assemblage, and than to merely sample the cul-
tural materials in the interests of exposing more wooden
architecture. An additional consideration was that as the
excavation grew in area, the volume of water to be
pumped increased proportionally. As the full 24 m2
reached the sub-water table level, the task of pumping
out the excavation each morning (the pits would natu-
rally fill with water during the night) became more time

FIGURE 4.5 A finely decorated, flat-bottomed Lapita dish, as exposed in situ in the 1985 Area B excavation.

77

THREE LAPITA VILLAGES

consuming, to a point where the entire 15 member team
was engaged in pumping and bailing for as much as
two to three hours before excavation could commence
(Fig. 4.8).

Simultaneously with the expanded Area B excava-
tions, we laid out a series of transect test units along the
W200 grid line which had been tested by the W200N100
(FP-13) and W200N150 (F1P-14) pits in 1985. Units were
laid out at 10 m intervals along the W200 line, begin-
ning at W200N1 10 and ending at W200N200 (see Fig.
4.1). A second series of four units were laid out along a
perpendicular axis following the N140 grid line, at 20-
m intervals. These two transects were positioned to re-
veal a detailed picture of the stratigraphic context sur-
rounding Area B over a distance of 100 m N-S and 80
m E-W. The N-S transect along the W200 line further-
more extended across a major geomorphological tran-
sition: a drop of - 0.8 m in elevation (from 2.01 asl at
W200N100 to 1.20 asl in the vicinity of Area B), which
we interpreted as the approximate location of the
former shoreline of Eloaua Island.

These transect excavations verified that the W200
transect indeed crossed a former shoreline feature, so
that at the time of the mid-Holocene higher sea-level

stand the Area B stilt-house had once stood on a reef
flat over open water. Unexpectedly, the W200 transect
pits lying between Area B and Unit W200N100 exposed
a deeply buried zone of fine, silty textured calcareous
sediment that contained a high quantity of anaerobically
preserved organic materials, especially wood fragments,
coconut husk and endocarp, and remarkably, a variety
of seeds, nut shells, and other floral remains. We came
to refer to this fine-textured deposit as the "muck zone."
As along the foreshore reef flat environment fronting
Eloaua Village today, there was in Lapita times a low-
energy tidal flat lying between the stilt-house and the
former beach slope, which had trapped organic refuse
discarded by the site's inhabitants. Here was another
unanticipated payoff of the waterlogged condition of
this Lapita site, for the abundant plant remains con-
sisted of a wide range of economic taxa, such as Tahi-
tian chestnut (Inocarpusfagiferus), vi apple (Spondiasdulcis),
Canarjum almond, and others (see Volume II). For the
first time, a Lapita site had yielded an abundance of
direct, macrofloral materials of ethnobotanical signifi-
cance.19 We closed the 1986 excavations at ECA on 25
September, although two more weeks of intensive work
were required to complete the processing of shell

FIGURE 4.6 View of the Area B excavation during the 1986 field season, with the E expansion being excavated
through Zone B. Note the use of pumps to keep the water table low, and to provide water for wet sieving.

78

THREE LAPITA VILLAGES

FIGURE 4.7 Excavating in Zone C of the S expansion of Area B, during the 1986 field season. Several wooden
posts can be seen emerging from the Zone C sediments.

midden, and to prepare all other materials for shipment
to Seattle.

THE 1988 SEASON

The third season at Talepakemalai began on 22 Sep-
tember 1988, with Dana Lepofsky andJason Tyler act-
ing as site supervisors. We were again ably assisted by a
crew of 10 to 15 Eloaua men and women. After re-
cutting the now overgrown W200 transect line and re-
locating our Area B datum, we cleared a transect run-
ning N-S along the W250 line, 50 m W of our 1986
transect and parallel to it. This new W250 transect started
on top of the former beach ridge and cut across the
former shoreline feature, out over the buried reef flat.

As with the 1986 transect, test units were laid out at 10
m intervals, commencing at N70 and running continu-
ously to N200, a total of 14 units.20 We expected this
transect to provide a second stratigraphic and geomor-
phological sequence to assist in the interpretation of the
site's depositional history, to reveal the extent of stilt-
house structures W and N of Area B, to further sample
the buried foreshore zone of anaerobically preserved
plant remains (the "muck zone"), and to provide samples
of shellfish remains for our shell technology and inver-
tebrate exploitation studies. In all these respects, the
W250 transect served admirably. As in 1986, hand-op-
erated bilge pumps were used to keep the pits (often
reaching depths of 1.8 m) from flooding, and to pro-
vide water for wet-screening.

79

THREE LAPITA VILLAGES

FIGURE 4.8 View of Area B during the 1986 excavation, with the S expansion partly excavated through Zone B. Note

the standing water in the previously excavated (1985) area, due to the Ghyben-Herzberg aquifier which continually
flooded the deeper portion of the excavation, and which had to be pumped dry every morning.

A further innovation of the 1988 excavations was
the use of volumetrically controlled samples for fine-
screening and flotation, to test whether our samples were
biased towards larger-sized items. While all excavated
sediment was routinely wet-screened through 5 mm
mesh,21 we took 1-liter bulk samples of sediment from
every level, and processed these in our field laboratory
by flotation with wet-sieving of the non-floating frac-
tion through 0. 125-inch mesh. Flotation involved pour-
ing the bulk sample into a bucket which had the bot-
tom removed and was covered with a fine mesh. The
floating fraction was skimmed off and air-dried; this
material was subsequently examined by Lepofsky for
small plant remains.22 The non-floating residue was then
wet-sieved through 0. 125-inch mesh and air-dried. The
small vertebrate remains recovered from fine-sieving
were subsequently examined by V. Butler to check for
size and recovery bias in her analysis of the ECA fish
remains.

As the W250 transect excavations proceeded, it
became evident that stilt or pole structures had formerly
been erected over the reef flat throughout the zone
sampled by the transect, as many test units N of the
former shoreline (situated at about N105) yielded one
or more preserved wooden posts or stakes in upright

position. Thus the stilt structure exposed in Area B
proved to be just one of a complex of pole-supported
structures that extended over a zone some 100 m from
the original beach line.

As our transect excavations progressed from S to
N, examination of the pottery assemblages revealed
changes in the relative frequency of vessel forms and
decorative techniques and motifs. Having completed a
preliminary analysis of the Area B ceramic assemblages
from the 1985-86 seasons, 1 now knew that differences
along the W250 transect reflected chronological changes
in the Lapita pottery ceramic complex (represented
stratigraphically in Area B). This pattern became appar-
ent as we approached the N end of the W250 transect,
where dentate-stamped pottery was replaced by incised
pottery. In short, there was an age-progression along
the W250 transect.23 The most likely hypothesis was that
the shoreline had rapidly prograded while the
Talepakemalai Site was being occupied, and that as the
shoreline migrated N, the zone of pole houses had gradu-
ally shifted to the N as well.

When Unit W250N190 came down upon two es-
pecially large wooden posts-associated exclusively with
incised pottery (which I already knew to be dominant
at the terminal end of the Mussau Lapita ceramic se-

80

THREE LAPITA VILLAGES

quence)-I decided to expand the excavations in this
vicinity. Because the sandy deposits here were completely
unconsolidated and susceptible to collapse (especially
as one penetrated below the water table), we were not
able to open up an extensive area.24 However, two blocks
of 4 m2 each, offset 1 m N-S and E-W, were gridded
out and designatedArea C (Fig. 4.9). These exposed an
array of no less than 20 posts and stakes, along with an
assemblage of incised pottery, obsidian, and other cul-
tural materials, providing an excellent sample of the late
end of the occupation sequence at Talepakemalai.

We also expanded excavations at the Area B local-
ity, with a new 4 m2 block defined by Units W200-201/
N144-145, separated by a 1-m baulk from the S edge
of the 1986 Area B excavation. Our aim in opening this
new block at Area B was primarily to obtain a closely
controlled sample of molluskan midden for analysis by
our marine biologist colleague Dr. Catterall. However,
this also permitted us to obtain an additional sample of
ceramics and other materials which, because we were
able to impose particularly good stratigraphic control
(having previous knowledge of the stratigraphic se-
quence from previous seasons), we were subsequently
able to use for detailed temporal analysis of the ce-
ramic sequence in Area B.

Finally, we carried out a series of corings using a
10-cm diameter bucket auger W along the N140 grid
line 50 m to W300, in order to track the subsurface
extent of the organic zone containing preserved plant
remains. These cores, combined with the transect op-
erations, revealed that the waterlogged part of the
Talepakemalai Site extends over an area of at least 10,000
m2 lying N of the old shoreline feature. All of this area
is also N of the zone of surface sherd distribution, with
the combined total area for cultural deposits at ECA
something in excess of 82,000 m2. This makes ECA the
largest Lapita site yet recorded in either Near or Re-
mote Oceania.25

The 1988 season at ECA ended on October 12,
and all units were back-filled. Over three field seasons,
a total of 84 m2 was excavated at the site, reflecting a
volume of - 95 m3 of sediment, 100% of which was
sieved (Table 4.1). Given a total area in excess of 82,000
m2, our excavations represent a mere 0.1% sample of
the site. Yet the acquisition of that sample was no small
feat, requiring the expenditure of 849 person-days of

eldwork, not to mention a much greater amount of

or in laboratory processing, cataloging, and analysis,

all supported by a small fortune in grant funds. While a
0. 1% sample is precisely one order of magnitude smaller
than the ?mystical" 10% figure cited as minimal by some
devotees of random sampling, the quantity of cultural
materials produced is daunting: 44,089 ceramic sherds,
94,749 vertebrate faunal remains, 1.43 metric tons of
molluskan shell midden (all of which was weighed and
processed), and hundreds of other artifacts and
manuports. Writing this monograph several years after
the close of our excavations, from the distant vantage
point of Berkeley, I regret that we were unable to ex-
pose at least one complete stilt-house structure. How-
ever, when I remind myself of the effort expended
simply to achieve what we did, I am brought back to
the reality that an excavation of sufficient magnitude to
expose a full structure-minimally 150 contiguous m2
or more-would have consumed financial resources,
time, labor, and equipment far beyond those available.

THE ECA SITE EXCAVATIONS

I turn now to a presentation of the main results, begin-
ning with the initial 1985 transects and progressing to
the later areal excavations and transects focused on the
N, waterlogged portion of the site. 1 concentrate on
the stratigraphy, features, spatial patterning, and chro-
nology of the various excavation units, with only brief
summaries of the cultural materials recovered, as these
latter are the subject are various specialist reports in
Volumes II and III.

THE AIRFIELD TRANSECTS (1 985)

The ?airfield transects" comprise the initial set of test
units excavated in 1985 along the main NO grid line
parallel to the landing strip, and along the perpendicular
W100, W200, and W400 grid lines. During excavation,
these were designated TP-2 through -12, and are iden-
tified as such on the site plan (Fig. 4. 1).2 Along the NO
line, TP-2 (E1OONO) and -8 (W500N0) both proved to
be culturally sterile, serving to define the E-W bound-
aries of the Lapita deposit. TP-3 through -7 all con-
tained cultural materials, but the deposits were uniformly
shallow and heavily reworked by gardening. The strati-
graphic section recorded at TP-6 (W300NO) is fairly
typical:

LAYER I: 0-25/27 cm bs. Color: dark brown (7.5YR

3-4/2). Sandy loam with a dense root mat in

THREE LAPITA VILLAGES

FIGURE 4.9 Excavation in progress in one of the two 4 m2 excavation units of Area C, during the 1988
field season.

the upper 5 cm. Thoroughly reworked and fea-
tureless. This deposit contains small, worn and
highly weathered potsherds along with some
bone, and marine shell which is "chalky" tex-
tured due to chemical decomposition (result-
ing from humid acids). The contact with Layer
II is gradational over a 3-4 cm zone.

LAYER II: 25/27-35 + cm bs. Color: pinkish white

(7.5YR 8/2). Medium-sand sized calcareous
sediment consisting of sand grains mixed with
reef detritus (water-rolled shells, branch coral
fingers, foraminifera tests, etc.). This deposit is
culturally sterile.

The Layer I deposit in all of these test units yielded
pottery, some with Lapita decorations, but in virtually
all cases the sherds were highly worn and weathered,
and of small size. The poor condition of the sherds

undoubtedly derives from the repeated reworking of
this deposit for root-crop gardening. The thorough re-
working of the Layer 1 deposit had destroyed any fea-
tures which might once have been present, except in
Unit 4 (W1O0NO), where two circular pit or postmold
features were found extending from the base of the
cultural deposit down into the underlying Layer II sand.

The limits of surface sherd and shell midden distri-
bution, as recorded by intensive surface walking, did
not extend farther S than about the S80 line, as shown
by the dashed line on Figure 4.1. TP-12, at W200S50,
was excavated to test this S periphery of the site, and as
with the units described above, yielded only a small quan-
tity of worn sherds in a thin Layer I deposit.

On the N side of the airfield, TP-10 and -11 were
excavated at W1OON100 and W1OON1 10 respectively.
Neither of these units revealed any features, and only
low densities of small potsherds and some shell and

82

THREE LAPITA VILLAGES

bone midden were recovered.

TP-9 was excavated at W400N72, also on the N
side of the landing strip, just S of a steep slope leading
down to a low-lying swampy depression which pre-
sumably represents the position of the Lapita-period
reef flat prior to coastal progradation. The unit itself
was excavated into the top of the mid-Holocene
paleobeach terrace. Because this unit yielded the oldest
14C age determination from ECA, its stratigraphy and
other details are potentially significant. The stratigraphy
of TP-9 follows:

LAYER IA: 0-12/15 cm bs. Color: very dark gray

(lOYR 3/1). This layer consisted of an organi-
cally enriched loam, the modern soil used for
gardening by the Eloaua villagers, and was
heavily penetrated by rootlets. The contact with
Layer IB was gradational.

LAYER IB: 12/15-30/35 cm bs. Color: grayish brown

(lOYR 5/2). A fine-grained sand, this layer is
stained grayish in color from the downwards
percolation of organic materials from Layer
IA. Both Layers IA and IB were midden de-
posits with substantial quantities of pottery and
shell midden.

LAYER II: 30/35-110 cm bs. Color: white (lOYR

8/1-2). Medium to fine-grained calcareous sand
at this layer represents the paleobeach deposit.
The water table was reached at 110 cm bs, and
excavation was terminated.

TABLE 4.1 Summary of the Talepakemalai

Site excavations.

Field     Location    Units   Volume
Season                  (2)      ()3

1985      Airfield         15     12.00

Transects

Area A            6      6.00
Area B           12     13.20
1986      Transects        13     16.65

(W200)

Area B           12     13.20
1988      Transects        14     18.70

(W250)

Area B            4      4.40
(extension)

AreaC             8     10.40
TOTALS                     84     94.55

Virtually all cultural materials were confined to Layers
IA-IB, although 9 sherds were found dispersed in the
upper part of Layer II. Layer IA yielded 210 sherds
and 44 pieces of obsidian, and Layer IB yielded 205
sherds and 16 pieces of obsidian. Layers IA-JB also
contained a high density of shell midden, with more
than 20 kg of mollusks recovered.

From the upper part of Layer II, in association
with the 9 dispersed sherds mentioned above, we ob-
tained a sample of wood charcoal which yielded a con-
ventional radiocarbon age of 3260 + 90 BP, calibrated
to 1687-1441 BC, one of the oldest radiocarbon dates
from ECA or any Near Oceanic Lapita site. The sig-
nificance of this radiocarbon date is discussed further
in Chapter 10.

In all of the airfield transect units, some dispersed
charcoal was present in the Layer I garden-soil deposit.
However, as the dark loam had been repeatedly turned
over and thus incorporated charcoal from burning the
second growth during gardening, it seemed likely that
most of this charcoal derived from post-occupation
activities, and would be inappropriate for assessing the
age of the Lapita occupation. Only in TP-9, with its
deeper cultural deposit, we were able to obtain a char-
coal sample directly associated with Lapita pottery and
midden. As noted above, this sample may indicate that
Lapita cultural activity had commenced on the
paleobeach terrace above the former reef flat as early
as 1500 BC, although an inbuilt "old wood" error factor
could also be present in this case.

AREA A EXCAVATIONS (1 985)

Area A was excavated to provide an expanded sample
of the cultural deposits on the paleobeach terrace (.- 1.9
m asl) located just S of the Lapita-period shoreline. This
was the same area tested by Bafmatuk et al. (1980) in
1978, and by us at TP-13 in 1985, the latter revealing
ceramics extending down into the basal calcareous sands
for some depth. Area A consists of six contiguous units
defined by Units W228-229/N100-102 (Fig. 4.10).
Area A Strat%igraphy

Lithologically, the stratigraphy of Area A did not differ
appreciably from that evidenced in the airfield transect
test pits, with the upper reworked garden soll overlying
a basal deposit of calcareous sand and reef detritus.
However, while cultural materials in the airfield tests

1

83

THREE LAPITA VILLAGES

0     10   20m

.1

Grid N

* Area C

- N1 80

TA L EPA KEB UA

x x   x  x   x  Z
Z X  X

Z

-     Approximate location of

shoreline at -3,500 B.P

TA L EPA KEMA LA 1

X    X     X      Area B

Area BZ
Extension

_ _

1 ~ ~ ~ ~ ~ ~ ~ -

Area A

1 978

National
Museum

Excavation

Z   OD

u

u)

*     _

0
LO)

00

Q1)
(1)

X = Auger test hole
* = Transect unit

W250

W200

FIGURE 4.10 Plan of the N portion of the ECA Site, showing the locations of the W200 and W250 transects, and
of Areas A, B, and C.

x

x

-N150
-N100
N50

l~~~~~~~~~~~~~~~~~~~                                      1

84

THREE LAPITA VILLAGES

were confined to the upper Layer I deposit, in Area A
(as in TP-13) they extended down into the Layer II beach
sands to a depth of 1 m bs. The stratigraphic section of
the S face of Area A was as follows:

LAYER I: 0-18/25 cm bs. Color: Very dark grayish

brown (lOYR 3/2). This sandy loam is exten-
sively reworked through gardening activity, with
many roots in the upper portion. Potsherds are
small and heavily worn. The contact with Layer
II is indistinct and gradational over a - 5 cm
zone.

LAYER IIA: 18/25-45/60 cm bs. Color: White to

very pale brown (1OYR 8/2-3). Coarse-grained
calcareous sand and reef detritus (small
waterworn shells, foraminifera tests, etc.), con-
taining pottery sherds, bone, and molluskan
midden. No charcoal flecking present; no evi-
dent organic staining of the deposit. The de-
posit is weakly cemented in places. A sharp
contact with Layer IIB was evident both in
color and in the degree of cementation.

LAYER IIB: 45/60-90 + cm bs. Color: White (lOYR

8/1). In terms of lithology and cultural con-
tent this deposit is identical to Layer IIA, but it
has been strongly cemented throughout, by
deposition of calcium carbonate bonds
(CaCO) between individual sand grains. It is
this calcium carbonate cement that gives the
deposit its brilliant white color.

As excavation proceeded from the base of Layer
IA into IIB, it became increasingly difficult-and fi-
nally impossible-to remove the concreted sediment
by trowelling. Because sherds, midden, and artifacts con-
iinued to be exposed, we resorted to the use of a heavy
iLeel crowbar to break up and prize out the deposit.
This method was far from ideal, resulting in inevitable
kerd breakage, but could not be avoided. Calcium car-
ionate cementation of beach sands as in the Area A
tlposit is a phenomenon noted for other early south-
*est Pacific coastal sites, such as the Lapita-age depos-

on Niuatoputapu Island (Rirch 1988a). The deposi-
n of CaCO3 cement between sand particles is pre-

ably due to leaching of the carbonates in the upper
ds by rainwater and downward percolation and

sition. At - 1 m bs the deposit became so strongly
mented that even the efforts of our strongest work-

with the steel crowbar were inadequate to break

up the sediment, and excavation had to be terminated.
Although the frequency of sherds had decreased at this
depth, they were still present, and therefore the base of
excavation is probably not the actual base of the cul-
tural materials.

Cultural Content at Area A

Area A yielded an assemblage of 1,493 ceramic sherds.
In this entire assemblage, only two sherds carried den-
tate-stamped decoration, and there were no instances
of incising. One decorated sherd is from a carinated,
wide-mouthed bowl and has a simple motif using De-
sign Element 5, with lime-infilling. The second sherd,
which is red-slipped, exhibits a single dentate-stamped
line with a right-angle bend.

The majority of the ceramic assemblage is quite
uniform, with a predominately red paste (Munsell color
ranges from lOR 4-5/6-8 and 2.5YR 4-5/6-8) and cal-
careous sand temper. Red slipping (color lOR 4/6) is
present on several sherds. The pottery is low-fired and
inoxidized carbon cores are common. The primary ves-
sel shape represented is a large jar with a restricted ori-
fice and strongly everted rim. Identical pottery was re-
covered from the waterlogged portions of the site
sampled by Area B and the W250 transect, but in those
localities it is a minor type found in association with a
variety of other paste-temper combinations as well as
vessel forms. What makes the Area A assemblage unique
is the restriction to a narrow range of variation.

Aside from ceramics, another artifact class recov-
ered in significant numbers from Area A is small-to-
medium diameter Conus-shell rings, of which some 18
were excavated. This is a high concentration, a second
feature that differentiates Area A from other parts of
the ECA Site.

Radiocarbon Dating

Because charcoal was wholly absent from the Area A
deposits, radiocarbon age assessment had to be based
on dating midden shell. Two samples of large bivalves,
not waterworn, which were clearly culturally deposited
midden items were selected: a Tridacna"gzgas valve from
Unit W228N102/3 (Level 3, 12-18 cm bs) and aHyotissa
hyotis valve from Unit W229N100/9 (Level 9, 72 cm
bs).27 As these samples came respectively from the top
and bottom of the Layer II cultural deposit, they should
indicate whether the deposition of Layer II occurred
over a significant time interval. The samples produced

85

THREE LAPITA VILLAGES

uncalibrated ages of 3190 + 80 BP (ANU-5084) and
3130 + 80 BP (ANU-5085), strongly overlapping at la.
The Layer II deposit therefore accumulated rapidly, and
the entire Area A assemblage may be regarded as a single
unit for temporal analysis. The calibration of these ra-
diocarbon ages-based on marine shell samples-to
calendrical ages is significantly influenced by the choice
of a AR correction factor for marine reservoir effect.
This is a complex issue, discussed at length in Chapter
10. Here I simply note that our preferred summed cali-
bration for the two dates (pooled sample) at la is 1500-
1290 BC.28

AREA B AND W200 TRANSECT EXCAVATIONS

(1985-88)

As described above, the excavation of transect Unit
W200N150 (TP-14) during the early days of the 1985
season led to the discovery of waterlogged sediments
containing well-preserved ceramics and other cultural
materials in association with the bases of wooden posts.
This test unit was expanded to a 12 m2 areal block and
was designated Area B; and in 1986 it was further in-
creased to 24 m2 in 1986. During the 1986 season, we
excavated a series of additional 1 m units along the
W200 transect line (from N100 to N200), thereby situ-
ating the Area B block within a detailed stratigraphic
context. Finally, in 1988 an additional 4 m2 extension to
Area B was excavated in order to procure tightly con-
trolled samples of shell midden and ceramics. The lay-
out of all of these excavations is shown in Figure 4.10.
Because Area B and the W200 transect are closely inte-
grated, I describe the results from both of these opera-
tions together in this section.

Area B and W200 Transect Stratigraphy

The generalized stratigraphy along the W200 transect is
summarized in Figure 4.11, proceeding from Unit
W200N100 at the approximate crest of the paleobeach
ridge (2.01 m asl), out to W200N200 (1.20 m asl). The
stratigraphy of the W200N100 pit was essentially the
same as described above for Area A, consisting of an
organically enriched, disturbed garden soil developed
on otherwise structureless and lithologically uniform
calcareous sands. Beginning in W200N1 10 and extend-
ing through W200N130, however, the stratigraphy be-
came more complex with the presence of a waterlogged
fme-silty deposit containing abundant anaerobically pre-

served plant remains such as coconut and Cananium en-
docarps. The stratigraphic section for W200N120 (IP-
18), shown in Figure 4.12, is fairly typical:

LAYER I: Color: Very dark brown (1OYR 2/2). Dis-

turbed garden soil, with numerous roots. The
sandy loam textured sediment contains much
dispersed charcoal (resulting from burning dur-
ing gardening), along with shell, pottery sherds,
bone, and obsidian. Land crab burrows were
observed at the interface of Layers I and II,
which was otherwise abrupt and clear.

LAYER II: Color: White (lOYR 8/1). Massive de-

posit of medium to coarse-grained calcareous
sand, with considerable CaCO3 concretions in
the upper 10-20 cm. Cultural material is present
throughout, but less densely concentrated than
in Layer I. The lower stratigraphic boundary is
fairly straight, and transitional over a - 3-5 cm
zone.

LAYER III: Color: Gray (lOYR 6/1). Fine-grained

sand to silt textured calcareous sediment, with
some clay-particle sized lenses in the upper 20
cm; very sticky and plastic. This deposit is en-
tirely waterlogged and contains abundant
anaerobically preserved wood fragments, co-
conut shell, and other seeds and seed cases. Also
present are sand-dwelling bivalves such as
Quidnipagus and Veneridae spp., in death posi-
tion with both valves intact, indicating that this
layer was deposited in a sub-tidal environment.
Finely dispersed charcoal flecks are also present
throughout the deposit. The basal contact into
Layer IV is gradual.

LAYER IV: Color: White (lOYR 8/1). Medium to

coarse-grained calcareous sand with reef detri-
tus, similar to Layer II in lithology and struc-
ture. Culturally sterile, this represents the sub-
tidal reef platform at the period prior to initial
Lapita occupation.

Moving N along the W200 transect, the "muck-
zone" deposits represented by Layer IH in W200N120
disappeared between N130 and N140, to be replaced
by a highly concentrated waterlogged deposit of pot-
sherds and other occupation debris, associated with
wooden post bases. The stratigraphy in Unit W200N140
?IP-16) is typical:

LAYERI: Color: Very dark gray (1OYR 3/1). A heavily

86

87

THREE LAPITA VILLAGES

Depth below Area B datum (cm)

o      8 0           8            8

0                                                              Co~~~~~~~~L)  ,  l

(1)

CJ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~C
0~~~~~~~~~~~~~~~~~~~~~~~~~7

-c
GL~~~~~0

(n~~~~~~~~~~1

LO~~~~~~~~~~~~~~~                                     0~~~~~~~DL

(L   X      1'  ' .  "' ''  '0  i.  .  '0 0'  F~~~~~~~~~~~~~~~~~~~~~~~~~~~~70 CI  >s

Q)

J ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- 4  o

3:
Co~~~~C

I 5 ~~~~~~~~~~r

z                C~~~~~~~~o

4-'     ' ) .'i

-~~~~~~~~~C -                     4-,8 c

uo   bo?)~    ~  ~~o

I;~~~~~~ -o            - a  ,       0=

|     ::     : 0  0   8                     D 0  0  C

W i- -, -,z                     CO
_  I   ,  .L. -. - .  .. :, -.  -~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~(

E ' t   ' t 41  ' X E                   X~~~~~~~~~~~~~~~~~~0

11~~~~~~~~~~-L

THREE LAPITA VILLAGES

10cm

_ Pottery         0 Root        < Shell

reworked, sandy loam garden soil, penetrated
by rootlets and incorporating recent charcoal
chunks and flecks. This deposit contains pot-
sherds, heavily weathered and eroded, along
with shell midden, obsidian, and other cultural
materials. The contact with Layer II is grada-

11

0
t

-n

311
0.
0

III

Iv

FIGURE 4.12 Stratigraphic section of
the E face of Unit W200N1 20.

tional over - 3 cm.

LAYER II: Color: Dark grayish brown (lOYR 4/2)

to grayish brown (lOYR 5/2) to light grayish
brown (lOYR 6/2) at the base. The sandy loam
textured sediment, with some gravel-sized coral
inclusions is low in cultural materials. The de-

88

THREE LAPITA VILLAGES

posit is fairly loose, and lacks the concretions
present in units further to the south along the
W200 transect. Land crab burrows were evi-
dent throughout this layer. The contact with
Layer HI is irregular but fairly distinct.

LAYER HI: Color: White (lQYR 8/1). Coarse-grained

calcareous sand and reef detritus, compact but
with no concretions. Some sherds are present,
but the density of cultural materials is fairly low.
The water table fluctuates over the basal con-
tact of Layer III and IV.

LAYFR IV: Color: White (1OYR 8/1). Lithologically,

this deposit is identical to Layer Im, but Layer
IV contains a dense concentration of large shell
midden (Trdacna, Chama, Lambis, Spondylus, and
other genera), coral cobbles, branch coral fin-
gers, and reef-detritus gravel. Pottery sherds are
most heavily concentrated in a - 15-cm zone at
the top of this layer, although cultural materials
extend down to the base of excavation. (This
deposit correlates to Analytic Zone C of the
Area B excavation.)

To the N of the Area B excavation, the stratigraphy

e simplified, consisting of the upper garden zone
llowed by a thick, structureless deposit of medium-

oarse grained calcareous sand, which eventually be-
e water-logged at - 1.45 cm below the Area B da-
* The density of cultural materials in the units ex-
nding from W200N160 to W200N200 decreased

y, with the latter unit being largely sterile.

The stratigraphy of the Area B excavation itself
as in most respects similar to that just described for
nit W200N140. Stratigraphic sections of the main
985-86 Area B block are shown in Figure 4.13. The
llowing stratigraphic description was recorded for the
profile of Units W200N144-145 (Area B Expan-
n)in 1988:

LAYER I: Color: Dark gray (5YR 4/1). Sandy loam,

reworked garden soil, penetrated by roots and
containing dispersed charcoal. Sherds are heavily
worn and eroded, and shell midden is chalky.
The basal contact with Layer II is gradational.
LAYFR H. Color: Reddish gray (5YR 5/2). Medium-

grained calcareous sand, with numerous land
crab burrows, containing a low density of cul-
tural materials. The contact with Layer III is
highly irregular although relatively sharp and

abrupt (< 1 cm transition).

LAYERIIIA: Color: White (5YR 8/1). Massive, struc-

tureless deposit of coarse-grained calcareous
sand. Some fine, dispersed charcoal flecking is
present. The Ghyben-Herzberg aquifer fluctu-
ates vertically through this layer depending upon
tidal range, so that the extent of water-logging
is variable. However, land crab burrows do
not extend into this deposit. It is in this layer
that the rotted tops of wooden post bases first
became evident during excavation. Much cul-
tural material is present, and large sherds ex-
hibiting incised decorations lay horizontally near
the top of this layer in the 1988 excavation.
The boundary with Layer IV is gradational and
not very clear.

LAYER IB: Color: White (5YR 8/1). Lithologically,

this layer consists of the same coarse-grained
sandy calcareous sediment as in Layer IHA, but
it is differentiated by the markedly heavier con-
centration of shell midden along with gravel-
sized coral reef detritus.

LAYER mHC: Color: White (5YR 8/1). Coarse-grained

calcareous sand, with a lower density of shell-
fish midden than in Layer IB, but an increased
frequency of Acropora branch coral fingers. Sig-
nificant quantities of anaerobically preserved
plant remains also appear in this deposit for
the first time. The frequency of pottery and
other artifacts is very low, however, and ap-
pears to cease completely at the base of the
layer.

Several observations regarding the Area B strati-
graphic section are critical to an understanding of the
depositional and post-depositional history in this part
of the site: (1) Layers HA to IHC are essentially identi-
cal lithologically, and represent changes in the nature of
cultural deposition into the same geomorphological
context. Deposition of these sediments occurred sub-
tidally, as indicated by the presence of large numbers
of complete shells of marine bivalves such as Quidnipagus
sp. and Veneridae spp. which live in continually sub-
tidal, sandy environments. (2) Given that cultural mate-
rials tend to lie horizontally within Layers hA to IHC,
and that many matching sherds forming large segments
of vessels were found together in undisturbed contexts
(Fig. 4.14), the deposition of cultural materials must have

89

THREE LAPITA VILLAGES

W199 N147

W200 N1 47

0                    50cm

FIGURE 4.13 Representative stratigraphic sections for Area B at Site ECA: (A) S face of units W1 99-
200N1 47; (B) E face of units W1 98N1 48-149.

A

11
111
BII

Be

II
III

90

THREE LAPITA VILLAGES

?. g'J'?'

iv W

:'''>',

d+4, ',"

, ?8R W-
h f *, '

- >; s s

t e.n: ".

z    W*F'''   aN
X ' ,:

k'ht4t,s.'. |

f "'') ' ,,

FIGURE 4.14 A cluster of sherds in situ in Zone C of Area B; several of these sherds came from a
pedestalled bowl decorated with fine-stamped dentate designs.

occurred in a sub-tidal situation that was not subsequently
disturbed by human activities, or subject to high-energy
natural geomorphic processes. Presumably, this is be-
cause the occupation at the site occurred on pole or
stilt-house platforms, with deposition of pottery,
midden, and other materials resulting from purposive
and/or accidental dumping of materials off of these
house platforms. (3) The Layer II deposit, and to some
extent the upper portion of Layer IIIA (the zone of
tidal fluctuation of the Ghyben-Herzberg lens), have
been heavily disturbed by land crabs (Cardisoma sp.)
which burrow down to the water table, but do not
penetrate the water-logged sands. Over many years, these

crabs have depleted Layer II of most of the larger-
sized cultural materials, depositing them in their bur-
row ?tips" on the ground surface. Pottery, shell midden,
obsidian, and other materials so removed by the crabs
became reworked into the upper Layer I garden soil
through the repeated gardening that took place over
this part of the site. Layer I thus represents a palimpsest
that incorporates cultural materials removed by post-
depositional activity from varying depths and time pe-
riods.

In sum, the water-logged Layers IIIA to IIIC de-
posits which were sub-tidally deposited have always
remained undisturbed, thanks to their continual aque-

91          i

m

THREE LAPITA VILLAGES

ous immersion, preventing later disturbance by land crab
burrowing. On the other hand, the upper deposits of
Layers II and I have been significantly disturbed by post-
depositional processes, resulting in the depletion of cul-
tural materials from Layer II (and some internal mixing
due to burrow collapse), and the addition and rework-
ing of cultural materials into Layer I.

Referring to the summary W200 transect diagram
(Fig. 4.11), we can now describe the geomorphological
context of the Area B stilt-house occupation. At the
time the stilt structures were occupied, they covered a
zone - 30 m wide, situated - 20 m offshore (N) of the
extant shoreline (this shoreline lay between Units
W200N110 and W200N100). Dumping of this shell
midden, pottery, and other debris off of the stilt plat-
forms onto the sub-tidal reef flat resulted in the accu-
mulation of a thick and highly concentrated cultural
deposit, which gradually became buried under calcare-
ous sands as the shoreline prograded northwards. At
the time of occupation, however, the shallow water
reef flat between the stilt sturctures and the shoreline
accumulated a deposit of fine-grained sands and silts
(the ?muck zone") trapping large quantities of organic
materials, particularly wood, coconut shells, and other
plant materials (Fig. 4.15). The very fine texture of the
sediments indicates a low-energy depositional environ-
ment, probably because the pole structures themselves
broke the force of wave action before the beach. This
pattern of a zone of stilt structures lying offshore, with
a ?muck zone" in the between the beach and houses,
was also replicated along the W250 transect.

Area B Anartic Zones

The cultural context of the Area B excavations is best
analyzed in a set of analyticzones (see Chapter 3) repre-
senting litho-stratigraphic units. The concordance be-
tween excavation levels (see Chapter 3) and these analyti-
cal zones is shown in Figure 4.16 for Area B, and in
Figure 4.17 for the 1988 Area B Expansion. Zone A
corresponds with Layer 1 as described above for the
Area B Expansion, and is the reworked garden soil de-
posit. Zone B, which is divided into B1 and B2 (upper
and lower respectively), corresponds with the crab bur-
row-disturbed sandy deposits underlying the garden soil
and extending into the zone of tidal fluctuation of the
Ghyben-Herzberg aquifer. Zone C, with three subdivi-
sions (C1, upper to C3, basal), corresponds to the per-
manently waterlogged deposits (below - 1.45 cm bd,

FIGURE 4.15 Close up view of anaerobically preserved

plant remains in the face of Unit W200N1 20, within the
'muck zone' sediments. Visible here are parts of a

coconut shell (Cocos nucifera), as well as the endocarp
of a seed of Pangium edule.

corresponding to 0.45 m asl) containing the highest den-
sity of cultural materials (Layers IIIA to IIIC). These
analytic zones are used throughout this chapter, as well
as in Volumes II and HI, to present the vertical distribu-
tion of materials excavated at Area B.

Architecture and Features

The 24 m2 excavation block at Area B exposed 33 pre-
served wooden posts or stakes, while the 4 m2 Area B
extension excavated in 1988 exposed an additional six
posts and stakes. A horizontal timber was also exposed
in units W198-199/N149. The details of these wooden
objects are reported in Table 4.2, while plans of the
features are shown in Figures 4.18 and 4.19. A perspec-
tive rendering of the completed Area B excavation with
standing posts exposed is shown in Figure 4.20. Varn-
ous views of the Area B and Area B expansion excava-
tions, showing the wooden posts, are provided in Fig-
ures 4.21 to 4.24.

Metrically, these upright wooden objects can be di-
vided into two size classes: (1) posts, with diameters be-
tween - 10-20 cm; and (2) stakes, with diameters be-
tween 3-10 cm. In Area B, the posts are aligned in two
rows running roughly N-S and E-W, which meet in a

92

THREE LAPITA VILLAGES

1986-88 Datum depths (m)
I   I

L-~~~~~~~~~~~~~~~L

N    t U)C.       L

N      _   t    U)~~~~~~~~~1

CN          t     U-

_Y                      _ _

CY~ ~         U)  r,

q    Co

g~~~~C    (1C,              cq t
E~~~~c     or

E~~~~~~~~l CQ             N  Ci]   LQ

E~~~  ~    ~   ~~~~~~ c4 Iq cq

198 Datm ephsm

ZONE

W1 95N1 51
W1 95N1 50
W196N151
W196N150
W197N151
W197N150
W198N151
W1 98N1 50
W198N149
W1 98N1 48
W1 98N1 47
W199N151
W1 99N1 50
W199N149
W1 99N1 48
W199N147
W200N1 51
W200N1 50
W200N 1 49
W200N1 48
W200N1 47
W201 N151
W201N150
W201N149

E

(U

-n

0

7o

>

0

X

m

._

0 O
C

>

C0

> .

CG)
0 ,>
o

Co )

0c0

CoC

._ _

bO

di>

J -.-

O u)

C X
CoC

>xC0

u C
(0 D

CO

93

THREE LAPITA VILLAGES

Area B Extension

W200        W200        W201
N144        N145        N144

W201     Analytical
N145     zones

A
B1
B2

1 6

5

L r nJ9

6        6    C1

C2

1         9    C3

Disturbed
garden
soil

Large incised sherds

Iying honzontally; finely
dispersed charcoal

Low density

Dense midden
concentration;
dentate sherds

Decreased midden
density; eroded

sherds, branch coral

FIGURE 4.17 Concordance diagram showing the correlation of excavation units, levels, and analytic zones in the 1988

Area B extension excavation. This diagram may be used to determine the proper zone assignment of any excavated
object, given its unit and level provenience.

corner at the conjunction of excavation units W200-
199/N150-15 1, marked by a cluster of three large posts
(see Fig. 4.20). Six small stakes lie immediately W of the
corner, and may have supported some kind of adja-
cent, ancillary construction (a walkway, racks, or similar
structures are likely possibilities). The two principal post
alignments may represent main supports for a substan-
tial structure; given that the N-S alignment continues
into the 1988 expansion, the sides of this stilt-house
structure were mmnimally > 8 m long.

Posts 1 and 2 from the main corner, and a smaller
stake (No. 30 in Table 4.2) were submitted for radio-
carbon dating; Figure 4.25 shows Post 1 with the samples
removed for dating. The dates all overlap at lcs, and
suggest that the stilt-house structure and associated fea-

tures at Area B were constructed between - 1390-1090
cal BC (see Chapter 10). Posts 1 and 2 were also exam-
ined by Prof. D. E. Yen (Australian National University)
prior to dating, who identified the wood as Intsia bijuga.
Other posts appeared to us, and to our informants
during excavation, as probably being Cordia subcordata.
Gerhard Peekel, a German priest and naturalist who
wrote an important flora of the Bismarck Archipelago,
observed that Intsia bijuga is a common foreshore tree,
whose "wood is extremely hard and is called ?ironwood'.
It is excellent for furniture- and building-timber, . . .
resistant to termites" (1984:214-16). Of Cordia subcordata,
Peekel writes that it is a "widespread, abundant beach
tree... The brown, black-veined heartwood is very
strong" (1984:471).

0.5 -
0.7 -

E 1.0-

F
3

_

00

0-

00

-Q

+-? 1.5

0:
0

2)

2.0-_

Area B
(1986)
W198
N1 47

A

B1

B2
c1
C2
C3

Z                     1

94

THREE LAPITA VILLAGES

Because permanent preservation of all of these
posts posed a significant conservation problem, we felt
it preferable to leave these architectural features in situ
where they could be re-exposed if an expanded areal
excavation is undertaken at Area B; most posts and
stakes were therefore not removed. However, Posts 1
and 2 were removed at the close of the 1985 excava-
tion and sent to the Prehistory Department at the Aus-
tralian National University, where they were conserved
by W. Ambrose. These posts both revealed sharpened,
pointed bases with adz marks, as seen in Figure 4.25.

Because the Zone C deposits were deposited sub-
tidally onto the sandy reef flat, there are no features in
the usual sense of hearths, ovens, or pits. However, as
deposition was within a low energy marine environ-
ment without significant disturbance, there are a num-
ber of clusters of sherds and other materials as shown
in Figure 4.18. For example, in Unit W201/N149 we
exposed an in situ, undisturbed arrangement of sherds
representing about two-thirds of a large, beautifully
decorated flat-bottomed dish (Fig. 4.14). This vessel had
evidently been tossed off the stilt-house platf?rm, pre-
sumably after breaking, where it had landed upside down
on the sandy lagoon floor. Although the impact of land-
ing on the sandy flat fractured the vessel into more than
one dozen sherds, these were not subsequently dislodged.
Similarly, much of a large plainware jar was found in a
similar condition in Unit W201/N150, and half of an
exquisite vessel or cylinder stand with anthropomor-
phic design was exposed in Unit W198-199/N148 (Fig.
4.26). Another feature of interest was a dense concen-
tration of large mollusk shells packed closely around
the bases of wooden posts 20, 21, and 29.
Spatial Patterns in Area B

As described above, Area B exposed what appears to
be two sides of a rectangular structure, originally sup-
ported on wooden stilts or posts, and situated over a
shallow reef flat. In order to determine whether the
distribution of cultural materials in Zone C of Area B
might reflect patterns associated with this structure (such
as differential density of materials to one side of the
post alignments), we plotted the distribution of a vari-
ety of classes of materials (both artifacts and faunal
remains) by grid unit.29 Figure 4.27 is a perspective den-
sity plot of the Area B excavation (oriented in the same
manner as the perspective drawing of posts shown in

Figure 4.20), showing the distribution of plain body
sherds, while Figure 4.28 depicts the distribution of
decorated body sherds. In Zone C3, there is a high con-
centration of plain sherds in the vicinity of the putative
corner of the stilt house, while the decorated sherds
display a fairly uniform distribution. In Zone C2, a rather
different pattern emerges, with both plain and deco-
rated sherds displaying higher concentrations in the S
portion of the main excavation, and extended beyond
in the 1988 Area B extension. Finally, in Zone C1, which
has the highest overall concentration of ceramics, there
are fairly dense "pockets" both in the S and E portions
of the main excavation, while the 1988 Area B exten-
sion displays a relatively low density.

The distribution of lithic flakes (primarily obsidian,
but also including some chert flakes) is shown in Figure
4.29. In Zone C3, two units to what was presumably
the exterior of the stilt-house structure (on the N and
W) display slightly elevated densities, while in Zone C2
there is a similar high-density unit on the W side of the
N-S post alignment. Zone C1 shows the highest overall
density of lithics (thus correlating with the vertical dis-
tribution of ceramics), with the highest concentrations
again to the N and W of what we interpret as the stilt-
house alignment. This is a pattern which might be an-
ticipated if knapping was taking place on the elevated
stilt house platform, with debris falling or being tossed
off the edges of the platform.

Cultural Content of Area B

Area B yielded a large assemblage of ceramics and non-
ceramic portable artifacts, as well as faunal materials
(Table 4.3). By far the greatest density of materials oc-
curs in Zone C, the undisturbed waterlogged deposit.
Ceramics, obsidian flakes, and vertebrate fauna increase
steadily in quantity from Zone C3to Zone C1, whereas
shell midden is fairly uniformly distributed. The Zone
C3-1 deposit is directly associated with the wooden stilt-
house architecture. Zone B, above the Ghyben-Herzberg
lens, has a considerably lower density of all classes of
cultural material. Zone A also has a low density, and the
materials contained in this reworked garden soll appear
to have been deposited largely through the burrowing
actions of land crabs which have removed materials
from lower depths (especially Zone B). Zone A should
therefore be entirely discounted as a primary occupa-
tion unit in any analysis of the Area B materials.

95

THREE LAPITA VILLAGES

TABLE 4.2 Data on wooden posts in Site ECA excavations.

AREA/UNIT   POST   DEPTH   MAX.   MIN.   LENGTH    IDENTIFICATION               COMMENTS

NO.    (B.D.)  DIAM.  DIAM.
AREA B

W200N150      B1
W199N151      B2

W199N150      B3

W199N150      B4
W200N1 51     B5
W200N1 51     B6
W198N151      B7

-         -       -      -      Intsia biuga
-        18       12     -      Intsia biuga

-        17       -      -      Intsiab#iuga ?

10

3
4
13

W200N149     B8
W199N149     B9

W199N151     B10
W198N151     B11

W198N149     B12

B13
W201N151     B14
W201N151     B15
W201N151     B16
W195N150     B17
W196N151     B18
W196N151     B19
W197N150     B20
W197N150     B21
W200N147     B22
W197N151     B23
W200N148     B24
W200N148     B25
W195N150     B26
W199N147     B27
W198N147     B28
W196N150     B29
W198N148     B30
W200N147     B31
W200N147     B32
W1 97N1 51   B33

12
20

13

1.59
1.61
1.63
1.27
1.36
1.30
1.68

1.64
1.66
1.68
1.60
1.64

6
1 1
12

7
8
15

7
10
14

9
2
3
10

15
10

5
10

4
1 1

5
17
12

Main corner post; 14C dated (ANU-5790)
Main corner post; base faceted and

sharpened; adz marks visible; inclined on
450 angle towards N; 14C dated (ANU-
5791)

Main corner post; squared off in cross

section, with adz facets clearly visible; dark
black wood

-      Intsia b#uga ?     Dark black wood, smaller post
-      -                  Small stake
-      -                  Small stake

-      -                  Inclined towards E; rounded in section;

probably structural

-      Cordia subcordata ?  Brownish wood; circular section
-      Cordia subcordata ?  Brownish wood; roughly squared

Soft brown wood

-      Intsia bijuga      Dark black wood; roughly squared in

section; probably structural
-      Cordia subcordata ?  Soft brown wood
-      Cordia subcordata ?  Soft brown wood

-      -                  Very soft brown wood; circular section
-      Cordia subcordata ?  Small oval stake; soft brown wood

16        11       -       -

18

3
8
13
1 1

1?
9
3
3
9
9
8

14

2
7
10

7
10

9
2
2
8
7
5

Inclined towards N

Probably structural post
Small stake
Small stake

Part of cluster with posts 21 and 29;
probably structural

Part of cluster with posts 20 and 29;
probably structural

Main structural post
Small stake

Small post/stake

Probably structural post
Possibly structural
Small stake
Small stake

Next to Post 21; probably structural
14C dated (Beta-20452)
Small post
Small post
Small post

continued on next page

THREE LAPITA VILLAGES

AREA/UNIT    POST    DEPTH    MAX.    MIN.   LENGTH       IDENTIFICATION                  COMMENTS

NO.     (B.D.)  DIAM.   DIAM.
AREA B

EXTENSION

W201 N145     B34     -           4        3     -      -

W200N145      B35     -          14       13     -      -                    Probably structural post
W200N1 44     B36     -           8        8     -      -                    Inclined towards NW
W201 N1 44    B37     -           5        4     -      -                    Poorly preserved

W201 N1 44    B38     -           7        6     -      -                    Inclined towards south
W201 N1 44    B39     -           8        7     -      -                    Inclined towards east

W200

TRANSECT

W200N120     -        -         10.5      -      -      -                    Extends from 129-151 cm bs
W200N1 30    -        -           4        4     33     -                    Small stake with adzed point
W200N160     -        -          14       13     -      -                    Top of post at 89 cm bs

W200N1 70    -        -          10       10     67     -                    Possibly structural; top at 81 cm bs

W250

TRANSECT

W250N11 0    1        -          4.5      -      71     -

W250N1 10    3        -           8       -      65     -                    Adzed to sharp point

W250N1 20    -        -           -       -      -      -                    14C dated (Beta-30681)
W250N1 40    -        -           -       _      -      -                    14C dated (Beta-30682)

W250N1 70    -        -           -       _      -      -                    Small stake; 14C dated (Beta-30684)
W250N181     1        -          15       -      69     -                    Base roughly adzed
W250N181     2        -          10       -      -      -                    Fragmentary

AREA C

W250N1 88    C1       1.81       18       16     -      -                    Main structural post

W250N1 88    C2       1.70       12      10      -      -                    Structural post, near Cl
W250N1 88    C3       1.85        6       5      -      Diospyros sp.        14C dated (Beta-30686)
W249N188     C4       1.83        5       4      -      -                    Stake

W249N1 88    C5       1.76        8       6      -      -                    Small post
> W249N1 88   C6       1.86        9       8      -      -                    Snall post
W249N1 87    C7       1.85        8       5      -      -                    Small post

'W250N1 90    C8       1.63       14      12     -       -                    Probably main structural post

W251 N1 90   C9       1.58        9       7      -      -                    Small post; probably structural
W250N1 91    cia      1.68        9       9      -      -                    Probably structural
W250N1 91    C1i      1 .76      1 1      9      -      -                    Probably structural
W250N191     C12      1.71        7        6     -      -                    Small post
W250N191    C13      1.76       11        9     -      -                    Small post
W250N1 91   C14      1.60        9        6     -      -                    Small post

250N191     C15     1.81       10        8     -      -                    Small post
251 N191    C16     1.72        6        5     -      -                    Stake
251 N191    C17     1.75        5        3     -      -                    Stake

251N191     C18     1.82        8       -      -      -                    Small post

50N190     C19     1.78       10        7     -      -                    Possibly structural

50N1 91    C20     1.56       20       18     70     Intsia bijuga        14C dated (Beta-30687); Main house post;

adzed from two sides to a sharpened
point

51N190     C21     -            3       2     -      -                    Small stake
51 N190    C22     -            7      -      -      -                    Small post

51N191     C23     -            4       3     -      -                    Small stake

97

o-         0          0o

UL         Ln         t

z       ~    ~z    z

98

THREE LAPITA VILLAGES

a)
y

(n
Co

0
0.

E
cn

(n
Co

0

._

.

E
c'J

0

oo
0'

o
c-J

4-
4,
c:
(1
4-

cn
0

(n)

0

c:5
0
0
0~
4n
a1)

Co
~0

c
Co
0

0
4

O

4-

0

OD
0
.o

g

0
a)

._

0

(1,

0
Co
0
4)

(n

()

Co

4-

4)
0
-O

C_
(1)
4)

(1)

Jn

N1 45

Gi

Grid N

0                                1m

1      1     1      1   -   _ l   1

N1 44

FIGURE 4.19 Plan of the Area B extension, showing the locations of all wooden posts
and stakes, as well as the postions of other features and artifacts.

FIGURE 4.20 Perspective rendering (from the NW) of the main Area B excavation, showing the positions of wooden
posts and stakes.

THREE LAPITA VILLAGES

W201

W200

THREE LAPITA VILLAGES

FIGURE 4.21 View of Area B, at the completion of excavations in 1986, with wooden posts exposed but still in situ.
Note that Posts B1 and B2 had been removed at the close of the 1985 season, and thus are absent from this
photograph.

FIGURE 4.22 Close up view of Posts B20
and B21 in the E part of the main

! ^ X *-                                               Area B excavation (view is from the E

looking W). The top of a smaller post
or stake is also visible.

100

THREE LAPITA VILLAGES

FIGURE 4.23 View of the S expansion of Area B, at the completion of excavations in 1986, with wooden posts
exposed but still in situ. Note horizontal timber.

FIGURE 4.24 The Area B Expansion at the close of excavations in 1988, with several
exposed posts and stakes in situ.

101

THREE LAPITA VILLAGES

FIGURE 4.25 Posts B1 and B2 after conservation by freeze-drying, and after sampling for
radiocarbon dating. Note that the bases of the posts have been tapered to a point by
adzing, presumably for ease of insertion into the reef flat. The samples for radiocarbon

dating were removed from the outermost growth rings of the posts, and the resulting dates
should therefore give a close approximation of the age of death of the trees, and of
construction of the stilt house.

As indicated above in the analysis of Area B stratig-
raphy, the Zone C materials were deposited sub-tidally
as a result of discard from and accumulation around
the stilt-house structure. Following the abandonment of
this structure, and initial progradation of the shoreline
to the N, the Zone B materials appear to have been
deposited inter-tidally or supra-tidally as a beach front
or foreshore microenvironment. Despite this change in
the environment and mode of deposition (and conse-
quent lack of waterlogging of the Zone B deposits),
the vertical stratigraphy in Area B does present an intact
temporal sequence, which is extremely important for
the analysis of change in ceramic style and other aspects
of Lapita material culture in Mussau.

Figure 4.30 displays the vertical distribution, by ana-
lytic zone, of several class of ceramic decoration on the
sherd assemblage from Unit W200N144 of the 1988
Area B extension. A clear pattern of temporal change is
evident, with a progression from an early emphasis on
fine dentate stamping, through a middle phase with an

increase in coarse-dentate stamping, to the late deposits
marked by a dominance of incised decoration. This
temporal sequence is mirrored by the horizontal strati-
graphic progression from S to N along the W250
transect (see below). The Area B sequence similarly re-
veals major shifts not only in decorative technique, but
in vessel form, temper, and other attributes. A full dis-
cussion of these trends is deferred to Volume III.

The Zone C deposits at Area B-associated with
the large stilt structure-are notable for the presence of
the variety of shell objects which are sometimes glossed
under the rubric "ornaments." These include many ex-
quisitely executed Spondylus-shell beads and pendants,
Nautilus-shell discs, Conus-shell and Tridacna-shell rings,
large Conus "rectangular units," and other objects. Else-
where (Kirch 1988c; 1997:236-38), I have argued that
such objects functioned as "exchange valuables" in the
complex long-distance exchange system into which the
Mussau Lapita communities were integrated. The con-
centration of such numbers of these "valuables" in Area

102

r-

j

THREE LAPITA VILLAGES

B may be noteworthy. This is especially so in light of
the presence, in Zone C, of a number of elaborately
decorated Lapita ceramic vessels, including pedestalled
bowls, and dishes bearing anthropomorphic designs.
Area B also yielded a unique bone sculpture, a small
anthropomorphic image carved from sea mammal bone
(probably porpoise). This concentration of special ob-
jects and ceramics-unparalleled elsewhere in the ECA
Site, or indeed at other Lapita sites excavated to date-
might support an argument that the Area B stilt-house
was a special-function structure. However, we must not
ignore the fact that the Zone C deposits also contain
heavy concentrations of shell and bone midden, oven
stones, and food-preparation equipment (scrapers and
peeling knives). Thus food preparation (and consump-
tion?) was also a major activity at this structure.

Radiocarbon Dat/ng

Eleven samples from Area B were radiocarbon dated,
and an additional sample was dated from Unit W200/
N120 (FP-18). The samples from Area B included char-
coal, shell, and wooden post bases, while that from
W200/N120 was of coconut endocarp. Details of all
samples, and a thorough discussion of the dates, the
problems associated with their calibration, and their in-
terpretations are given in Chapter 10. Here I note sim-
ply that this suite of radiocarbon dates indicates that the
Area B stilt house was likely to have been constructed
not earlier than - 1300 cal BC, and not later than - 1200
cal BC. Although the house itself was unlikely to have
been occupied or used for more than a century or so,
deposition of cultural materials at the Area B locus con-
tinued until - 1000-900 cal BC, with the later part of this
sequence marked by a shift from dentate-stamped to
incised ceramics, as discussed above. Thus the entire
depositional sequence at Area B spans a period of no
less than - 200 and no more than - 400 years.

THE W250 TRANSECT (1 988)

The W250 transect was excavated to meet certain ob-
jectives outlined for the 1988 field season. The transect
lies 50 m W of Area B and, as with the W200 transect,
extends from the Lapita-period beach terrace (1.73 m
asl) marked by a - 0.7 m elevation drop N across a low
flat to N200 at 1.03 m asl, a total distance of 130 m
(see Fig. 4.10). The S portion of the transect, between
W250N70 and W250N110 (situated on the higher

FIGURE 4.26 Finely decorated vessel or cylinder stand as
exposed in situ in Zone C of the Area B excavation.

ground of the paleobeach terrace) consists of dark gray-
brown loam littered with shell midden and potsherds,
supporting a vegetation association of second-growth
scrub Pandanus, Morinda, and other shrubs, ferns, and
grasses. This vegetation results from repeated root-crop
gardening. Proceeding N beyond W250N1 10 and across
the former shoreline, the soil becomes a lighter brown

103

?,2

THREE LAPITA VILLAGES

Zone C1
Zone C2

C_unts  200  4q 250  500  MM= -3 0
Counts  4M-z-- IMMM  450 _ i  5ffl

Zone C3

FIGURE 4.27 Spatial distribution
(density) of plain body sherds
in Zone C of the Area B and
the Area B extension;

perspective is from the NW

g 1   n   T50  f    s 1 C00    1 50
12_          250 _1 29   300) a= 35Cw

Counts im=;       EH _1 4S0   _   50

q

104

- gab

Esmm       1 ,         50     cm=       1 00        150
Counts           2M     eemommam 2W            300    ?   ?  3w

4ffl          4W      ii? -CM

THREE LAPITA VILLAGES

Zone C1

Counts r7z77 8                 1 6        2 4     1     2    _   403

Zone C2

|R 1  I  rl 8         16   2I     4          240 24

Counts    e7 48                      64 n n n      72 _0

Zone C3

FIGURE 4.28 Spatial distribution
(density) of decorated body
sherds in Zone C of the Area B

and the Area B extension;
perspective is from the NW

1   (T    8    F=   16  (fl   24        32 w      40
Counts     = 48    _M-5     o=      64  _     72 _     80

105

106                                                  THREE LAPITA VILLAGES

Counts rZZJ2 2

IllI    5            0           15   X

E9 3s        Z  ~     40         45     MM-

,' ?fff'Ti

X-IH

Cn   J

Counts f77 30

_      _  10    15   X

r9=  5   =Z 40 _ME=I- b~        50

20 ~~~ 45

0      .90

Counts  2 =  1035 _ o

Zone C3

FIGURE 4.29 Spatial distribution
(density) of lithic flakes

(primarily obsidian) in Zone C
of Area B and the Area B
extension; perspective is
from the NW

1?

Zone C1

Zone C2

mm= 25

THREE LAPITA VILLAGES

TABLE 4.3 Summary of cultural content of Area B, Site ECA (by analytic zones).

CLASS
Ceramics

Plain Body Sherds

Decorated Body Sherds
Plain Rim Sherds

Decorated Rim Sherds
Ceramic Disks

Non-Ceramic Artifacts

Obsidian Flakes

Adz/Adz Fragment
Abraders

Scrapers & Peelers
Fishhooks

Fishhook Preforms
Shell Rings

Other Ornaments
Coral Oven Stones

Manuport/Volcanic Stone

A       B,      B2      CC c2           C3     TOTAL

1,512

96
21
96

2,882

222

38
157

2

2,580

230

53
207

399    422      309

1
1               1
5     22       28
2      4        3
25     20       27

6      8       11
2      5        4
5     69        7
9     18       49

5,604

655

91
504

483

2
2
46

6
26
28
20
12
103

4,481

417

56
376

3

244

1
3
29
26
20
17
108
39

3,949

240

84
221

9

21,008

1,860

343
1,561

14

sandy loam, lacking surface shell midden and sherds,
covered in climax forest vegetation. Because the soil
here has not been organically enriched from human oc-
cupation, the Eloaua people regard it as infertile terrain
for gardening.

The W250 transect was cleared of vegetation with
bush knives, and a series of profile elevations were taken
with a telescopic level. These elevations were tied to the
Area B datum, and a new permanent datum peg was
established for use during the W250 transect excava-
tions. As with the 1986 Area B excavations, our hand-
operated bilge pumps kept the W250 pits dry, and the
pump out-flow was utilized for wet-sieving the exca-
vated sediment. During the W250 transect excavations
we also took systematic 1-liter sediment samples from
each excavation level for flotation and fine wet-sieving
through 0. 125-inch mesh screens in the field laboratory.

W250 Transect Stratigraphy

The W250 transect provides a stratigraphic profile ex-
tending 130 m from the beach terrace and Lapita-pe-
riod shoreline in the S, to the former reef flat on the N.
A diagrammatic overview of this stratigraphic transect
is shown in Figure 4.31. Detailed sections of all transect
pits were drawn during fieldwork, accompanied by

profile descriptions. Here we provide descriptions of
four units that represent the range of stratigraphic varia-
tion across the transect: W250N80, W250N120,
W250N150, and W250N200.

Unit W250N80. This unit lies at the highest point
of the paleobeach terrace (1.73 m asl), and therefore
occupies the same geomorphic zone as Area A, and as
TP-9 (W400N72). As with Area A, the stratigraphy in
W250N80 was simple, consisting of only two sedimen-
tary deposits:

LAYER 1:0-20/25 cm bs. Color: dark brown (7.5YR

4/2). Sandy loam with a thick root/fern rhi-
zome mat in the upper 5-10 cm, heavily re-
worked by gardening. Sherds in this layer are
typically small and weathered. The contact with
the underlying sand deposit is gradational over
a 3-4 cm zone.

LAYER II: 20/25-75 + cm bs. Color: white (5YR 8/

1). Structureless deposit of coarse-grained cal-
careous sand and reef detritus, becoming more
gravelly textured with increasing depth. Sherds
and shell midden become less frequent with
depth, and were not recovered below 50-60
cm, although this does not correspond with
any evident lithological break. Layer II becomes

213    2,070

4
1       8
22     152

3      18
8     132
14      87
5      53
18     219
87     305

-

107

THREE LAPITA VILLAGES

25
20

c
0

15

10

A1     B1      B2     C1     C2      C3

Zone

strongly cemented at the base of excavation at
-75 cmbs.

The stratigraphy of Unit W250N80 is in all respects
highly similar to that of Area A. Furthermore, the ce-
ramic assemblage recovered from Units W250N70
through W250N100 also closely matches that from Area
A, being dominated by large, plainware jars with re-
stricted orifices and strongly everted rims.

Unit W250N120. The ground surface here is 54 cm
lower than at W250N80, suggesting that the former
unit was originally N of the Lapita-period shoreline;
this was confirmed by excavation. The E face of
W250N120 is shown in Figure 4.32, and the stratigraphic
profile was as follows:

LAYER IA: Color: dark reddish brown (5YR 3/2).

Sandy loam, reworked by gardening. Many
rootlets present in the upper part of this layer
which grades into Layer IB.

LAYER IB: Color: reddish brown (5YR 5/3). A zone

of leaching at the base of Layer I; coarse-
grained calcareous sand, stained gray.

LAYER II: Color: white (5YR 8/1). Massive, struc-

tureless deposit of coarse-grained calcareous
sand, containing a high density of shellfish
midden. The deposit is heavily cemented with
calcium carbonate bonds between individual
sand particles in the upper part, but becomes

FIGURE 4.30 Vertical distribution

(by analytic zone) of several main
ceramic decorative techniques in
Unit W200N1 44 of the 1988
Area B extension.

less cemented to loose with increasing depth.
The boundary with Layer III is gradational.

LAYER III: Color: light gray (5YR 7/1). Coarse-

grained calcareous sand with some fine silt-clay
sized particles mixed in. This deposit contains
anaerobically preserved plant remains as well
as cultural materials such as pottery and bone.
The contact with Layer IV is very abrupt and
smooth.

LAYER IV: Color: white (5YR 8/1). A thin deposit

of coarse-grained, loose calcareous sand con-
taining cultural material. The contact with Layer
V is very abrupt and sharp.

LAYER V: Color: light gray (5YR 7/1). A densely

compacted layer of fine clay-silt textured cal-
careous sediment, lacking any larger-sized in-
clusions. The deposit is full of preserved wood
and a variety of anaerobically preserved plant
remains. Two wooden posts passed through
this layer down into Layer VI, and were sur-
rounded by a mass of rattan-like tough, flex-
ible sticks (- 1.5-2 cm diameter), as shown in
Figure 4.33. These sticks and the posts trapped
a dense accumulation of plant remains (seed
cases, nuts) in a pocket of fine-grained silt-clay.
The thick deposit of plant remains accumu-
lated around the posts in situ. Layer V slopes
steeply from the southeast to northwest; the

108

THREE LAPITA VILLAGES

Depth below Area B datum (cm)

o     8     ?    8     ?     8

0     O        urcl          c,4  cn

0

z ,

0

- 0

z

0

z

8

z

-0
a)
o        41

z        (U

-O

a
Qi

O         E

(n)

-c

00
-   ??) z

Z

-a

Q)
Co
(-)
Co  0

0     (0

0 )   C o
-O    -o

o     0
o  0

gw>

0

z

-  2 o

0)  .)   z

~z  <     0)

L    0  T>C

0) ~ ~ N

1  =*

o~~~~b

z ~ ~ ~

Z0

-

8

r-

Q) 0)

0)'-
0-

0)

cn

CO

x

0)

cU

00

Co
0

CO

4-

00

.

-a

x
0)

Co

0)

0)

0n

OO

(-),

0)

o~

E (

o~ ~ ~~~~~~~~-

LL

1 1 1 1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

2c Z )          Z - EIg ?

m

109

THREE LAPITA VILLAGES

-   Plant remains

FIGURE 4.32 Stratigraphic section of the E face of Unit
W250N1 20.

lower stratigraphic boundary is very abrupt and
sharp.

LAYER VI: Color: white (5YR 8/1). A loose deposit

of coarse-grained calcareous sand containing
large midden shells but very little bone, and only
a few water-rolled sherds. Post 2 continued
down through this layer.

Excavation was terminated in W250N120 at 140
cm bs, when the high volume of water flowing into the
pit proved too much to be removed with the hand
pumps, and threatened to collapse the sidewalls. The
tops of the two wooden posts were removed for ra-
diocarbon dating and wood identification.

The depositional sequence in W250N120 began with
deposition on the inter-tidal flat immediately seaward

of the shoreline at the time of initial Lapita occupation
(Layer VI). This inter-tidal zone incorporated a few
water-rolled potsherds and some shell midden. The
construction of a pole or stilt structure is associated
with the next depositional unit (Layer V), and a mass of
organic material was trapped and buried in the fine sedi-
ment which accumulated around the post bases. Aban-
donment of the stilt structure probably occurred dur-
ing or after the deposition of the next two, relatively
thin sedimentary layers (IV and III). Layer II, with its
high density of shell midden, no longer represents an
inter-tidal depositional environment, but rather a beach
ridge accumulation unit, and post-dates the progradation
of the shoreline in a northerly direction.

Unit W250N150. This unit lies - 45 m N (and sea-
ward) of the mid-Holocene shoreline associated with
initial Lapita occupation of ECA, at an elevation of
1.02 m asl. The stratigraphic profile of the E face, shown
in Figure 4.34, was as follows:

LAYER 1: Color: Dark brown (7.5YR 3/2). Silty-

loam, A horizon, heavily penetrated with root-
lets; somewhat reworked through gardening.
The contact with the underlying Layer IIA is
abrupt and sharp.

LAYER IIA: Color: Reddish yellow (7.5YR 7/6). This

appears to be a zone of organic staining due to
the leaching of the A horizon and downward
percolation of organic materials.

LAYER IIB: Color: White (7.5YR N/8). A massive,

structureless deposit of medium-to-coarse
grained calcareous sand, heavily cemented. The
contact with Layer Im is gradational over - 1.5
cm.

LAYER IA: Color: White (7.5YR N/8). Structure-

less deposit of coarse-grained calcareous sand
with a slight fine-fraction admixture of silt-clay.
The contact with Layer IIIB is abrupt.

LAYER ilIB: Color: Light grey (7.5YR N/7). A thin

lens of very fine, silt-clay sized calcareous sedi-
ment, sloping gently towards the north. Abrupt
contact with Layer IHc.

LAYER IIIC: Color: White (7.5YR N/8). Thick de-

posit of coarse-grained calcareous sand, con-
taining considerable cultural material, including
anaerobically preserved plant remains in the
lower portion of the deposit. The contact with
Layer IV is gradational over a 3-5 cm zone.

LAYER IV: Color: White (7.5YR N/8). Coarse-

THREE LAPITA VILLAGES

FIGURE 4.33 Concentration of sticks and anaerobically preserved plant remains, adjacent to a vertical wooden post, in
the 'muck zone' sediments in Unit W250N1 20.

grained calcareous sand with numerous water-
rolledAcropora branch coral fingers. Various
entire sand-dwelling bivalves, such as Quidnipagus
sp. and Veneridae spp., were recovered in death
position, indicating a sub-tidal, reef platform
depositional environment.

The depositional sequence in W250N150 suggests
initial occupation on a sub-tidal reef flat composed of
coarse-grained sands with reef detritus includingAcropora
branch coral (Layer IV). A thick deposit of cultural
material including sherds, shell midden, and plant re-
mains (Layer IIIC) subsequently accumulated on top
of the reef platform, with the main source of non-
cultural sediment being calcareous sands. Further sand
accumulation occurred with progradation of the shore-
line past the N150 line and towards the present shore-
lne.

Unit W250N200. This unit marks the N end of the
W250 transect, and was situated 10 m N of the Area C
excavation, in dense climax forest not presently utilized
for gardening. The stratigraphic profile of the E face,
shown in Figure 4.35, was as follows:

LAYER I: Color: Dark grayish brown (1OYR 4/2).

Present A horizon with numerous rootlets. The
contact with Layer II is gradational.

LAYER II: Color: Light gray (lOYR 7/1). This

mottled zone has been heavily reworked by the
burrowing of land crabs, whose burrows were
evident during excavation. The deposit of
coarse-grained calcareous sand was essentially
culturally sterile. The contact with Layer Im was
abrupt but irregular.

LAYER III: Color: White (lOYR 8/1). Coarse-to-

medium grained, compact, calcareous sand,
containing many entire Quidnipagus bivalves in

11 1

THREE LAPITA VILLAGES

IIIA          7.5YRN/8

IIic

7.5YR N/8

iv              7.5YR N/8C
.Unexcavated..- ..

Acropora sp. branch coral

_   Pangium edule and other plant remains

death position, indicative of a sub-tidal depo-
sitional environment. Limited cultural material
was recovered from the upper part of the de-
posit, and the frequency of sherds and other
cultural debris declined with increasing depth.

The W250N200 unit appeared to be near the N
limit of cultural activity over the former reef platform.
The low density of sherds and other materials appears
to mark the outer fringe of midden dumping from the
stilt-house structure exposed in Area C.

FIGURE 4.34 Stratigraphic
section of the E face of
Unit W250N1 50.

Wooden Posts

Wooden posts were discovered in five of the W250
transect units, schematically indicated on the transect
diagram (Fig. 4.31). Dimensions and other details of
these posts are provided in Table 4.2. The posts are
discretely distributed into two areas or zones. The first
of these is in Units W250N110 and W250N120, im-
mediately N of the mid-Holocene shoreline; five posts
were exposed in these two units. Their stratigraphic
position at the base of the cultural deposits-set into
the underlying inter-tidal reef platform-indicates that

o cm
10
20
30

40     8)

u

(?3

D
(1)

50       z

0
6)
-a
4-_C
60     0.

6)
0)

-70
- 80
-90
-100
- 110

7

1 12

I

11                 7.5YR N/7
..-1 -

THREE LAPITA VILLAGES

111

Unexcavated-

o cm
-10
20
30

-40   Q)

-50 Q

(n

- 60   0

-70

Q

._Qn  0

- u
- 90
- 100
-110

j= 120

Sediment sample

FIGURE 4.35 Stratigraphic section of the E face of Unit
W250N200.

at the time of early Lapita occupation they supported
one or more pole structures situated - 10-20 m off-
shore. One post from W250N120 was radiocarbon
dated, yielding a conventional age of 2860 + 60 BP,
which calibrates to 1209-933 cal BC at la. These posts
correspond with a peak in the density of ceramics, ob-
sidian, oven stones, and other indicators of occupation
(see Cultural Content, below), indicating that they sup-
ported habitation structures. The ceramic assemblage is
characterized by a mix of fine and coarse-dentate
stamped sherds, as well as pedestal sherds with cut-
outs. In general, the ceramic assemblage associated with
the W250N1 10-120 posts is similar to that from Zones
C3-C2 of Area B, and the house(s) that stood along this
part of the W250 transect were probably a part of the
same settlement pattern as the Area B structure. Most
likely there was an array or zone of pole structures par-
alleling the original shoreline at a distance of 10-40 m
offshore.

The W250N130 through W250N150 units revealed
no posts, however another set of wooden post bases
occurs from N160 to N191.3? A total of five substan-

tial posts, each in vertical position, was exposed by this
set of units, while expansion of the W250N191 unit
into Area C (described below) exposed an additional
23 posts and stakes. These posts were set in coarse-
grained sands containing water-rolled branch coral
(Acropora sp.) along with many sand-burrowing marine
bivalves (Quidnipagus and other species in the family
Veneridae). These bivalves were not midden compo-
nents but rather occurred as sets of intact valves in up-
right death position, indicating that they had died in situ
in the sandy matrix surrounding the posts. These shells
indicate that the depositional environment at the time
the wooden post structures were constructed was con-
stantly sub-tidal.

The cultural materials that accumulated around the
wooden post bases in units W250N160 to N191 dis-
play another peak in density (see Cultural Content, be-
low). The high frequency of oven stones, obsidian, ce-
ramics, and other materials corresponding with the posts
thus indicates another set of habitation structures. The
nature of the ceramic assemblage, however, differs from
that associated with the structures in the N1 10-120 zone
described above. Here the ceramics are dominated by
thin-walled jars with everted rims, decorated primarily
by incising and by notching or pinching of the rims.
This pottery corresponds to the upper part of the Area
B sequence (Zone B). We thus interpret the pole or stilt-
house structure(s) in the area of W250N160-191 as a
somewhat later construction, probably after the origi-
nal shoreline had begun to prograde to the N. This in-
terpretation is also supported by the radiocarbon dates
from the W250 transect and from Area C (see below).

Cultural Content alon,g the W250 Transect

The W250 transect yielded a large collection of cultural
materials; patterns of distribution of sherds, midden,
oven stones, obsidian, and other materials along the
transect yield evidence significant for site interpretation.

The spatial distribution of several classes of arti-
facts exhibit similar patterns along the S-to-N course
of the 130-m long transect. In Figure 4.36 we show the
density distributions of undecorated sherds, lithic flakes
(primarily obsidian), and coral oven stones. All of these
materials exhibit a bi-modal distribution pattern, with
density peaks centered on Units W250N1 10-120 and
on Units W250N181-190. These peaks correspond to
the two sets of preserved wooden post bases that we
had interpreted as representing temporally distinct phases

L

1 13

THREE LAPITA VILLAGES

FIGURE 4.36 Distribution of

ceramic sherds, obsidian flakes, and
coral oven stones along the
W250 transect.

800-

Ceramic Sherds

'''''''''''''''''''''''''''''''''''''''''''''''''..'.........'''..........    ----------------------------

zo Z C , , , _ ~~~~~~~~~~~------------------

7C

50

40
30
20
10

0  o-   ,,------ ---------- -- - -- - -- - -- - -- -- - -- - -- - -- -

)0 .   ..                                    ..

0 -  ---   --    .       .      :.                                 .

Obsidian Flakes

...........................................................  ........................................

. . . . . . . . . . . .....  ...... ......       l...             . . .
......... ... _  ......   .............          _..__...

I I  I I .I I I I  I I I

12 -2
100

0.
80

40
20

100

90
80
70

Coral Oven Stones

,............................................................                      .......................................

....... ,..................................                       ....             .......    .................
............................................................                              ....................

.............................................. , .                                        .............................         ...

50   ., ....................................... ...............  ........-. . .......... ...

J0   s*@__Z-*Oo*?**s*svo _ .........  .................................  ...............................__*b****wwee

$0'.

20?#,vv ........*                   ww             w..*..............

0  .

0 ~  ~      ~     ~      --      -        ~                    ==

j

........ _    -_-_-_                                             _..__...

--------

,--------

--------

--------

gm     .                                                   W                      f

1

114

1 ,

? 9 C-0 ro- 2 2 i ? 9 9 08 9 2 llrl_

1- - - - - - - Z Z Z
Z Z Z Z Z Z Z Z Z Z Z

5

2

s' 0-0 rcl -, 3 2 ? i 9 9 11-1 8 9 9i C ,

T- -Ir- - r- - - - -- Z Z Z
Z Z Z Z Z Z Z Z Z Z Z

THREE LAPITA VILLAGES

of stilt-house construction and occupation. As would
be expected from an over-water platform habitation-
with midden and trash disposal directly off the plat-
forms into the surrounding shallow water-the density
of cultural materials is highest immediately around the
posts, tapering off to either side. These patterns are
strong evidence that the preserved post bases supported
dwelling structures, upon which domestic activities such
as oven use and food preparation, use of pottery, shell
artifact manufacture, and other tasks took place.

The distribution of anaerobically preserved plant
remains along the W250 transect units is shown in Fig-
ure 4.37. This distribution differs from those examined
above in having a single zone of high concentration,
between units N110-130, corresponding to the main
area of fine silt-clay sediment (the "muck zone") which
was so rich in organic remains. Unit W250N130 espe-
cially, produced a large assemblage of preserved seeds
and nuts, including large numbers of Canarium sp. al-
mond, Dracontomelon dao, Spondias dulcis, Pandanus spp.,

and Cocos nuczfera. All of this plant material accumulated
in the fine sediment around the stilt-house posts, pre-
sumably as cultural refuse of the house's occupants.

Figure 4.38 displays the distribution of worked shell
of Trochus niloticus and two large cones, Conus litteratus
and C. leopardus. These large marine gastropods pro-
vided raw material for the manufacture of fishhooks
(Trochus) and of several kinds of ornaments or exchange
valuables (Conus).31 While worked shell of both genera
is distributed throughout the length of the transect, it is
especially concentrated in the N100-120 section, again
associated with the earlier phase of stilt-houses.

The W250 ceramic assemblage also displays a
marked distribution along the course of the transect, as
seen in Figure 4.39. At the S end of the transect, sherds
with fine-dentate stamping dominate the ceramic as-
semblage. A considerable increase in the frequency of
coarse-dentate stamping occurs in the middle of the
transect, while the N end in the vicinity of the Area C
stilt-house, is dominated by sherds decorated with in-

FIGURE 4.37 Distribution of anaerobically preserved plant remains along the W250 transect.

1 400.

1 200

1000.

800.
0

600 -

400 _ -------------.-----*---.  - * _ - - _ _ _ _ - _ - - _ _ - _ -

200 4+.~~~~~~~~~~~~~~~~~~~~~..

0

II       - --~-

I

Ld9   -- 0 rl   0 0       -           -
-~~~~~~~~~~~~~~~      1

03)  03  0'  0

z     z  z   z

Area C

__1._

0

z

z3   r-    r   l l   z  z

OD   zl   z    z    z   zO

o o

z z

...............     ......................* w-- - w

b-* 4-         *  - .  ....  *- .  * .- . * . - . - . .  ....

**? . ..    .  .. ..  .  .. ..  .  .. ..  .  .. ..

?

o o

cZJ

z z

o 0

o 0'

z

z       ,

o 0
z z

r

................................................................ m

................... ................................. e...o .......0

.................................

................................................

1

900
. 0

. o ,

1

-i-

4

1 15

4

THREE LAPITA VILLAGES

1 801
1 60 -
1 40
1 20

3
0
(-

100

80

.                                               0

0           0                   -

CO   O ~   O D ~

oo  Tr   h        Ln

Z     Z     Z     Z

O  Trochus niloticus

.                          . . . . . . .  . .  . . . . . ..-1I R- ----

-----...                           ....................

0

z

0
(Y)

z

o     o     0     0      0     0

C-4J   -    0     0'1    Co    r~-

z     z      Z    Z           z

* Conus litteratus/leopardus

FIGURE 4.38 Distribution of worked shell along the W250 transect.

cised designs. The changing frequency distributions of
decorative technique and vessel form along the W250
transect correspond closely with the vertical, stratigraphic
sequence revealed by the Area B excavation, and con-
firm that a real temporal progression is being tracked
along the course of the W250 transect.

Radiocarbon Dating

Ten samples (four wood, six shell) from the W250
transect were submitted for radiocarbon dating; full
details are reported in Chapter 10. The conventional
14C ages are plotted on the transect diagram (Fig. 4.31).
A wooden post in W250N120 (Beta-30681) yielded a
calibrated age of 1209-933 cal BC. A wooden timber
lying horizontally in W250N140 (Beta-30682) dated to
1316-1129 cal BC, and a second horizontal timber in
W250N170 (Beta-30684) dated to 1490-1222 cal BC.
An upright post in W250N191 (Beta-30687) was dated
to 826-664 cal BC, which was confirmed by other dates
from the expanded Area C excavation, described be-
low. The calibration of six shell samples from the W250
transect depends upon the choice of a AR correction
factor, which for reasons to be discussed in Chapter 10,

is complicated. Full discussion of these shell dates is
therefore deferred to Chapter 10.

When all radiocarbon dates-including calibrated
shell samples-are taken into consideration, the radio-
carbon suite from the W250 transect suggests a chro-
nology of occupation commencing as early as - 1500
BC, with initial cultural activity focused on the paleobeach
terrace at the S end of the transect. One or more pole-
supported structures were constructed immediately sea-
ward of the mid-Holocene shoreline, and probably date
to the period - 1300-1000 BC. These structures were
thus contemporaneous with the Area B stilt house, pre-
sumably forming part of a single village. Following
abandonment of these structures-and progradation
of the shoreline to the N-a new set of pole structures
was erected near the N end of the W250 transect; these
date to an interval of - 1100-800 BC (see below for Area C
dates).

AREA C EXCAVATIONS (1988)

When Unit W250N190 was excavated into the water-
logged Ghyben-Herzberg lens, four wooden posts were

60

40 -
20

0 -

0
0
C\
z

0'
z

..w                                                                                                                                             -"r

-

.~ 1

a  .

0

1 16

. . . . . . . . . . .

THREE LAPITA VILLAGES

0

:3  - c

0

l 3

100

90
80
7 0

Co

0-

,F,. 6 0

0  50O

40
30
2 0
1 0

N: 6 9    N  6   60  N 1 ?   O  N 1     N 1 3 3   N 1

Transect Units

1

A:

D:D

d::: f:::
S:
::

-

1

5

N200  Nl90  N181  N 171  N170  Ni69  N' 6  N150  N14 0  N  N     NO    N100  N90   N 0

Transect Units

FIGURE 4.39 Distribution of primary ceramic decorative techniques along the W250 transect: A, sherd counts; B,

percentage frequencies. Note the clear transition from a predominance of fine dentate stamping in the S portion of
the transect (Area C).

exposed in deposits yielding thin-walled potsherds char-
acterized by notched rims and fine incised decoration.
Based on our knowledge of the ceramic sequence in
Area B (from the 1986 excavations), it appeared that
we had encountered a relatively late Lapita occupation,
post-dating the ceramic transition from primarily den-
tate-stamped ceramics to incised pottery. Given the con-
centration of posts (two of them quite massive), it also

seemed likely that there had been a substantial pole-house
structure in this locality. We therefore decided to open up a
number of contiguous units, designated Area CI?

Because the unconsolidated sands in Area C were
very loose and continually threatened to collapse as we
penetrated into the waterlogged levels, it was not pos-
sible to expose a large area with long sidewalls. Area C

thus consists of two 4 m2 blocks separated by a 1-m

1

dom?

11 7

N 9 0    N S

THREE LAPITA VILLAGES

baulk, as shown in the plan (Fig. 4.40). Furthermore,
because of the continual danger of sidewall collapse
(which indeed occurred several times), excavation and
pumping had to proceed rapidly. Although these were
less than ideal conditions in which to excavate, we man-
aged to take all eight units down to 1.25 m bs, at which
depth basal reef detritus was exposed.

Area C Stratigraphy

In Area C the stratigraphic section was similar to that
described for Unit W250N200 (see above), with a dark
brown A horizon formed under the heavy forest cover,
overlying light gray to white calcareous sands. Between
20-60 cm bs the sands were weakly cemented, but be-
came loose and unconsolidated once the Ghyben-
Herzberg aquifer was reached. The excavation of
W250N190 revealed that the deposits between the sur-
face and - 60 cm bs were essentially culturally sterile.
Therefore, in the expansion units comprising Area C,
these upper sands were removed by shoveling and were
only cursorily sieved to ensure that no cultural materials
were missed. Below 60 cm, excavation proceeded ac-
cording to standard procedures.

As there were no visible lithological, textural, or color
distinctions in the calcareous sandy deposit, excavation
was conducted by arbitrary levels. During excavation it
appeared that the levels between - 60-80 cm bs yielded
a high density of obsidian flakes (many of large size),
along with incised and notched-rim pottery and shell
midden. From 80-100 cm bs the density of cultural
materials decreased notably. Then from 100-120 cm
there was another concentration of cultural materials
including incised and notched rim pottery (along with a
few classic dentate-stamped Lapita sherds), along with
large midden shells clustered around the bases of the
many wooden posts and stakes. This lower zone also
included a high density of anaerobically preserved plant
materials. For analytical purposes, we have divided the
excavation levels into four zones, as follows: Zone A,
the sterile upper sands to 60 cm; Zone B, from 60-80
cm; Zone C, from 80-100 cm; and, Zone D, from 100-
125 cm (Table 4.4).
Wooden Posts

A total of 20 posts or stakes was exposed in the eight
excavation units comprising Area C. Data on these posts
are provided in Table 4.2. Although no alignment or
patterning could be detected in the limited area exposed,

several of the posts are quite large, with diameters of
15-20 cm. These certainly were structural timbers de-
signed to support some kind of pole house. Two of
these posts were radiocarbon dated (see below), with a
difference in age of 250 years, an age differential matched
by the age range in three shell dates from Area C. One
interpretation is that the posts exposed in the Area C
blocks may have been cut and set in position at differ-
ent times. Given that the vertical distribution of cultural
materials observed during excavation suggested two
phases of deposition, it may be that the posts also rep-
resent two phases of stilt-house construction.
Cultural Content

The vertical distribution of cultural materials in analyti-
cal zones B, C, and D reveals significant differences.
The basal zone D is marked by a high frequency of
anaerobically preserved plant remains (886 itemized
specimens, not including Cocos nucifera endocarp and
wood fragments), and by high densities of both verte-
brate and invertebrate faunal materials (Fig. 4.41). Zone
B, on the other hand, shows the highest densities of
ceramics (more than twice the total in Zones C and D
combined), obsidian flakes, and coral oven stones. These
differences suggest that the activities occurring on the
stilt-house structures in the vicinity of Area C may have
changed over time. The earlier phase is marked largely
by midden dumping, whereas the later phase is repre-
sented by a broader range of artifacts indicative of
pottery use, shell artifact manufacture, lithic use, and
cooking (oven stones).

As a whole, the ceramic assemblage from Area C
(all zones) is fairly uniform, and is marked overall by a
high percentage of thin-walled sherds, deriving prima-
rily from large jars with constricted orifices and everted
rims. The rim lips are usually decorated with fine tool
notching or finger-pinching (crenate), and red slip is
common, especially on the upper parts of the vessels.
Fine incised decorations also dominate. Despite the
overall consistency of the assemblage, some differences
between zones appear to be significant. For example,
whereas jars predominate throughout the Area C se-
quence, a few bowls are also represented in the deeper
zones C and D. The only two sherds with dentate-stamp-
ing from Area C were both recovered in Zone C, and
both were water-rolled and eroded, in sharp contrast
with the non-eroded incised and notched-rim sherds
from the same zone. This suggests that the dentate sherds

1 18

THREE LAPITA VILLAGES

N

Grid N

0                            im
W249

N1 88
N1 87

FIGURE 4.40 Plan of the 1988 Area C excavations, showing the positions of wooden posts and
stakes, and other features.

1

W251

W250

N1 91

-WC1 5
* C17                            C14

De      4*C16                *C13         Igneous

-- v                           ~~~~~~~~~~~~~stonet

*C12

Coconut

C18      endocarps            ed

^C 1

Horizontal
timber

C23    C20
C24

LC9

Cassis rufa

.         C19
C24     C8
C22                 ,

N1 90

W250

*C4

?C1                             *C5

Tndacna       ~~~Large Cassis     CD6

T na        C     a crnuta shells  Coconut

,gi,gas                             endocarp

(  Tridacna      C7
(:   n   U maxima shell

Hyotissa
shell

1

1

11 9

THREE LAPITA VILLAGES

had been deposited on the reef platform and had been
subjected to water action for some period prior to the
deposition of the incised pottery. While the Area C pottery
is almost exclusively decorated with incised designs, it is
noteworthy that the motifi displayed with the incised tech-
nique are in most cases the same as those on the den-
tate-stamped pottery from other parts of the site. Two
examples of shell-impressed decoration are also repre-
sented in Area C, both from the upper Zone B context.

The Area C pottery assemblage displays close simi-
larities with that from Site EKQ on the NE part of
Mussau Island. Both assemblages appear to represent a
late phase in the local evolution of Lapita ceramics,
marked by a reduction in the range of vessel forms
(almost exclusively confined to large, thin-walled jars),
and by the restriction of decorative technique to incis-
ing. These changes will be fully described in Volume mI.
Radiocarbon Dates

Three shell and two wooden post samples from Area
C were submitted for radiocarbon dating; details are
presented in Chapter 10. The two posts yielded cali-
brated ages of 1209-924 cal BC (Beta-30686) and 826-
664 cal BC (Beta-30687), suggesting that more than one
phase of pole structure construction may be represented.
Calibration of the shell sample dates, which depends
upon the AR correction factor, is discussed in Chapter
10. For reasons given there, our best overall estimate
for the occupation period represented by the Area C
assemblage is - 1100-800 cal BC.

SEDIMENTOt OGY OF THE TALEPAKEMALAI SITE

With the discovery in 1985 that the ECA Site had both
paleobeach terrace as well as lagoonal depositional en-
vironments (the latter now containing the anaerobically
preserved stilt-house remains), it became necessary to
understand in some detail the site's complex and dy-
namic geomorphological history. In part, this history
could be reconstructed through our field observations,
including paleo-shoreline indicators as weil as the stratig-
raphy of excavated units, especially those along the W200
and W250 transects. An additional source of data po-
tentially bearing on depositional processes was that to
be obtained from laboratory analysis of sediment
samples collected from key stratigraphic layers. Such
samples, which could be compared with modern con-
trol samples taken from active lagoon and seaward

TABLE 4.4 Summary of cultural content of Area C,

Site ECA (by analytic zones).

CLASS           B     C     D    TOTAL
Ceramics

Plain Body Sherds     522   209   202    933
Decorated Body Sherds   7    24    15     46
Plain Rim Sherds        5          3       8
Decorated Rim Sherds   35    13    12     60
Non-Ceramic Artifacts

Obsidian Flakes       127    41    38    206
Abrader                 2                  2
Fishhooks               1     1            2
Fishhook Preform        1                  1
Coral Oven Stones        80     6    13     99
ManuportVolcanic Stone    1     1    1       3

beaches, may yield information on the mode of depo-
sition and critical energy levels at the time that the ECA
deposits were laid down. For this purpose, 31 archaeo-
logical sediment samples were collected over the course
of the three field seasons, along with 16 modern beach
and reef/lagoon flat control samples. Laboratory analyses
were carried out largely by Melinda S. Allen in the Uni-
versity of Washington geoarchaeology laboratory, fol-
lowing a protocol we jointly developed.33

FIELD AND LABORATORY METHODS

Field sampling of archaeological sediments followed
methods outlined by Stein (1985), with samples taken
from discrete depositional layers (i.e., not crossing strati-
graphic boundaries). Samples were collected into ster-
ile, heavy polyethylene bags after the stratigraphic sec-
tion of particular excavation unit had been drawn and
interpreted; sample locations were indicated by depth
and notated on the stratigraphic drawings.

Modern Control Samples

Four sets of modern control samples were collected,
all from Eloaua Island. (1) In 1985 we obtained general
samples (GS-1 to -3) from the foot of a seaward beach,
from the top of a lagoon beach, and from the foot of
a lagoon beach, respectively. (2) In 1986, we collected
four samples in conjunction with a measured transect
across the seaward reef flat in front of the Talubagalim

120

THREE LAPITA VILLAGES

A           ,         ,       A) Ceramics
c

D

0        100       200      300       400       500       600

Count

C) Bone (NISP)

A

8)

r0 t

A

8)

L
r-

0 C
N

c

B

(3

40

60     80

Count

100     120      140

N)
0
N

0       100       200      300

Count

400       500       600

Count

A           E) Plant remains

8

c

*I            I

0    100   200    300    400   500    600    700

Count

800

A

F) Coral oven stones .

8)

0
N

900

Count

FIGURE 4.41 Distribution of selected categories of cultural materials, by analytic zones, in Area C: (A) decorated and
plain ceramics; (B) lithic flakes; (C) bone (NISP); (D) molluscs (NISP); (E) plant remains; (F) coral oven stones.

School, located due W of the ECA Site and fronting
the Malle Channel. These samples (designated RT-1 to -
4) were obtained from the seaward beach slope (RT-1),
the "muck zone" with considerable organic materials
15 m from the beach (RT-2), the "seagrass zone" 50 m
from the beach (RT-3), and the outer zone marked by
the presence of much dead branch coral, 80 m from
the beach (RT-4). (3) In 1988, we took four samples
along the slope of the seaward beach in front of Ave
Male's hamlet, with the elevation of samples above the
reef flat precisely measured by Leitz telescopic level.

The positions of these samples (designated SB-1 to -4),
and the topographic profile of this seaward beach, are
shown in Figure 4.42A. (5) Also in 1988, we took five
samples along the modern Eloaua lagoon beach on the
W side of the island near Site ECA. As with the previ-
ous sample set, these lagoon beach samples (designated
LB-1 to -5) were precisely plotted along a topographic
transect, as depicted in Figure 4.42B. These control
samples, from a seaward beach, seaward reef flat, and
lagoon beach, provide critical comparative data for the
interpretation of the archaeological sediment samples.

N)
0
N

:

B) Lithic flakes

. .

: :

. .
. .

: :

8)
0
0
N

IG              %4                e                 4        -- -- -.

B

s . . . . . .

121

C

.I

w

122                                      ~~~~~~~~~~~~~THREE LAPITA VILLAGES

0

Coarse-      CsatFine

graned          beach -*1--l - silt/                     zon- ears
0.5  ~ ~ ~~~~>sand                       front          clay                      zn

beach                       mc

1.0

1.5

2.0-

?SEAWARDB13E A CHQ.                                                 ..*. ~. . . ...

2.5

0                          10                          20                         30m

0

0.5             ?        Beach ________Silty sand

slope                  platform

1.0

1.5    . . . . . . . . . .

Low

2.5 -.

0                          10                          20                          30 m

Q    Sediment sample

FIGURE 4.42 Topographic profiles of modern Eloaua Island beaches, showing the locations of control

sediment samples: (A) seaward beach and location of samples SB-1 to -4; (B) lagoon beach and Iocation
of samples LB-1 to -5.

122

THREE LAPITA VILLAGES

Anattical Methods

Both archaeological samples and modern control
samples were subjected to the same analytical protocol
in the laboratory. Sediment color, which may be useful
for indicating the presence of organics, iron, or gleyed
properties, was described using the Munsell system (both
moist and dry determinations were made). Hydrogen
ion content (pH), which indicates relative acidity or al-
kalinity, was determined using an automatic Altex 70
pH meter, following procedures outlined by jackson
(1958). Samples were prepared by adding distilled wa-
ter to 20 ml of sediment, allowing the sample to stabi-
lize for 1 hour, followed by taking a minimum of three
pH readings (reported values were calculated as aver-
ages of these readings). pH readings were taken for the
1985 samples only.

The following pretreatment procedure was used
to remove organic matter which might adversely influ-
ence grain size analysis. This procedure is based onJack-
son (1958), with the substitution of sodium hydrochlorite
for hydrogen peroxide (the former is both less expen-
sive and safer to use). The sodium hydrochlorite (com-
mercial chlorox beach) was diluted with hydrochloric
acid to a pH of 9.5, to avoid the possibility of diges-
tion of clay minerals. Sediment samples were immersed
in the dilute solution, and kept over low heat in a fume
hood for several days, with occasional decanting and
centrifuging the liquid portion, and addition of fresh
sodium hydrochlorite. This procedure was continued
until all organics had been removed; samples were then
rinsed in distilled water until all sodium hydrochlorite
was removed.

After this pretreatment was completed, samples
were oven dried at 100?C. A solution of sodium
hexametaphosphate (a dispersing agent) and distilled
water was added to the dried sediment, flasks were
mechanically shaken for 5 minutes, and then allowed to
sit overnight to insure that flocculation was not a prob-
lem. Samples were then wet screened using a 4f geo-
logic sieve to separate the sand fraction from the silt
and clay fractions. The sand fractions were oven dried
and mechanically shaken for 15 minutes through nested
geologic sieves with mesh sizes of -2, -1, 0, 1, 2, 3, and
4Q. The resulting sand fractions were weighed on an
analytic balance to the nearest 1000th g.

The silt-clay fraction was placed in 1000 ml cylin-
ders and sufficient sodium hexametaphosphate solu-
tion was added to fill the cylinders. The pipette method

(Shackley 1975) was used to determine grain size for
this finer fraction. Timed pipette withdrawals were
placed in premeasured beakers, oven dried, and weighed
to the nearest 1000th g.34

Statistical analysis of grain size data were performed
using a computer program developed by the Univer-
sity of Washington geoarchaeology laboratory, which
calculated and plotted frequency and cumulative per-
centage plots by 0 size, as well as providing summary
textural designations, sample kurtosis, and sorting char-
acteristics (based on Folk 1974).

ANALYSIS OF MODERN CONTROL SAMPLES

Analytic data on the four sets of modern control samples
are presented in Table 4.5, and frequency and cumula-
tive plots of grain size distributions for these samples
are shown in Figure 4.43.

General Sediment Samples

The three 1985 control samples are fairly similar in many
characteristics (e.g., white color, alkaline pH averaging
8.4), and in textural terms can be described as gravelly
or slightly gravelly sands, with moderate to very poor
sorting. The coarsest sample is that from the foot of
the lagoon beach which was very poorly sorted and
had the highest percentage of gravel (12.03%).

Eloaua Reef Flat Transect Samples

The seaward reef flat transect samples collected in 1986
provide a series extending from the seaward beach out
across the reef flat, a distance of 80 m. These samples
show much more variation than in the general sediment
samples from 1985. For example, sample RT-2 from
the modern "muck zone" is characterized by relatively
higher quantities of silt and clay, and is a gravelly muddy
sand (the organic content is also reflected in its light
gray color). Sample RT-3, from the seagrass zone, lacks
any gravel-sized component and is largely made up of
silt and clay. Sample RT-4, from the zone of dead coral
near the active reef margin, has the highest gravel com-
ponent (25.12%) of any sample, although it also has
significant quantities of silt and clay; not surprisingly,
then, it is very poorly sorted.

Seaward Beach Samples

The 1988 seaward beach samples provide data on the
variation across an active seaward beach front. Textur-

123

THREE LAPITA VILLAGES

ally, all four samples are gravelly sands, and all of their
distributions are marked by leptokurtic kurtosis. The
greatest variation is between the sediment at the foot
of the beach, which contains the highest percentage of
gravel (16.46%) and very little silt-clay, and the sediment
from the sandy flat fronting the beach, which has the
lowest amount of gravel (5.69%) and the most silt-clay
(5.51%).

Lagoon Beach Samples

The five 1988 lagoon beach and adjacent sand flat se-
ries provides data on variation across an active lagoon
shore. This series differs from the seaward beach series
in the overall lower quantity of gravel, and higher quan-
tity of fine particles. Thus only the sample from the
foot of the lagoon beach (LB-3) is characterized as a
gravelly sand, while those samples (LB-1 and -2) from
the higher beach slopes are slightly gravelly sands. In
contrast, the samples from the lagoon flat fronting the
beach are slightly gravelly muddy sands, with extremely
little gravel and reasonably high quantities of silt-clay
(11.76 and 15.13%).

Although the range of variation in the modern con-
trol samples is not extreme, there are sufficient differ-
ences to be able to distinguish between the depositional
environments of seaward and lagoon beaches and their
adjacent sandy flats. Seaward beach sediments are char-
acterized by greater amounts of larger clastics (espe-
cially in the sand and gravel size ranges) than lagoon
beach sediments, reflecting the higher energy levels as-
sociated with wave action along the seaward coastlines.
Also characteristic are the sediments of the seagrass flats
with their very high silt-clay fractions, and those of the
dead coral outer reef flat zone with the highest per-
centages of gravel (and very poor sorting overall). The
analyses of our Eloaua control samples presented above
also fit well with the range of variation in sediment sizes
described for other atoll systems, such as that of Ifaluk
described by Tracey et al. (1961; see also Wiens 1962).
They provide a set of modern analogues that may be
used to interpret the archaeological samples from Site
ECA.

ANALYSIS OF ARCHAEOLOGICAL SEDIMENTS

The archaeological sediment samples from Site ECA
were taken from a series of excavation units, both along
the W200 transect line (sampled in 1985 and 1986), and

along the W250 transect line (sampled in 1988). In in-
terpreting these samples, it must first be recalled that
both the W200 and W250 transects are oriented ap-
proximately N-S and are perpendicular to the orienta-
tion of the slope up to the paleobeach ridge (1.75-2.35
m asl) which is interpreted as the shoreline at the time
that the ECA Site was first occupied. Thus as one
progresses from S to N along either transect, one moves
from the paleobeach ridge down across the former
shoreline feature, and out onto what was formerly a
sandy flat. At the same time, it must also be kept in
mind that there is a time-transgressive facies within each
vertical excavation unit, which will reflect the dynamic
changes occurring over time at each micro-locality.
The W200 Transect Series

Analytic data for 13 samples collected along the W200
transect are given in Table 4.6, and selected frequency
and cumulative distribution plots are provided in Fig-
ure 4.44. In Table 4.6, the samples are arranged by ex-
cavation unit and by stratum, proceeding from S to N
along the W200 transect.

The first three samples are from Unit W200N100,
which was excavated into the crest of the paleobeach
terrace (elevation 2.01 m asl). Here the deeper Layer II
sediment, with 16.49% gravel but very little silt or clay
(and hence a gravelly sand), is a good match with a
modern seaward beach sediment (such as control sample
SB-3), consistent with our interpretation of this deposit
as the active beach zone at the time of deposition. The
upper deposits in W200N100, especially Layer IA, have
a much greater percentage of fine particles, reflecting
the later development of an organic soil on top of the
paleobeach ridge (note also the dark color and lower
pH). Layer IA is essentially the modern garden soil cul-
tivated by the Eloaua villagers for sweet potatoes and
manioc.

The next set of samples come from Layer II of
Unit W200N110 (TP19), at 50 cm bs (upper) and 90
cm bs (lower). This excavation unit is situated near the
foreslope of the former paleobeach terrace, and the
sediments are again quite coarse, reflecting a fairly ac-
tive zone of beach deposition. The lower sample, in
particular, is characterized by a high percentage of gravel
(46.30%), indeed higher than that in any of the modern
control samples (note that it does not have as much silt-
clay, however, as in the modern "dead coral" zone
sample, RT-4). This may reflect shoreline conditions

124

THREE LAPITA VILLAGES

125

-CI Xi- Xi       1- S   i  -
I  | ^  2  2~~~~~~~~~r

0f >e  >  >

u ~ ~ ~ ~~~~~~~~~~~~ ~  ~    ~~~~~~~ c Q  O_  c, f ,4

o  | 12  Q  @ ~~~~~~~~E  E 3@Q  E 3E  E 3?E  E  E E - ol

E~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ E M

w5 t c~~~~~~~~~~~~~~~~~~~~~~~o  bo  b o t  o  F   o @c

E~~~~~~~~~~c  us  ci o~  o        ci   ci oi o- oi t  oo

2~~~~~~~~~~~~~~~~~~~~~%  OD  (I  X0oN  C 4  0  F   u

lx             10              c. )  0C Cl U ,.0

Q ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~ 0  0  CY       ei

> OD g    wO       rl  0  ci

e~~~~~~~L  ur  Co  O  0O  r) ur    0s O D        00Cr) %O

J~~~~~~~~, 0              --                        lo

0%  00  _    _  CY  O  e     C o  CoO%  e o  w  0
X U~~~~~C)  o  F  0  C4  %0  V)  U)  O% 0  ONC)

?~~~~~~~~~~~,  I 2  8  ? r  C N  ^ 0?  qe  %  ?O r_  e
? becq  CQ  oi  cq  o;  o  ui  1o  K  16  ui ?i  ?

@~~~~~~~~~~~~~~~~~~~~~~ E

3~~~~~c ?  ) 2 C}        O m   O  m

THREE LAPITA VILLAGES

80

70 r

60 .

Eloaua
SB-1

8)

D

0E

-0

.5

._

-F

-2   -1    0     1    2     3     4    5

4 intervals

Eoaua
SB-4

1-

.*

50
45
40
35
30
25
20
15
10

5
0

Eloaua
SB-2

8)

E

8)

en

-0

.5

:_

-O

-2    -1     0      1     2

4 intervals

60 r

50

8)

E

U 40

8)

en

8 30

> 20

-   10 .

o

-2   1   0    1   2    3    4    5

4 intervals

70
60
50
40

/1

Eoaua
SB-3

30 .

20 .

10
0

7/  X

3    4   5         -2  1    0   1    2   3

4) intervals

Eloaua
LB-2 r

8)

E

8C)

(1)

n

F0

.5

-O

>0

30 r

Eloaua
LB-3

I' \
i\,

25 F

20 p

15 P

10 F

5

0

-2    -1      0      1     2      3

4 intervals

4    5           -2   1     0    1    2     3    4

4 intervals

Eloaua
LB-4

60 r

50 .

8)
E

7:P 40 .

8)

8& 30 .
.X5 20 .
C0

r- 10

0

-2   -1    0     1    2     3     4    5

4 intervals

Eloaua
LB-5

-2     1     0      1     2

4 intervals

3   4    5

FIGURE 4.43 Representative plots of
sediment grain size distributions

from modern control samples, from
the seaward beach (SB1 -4), and
lagoon beach (LB2-5), of Eloaua
Island.

during the mid-Holocene higher sea-level stand, when
the reef flat fronting the ECA Site would have been
exposed to a slightly higher water level and greater wave
energy at high tide, than is the case at present.

Another set of four samples were taken from the
face of Unit W200N147, part of the 1986 Area B ex-
cavations, and are representative of the sequence of
sediments within the Area B locality. Layer 1, a gravelly
muddy sand, is the organic, reworked garden soil which
caps the area today (note the very dark gray color). Layer

II, a gravelly muddy sand but with less gravel-sized par-
ticles and a lighter gray color, shows a grain-size distri-
bution characteristic of an active beach zone (similar to
control sample GS-2). The two samples from Layer III
(the waterlogged zone of Area B) are both gravelly
muddy sands that are very poorly sorted, with a con-
siderable range in particle sizes from gravels through
silt-clays. The Layer IIIA sample is a good match with
control sample RT-2, from the modern "muck zone"
in front of Eloaua village, while the IIIB sample is a

E

-_

V

8)

-o

.F5

:-

-0

50
40
30
20

10
0

30 r

4    5

4-. 25

8)

E 20

8)

(n

bR 15

FD

.p 10

- 5

0

70
60
50
40
30
20
10

4

._

8)
E

-O

8)
u)

>

-0

-5

r_

5

0

1:

?

i
q

,E cI

127

THREE LAPITA VILLAGES

,S D 1 {E D!  Q>'SED- , X5  :2 _  )  -F:  D

ei  (4 3i ci  0  >    4

O,   00> C          0>>   .

z~~~~~~

Z      - .   r--

o   OO @ 8     8 o o ? 8 o       ? 8 8

10  Ac> '                  >

_         m            O>0  0>O 1,  (
F~~~~~2 i>r     t iwwr

4-

0~~~~

in ooca  o0>          0       >   w       O

8       ~~   ~     ~ . r_ c c  c c     r

_q         0  0  (  0  0  0  0 0  0   0   0

0C                   C1   Cv  Ci  ?)

O   X  g  w  o ~bo  No

E o  CO     0            0   00

6~0     0.     >  0.  0.  0.  0.0  0.  0.0  0. q

E

~~~~            e  r,  N v C,4  r,   s  O  S0
(0

0          lY   LO N0 U)i Os  C0 C) 0 C  1   N

4-  11       Co Cnb ? ? ? X o?   O   Oo  U)<) C

O)  O O   OO F F F F O       OO C
cac

(/)                    0  ?0       ?1   O

- ~ ~ o_ o     o b  o 8  o  o  o oe  O.'  eO O O

70

Co~~~~C

Co   ~~~~~  COo (V)       r  C

_          O~~~%  CM-  CvO  %0' C ? 0 M?  00'to  C

c' 04   Q  r  Mi4  r C  )   vCv

(0 ~ ~  ~ ~~ cvi        -

(v) 80   C)  0  C)  0O  v)  00 Cv
?- 0 a.  Y 00 cy   Mi a   C- ?M  1'-  V-'

w ~ ~ 3  l3  3.b :t  3     :31 3.  3  : 3

1

THREE LAPITA VILLAGES

ECA, Layer IA
W200 N1 00

35 r

F, 30
8)

E 25

8)

Jn 20

-F 1 5
:3
-0

.? 10
.2 0

5

0

-2   0    2   4   6    8   10   12

+ intervals

ECA, Layer IB
W200 N 1 00

-2   0    2   4    6    8   10  12

+ intervals

40
35

) 30

E

D 25
8)
(1)

& 20

3 15

.5_

-o 10
c:

5
0

ECA, Layer II
W200 N1 00

-2   0    2   4    6    8   10  12

+ intervals

25

r 20
8)

E

8)  1 5
(11

3 10
-o

.5
-o

0

ECA-B, S face,

Layer lll (upper)
W200 N1 47

-2   0   2    4   6    8   10  1 2       -2  0    2   4    6    8

+ intervals                              f intervals

10   12

25 r

- 20
8)

E

8) 15
cn

3 10
7U

55
-o

C 5

0

ECA-B, S face,
Layer lll (lower)
W200 N1 47

-2   0    2    4    6    8   10   12

j intervals

ECA-T,

Layer II

W250 N200

2    4   6    8   10

1 intervals

30

a 25

. _

8)
E

0 20
8)

(n

- 15

:3
70

*5 10

5

0

ECA-T,

Layer lll

W250 N200

.*

1-

12        -2   0   2    4    6   8

4 intervals

10   12

FIGURE 4.44 Representative plots of

sediment grain size distributions from
archaeological sediment samples,
ECA Site.

relatively close fit with RT-4, the "dead coral" zone.
Indeed, these samples reinforce our stratigraphic inter-
pretation that the Area B stilt house was originally con-
structed over an open reef flat (marked by the presence
of much dead Acropora branch coral). Over time there
was a facies change as the "muck zone" migrated sea-
ward, probably in consort with a gradual seaward (S to
N) progression of the active beach front. Thus the
samples from W200N147 show a local micro-facies
progression over time from (1) open reef flat, to (2) a

?muck zone" characterized by a higher percentage of

fine particles (incorporating organics and anaerobically
preserved plant remains), to (3) an actively prograding
beach slope, and finally to (4) an organically enriched
soil developing on the old beach terrace.

The final set of three samples comes from
W200N170, some 20 m N of the Area B excavations.
Here again the transition from Layer III to Layer II is
best interpreted as a shift over time from a more ex-
posed, open reef flat (Layer III) to a prograding beach
deposit (Layer II), and eventually to the formation of
an organically enriched soil deposit (Layer I), correlat-

11
10
8)9

8

(n 7

3

2p

_g  6

3
5

r_4

3
2

25 r

ECA-B, S face,
Layer 1

W200 N1 47

C 20
8)

E

?1 5

3 10
-5

. 5

0

F

30 r

25 .

8)

E

70
8)
O

:3

-O

.r

.2

20 1

15 F

10 1

5

0

-2   0

-   -                         .  --- ?----i

- - s - - s -

2

s   *   E ~~~~~~~~~~~~~~~~~~~~~~~~

- - - - - -

tr-
1

128

f\

1

THREE LAPITA VILLAGES

ing with the gradual S to N progradation of the shore-
line through this locality.

The W250 Transect Series

Analytic data for 17 samples collected along the W250
transect are given in Table 4.7, and selected frequency
and cumulative distribution plots are provided in Fig-
ure 4.44. In Table 4.7, the samples are arranged by ex-
cavation unit and by stratum, proceeding from S to N
along the W250 transect.

Unit W250N80 is situated on the paleobeach ter-
race, S of the slope representing the former shoreline.
Layer II, a structureless sand deposit, displays sediment
characteristics that compare favorably with a lagoon
beach sediment (such as control sample GS-2). Some
30 m to the N, Unit W250N1 10 is immediately N of
the former shoreline feature. The deepest stratum, Layer

IV (- 120-170 cm bs), was described in the field as a
"fine silt-clay with anaerobically preserved plant remains
and upright [in situ] posts." Thus it is not surprising that
the sediment sample from Layer IV is a gravelly
muddy sand with characteristics not unlike the mod-
ern ?muck zone" control sample (RT-2). Layer III
(- 100-120 cm bs) is similar, with a slightly lower fre-
quency of gravel-sized particles. Layer II (- 25-100 cm
bs), however, displays quite a different set of sediment
characteristics, being a gravelly sand with a much lower
frequency of silt-clay. Thus in the W250N1 10 unit one
can see a facies change over time from a typical "muck
zone," which would have fronted the original foreshore
when this was perhaps 20-30 m to the S, to a beach
foreshore at a slightly later point in time. This is in keep-
ing with our geomorphological interpretation of a S to
N progradation of the shoreline along the axis of the

TABLE 4.7 Analytical data for archaeological sediment samples from the W250 Transect, Site ECA.

Unit      Layer    Gravel  Sand    Silt    X     SD    Textural  Sorting  Skewness    Kurtosis

%       %      %                   Class

W250N80    II          12.36  86.89   0.75   -0.01  1.01    gS     poor    coarse      leptokurtic
W250N110   II          14.39  81.19   4.42   -0.20  1.18    gS     poor     stronglyflne  very

leptokurdc
W250N110   lll         9.00   72.18   13.14  1.04   2.20   gmS     very    strongly fine  platykurtc

poor

W250N1 10  IV          16.63  69.10   14.26  1.15   2.30    gmS    very    strongly fine  platykurtic

poor

W250N1 30  lll        13.17   84.23   2.60   -0.05  1.34    gS     poor     strongly fine  very

leptokurtc
W250N1 30  IX          3.98   75.60   20.42  1.91   1.99   (g)mS   poor     near        platykurtlc

symmetrlcal

W250N140   lll         0.31   51.62   48.07  3.67   1.15   (g)mS   poor     strongly   leptokurtlc

coarse

W250N1 40  IV          14.03  74.27   11.70  1.08   2.10    gmS    very    fine         platykurtlc

poor

W250N1 50  II         15.07   73.78   11.15  0.82   2.11    gmS    very    strongly fine  mesokurtlc

poor

W250N150   IIIA       12.09   71.06   16.85  1.15   2.21    gmS    very    strongly fine  very

poor                 patykurtc
W250N150   IIIC        3.25   86.20   10.55  1.21   1.77   (g)ms   poor     strongly fine  mesokurtc
W250N1 50  IV          19.35  72.68   7.97   0.49   1.88    gS     poor     strongly fine  mesokurlc
W250N170   111         2.65   66.13   31.22  2.64   1.95   (g)mS   poor     strongly    platykurtic

coarse

W250N1 70  IV          15.74  70.87   13.39  1.21   2.19    gmS    very    fine         platykurtic

poor

W250N190    80-85 cm   9.11   81.55   9.35   1.23   2.00    gmS    poor     fine        platykurtlc
W250N200   II          7.68   90.49   1.83   0.79   1.48    gS     poor     fine        mesokurtlc
W250N200   lll         5.82   84.64   9.53   1.20   1.88   gmS     poor     strongly fine  platykurtle

129

THREE LAPITA VILLAGES

W250 transect.

A similar time transgressive change can be detected
in the samples from Unit W250N130. The deeper Layer
IX sample (- 150-180 cm bs) represents a slightly
gravelly muddy sand with considerable silt-clay content,
whereas the higher Layer HI sample ( - 35 cm bs) shows
a change to a gravelly sand with very little silt and a
higher gravel content, typical of a beach slope (com-
pare control samples SB-3 and LB-3). Thus again, we
see a temporal shift from reef flat to beach foreshore.

In Unit W250N140, the two analyzed samples
show a temporal progression from a gravelly muddy
sand in Layer IV (- 110-150 cm bs), to a slightly gravelly
muddy sand with much higher silt-clay content in over-
lying Layer III (- 60-110 cm bs). This is then overlain
by Layer II, described in the field as a "structureless
massive deposit of coarse-grained calcareous sand"
(Layer II was not sampled for laboratory analysis). The
W250N140 temporal progression would then seem to
reflect the gradual "migration" of a "muck zone" along
a S to N course, with Layer III representing this zone
when it was concentrated in the vicinity of this unit, and
Layer II representing the beach foreshore which even-
tually replaced and covered it over.

Moving farther to the N along the transect, we took
a set of four samples from the W250N150 pit, which
perhaps even more clearly than in W250N140 show the
temporal migration of the "muck zone" along a S to N
track, as reflected in the micro-facies of this unit. The
deepest deposit exposed in excavation, Layer IV (- 100-
120 cm bs), was described in the field as a coarse-grained
sand "containing numerous water-rolledAcropora branch
coral fingers," and as having Quidnipagus and Veneridae
bivalves in death position. This description is matched
by the sediment texture with a high percentage (19.35%)
of gravel, indeed the highest percentage of gravel from
any sediment sample taken along the W250 transect.
The closest analog here would seem to be the "dead
coral" zone from the seaward reef transect (control
sample RT-4). Both our field description and the ana-
lytical data strongly suggest that Layer IV represents the
original reef flat. Layer IIIC (- 60-100 cm bs), which
was subsequently deposited over the Layer IV reef flat,
has a significantly reduced gravel component (only
3.25%) and increased silt-clay content. In the field we
noted that this stratum contained anaerobically preserved
plant remains, and thus represents the onset of the "muck
zone" as it moved gradually from S to N. Layer IIIA

( - 40-60 cm bs) is an even more poorly sorted sedi-
ment, with a greater range of gravels to silt-clays. This is
then capped by Layer II (.- 20-40 cm bs) which while
also very poorly sorted, shows a shift towards more
gravel and less silt-clay fraction.

With Unit W250N170 one is now - 60 m N of the
original position of the shoreline associated with the
mid-Holocene higher sea-level stand. Here we sampled
the two deepest strata, which once again show the pro-
gression of the muck zone over the underlying reef
flat. Layer IV ( - 90-150 cm bs), a very poorly sorted,
gravelly muddy sand, represents the original reef flat in
this locality. Layer III (- 75-90 cm bs), which is quite
thin in this unit and which contained anaerobically pre-
served wood, coconuts, and other plant remains, shows
averyhigh silt-clay content (31.22%) andvery little gravel.
In these characteristics, it seems most similar to the
seagrass zone of the seaward reef transect (control
sample RT-3) and may indeed have been a patch of
reef flat dominated by seagrasses. This was later cov-
ered by a thick deposit of structureless coarse-grained
sand (Layer II, not analyzed), which represents the final
stage of beach progradation.

The final three samples from the W250 transect are
from Area C and from a unit 10 m N of Area C. These
are all either gravelly muddy sands or gravelly sands,
with characteristics not unlike the modern sand flat
samples, especially that from the seaward beach transect
(sample SB-4). This suggests that the Area C vicinity
was a sandy, tidally exposed flat at the time of stilt-
house occupation, an interpretation consistent with our
field observations.

SEA-ZEVEL CHANGEAND DEPOSITIONAZ
HISTORYATECA

The 1985 discovery of preserved wooden post bases
and other organic materials in a waterlogged deposi-
tional context at Area B of the ECA Site was the first
indication that Lapita settlements in the Bismarck Ar-
chipelago had incorporated stilt or pole-structure ar-
chitecture constructed over shallow reef environments.
Subsequently, other sites such as Apalo (FOJ) in the
Arawe Islands, the Kreslo Site (FNT> off New Britain,
and the reef flat pottery sites on Buka, confirm that this
was a widespread Lapita settlement pattern in Near
Oceania (Gosden and Webb 1994; Specht 1991; Wickler
1995; see also Kirch 1997:162-88). Understanding the

130

THREE LAPITA VILLAGES

depositional history of such sites requires a geomor-
phologically-informed research and excavation strategy,
along with a careful consideration of evidence for local
changes in relative sea level and sediment budgets. The
transect excavation strategy utilized at ECA permitted
a detailed reconstruction of the micro-geomorphology
and depositional sequence for the period from the mid-
second millennium BC until about two thousand years
ago. Initially, we attempted to interpret the field evi-
dence in terms of a model of relative sea level stability
during this time period. Increasingly, however, our field
evidence (relative elevation levels of the paleobeach ter-
races in relation to modern sea level, evidence of wave
cut notches, etc.) convinced us that a drop in sea level
must have been responsible for the progradation of
the coastal terrace on this part of Eloaua Island, result-
ing in the burial and preservation of the ECA stilt-house
occupation deposits. This view was reinforced by a spate
of studies during the late 1980s indicative of a mid-to-
late Holocene higher sea level in the SW Pacific.

A rapid rise in sea levels following the end of the
Pleistocene is a global phenomenon that has been widely
recognized for some time (Fairbridge 1961; Shepard
1963). More controversial-because they depend upon
a complexity of local conditions and processes-have
been the details of the eustatic sea level curve in the
mid-to-late Holocene, especially the matter of whether
there have been higher-than-present stands. Bloom (1980,
1983) modeled some of the global diversity in these
Holocene curves, and suggested that a + 1-2 m stand
existed in the South Pacific during this period. Substan-
tial geomorphic and radiometric evidence from a vari-
ety of islands now supports this interpretation of a + 1-
2 m high sea level during the period between - 4-2 kyr
BP. In Fiji, Nunn (1990:304) concluded that the coasts
?experienced a middle to late Holocene sea-level maxi-
mum some 1-2 m above present mean sea level." This
conclusion is supported by work by Miyata et aL (1990),
and similar results are presented by Ash (1987) for Viti
Levu island. On Mangaia in the Southern Cook Islands,
Yonekura et al. (1988) report evidence for a + 1.7 m
stand between 3400-2900 BP. In French Polynesia,
Pirazzoli and Montaggioni (1986, 1988; Montaggioni
and Pirazzoli 1984) describe evidence from various is-
lands for a MSL between +0.8 and 1.0 m beginning

6-5.5 kyr BP and lasting as late as 1.2 kyr BP. In Western
Samoa, Rodda and his colleagues summarize various
evidence for Holocene higher stands, while Isla

(1989:361-363) reviews a comparable range of evidence
for several Pacific Islands. In developing a
morphodynamic model for landscape change and coastal
terrace formation at the To'aga Site on Ofu Island,
American Samoa, Kirch (1993) synthesized these and
other data relating to this Holocene high-stand. The ra-
diocarbon chronology for To'aga suggests that the
drop in sea level from the mid-Holocene high stand
down to the modern level occurred sometime around
2000 BP. Recent and continuing geomorphological re-
search has added additional evidence to support such
an interpretation for a mid-Holocene higher sea level
stand throughout much of the SW Pacific (Allen 1998;
Dickinson 1998, 1999, in press; Dickinson and Green
1998; Dickinson et al. 1994, 1998; Fujimoto et al. 1996).

While we have no directly dated evidence from
Mussau for a higher sea level during the mid Holocene,
the presence of wave-cut notches along the exposed
limestone cliffs of Eloaua, Boliu, and other islands of
the group corresponds well with the + 1-2 m higher
stand reported by the investigators cited above. In 1988,
we measured several profiles of such wave-cut notches,
discussed and illustrated in Chapter 2. All of these are
indicative of a higher sea level of - + 1-1.5 m. Gosden
and Webb (1994) similarly discuss coastal cliff profiles
in the Arawe Islands off New Britain, which have an
older solution or wave-cut notch  1 m above the
present high water mark. While there is no means to
directly date these notches in either Mussau or the Arawe
islands, they are no older than Holocene in age, given
coastal erosion rates.

In sum, a widespread and consistent pattern of ra-
diometrically dated shoreline features provides strong
evidence for a higher sea level ranging between - + 1-2
m over the SW Pacific, from at least 5 kyr BP, and last-
ing until sometime between 2-1 kyr BP. This is matched
by the evidence of wave-cut notches along the Mussau
islands. After - 2 kyr BP sea level fell (perhaps fairly rap-
idly) to its present position. This mid-to-late Holocene
sea level curve thus provides an important dimension
for a model of coastal terrace formation at the
Talepakemalai Site.

Progradation, resulting in the formation of a coastal
terrace such as that at ECA, can be defined as the "pro-
gressive formation of new land by sedimentation irre-
spective of the tendency of sea level movement"
(Chappell 1982:71). Sea level changes, whether due to
glacio-eustacy or tectonic movements, or both, are im-

131

THREE LAPITA VILLAGES

portant as controlling factors for the sediment budget,
but alone they do not provide a sufficient model of
progradation. As Chappell emphasizes, ". . . sea level
changes alone cannot be used to account for coastal
changes. In fact, for the last 6000 years, the sedimentary
budget is the more important factor" (1982:71). The
sediment budget can be thought of as the net sum of
sediment input, minus the loss of sediment from trans-
port. In the case of the ECA Site, there is no significant
source of terrestrial sediment. Rather, the terrace has
been formed wholly from calcareous sands and larger
clastics of biogenic reef origin (coral heads and shingles,
shell fragments, branch coral fingers, etc.). To explain
the rapid progradation of the Eloaua coastal flat dur-
ing the first millennium BC, resulting in the burial of the
stilt-house occupation deposits, requires a mechanzsm for
a substantial increase in marine biogenic sediment load.
This mechanism is provided by the rapid drop in sea
level at this time, which exposed the outer reef edge
and reef platform to wave erosion, thus generating an
increased quantity of calcareous sediment.

The occupation sequence and depositional history
of the ECA Site, in relation to the controlling geomor-
phic mechanisms of relative sea level change and in-
creased sediment budget, can be summarized as fol-
lows (also depicted diagrammatically in Figure 4.45).

STAGE 1, -3,500 BP

At the time of initial settlement of the Talepakemalai
Site, the active beach ridge ranged from 0.70 to 1.35 m
higher than the modern beach ridge and was situated in
the vicinity of units N80-N100 between the W200 and
W250 transects. This paleobeach terrace reflected the
then higher sea level stand of - + 1 m. Construction and
occupation of stilt-structures over the sub-tidal, sandy
reef flat, which began at about N1 10 and continued N,
resulted in the deposition of a dense accumulation of
shell midden, ceramics, oven stones, and other cultural
materials underneath and surrounding these structures.
Furthermore, a zone of fine-grained sediments depos-
ited in the low-energy sub-tidal environment between
the beach and the stilt houses trapped organic remains
such as wood, coconut, and seeds. A modern analog
for this kind of low energy sub-tidal depositional envi-
ronment that traps organic materials such as discarded
coconut half shells is provided by the seaward fore-
shore fronting the modern Eloaua Village (Fig. 4.46).

On the paleobeach terrace itself, cultural activities re-
sulted in the incorporation of plainware sherds, midden,
and other materials into the loose calcareous sands (ex-
emplified by the Area A excavations).

STAGE 2, -2,800 BP

By this date, which represents the later occupation phase
at ECA as marked culturally by the dominance of in-
cised ceramics, coastal progradation had commenced,
and the beach had migrated N to cover the remains of
the original zone of stilt houses. A new zone of stilt
structures now stood between about N160 to N190,
and a similar pattern of midden and ceramic deposi-
tion under the structures resulted. A thin "muck zone"
organic deposit accumulated between the beach and
the stilt structures. Cultural activities along the now-
prograded beach resulted in the deposition of incised
ceramics into the beach sands over the now buried ear-
lier (Area B) occupation zone (as reflected in the vertical
distribution of cultural materials in the Area B sequence).

STAGE 3, POST-2,500 BP

Around the mid-first millennium BC, the stilt-houses at
ECA seem to have been abandoned altogether, as
progradation increased in tempo, and the sand spit con-
necting the formerly separated N and S parts of Eloaua
Island became joined. As progradation advanced, the
materials on the former reef flat which had been sub-
tidally deposited were flooded by the Ghyben-Herzberg
aquifer, thus preserving them in a waterlogged, anaero-
bic condition. This low-lying, newly prograded terrain
was colonized by typical coastal vegetation such as Pan-
danus, Pisonia, Calophyllum, and othertrees. Disturbance
of the buried cultural deposits was limited to the bur-
rowing actions of land crabs which, fortunately, did
not penetrate into the Ghyben-Herzberg aquifer. The
organically enriched cultural deposits situated on the
older, higher beach terrace (in the vicinity of the mod-
ern airfield), however, were cultivated in a pattern of
relatively short-falbow shifting cultivation. This repeated
shifting cultivation resulted in extensive disturbance and
mixing of the occupation materials situated on the higher,
original beach terrace.

In addition to portraying the depositional sequence
at Talepakemalai along a series of diagrammatic
transects, as in Figure 4.45, we can suggest what a map

132

THREE LAPITA VILLAGES

E

G)
E

E

E

(n

FIGURE 4.45
A diagrammatic
representation of the

probable stages of

geomorphological

evolution of the
Talepakemalai Site,

from -4,000 BP to
Modern. Vertical scale
exaggerated 10 times.

3-49000BP                                              1--,Z---~,-           i~

3  4, O O O B P                        ~~~~~~~~~~~~~Forest  .

2 -                     High water line

.    . *. .   *.   .   .   .   . . .

Reef platform ... .;;... .

O1-.......

3 -3,500 BP

2 -                          High water line

1 -                         Dense midden

accumulation
0 -                                  .

.   .   *.   . .;   .,  ., .   ..

kf' r Z n'  '  ,  . .r : : ,
latreansi

3-
2 -
1 -
0-

3 ,

0)

1

0

.,   .   . ..  .,   ..  .. .   .   .   .   .. .. ..   ..   .   .   .  . .   ..   ..   ...
,:   , .   ...   .   .   .   .   .:.   .G    y    H e r z b g .

.Top~~~~~~~~~~~~o                 of Ghyben-Herzberg-'

Area C                      Area B

0                      50                     100 m

133

0                                       50

a       a       3       1        1       1

100 m

THREE LAPITA VILLAGES

FIGURE 4.46 View of the seaward foreshore fronting Eloaua Village, at low tide, with the fine-grained 'muck

zone' sediments exposed (Emussau Island visible in the distance). The vertical posts are the remains of two
abandoned stilt house dwellings. Note the presence of numerous coconut half shells (endocarps), wooden
timbers, and other organic remains Iying on and in the fine-grained sediments at the foot of the gentle

beach slope. A similar depositional environment presumably characterized the ECA Site at the time of its
Lapita occupation.

of the original coastal topography of this part of
Eloaua Island would have looked like at - 3,500 BP. As
seen in Figure 4.47, Eloaua was actually divided into
two separate islands, with a broad, sub-tidal sandy reef
flat lying between. The northern end of the main Eloaua
Island was occupied by the Talepakemalai Lapita settle-
ment, while across the channel on the southern shore of
the smaller island was another Lapita settlement, repre-
sented by the Etakosarai (ECB) Site (see below).

THE ETAKOSARAI SITE (ECB)

The ECB Site was first reported by Egloff (1975:15),
who excavated two 1 by 2 m tests there in May, 1974.
Egloff's report on the site was minimal:

The excavations at ECB indicate that the deposits are

quite shallow, having a depth in most places of no

greater than 15 to 25 cm. What deposits there are at

ECB appear to be restricted to an irregular midden - 20
metres in diametre. Also, the sherds recovered at the site
are quite battered due to the shallow and exposed
nature of the deposit (1975:15).

Egloff reported that the two test pits yielded a total of
359 potsherds plus 24 "special" sherds, 0.7 g of bone,
and 40.7 g of obsidian (1975, table 7). He combined
the ceramics from ECA and ECB in his analysis, noting
no evident differences between the two assemblages,
describing both as falling "weil within the accepted range
for the Lapita style" as then known from the Watom
and Ambitle Island sites (1975:25).

In 1985 I relocated the ECB Site with Ave Male's
assistance and determined that it was substantially larger

o

134

THREE LAPITA VILLAGES

than the "20 m diameter" mound described by Egloff.
Although the deposits did appear to be relatively shal-
low and somewhat disturbed by gardening, additional
test excavations seemed warranted in order to better
understand the site's stratigraphy, determine its areal
extent and size, obtain materials for radiocarbon dat-
ing, and generally to increase the sample of cultural
materials. During the 1985 season I excavated six 1 m2
units along a single transect bisecting the site. In 1986,
Hunt was assigned the task of extending the excava-
tions at ECB and dug an additional 13 units, bringing
the total sample size up to 19 m2. The following ac-
count combines the results of the 1985 and 1986 sea-
sons at ECB.35

THE SE TTING

Site ECB is situated in the W part of Eloaua Island, at a
place known as Etakosarai ("place of red soil"), just W
of a trail that crosses the island at its narrowest point.
The traditional landowner is Mr. Aimalo Lavatea of
Lomakunauru, whose family has made gardens there
for many years. At the time we worked there, ECB was
largely planted in manioc, and the site's surface was lit-
tered with large midden shells, obsidian flakes, and sherds
(including many bearing both dentate-stamped and in-
cised decorations) which had been brought to the sur-
face by the action of digging sticks (Fig. 4.48). The gar-
dened area was - 50 by 60 m, and the surrounding veg-

FIGURE 4.47 Suggested configuration of the NW part of Eioaua Island during the period of Lapita
occupation at Talepakemalai (-3,500-2,800 BP). Note that at this time there were two separate
islands, with an area of shallow reef flat separating them. The Talepakemalai (ECA) and
Etakosarai (ECB) sites would have been intervisible across this reef flat.

135

THREE LAPITA VILLAGES

FIGURE 4.48 View of the ECB Site during the 1986 excavations.

etation consisted of second growth with Pandanus, Ca-
suarina, Morinda, Macaranga, and some immature
Calophyllum. One of Egloff's 1974 test excavations had
not been backfilled so we were able to determine the
precise location of at least one of his units.

The site occupies gently sloping terrain which rises
from the cross-island trail for an elevation gain of - 1
m. The trail itself is situated on low-lying sandy ground,
and we suspected that the slope on which the site lies
was probably a paleobeach terrace, which formed at
the mid-Holocene sea level stand. To the W of the site
this terrace adjoins an uplifted block of limestone which
rises to an elevation of 51 m asl. Our geomorphologi-
cal interpretation is that at the time ECB was occupied,
the low-lying marshy terrain between ECB and ECA
(- 1 km to the E) consisted of an open reef or sand
flat, certainly awash at high tide if not continually under
water. Thus the ECB and ECA sites faced each other
across this reef or sand flat, and Eloaua comprised two

separate islands at that time. With sea level fall to the
modern level (probably - 2000 BP), this flat accumu-
lated a deposit of calcareous sand between 0.5-1 m asl
and was colonized by a typical halophytic vegetation
aciation (Pandanus, Tenninalia, Calpyllum,Baningtonia,
and other taxa).

EXCAVATIONS AND STRA TIGRAPHY

A contour map and plan of the excavated units at ECB
is provided in Figure 4.49. We estimate that the area
containing cultural materials covers - 3,000 m2.

THE 1985 TRANSECT

From August 25-27, 1985, six 1 m2 tests were exca-
vated along a single E-W oriented transect which ran
from the low-lying terrain E of the area of surface
sherd scatter and up across the sloping garden. The

136

THREE LAPITA VILLAGES

transect units were not equally spaced, because we had
to work around the active manioc plantings, but from
E to W the units spanned a total distance of 90 m. Our
excavations demonstrated that cultural materials were
concentrated on the higher ground W of the trail, with
almost nothing recovered in the test unit at E40N0,
which was positioned on the low-lying terrain that we
inferred to be the old reef flat.

THE 1986 EXCAVATIONS

In 1986 I oriented Hunt to the site, and we reestab-
lished the 1985 E-W transect line so that the additional
test units were positioned on the same grid. As seen in
the site plan (Fig. 4.49), these were spaced at regular 10-
m intervals running N and S of the baseline, to form a
grid of units that we hoped would allow us to assess
spatial patterns of artifact and midden distribution within
the main area over which cultural materials were con-
centrated.

STRATIGRAPHY

Although there were minor variations from unit to unit,
including the overall depth of cultural deposit, the stratig-
raphy was more-or-less consistent over the entire area
excavated. A representative profile of the unit situated
at E1S21, is illustrated in Figure 4.50. The four main
layers distinguished in this and other sections are as fol-
lows:

LAYER 1: Typically 10-15 cm thick. Dark brown

(7.5YR 4/2) sandy loam, A horizon, much re-
worked by gardening with numerous rootlets
and scattered charcoal from garden burning.
Sherds in this deposit are typically small and
heavily worn or eroded. The boundary with
Layer II is diffuse.

LAYER II: Ranges from 5-15 cm thick. Reddish-yel-

low (7.5YR 6/6) sandy loam, fine-grained with
considerable shell midden. The boundary with
Layer mI is fairly sharply marked.

LAYER III: Thickness varies considerably over the

site, but is typically from 30-60 cm thick in the
central part of the tested area. The sediment is
light gray (lOYR 7/2), medium-grained sand
and contains abundant shell midden. This de-
posit is frequently heavily cemented with CaCO3
concretions, apparently due to repeated wet-

ting and drying from the fluctuating water table.
In several units it was necessary to use an iron
crowbar to break up the concretions.

LAYER IV: This is the basal, culturally sterile deposit,

consisting of white (1OYR 8/1), coarse-grained
calcareous sand with some waterworn shells
and coral rubble. Towards the E end of the
transect, this deposit also included numerous
Tellinidae shells with both valves intact, indicat-
ing a sub-tidal depositional environment.

Layer IV clearly represents a paleobeach terrace
deposit that accumulated during the mid-Holocene
higher sea level stand and represents the original surface
on which the Lapita occupation at ECB was established.
Layer III seems to be a largely intact cultural deposit
which accumulated on this beach slope, while Layer II
is best interpreted as a post-occupation accumulation
of beach sand that was probably under a stable vegeta-
tion cover for some time. This was subsequently dis-
turbed and reworked by gardening, creating the Layer
I/11 distinction.

No features of any kind (such as earth ovens or
pits) were encountered during our excavations, leading
us to conclude that the Lapita occupation at ECB prob-
ably consisted of stilt-houses situated on a sloping beach
fronting the sandy reef flat to the E. The Layer III de-
posit thus accumulated as detritus dropped or was dis-
carded from these houses. Unlike the situation at ECA,
no wooden posts or other non-carbonized plant mate-
rials have survived at ECB, because the deposits lie
above the Ghyben-Herzberg aquifer which created the
anaerobic conditions at ECA.

SEDIMENT ANALYSIS

Four sediment samples were collected during the 1985
test excavations at ECB, three samples from Test Unit 1
(Layers IA, IB, and If) and one sample from Test Unit 2
(Layer II). Samples were analyzed using the same meth-
ods described above for sedimentology of the ECA
Site. The samples from Test Unit 1, located near what
was probably the paleobeach ridge, display a temporal
progression from a very poorly sorted, slightly gravelly
muddy sand (mean 0 = 1.77 + 2.59), white (lOYR 8/
2) in color represented by Layer II, to a very poorly
sorted, gravelly muddy sand (mean 4 = 1.64 _ 3.15),
yellowish brown (1OYR 5/4) in color. This is capped

137

THREE LAPITA VILLAGES

* *TP5

W4ONO

Secondary Forest

tP4

- - -  W2OQ! -

19  17  15  13

20    1.8  11   12

* C\TP18        eTP16

\  W1 OS21        W1 OSi 1

,  * TP1 3 i

Wl ONO

W TPi 4 /

S W1ON10

1 985
, datum

x   A i'

\, i

El S21$

e , \

! )

J     6!  Il  l  \  ?,

* TP9   `   1986   EONO

El Sl    datum

I         ? 19     *TP15

, E1 1S21           El E1 Si1 1

k~~   ~~  ",   ',~  11.1.-  ... p

* TP7          * TP8

E1N10         E1 N20

* ITP12

s',' E1 NO

* TP2

E20NO

0   2   4    6   8  10m

Grid N o   e    10

Contour interval 1 0 cm

S20

S10

* TP3

E35N0

0

N10

0 -

N30

N20

E40

FIGURE 4.49 Plan of the Etakosarai Site (ECB), showing the location of units excavated during the 1985 and
1986 field seasons.

- W40
- W30
- W20

- w1O

7/  *--'//

1

1.1

* TP17

W1lON20

T l  :

' El S30

1

1

0

-E10
-E20
-E30

S30

i

138

E

THREE LAPITA VILLAGES

by the organically enriched gardened soil covering the
site today, Layer 1. Texturally, this layer is also a gravelly
muddy sand (mean o = 4.24 + 4.5 1) that is extremely
poorly sorted and incorporates much more silt-clay than
the underlying sands; Layer 1 is also much darker in
color (lOYR 3/2, very dark grayish brown). lt has a pH
of 7.74, whereas the underlying sand deposits have pH
levels of 8.09 and 8.14 respectively. The Layer II de-
posit in Test Unit 2, situated at the foot of the former
paleobeach ridge and therefore probably representing
a tidally exposed sand flat at the time of deposition, is
characterized as a very poorly sorted, gravelly muddy
sand (mean 0 = 2.10 + 3.42), white in color (1OYR 8/
2), with a pH of 8.03. This sample, with a relatively
high percentage of silt and clay fractions, compares well
with the sand flat modern control samples (e.g., sample
LB-4). In sum, the sedimentological analysis of samples
from ECB is entirely consistent with our field geomor-
phological interpretations.

RADIOCARBON DATING

Three radiocarbon dates were obtained for Site ECB
(see Chapter 10 for sample details). Two of these (ANU-
5086, -5087) each consist of valves of Hyotissa hyotis
midden shell from the Layer I and Layer II deposits
obtained in 1985. These shell samples yielded nearly iden-
tical conventional ages of 3120 + 80 (ANU-5086) and
3150 + 80 BP (ANU-5087); calibration of these dates is
discussed in Chapter 10. The third sample (Beta-20453)
was obtained in 1986 from Unit ElS11 (TP-9) from
the undisturbed Layer III and consisted of several large
pieces of wood charcoal which were dispersed through
the deposit at a depth of - 71 cm bs. This sample yielded
a conventional age of 3200 + 70 BP, which when cali-
brated yields an age range of 1597-1414 BC. For rea-
sons given in Chapter 10, we believe that the best over-
all estimate for the age of the ECB Site is - 1520-1300
cal BC, and that the site is penecontemporaneous with
Zone C1 of Area B at Site ECA.

CULTURAL CONTENT

The ECB excavations yielded a cultural assemblage, in-
cluding ceramics with both dentate-stamped and incised
decorations. Non-ceramic artifacts include several abrad-
ers, a number of scrapers/peelers, three shell fishhooks
and several Trochus-shell hook preforms, 16 Conus-shell

1I

111        Tridacna

O

El

Iv

Unexcavated

5 Sediment samples

- 0 cm
- 10
- 20
- 30

_ 40
-50
-60

- 70
_ 80

90
- 100
_ 110

FIGURE 4.50 Stratigraphic profile of the N face of Unit
El S21 at the ECB Site.

rings, and a number of other items. There was a fairly
large number of manuports at ECB, totaling 244 for
an average density of 12.8/m2, almost exactly the same
density as at Area B of ECA (12.7/m2). The faunal
assemblage is dominated by fish and turtle bone, with a
few bird bones, one Odontocete bone, and three pig
bones.

THE ETAPAKENGAROASA SITE (EHB)

The EHB Site, on Emananus Island, was first reported
by Allen et al. (1984:9-10) who were taken there by "a
local resident [Saupa] who produced a classically deco-
rated, large Lapita rim sherd, which he had found in the
hole left by a fallen tree." They reported the locality
name as Karasa, and in 1985 we were told the name
was Erauwa, but we were later informed that the cor-
rect toponym is Etapakengaroasa. Allen et al. (1984:10)
found a second Lapita sherd as well as some sherds
with incised decoration, and obsidian, and they specu-
lated that excavations might "clarify the relationship of

i
1

1

j

L

139

I

THREE LAPITA VILLAGES

the incised and dentate stamped lapita pottery." I un-
dertook test excavations at EHB in 1985, and further
transect units were excavated by Hunt in 1986.

volcanic oven stones, shell midden, sherds, and obsid-
ian flakes that had been brought to the surface by the
crabs. Based on the surface distribution of cultural ma-
terials, we estimated that the EHB Site covers - 1,150 m2.

THE SETTING

EXCAVATIONS AND STM& TIGPAPHY

Etapakengaroasa is situated on the NE side of
Emananus Island, on the lagoon shore, a short distance
E of a small hamlet occupied by Pastor Ororea and his
family (UTM coordinates GU883255). Geomorpho-
logically, the site consists of a narrow (- 55 m wide)
paleobeach terrace which lies at the base of a steep slope
rising up to the island's upraised limestone plateau. This
terrace, which has a slightly higher berm (1.35 m asl) on
its seaward edge, presumably formed during the mid-
Holocene higher sea-level stand, and indeed, the depo-
sition of calcareous sand here may have been encour-
aged by the presence of Lapita stilt-houses. Today the
lagoon shore in front of the terrace consists of a man-
grove swamp, but this is unlikely to have existed at the
time the site was occupied. An elevation profile across
the site (Fig. 4.51) shows the structure of the beach ter-
race and berm in relation to the mangrove swamp.

The terrace is presently in old second growth and
coconut, but has been repeatedly gardened in the past; a
new garden had been cleared in 1985 - 75 m E of our
test pits. In addition to gardening disturbance, land crabs
(Cardisoma sp.) have extensively bioturbated the site's
deposits. Crab burrows were everywhere across the site,
and the ground surface was littered with fire-cracked

Although the extensive crab burrowing did not bode
well for undisturbed stratigraphy, we nonetheless de-
cided to test EHB during the 1985 season in order to
obtain a sample of pottery and other cultural materials
for comparison with those from ECA. Two test pits
were excavated from 1-2 September. Two radiocar-
bon dates obtained from these 1985 tests indicated that
the Lapita occupation at EHB was relatively early, and
this was confirmed by an analysis of the ceramics, which
are marked by a high frequency of very fine dentate-
stamping. 1 therefore decided that additional excava-
tions should be undertaken in 1986, primarily to ex-
pand our sample size of the ceramic assemblage and
associated cultural materials. T. Hunt excavated seven
more 1 m2 units between 28 October and 4 Novem-
ber, 1986, bringing the total excavated area to 9 m2, a
roughly 0.8% sample of the total area with surface cul-
tural material.

1985 TEST EXCAVATIONS

We first cleared a transect through the second growth
covering the site, running from the mangroves across

FIGURE 4.51 Elevation profile across the EHB Site, as measured by telescopic leveling in 1985.

S18         0           N20          N40         N60

.          1            1            1

N85

.\                                           ~~~~~~~~~~~~Mangrove'

. . ".\ ~~~~~~Beach

.. 91 ..... TP2 ~Ridge Xr

.    . .   .   .   .   . ,   .   .   .   .   .   .   .   . .   .   .   _ .   .   .S

,~~~.          .   .   .   .   .   .   .   .   '. \p y

g.  . . .. . ..... ..   . . .. . .   ., ., . . . . .   ., . .

t .~~.        . .   .   , .   .   . -   . -   . -   . -   .   .   ...--.*** ?

.      t . e . t .   . s .   . * . s . l . X . e . a . b .< s x8

.       . . . . . . . . . . . . . . . .-r O ss!w fU
. . . . . . . . .   .   . . . . . . . . t

.   .   . .   . .   .     . . . . . . . t *   *   , n s   u   b   w   t   e   %r b bu   r &

140

um

3m -

o 2 -
x
0G)
(-

.'t

> 1-

0-

Swamp

9          ,

THREE LAPITA VILLAGES

the beach terrace to the slope of the limestone escarp-
ment, a total distance of 85 m. The transect was staked,
and an elevation profile taken with a Leitz level and
stadia rod (Fig. 4.5 1). Two test units were gridded out,
20 m apart, along this transect line. Unit 1, - 50 m in-
land of the terrace berm, was excavated to a depth of
95 cm bs, when the water table was reached and the
mucky sediment could no longer be screened. The en-
tire depth of deposit here, which consisted of a dark
grayish brown sand loam (lOYR 4/2), appeared to have
been penetrated and disturbed by land crab burrows.
Unit 2, closer to the berm, was excavated to 65 cm bs
with great difficulty, due to considerable calcium car-
bonate cementation of the calcareous sand below 20
cm. We were able to break through this cemented de-
posit only by using a 2 m long iron crowbar. At 65 cm
bs the cemented sand became coarse grained and was
mixed with coral rubble (Acropora branch coral fingers,
etc.), and showed signs of recrystalization, probably due
to continual wetting by the fluctuating Ghyben-Herzberg
aquifer. Even with the crowbar, this deposit was virtu-
ally impossible to excavate, and the unit was terminated
at 70 cm bs.

1986 TRANSECT EXCAVATIONS

In late October, 1986, Hunt set up his datum on the
coastal trail which runs along the terrace berm, and
gridded out seven test units along two intersecting
transects. His N-S transect (which ran along the crest of
the berm) had four units spaced at 20 m intervals, while
his E-W transect (which intersected the N-S transect at
Unit ON2OW) had three units spaced at 10 m intervals.36

STRATIGRAPHY

Hunt reported the stratigraphy as being more-or-less
uniform over the entire 40 by 60 m area covered by his
two transects. Representative profiles of the ON1OW
and ON2OW transect units are shown in Figure 4.52.
Three layers were defined:

LAYER I: 0-25/40 cm bs. A dark grayish brown

sandy loam (1OYR 3/2), organically enriched
with abundant rootlets. This layer has been ex-
tensively reworked due to gardening activity.
The contact with Layer II is diffuse and varies
in depth.

LAYER II: 25/40-100/125 cm bs. A gray to dark

grayish brown (1OYR 5/1 to lOYR 4/2) cal-
careous sand, with frequent CaCO3 concretions
in the upper part of the layer, in some cases
covering almost the entire unit. The concretions
decrease with depth, and the lower part of the
Layer II deposit is within the zone of tidal fluc-
tuation of the water table, and thus continually
moist. Most of the cultural material was de-
rived from Layer II. This layer is also exten-
sively penetrated by crab burrows.

LAYER III: 100/125 + cm bs. A coarse to medium-

grained white (1OYR 8/1) to light gray (1OYR
7/1) sand, not cemented and continually wet.
Mixed with the sand are Acropora branch coral
fingers, waterworn coral pebbles, and numer-
ous Quidnipagusandsmall Tellina bivalves indeath
position (both valves intact), indicating that Layer
III was deposited sub-tidally.

From this relatively simple stratigraphy, we would
infer that the Lapita occupation began over a tidal la-
goon flat just offshore from the upraised limestone slope
of Emananus Island. Presumably, this settlement con-
sisted of stilt houses as at ECA although no preserved
posts were found at EHB. Potsherds, shell midden, and
other cultural debris began to accumulate under and
around the houses, gradually accumulating the Layer II
deposit and also trapping fine lagoonal sands which
form the matrix of this sediment. Layer 1 probably
developed through the post-depositional reworking of
the Layer II midden deposit by gardening.

Because there is no meaningful internal stratifica-
tion within Layer II, and also because of the extensive
land crab disturbance which has doubtless led to con-
siderable mixing, we have treated the cultural deposit at
EHB as a single assemblage. We did, nonetheless, check
for any significant trends in ceramic change by depth,
with a negative result.

RADIOCARBON DATING

Two marine shell samples (a Tridacna gigas and aHyotissa
hyotis valve), one from each of the two 1985 test pits,
were radiocarbon dated (see Chapter 10 for details),
yielding conventional ages of 3470 + 90 BP (ANU-
5088) and 3380 + 90 BP (ANU-5089). These suggest
that the ECB Site was occupied contemporaneously with
the earlier Area B settlement at ECA. Calibration of

141

THREE LAPITA VILLAGES

0

L

100cm      0

L

% Trochus

shell

Shell

6                               11

- 50

Q)

Q

t-)

tu

3:

ca

0
c-

e-c

01
Q)

- 100

O     White concretions (CaC03)
* e   Shell

FIGURE 4.52 Stratigraphic section of Units ON1 OW and ON2OW at the EHB Site.

these dates depends upon the AR correction factor, and
is discussed further in Chapter 10. Based upon the na-
ture of the EHB ceramic assemblage, the site is likely to
date to the early part of the Lapita sequence in Mussau,

and an age estimate of - 1500 cal BC is preferred.

CULTURAL CONTENT

The two seasons of excavations at EHB yielded a mod-
est sample of the site's material culture, including 7,552
ceramics (of which 523 were diagnostic sherds), 783
non-ceramic objects, and 553 vertebrate faunal speci-
mens. The pottery includes a significant quantity of
sherds with fine dentate-stamping, as well as pedestal
feet, both traits indicative of the earlier part of the
Mussau ceramic sequence, and thus consistent with the
radiocarbon dates. The majority of non-ceramic arti-
facts are comprised of flakes and cores (395 specimens),
and of manuports (336 specimens).

CONCLUDING REMARKS

The ECA, ECB, and EHB sites encapsulate a consider-
able range of variation, in size, structure, and cultural
content. All three sites appear to have incorporated stilt-
house architecture, and all were situated in similar geo-
morphic settings on paleobeach terraces or over shal-
low tidal reef flats. However, EHB at a little over 1,000
m2area could hardly have been more than a small ham-
let, about the same size as the Nenumbo (RiF-2) Site in
the eastern Solomon Islands, intensively studied by Green
(Green 1976; Sheppard and Green 1991). Site ECB is

slightly larger at around 3,000 m2 in area, but again within

the range that is best identified as a hamlet. Neither EHB
nor ECB is likely to have been occupied by more than
a few households, perhaps only a single extended house-
hold in each case. In contrast, the ECA Site which cov-
ers at least 82,000 m2 can properly be designated a "vil-
lage," and indeed is the largest Lapita site yet on record.

1 00 cm

Shells'

_     O,0

--O I-

(moist)

|9J S pondylus  c.

11    ~   Q)Chama .....

50

0)

100

tu

t

:3

0)

0)
-0

100

Unexcavated

142

1-

THREE LAPITA VILLAGES

There is reasonably good evidence as well that ECA
has some significant internal spatial differentiation, es-
pecially between activity areas on the paleobeach ter-
race (Area A) and those of the stilt-houses over the
tidal flat (Area B). These spatial patterns will be ana-
lyzed and discussed in greater detail in the conclusion to
Volume III, after the presentation of faunal and floral
materials, ceramics, and other portable artifacts.

Chronologically, EHB and ECB were probably
occupied for relatively short time periods, perhaps 100
years or less in both cases, and in this respect they also

contrast with ECA which seems to have had continuous
occupation within its overall boundaries for a period of
at least 500 years, and possibly as long as 700-800 years.37
ECA was first settled around the mid-second millen-
nium BC. EHB was probably in use during the period
between - 1400-1200 BC, and thus contemporaneous
with the earlier part of the occupation sequence at ECA;
this is indicated both by the radiocarbon dates and by
comparison of the ceramic assemblages from these sites.
The ECB Site seems to have been inhabited at a slightly
later date, probably sometime between - 1000-800 BC.

NOTES TO CHAPTER 4

? In this regard, it is important to note that the airfield was
constructed manually, without the aid of earthmoving equip-
ment, and follows the natural topography.

2 For these transect surveys, "mean sea level" was taken to be
the high tide as measured on September 9, 1988 at 2:30 pm.
The difference between this high water and the reef platform
exposed at the foot of the seaward beach during low water is
0.52 m.

'These low-level aerial photos were flown in March 1944, evi-
dently as a U. S. armed forces reconnaissance mission. Prints
(but unfortunately no negatives) were obtained after the war's
end by Edwin G. Bryan of the B. P. Bishop Museum (Hono-
lulu), and curated in the Pacific Scientific Information Center
formerly under Bryan's direction.

4 Indeed, the Eloaua villagers expressed considerable resent-
ment that the Seventh Day Adventist Mission had preempted
the use of a large part of this prime gardening land for the
airplane landing strip, apparently without any compensation.

5 Use of the organically enriched coastal soils of Lapita occupa-
tions by islanders for intensive root-crop cultivation has in most
cases resulted in severe archaeological site degradation. Many
Lapita sites, such as RL-2 in the Reef Islands (Green 1976) or
NT-90 on Niuatoputapu (Kirch 1988a), have been thoroughly
reworked in their upper portions, with intact features remain-
ing only at the basal contact with underlying beach terrace sands.
6These technical details of airstrip construction are important
from an archaeological viewpoint, because the relatively minor

extent of landscape rearrangement means that the site's depos-
its were not massively disturbed in the 1973 construction ef-
forts. Disturbance appears to have been limited primarily to the
airstrip path itself, and then only to the upper 20-30 cm of
deposit. The major impact of the airstrip on the site has been to
seal in a zone 30 m wide by 600 m long under crushed coral,
which for obvious reasons remains off limits to archaeological
excavations.

7Egloff's remark pertains to the absence of charcoal, since it
appears that he did not regard midden shell, or bone, as suit-
able materials for radiocarbon dating.

8 Regrettably, the proposed thesis on the ECA excavations was
never completed. In 1985, I met with Resonga Kaiku at the
PNG National Museum, who showed me the 1974 and 1978
collections which, according to the then Curator of Prehistory,
Pamela Swadling, had become mixed and impossible to cor-
rectly provenience. It is now unlikely that a final report on the
1978 materials will ever be completed.

9 The onlypublished report concerningthe 1978 PNG National
Museum expedition (Bafmatuk et al. 1980) is a popular ac-
count, giving no maps, plans, or other details of the excavation.
In 1985, Ave Male showed us two wooden stakes which had
remained standing from the 1978 excavations, with a shallow
depression between them. These stakes were - 9 m apart, sug-
gesting that this was the size of the excavation partly depicted
by photos in Bafmatuk et al. (1980). This National Museum
trench lay - 7 m W of our W200 transect line, and the trench
would have been situated between about W207N97 and
W207N106 on our grid system.

143

THREE LAPITA VILLAGES

10 Given the lack of a formal site report, and no details on this
"coral oven", it is difficult to determine exactly what kind of
feature this was. In our far more extensive excavations in 1985-
86, we never encountered any formal oven or hearth features.
We did, however, frequently find patches of calcareous sand
which had been partly cemented as a result of post-deposi-
tional precipitation of CaCO3. I strongly suspect that the pur-
ported "oven" feature was not a combustion feature at all, but
such a pocket of concreted sand, as this matches well with the
description by Bafmatuk et al. (1980:78) of a "single rock-hard
clump" of consolidated "coral."

Anson also erroneously reports the GX-5498 date as 3300 +
180 BP (1986:162).

12 The first set of transect units excavated in 1985 was initially
numbered TP-1 to 15, and are referred to in field notes and
earlier reports by these sequential numbers rather than by their
coordinate grid locations.

13 While no detailed report on the 1978 PNG National Mu-
seum excavations was ever prepared, the photos published in
Bafmatuk, Egloff, and Kaiku (1980:8 1) reveal that their trench
did not penetrate below the dark loam garden soil. We presume
that the white calcareous sand underlying this dark loam was
interpreted as culturally "sterile", and no attempt was made to
test into these sands for the presence of cultural materials.

14 Up to this point, all depths in test excavation units had been
recorded with respect to local ground surface. With the com-
mencement of the Area B excavation, we established a perma
nent datum (at 1.91 m asl), and began taking all depth measure-
ments with the use of a surveyor's level and stadia rod. This
datum was subsequently maintained throughout all three exca-
vation seasons.

15 The water saturation of the ECA sherds proved to be a
continuing problem beyond the drying stage, as in the labora-
tory we found that salts left within the sherds began to migrate
towards the surface, causing exfoliation and cracking. This ne-
cessitated a major time and money investment in intensive labo-
ratory conservation, described in full in Chapter 1.

16 The archaeologists were excited by the prospects of recover-
ing-forthe first time-actual perishable architecture of Lapita
age. The Eloaua people, on the other hand, saw the discovery
of the preserved post as confirmation of one of their oral
traditions that a special "haus matmat" or burial house had
formerly stood in the area known to them as Talepakemalai.

17 The pumps were mounted on sturdy wooden blocks for
stability during operation. Both pumps worked very weil, but
abrasion by the sandy sediment-laden water caused the rubber

diaphragms to give out, and it was vital that we had replace-
ment diaphragms. The greatest problems were encountered
with the 1.5-inch corrugated plastic hose which had been sup-
plied to us for use with the pumps by the marine outfitters in
Seattle. Pieces of sand and coral rubble continually became caught
in the corrugations and caused the hose to split, immediately
resulting in a loss of suction. Only an ample supply of fabric
"duct tape" saved the day.

1 In 1985 we had employed men exclusively in the excavations,
but when I asked for excavation assistants at the start of the
1986 season, we were approached by a delegation from the
village Women's Committee to ask whether archaeologists had
a stricture against equal opportunity employment! Surprised
though by no means upset over this feminist protest, we hired
equally by gender, a practice we continued in the 1988 field sea-
son as well. I can attest that the work benefited as a result, and
thank the Eloaua Women's Committee for their contribution
to the curtailment of gender bias in Oceanic archaeology.

19 The recovery of these plant remains coincided with the visit
to our field camp of Hoily McEldowney, then a Ph.D. student
at the Australian National University on her way to Manus to
commence field work on traditional subsistence practices. Hoily
and I both recognized the potential significance of the plant
remains being recovered, and I here wish to acknowledge the
stimulus of having her present to discuss these matters, and to
assist me in making modern botanical reference coilections and
in commencing ethnographic inquiries relating to traditional
arboricultural practices on Eloaua, work that was to be extended
in 1988 byDana Lepofsky.

20 Later in the season, one additional transect unit, W225N150,
lying half-way between the W200 and W250 transects, was exca-
vated under the supervision of N. Araho of the PNG National
Museum.

21 Exception was made for the upper, dry zone in the transect
pits, which prior experience had proved to be very low in den-
sity of cultural materials. These upper deposits were dry-screened
through 7 mm mesh. A complete record of dry/wet method
and of specific mesh size used was made for each excavation
level, and forms part of the permanent site records, to assist any
future analyses of the coilections.

22 This flotation work, extremely time-consuming and tedious,
yielded little in the way of plant remains, evidently because
most of the preserved seeds and nut cases were large enough to
be captured by the 5 mm mesh. Nonetheless, we consider the
effort expended in this work worthwhile since it eliminates any
doubt concerning potential size bias in our paleoethno-botami-
cal collections.

144

THREE LAPITA VILLAGES

23 Of course, age is represented not only along the horizontal
axis, from N to S along the transect, but vertically within any
given test unit.

24 To open up a large block, as at Area B, would have necessi-
tated the use of wooden or steel bulkheads to prevent the walls
from collapsing, as well as more powerful pumps. This equip-
ment was not available to us, and would be essential in any
future efforts to open up this part of the ECA Site.

While I am reasonably confident that the W250 transect reached
the N extent of subsurface cultural deposits in the vicinity of
Unit W250N200, I would be the first to point out that the full
extent of these waterlogged deposits to the E and W remains
undetermined. Further sets of intensive transects will be re-
quired to achieve this objective, well beyond the resources of
our 1988 expedition. Suffice it to say that sufficient questions
remain unanswered to engage more than one future generation
of archaeologists at Talepakemalai.

26 TP-1 was intended to be excavated at the E200NO grid local-
ity. When TP-2 at E1O0NO proved to be sterile, however, a
simple shovel test was put down at TP-1, also confirming a
complete absence of cultural material.

27 In earlier reports (e.g., Kirch and Hunt 1988b: 167) the Hyotissa
was identified as Pycnodonta, a taxonomic error that was subse-
quentlycorrectedL

28 Here 1 must correct the interpretation presented in the pre-
liminary report on the 1985 ECA excavations (Kirch 1987:167,
178) that the Area A assemblage reflected a late phase of Lapita
pottery, post-dating that recovered from Area B. This interpre-
tation was made, largely on the basis of comparisons with
other "Lapitoid" assemblages from the southwest Pacific, be-
fore radiocarbon dates were obtained from Area A. At the time
it seemed the most likely hypothesis to explain the Area A
assemblage, but clearly cannot be sustained.

29 Much of this work was undertaken by Steve Midgely, under
my direction, and reported in his Anthropology Senior Honors

Thesis at the University of California, Berkeley (Midgely 1992).
Figures 4.27 to 4.29 are adapted from his thesis.

30 Unit W250N170 did not contain avertically-positionedpost,
but did yield a large horizontal timber which may have derived
from the collapse of a stilt-house.

31 The counts displayed in Figure 4.38 are of actual worked
pieces, the detritus of shell artifact manufacture, which were
sorted and separated from other non-modified shell midden
of these taxa during molluskan midden analysis.

32 Note that Unit W250N190 is part of both the W250 transect
and Area C. Therefore, in tables and graphs reporting materials
from the transect and Area C, this unit is counted in both cases.
However, in summary statistics for the ECA Site as a whole, the
W250N190 materials are counted only once.

33 I thank Dr. Allen for her assistance in this analysis, although
all responsibiity for interpretation of the results rests with me.
34 The time-consuming pipette method was used only for the
1985 and 1986 samples. The 1988 samples were analyzed only
by mechanical sieving.

35 In 1986 T. L. Hunt accepted the responsibility of excavating
and reporting on Site ECB. Despite repeated requests, after 12
years Hunt had not produced a site report. Fortunately, I had
retained copies of his 1986 fieldnotes, and from these (as well
as my own notes from 1985) have written the present account.
The same situation pertains to Site EHB.

36 Unfortunately, Hunt's site map only shows the excavation
units, and no other features of topography, nor does it precisely
link his transects to my 1985 test pits. However, his notes indi-
cate that his unit 0N20W was positioned 6.10 m S of my Unit
2, so it is evident that our excavations were made in the same
part of the site.

37 The uncertainty here over the total occupation span for ECA
derives from the problems of calibrating radiocarbon ages that

date to the first millennium BC, as discussed further in Chapter
10.

145