Archaeological Research in Minahasa and The Talaud Islands, Northeastern Indonesia

Archaeological Research in
Minahasa and the Talaud Islands,

Northeastern Indonesia
Received 30 June 1975
PETER BELLWOOD
HE archaeological work described in this paper was undertaken by the author
T with the Indonesian archaeologist 1. M. Sutayasa in the Indonesian province

of Sulawesi Utara (northern Sulawesi). The two districts within the
province which received most attention were Minahasa (the northern tip of Sulawesi)
and the Talaud Islands. This report deals mainly with the results from excavations
on four major sites, one in Minahasa and three in Talaud. It is from these four sites
that a prehistoric sequence going back 8000 years can be reconstructed with some
precision.
The province of Sulawesi Utara was chosen for research because of its strategic
position: the Sangihe-Talaud Islands and Minahasa are located in a junction-zone
between the island chains of the Philippines, northern Indonesia (Borneo, Sulawesi,
and Halmahera), and western Micronesia and Melanesia. While the region remained
archaeologically blank before our work, with the exception of brief mentions by
Beyer (1947: 346-347) and van Heekeren (1972: 170), good sequences extending
back into Pleistocene times were available for a number of surrounding areas. These
included the Niah Cave in Sarawak (Harrisson 1970), the Tabon Caves on Palawan
(Fox 1970), the caves in eastern Timor excavated by Glover (1972), and, with less
certainty, the Toalean region of southwestern Sulawesi (van Heekeren 1972; Glover
1975). In Melanesia, a detailed sequence extending back into Pleistocene times was
available for the Highlands of Papua New Guinea (Bulmer 1975), and the wide-
spread Lapita ceramics of the period 1500 B.C. to 1 were believed to have ultimate
origins somewhere in Island Southeast Asia. While these facts were available,
however, they were not allowed to direct the course of fieldwork into a search for
Peter Bellwood is affiliated with the Department of Prehistory and Anthropology, School of General
Studies, the Australian National University.
BELL WOOD: l\1inahasa and the Talaud Islands 241
particular kinds of sites or remains. The northern Sulawesi sequence has been
determined from well-stratified and dated sites, and stands as an entity in its own
right.
The Physiographic Background
Figure 1 shows the geographical setting of the research area. The Sangihe
Islands extend north from Minahasa for about 230 km, and comprise the two major
islands of Siau and Sangihe, together with a number of smaller ones. The Talaud
Islands lie to the northeast of Sangihe and are separated from it by about 120km
of open sea. The Minahasa district, which forms the northern tip of Sulawesi, is
120 km long with a maximum breadth of 55 km.
The region under consideration forms part of the Northern Moluccan physiogra-
phic zone of van Bemmelen (1949: 4+4-8). There are two island arc systems
within this zone, which van Bemmelen terms the Sangihe and Ternate systems.
Only the former is of immediate concern. This comprises a volcanic inner arc, which
runs from Minahasa through the Sangihe, Kawio, and Sarangani islands to
Mindanao, and a nonvolcanic outer are, which supports the Talaud Islands. The
inner arc today reaches a maximum depth of 1700 m below sea level between Siau
and Tahulandang islands, and it supports active volcanoes on Minahasa, Siau, and
Sangihe. This inner arc is separated by the Sangihe Trough from the outer are,
which has no volcanic activity. However, the Talaud Islands on the outer arc are
clearly rising relative to sea level, for raised corals which grade into present living
reefs are found up to 50 m above sea level in coastal regions (van Bemmelen 1949:
378; see also EXCAVATED SITES IN THE TALAUD ISLANDS in this report).
Fauna and the Question of Land Bridges

The modern fauna of Sulawesi has long been noted for its high number of endemic
forms, a situation that reflects the position of the island to the east of the Wallace
Line, in the faunal zone of Wallacea. Elements of both the Asian placental and the
Australian marsupial faunas are present, the latter being represented only by
members of the genus Phalanger. Placentals include species of shrews, bats, rats)
monkeys, civets, a tarsier, two species of suids (Sus celebensis and Babyrousa
babyrussa), a deer (Cervus timorensis), and a bovid ( Anoa depressicornis). The only
well-described (but poorly dated) Pleistocene fauna from Sulawesi is that from
Cabenge on the southwestern arm, which comprises a remarkable suite of extinct
forms including elephant, stegodon, giant pig, and giant land tortoise, with Anoa
depressicornis and Sus celebensis being the only modern forms present (Hooijer
1958, 1969).
In a recent analysis of the Sulawesi fauna, Colin Groves (1976) has suggested a
Siva-Malayan origin in Sundaland, with land access to Sulawesi perhaps provided
through the Pulau Laut Centre of Diastrophism during the Pliocene (Groves 1976;
see also Audley-Charles and Hooijer 1973). The modern fauna may in part be
derived from this early radiation, although certain species, such as some of the
civets and the deer, have probably been introduced in more recent times by man.
In this regard, Glover has shown that pig, dog, goat, a bovid, civet, cat, macaque
monkey, and Phalanger have all made an appearance in the cave record in East
Timor since 3000 B.C. (Glover 1972).
Asian Perspectives XIX(2), I976 J
.
.•.
",0
TALAUD Is.
~RANGKA'A
KAWIO Is.o Karakeliang /.
N
"ff'AlNIS
0" Salebabu I~ ~UNG
TAHUN~005angihe I. Kabaruang I.
MANALU
<II
o
o
SANGIHE Is.
o
BEONG ~ Siau I.
.. <:::::) Tahulandang I.
o
SULAWESI
~
: TALAUD IS
~ I
Fig. 1 Map of Minahasa and the Sangihe and Talaud Islands. 1, Tanjong Rarangunusa;
2, Leang Tuwo Mane'e; 3, Leang Balangingi; 4, Leang Totonbatu; 5, Leang Buidane; 6, Paso.
BELLWOOD: Minahasa and the Talaud Islands 243
The 1974 excavations did not go back far enough in time to throw any light on
the questions of Pleistocene faunas and land bridges, but the still-continuing identi-
fications by A. T. Clason for the Paso site, dating from ca. 6000 B.C. (described fully
below), suggest that Anoa, Sus, Babyrousa, and Phalanger, plus a range of bats,
rodents; and reptiles, were present. Dog and deer were not present, and the macaque
monkey is uncertain. Elephant, stegodon, and the other Pleistocene species (except
for Anoa and the suids) were absent as welL The significance of these observations
should be more certain when Clason has finished her report.
The faunas of Sangihe and the Talaud Islands are impoverished versions of that
on Sulawesi, a circumstance that may reflect the small sizes of the islands (Karakel-
lang, the largest, is only 60 km long), together with isolation. No Pleistocene faunas
have ever been reported, and since a Pleistocene land bridge joining Sulawesi with
Mindanao via the Sangihe Islands has never been substantiated, then we may have
to accept that these islands have been isolated possibly for the whole three million
years of this epoch, and perhaps longer. The Sangihe-Talaud faunas comprise
rodents and bats, together with the marsupial Phalanger, and Sangihe also has a
tarsier. But at present there seems little evidence to suggest that these islands were
settled by any other animal species during the Pleistocene. A Spanish account of
Talaud dating from 1525 (Urdaneta 1911: 52) lists pigs, goats, and chickens as
present in the group, but these are surely recent introductions by man, as they are
in Timor.
ARCHAEOLOGICAL SURVEY AND EXCAVATION
The first two weeks of fieldwork were spent on site location in Minahasa, on
Sangihe Island, and on Karakellang and Salebabu islands in Talaud. Local
informants proved indispensable for this work, and we have undoubtedly discovered
only a small proportion of the total number of sites which may exist. Since time was
restricted, it was felt that excavation should begin in favored sites as soon as possible.
A total of twenty-one sites was located in the field. Of this total, four major and
three minor sites were chosen for excavation. All excavated sites were surveyed,
and excavated by natural layers where possible, or by 10 cm spits in cases of
uncertainty. Excavations were based on 1 m grids covering each site, and trenches
were initially 1 m 2, but in many cases were later extended or amalgamated. All
deposits were screened, and positions of important objects were recorded three-
dimensionally.
The overall date-range for the excavated sites runs from 6000 B.C. to the ethno-
graphic present. Three brief reports on the work have been published previously
(Bellwood 1975, 1976a, 1976b). Even the present report should not be considered
complete, since some soil and faunal analyses remain to be done, although most of
the artifacts have now been analyzed.
MINAHASA-THE PASO SITE
The major work in Minahasa was undertaken in the vicinity of Lake Tondano
(Fig. 2, top), particularly at the Paso shell-mound. Lake Tondano covers an area of
about 5 by 12 km, and its surface lies about 690 m above sea level. It has been
formed by ponded drainage between the Lembean Mountains, which rise to a little
Asian Perspectives, XIX(2), 1976
church ard terrace wall
/'
/' R5
I
'28 o
I
I o
I
-steps-I
137
o H29
C2J N
•
028/9
D22
0 C17 C15 ID6
\
ITIJ ,0
\ /
/
\ /
WEST
\
\ /
/
/
EAST
-- -
0 5M
C15 Cl6 Cli
topsoil 1
2
3
4
5
6
0
-- - CM
SO
Fig. 2 The Paso site: top, Lake Tondano and environs; center, plan of the site, showing excavation
trenches, church foundation and churchyard; bottom, section through trenches C15-17, north wall.
Layers are as follows: 1. loose shell; 2. broken shell, black soil, and charcoal; 3. charcoal; 4. ash;
5. gray soil 5Y5/3; 6. light gray lake sand 5Y7/2.
over 1000 m above sea level to the east, and a group of young and active volcanoes
rising to 1260 m to the west. The village of Paso, on the southwestern shore of the
lake, is built around natural hot springs, and these may have attracted settlers to
the site in the first place, some 8000 years ago.
A surface scatter of potsherds and obsidian tools led to discovery of the Paso
shell-mound during initial reconnaissance in the area. The true significance of the
site was not realized until after excavation began, since it had been buried, fortun-
ately without disturbance, beneath the foundation of the village church and the high
terraced churchyard which surrounds it. The site was thus encased in later deposits,
which produced the ceramics to be described below, although the shell-mound
itself is entirely preceramic.
The shell-mound is estimated to be some 30 m in diameter, from the evidence
of excavations and surface indications, and it attains about 1 m in maximum
thickness in its central part. The northern perimeter of the mound would appear
to have been destroyed by the construction of a 1 m high terrace wall forming the
northern boundary of the raised churchyard, and by an unmetalled road adjacent
to and below it (see Fig. 2, center).
For excavation, the site was laid out under a 1 m grid, and trenches were laid out
as shown on the plan (Fig. 2, center). A total of 10 m 2 was excavated, and three
trenches (CIS, C16, and C17) were excavated as one unit to give the section shown
in Figure 2, bottcmz. This section shows the general nature of the site, which com-
prises small discrete lenses averaging 10 cm in thickness, either of pure shell or of
occupation debris in the form of ash, charcoal, black soil, and broken shell. Purer
concentrations of ash or charcoal indicate specific hearth positions. The shells from
the site are unidentified according to scientific names, but the major species is a
large gastropod from the lake, called renga in the Minahasan language. Many shell
lenses consist entirely of these shells, which seem to be the only edible species
produced by the lake. However, some lenses have small numbers of marine bivalves
of a single species about 2 cm wide called wulele. Certain small lenses consist
entirely of these shells, but their overall frequency is low, and they had to be
brought about 20 km to the site from the sea. Although tiny, they might have added
some variation to an otherwise monotonous diet.
The ecological situation and method of buildup of the site can be reconstructed
with some certainty. When it was first occupied, around 6000 B.C., the new settlers
would have found a beach of light gray lake sand, perhaps covered with light soil
and vegetation, away from the edge of the lake. This beach sand sloped down toward
the lake for a vertical distance of about 1.50 m across the 30 m long east-west axis
of the site. The first settlers camped in one spot, perhaps quite close to the water,
and proceeded to dump food shells in lenses outside and around their living area.
In time, they decided to move their dwellings to a new position on top of one or
more of these shell lenses, and used their former dwelling area for further shell
dumping. This type of rotation, in which one specific area would be used alter-
natively for living and dumping, clearly went on for some centuries, until the midden
reached its present height of 1 m. It was then abandoned for several millennia,
during which time the lake has receded considerably. Today the edge of Lake
Tondano lies about 390 m from the center of the mound, and its surface is about 6 m
below the lowest midden layer. However, the shoreline of the lake is now under rice
Asian Perspectives, XIX(Z), 1976
fields, which actually commence about 230 m from the mound, so while the lake
has clearly receded from the site since it was occupied, the exact amount of the
recession is hard to estimate owing to the interference caused by the rice field
construction. It can only be placed between the two distances given above.
The date for the shell-mound is established from two carbon-14 determinations
from charcoal samples. The first, from an upper lens in trench C16, gave a date of
7530 ± 450 B.P. (ANU 1517). The second, from a lens near the base of CIS, gave
a date of 7360 310 B.P. (ANU 1518). Although these two dates are slightly
inverted, they do overlap within one standard deviation from each mean, and they
indicate a short period of use of the site, probably less than 3-500 years. This
conclusion is supported by the stratigraphic profile, which shows no sign of site
abandonment while the midden was forming. If the two dates are fitted to the bris-
tlecone calibration curves presently available, then they fall around 6000 B.C. within
quite narrow limits.
The lenses of pure dumped shell naturally contain few artifacts, but there are
sometimes a few animal bones present. It is the lenses of occupation which are of
the most interest, for these contain quite surprising densities of flaked stone and
bone tools, animal bones, and hematite. The obsidian, of a coarse vesicular type
which produces a very rough fracture surface, was quite definitely worked on the
site. It comes from the local volcanic region west of the lake, and a specific source
is known to occur in the vicinity of Tataaran village. Hematite is available near the
village of Pangolombian (see map, Fig. 2, top). Neither of these sources could be
examined carefully in the time available to us.
Over the whole of its area, the shell-mound is covered by 10 to 20 em of topsoil
which contains obsidian and pottery, mixed by recent gardening activity. The
pottery, described below, was found on the surface all over the churchyard. Of the
three trenches dug outside the limits of the shell-mound, 137 was totally disturbed
by churchyard construction, D6 was right on the edge of the mound where the
shell lenses terminated; and R5 was beyond the mound on the inland side, away
from the lake. R5 merits a brief discussion.
The deposits in R5 consisted entirely of a homogeneous dark soil, of Munsell
color lOYR3j2. Obsidian was found from the surface to a depth of 90 cm, while
red-slipped pottery, described below, was found to a depth of 70 cm. There are no
shell lenses. It appears that the R5 deposits, which have a high density of artifacts,
were built up in a slight hollow behind the shell mound. Whether the obsidian is
directly associated with the pottery, or whether it has been derived from some
adjacent and older deposit, can only be determined by further work. The obsidian
is unlikely to be derived from the shell-mound owing to the absence of shell in the
R5 deposits, but it is technologically identical to the shell-mound industry.
ARTIFACTS FROM THE PASO SITE
The Obsidian Industry (Plate la-e)

The coarse vesicular obsidian used at Paso was worked with volcanic stone
hammers and anvils to produce flakes and "chunks." The former show striking
platforms and bulbar surfaces, which are always recognizable despite the uneven
pockmarked surfaces of the stone. The latter were produced by the smashing of
e
d
f
h
9 -=-
o 3CM
Plate I Stone and bone tools from Sulawesi Utara: a, Paso: high-backed obsidian chunk (top and
side views); b, Paso: thick flake of obsidian heavily retouched on both sides; c, d, e, Paso: flakes of
obsidian (d is notched, e has ventral retouch); j, g, h. bone tools from Paso; i, L TM: bipolar blade
core of chert (front and side views); j, k, m, L'I'M: chert blades (m has slight edge-gloss); t. L'I'M:
chert flake; n, unifacially flaked chert pebble tool, from track between Beo and Rainis.
cores and nodules against anvils (cf. Crabtree 1972: 9), and do not show striking
platfonns or complete bulbar surfaces. Manufacture therefore involved a fairly
random process of striking and smashing, with a high rate of wastage. Owing to the
free availability of raw material, the rate of wastage was probably considered
insignificant by the manufacturers. This type of core-smashing technique is also
reported for a possibly contemporary chert industry from the Sohoton Cave on
Samar (Tuggle et a1. 1972: 76), and it has been used to the present by the isolated
Tasadays of southern Mindanao (Fernandez and Lynch 1972).
The fonowing statistics are derived from a total of 807 tools excavated from the
shell-mound in trench CIS. They were bagged during excavation according to lens,
and analysis shows no overall significant change from top to bottom of the trench.
The shell-mound industry represents a technological unity.
1. Cores form O-S % of total obsidians per lens. This percentage is quite low,
and most cores are in fact split or incomplete.
2. Flakes (Plate Ie-e) form 36-45 % of obsidians per lens.
3. Chunks, probably the result of core smashing, form 52--64% ot obsidians per
lens. This a very high percentage. especially when compared to that for the
Talaud chert industry described later in this paper.
4. Blades, blade cores, pebble-tools, and core-tools flaked from unbroken
primary nodules are absent.
5. Only 18 % of obsidians produced show signs of use-wear, and out of this 18 %
a total of 32% (about 6% of the total assemblage) were retouched.
Length, breadth, and thickness were measured on all flakes from CI5; the results,
plotted on triangular coordinate graph paper, showed low correlations between these
three measurements. This distinguished the Paso obsidians from the Talaud cherts,
which achieved more regular proportions. Lengths of Paso flakes ranged from 10 to
35 mm for unused specimens, with even scatter within the range. Used obsidians
ranged from 20 to 45 mm in length, and thus showed a definite tendency to be
larger than the unused. Figures for breadths differ little from those for length, and
length : breadth and breadth : length ratios seldom exceeded 2 : 1.
Although only the CI5 obsidians have been analyzed in detail, all other samples
appear to be virtually identical in morphology, and this also applies to the obsidians
associated with the upper layers with ceramics, particularly in trench R5. The
material itself is so coarse that it would probably be unsuited to blade production
anyway, and the manufacturing techniques described above would probably have
changed little over time.
One attribute of the Paso industry does stand out, particularly in comparison
with the Talaud cherts. This attribute is difficult to quantify objectively, but it
concerns the selection of flat-based, high-backed, and steep-sided flakes and chunks
(see Plate Ia) for use as tools. I would estimate that perhaps one-quarter of the
Paso tools would fall into such a subjective category, and this tendency is paralleled
in the pre-Toalean assemblage excavated by Ian Glover from the base of the Ulu
Leang cave in southwestern Sulawesi (Glover 1975). This assemblage is older than
8000 years, and could be roughly contemporary with Paso. Similar high-ridged
tools are reported from the Guri Cave assemblage of early Holocene age from
Palawan (Fox 1970; personal observations, National Museum, Manila), and they
are also present in the Pleistocene assemblages from Tabon Cave, although tools
from the latter site are generally much larger than those from Paso; Fox (1970: 37)
states that over 80% of the Tabon flakes are more than 50 mm long. High-backed
tools have also long been known to characterize early Australian and New Guinea
assemblages (Mulvaney 1975: chap. 6; White 1969), although horsehoof cores of
the types found in Australia or in the Cagayan Valley on Luzon are not present at
Paso. Nevertheless, the Paso obsidian industry fits quite well into the general
pattern of early Holocene industries predating the development of a flake-blade
technology from about 7000 years ago in the eastern regions of Island Southeast
Asia. The obsidian industries reported from Sumatra and Java (Bandi 1951; van
Heekeren 1972: 133-137) do not appear to be closely related to Paso, and they could
be more recent in time.
Tool Function in the Paso Assemblage

The Paso tools do not fall into "types" in the sense used by European Palaeolithic
prehistorians, and in this regard they are typical of many of the stone industries of
Southeast Asia and Australasia. This situation has been noted specifically for New
Guinea by White (1969; White and Thomas 1972), whose strategy has been to
analyze used edges rather than overall form. A code for recording the positions of
used edges on tools has been developed by Wilmsen (1970), but I have not so far
attempted to apply this to the Paso assemblage, and in this instance the used edges
are each considered in spatial isolation.
Two general points should be noted concerning used edges on stone tools: firstly,
the manufacturing process can cause heavy flaking damage around the striking
platforms on cores, and this is very easy to confuse with the concept of <\ "retouched
high-angled scraper edge." Since high-angled scrapers are common elements of
Southeast Asian and Australasian assemblages, I feel there may be dangers here
for the unwary. Some Paso tools fall into this category, and I have omitted them
from use-wear analyses where ambiguity is present. Secondly, short-term use on
soft materials may produce little identifiable damage (Tringham et a1. 1974), and
this may tend to reduce counts of used tools from prehistoric assemblages. On the
other hand, trampling and dropping can produce edge-damage not caused by use
(Keeley 1974). I know of no way to control these problems for assemblages such as
that from Paso; it is one thing to carry out experiments, but another to apply the
results of such experiments to specific archaeological contexts.
In the Paso assemblage, angles of used edges range from 30 to 90 degrees. There
are no clusters within the range, such as those claimed by Wilmsen for Palaeo-
Indian assemblages (Wilmsen 1970). Edge shapes were classified as convex, straight,
or concave; none of these shapes is associated with a particular range of edge-angles,
although concave edges tend on the average to have higher angles. Straight and
concave edges form about 30-50 % each of the edges in each lens, while convex
edges are rarer at about 15-25 %. There are no clear changes over time through the
lenses. Retouch, always unifacial and most often on the dorsal surfaces of flakes,
was present on 32 % of the used edges. Concave edges were most commonly
BELL WOOD: Minahasa and the Talaud Islands 249
retouched (50%), as were convex (43 %), but straight edges were retouched rarely
(15%). Of the total of 169 used obsidians from trench CIS, 31 (18%)had two used
edges, while only one had three. On the tools with more than one used edge there
is no correlation between different edge-shapes.
The Paso tools are technically scrapers: all use-wear is unifacial, and, like the
retouch, is generally on the dorsal surface of flakes and flat-based chunks. It
consists of small flake scars distinguishable from the larger scars of intentional
retouch, and no gloss of the type commonly reported on flake-blade industries in
Island Southeast Asia (see below) was observed. In use, the tools appear to have
been pulled over the surface to be worked, with their long axes at right angles to
this surface. The upper surface (usually the bulbar surface on flakes) was tilted
away from the worker at an angle greater than 90 degrees. I have replicated these
motor habits on obsidians flaked at the University of Hawaii (materials courtesy of
Richard Gould) and believe the reconstruction to be substantially correct. There
are no tools with bifacial use-wear in the knife or saw categories, and to-and-fro
cutting may have been done with bamboo knives. Knives and scrapers require quite
distinct motor habits, and they are likely to have been regarded as separate tool
categories by the Paso tool-users (cf. Gould, Koster, and Sontz 1971 for central
Australia).
Apart from scrapers, other tool categories in the Paso assemblage include awls,
tanged flakes, and rejuvenation slugs struck from blunt retouched scraper edges.
These categories are numerically insignificant, however. Four thick flakes with
retouched scraper edges on opposed sides (Plate Ib) could have been used like the
flat adzes described by Gould and Quilter (1972) from southwestern Australia,
possibly hafted.
The Bone Industry (Plate If-h)
A total of 17 long pointed tools of bone was recovered from the midden lenses,
together with 5 split and grooved long-bone fragments which were probably
associated with their manufacture. The longest specimen is 90 mm, but since many
were discarded broken it is not possible to give a range of lengths. The tools were
made either on sections split from large mammal long bones, or on smaller bones
which were not split. The animal species are not yet identified. The pointed ends
are either polished into a rounded taper, or facetted on the larger specimens. There
is one definite bi-point SO mm long (Plate Ig) and three possible broken bi-points.
One specimen (a uni-point) has a thick coating of hematite.
These bone tools, which presumably served as awls or needles were not found
in the upper ceramic layers and belong entirely to the shell-mound deposits. They
are well paralleled in Toalean assemblages in southwestern Sulawesi (Glover 1975;
van Heekeren 1972: Fig. 24), and in the Gua Lawa bone industry on Java (van
Heekeren 1972: PI. 48; the Gua Lawa spatulae are not paralleled at Paso). Similar
bone tools are commonly reported from Australian sites.
Hematite
Small pieces of hematite were found throughout the shell-midden, and a number
were clearly associated with an obsidian flaking floor in trench C16. A volcanic stone
with a coating of hematite on one surface, found in trench C 17, may have been used
as a surface for preparing the pigment.
The Ceramics
The Paso ceramics can be described only in general terms here, as analyses are
still in progress. There are two ceramic styles in the site, both of which are sepal ated
spatially and stratigraphically. One comprises a hard gray ware which is often
paddle-impressed, and the other comprises a softer buff or gray ware which is often
red-slipped. In the discussion below I will refer to the latter as "Paso Ware."
Tempers and pastes have not yet been analyzed, so the following descriptions
concentrate mainly on form and decoration.
The Paso Ware will be described first. This is concentrated in the deposits in
trench RS and dominates the sherd counts from the layer sealing the shell mound
in trenches D6, CIS, C16, and C17. It also dominates surface collections made to
the south of the church and thus concentrates heavily in the inland half of the
churchyard area. The lakeward half, including the upper layers in trenches D22,
G28j9, H29, and P28, is heavily dominated by the gray ware.
The Paso Ware excavated from RS shows considerable homogeneity through the
70 cm depth in which it occurs, and seems to have been deposited within a relatively
short time. Its surprising density (2600 sherds in 0.7 cmS) suggests that the loci of
both manufacture and use were close by, and numerous lumps ot lightly fired clay
indicate that firing was carried out somewhere in the vicinity. No stone or metal
tools (apart from one stone adze) were found in RS at all, which suggests fairly
concentrated and specialized dumping of sherds.
The forms of the Paso Ware are fairly standardized and occur in relatively
unvarying frequencie<> throughout all levels. They are illustrated in Figure 3a-r, and
may be described as follows:
Type 1. Jars with restricted necks and everted rims (Fig. 3a-e, h, m). No examples
are fully reconstructible, but diameters at lips range from 10 to 22 cm,
and bases are probably rounded rather than fiat. Orientation angles of
rims (Fig. 3a, angle a) range most commonly between 30 and 45 degrees,
while inclination angles (angle ~) range between 105 and 120 degrees.
(I wish to thank Geoffrey Irwin of the Department of Anthropology at
Auckland University for suggesting the two terms "orientation angle"
and "inclination angle" to me.) There is much variation within the
group. Fewer than 30 %of these vessels are red-slipped, but percentages
are hard to calculate since the slip is not strong and may often have
eroded away. Slips are generally on interior surfaces, but exteriors and
sometimes both surfaces may be slipped as well. Rim notching (Fig.
3h, m) occurs very rarely, as does simple-tool incision in straight-line
or zigzag motifs (Fig. 3m). Type 1 vessels account for 45 % of the Paso
Ware rim material.
Type 2. Unrestricted bowls. This category consists of three main forms referred
to as variants 2A, 2B, and 2C. Type 2A bowls have direct flattened rims
(Fig. 3i, j) and are highly standardized in shape. Almost all are red-
BELL WOOD: Minahasa and the Talaud Islands
o '.' ~ .... '

o
........
••• , ,* • • '
· . , ... ..
• '
,
••••••
~
• w
••••••
·..... ~. ... " ..

.. .. " " ..., " ......
· ......... e •••••••••••••
...... ·.·.".0 .....

; .. , .. "
... · . . ~
.
· . .. ........ . , ..
.................... ',~ "
...... ............
•• •• 0 ••••• •• •• • •••• •
perforat ion
\ .----r.~~_".,,...---,..,....
,
Fig. 3 Pottery from Paso: a-r, Paso 'Ware; $, gray ware. Stippled halo denotes red slip.
Asian Perspectives, XIX(Z), I976
slipped, often on both surfaces. Vessels of type 2A account for 31 % of

the Paso Ware rim material. Variants 2B and 2C have interior flanges
and are never slipped. Flanges of 2B bowls (Fig. 3k) are from 10 to
18 mm wide; those of 2C bowls (Fig. 31) are from 25 to 45 mm wide.
There is no overlap between the two variants in flange width, and the
flanges often have transverse grooves across their upper surfaces. Type
2B and 2C vessels are of coarse thick fabric and may have served as
mortars. Together they account for 6% of total Paso Ware rims. The
rims shown in Figure 31 andg are intermediate between types 1 and 2A
in having unrestricted orifices with indirect rims. These forms are very
rare and have not been considered statistically.
Type 3. Pedestals. Pedestals most probably belonged to type 2A bowls, although
no complete examples can be reconstructed. They are of two kinds-
simple and carinated (Fig. 3n, Q,p). They are rarely slipped, but usually
on exteriors if so treated. A few simple pedestals have single circular
perforations and some carinated examples (Fig. 3p) have simple incision
and lip-notching. Pedestals form 16% of the Paso Ware rim material.
Type 4. Pedestalled platters. This form is described separately, and is the only
one which shows any temporal distribution, being restricted to the
lower layers of R5. A fully reconstructed example is shown in Figure 3q;
this has a central perforation and lip incision. Type 4 vessels account
for only 2 % of the Paso Ware rim material.
Virtually all of the Paso Ware rim material fits the foregoing categories, and the
degree of standardization is quite remarkable. Three spouts, one perforated lug,
and two bottle necks were also found in R5, but these frequencies are insignificant.
There is also one thickened vessel foot (Fig. 3r). There appear to be no lids or
tripods present. Decoration, extremely rare, is only of the simple-tool incision type
shown in Figure 3, and about 2 % of body sherds have simple parallel impressions
made with a carved paddle.
The gray ware found over the eastern side of the shell midden is very fragmentary,
and no forms have been reconstructed. A few rim sherds are illustrated in Figure 3$;
their profiles are entirely different from those of the Paso Ware. The gray ware is
not slipped, but about 20 % of sherds have parallel or crossed-line paddle impression.
A complete gray ware globular vessel with an everted rim and paddle-impressed
sides was placed in a pit dug into the top of the shell-mound in trench D22, but this
had unfortunately collapsed before the cavity was filled with soil. A few fragmentary
bones were found in the pot, so it may have been a jar-burial. In trench G28/9 the
lower legs and feet of an inhumation burial were found associated with gray ware
sherds, but no definite statements can be made about burial posture. The rest of
the burial remains unexcavated.
In trenches H29, D22, and P28, sherds of gray ware were sometimes found
stratigraphically below the Paso Ware, but most of these contexts have been
disturbed by recent cultivation and churchyard construction. No datable material
was recovered with any of the ceramics, and I am awaiting the results of thermolumi-
nescence analyses from A. J. Mortlock at the Australian National University. A
number of problems do arise concerning dates, for the gray ware is similar to modern
impressed pottery made around Tondano, and it is associated with a few Chinese
sherds. The Paso Ware is unlike anything made in the area at present, and it was
found with only a stone adze (possibly nephrite) of quadrangular cross-section. I am
inclined to accept it as older than the gray ware, but can draw no firm conclusions
at present. The associations of the gray ware would suggest that it is not over 500
years old.
At the present time I have located no definite archaeological parallels for the
gray ware, but the situation is much more interesting with respect to the Paso Ware.
This has two very different sets of parallels, one being Neolithic, the other Late
Period, postdating A.D. 1300. The two alternatives are discussed in tum.
1. Neolithic Period
The site of Dimolit, on Palanan Bay in northeastern Luzon, has produced
ceramics very similar to those from Paso, in association with the post settings of
small square houses dating between 1300 and 3500 B.C. (Peterson 1974a, 1974b).
The Dimolit material comes from globular vessels with everted rims, open bowls
on pedestals, and some carinated forms. About 30% of the material is red-slipped,
mainly the open bowls, and there is no other type of decoration. Some of the
pedestals have circular perforations cut through, smaller but more numerous than
those on the Paso pedestals. Basically. the Dimolit pot forms are very similar to
those from Paso, despite specific rim differences, which may be expected considering
that the distance between the two sites is about 1600 km. I have been able to
examine the Dimolit material stored in the Department of Anthropology at the
University of Hawaii, and I feel the parallels reflect more than mere coincidence.
The Paso and Dimolit assemblages are much more similar to each other than either
is to the broad range of other reported ceramics of Neolithic and Early Metal date
from Island Southeast Asia. As we will see, plain red-slipped pottery is also charac-
teristic of the earliest Neolithic ceramics in the Talaud Islands, but the forms here
are rather different in that they lack pedestals and carinations.
2. Late Period
At the more recent end of the time-scale, the Paso Ware does have more tenuous
parallels with certain ceramics from Zamboanga and Jolo published by Spoehr
(1973). These ceramics postdate Spanish contact in the southern Philippines. The
Paso pottery may be compared with some of the red-slipped rim profiles from Fort
Pilar and with some of the notched rims from the Tausug walled enclosures on J 010
(Spoehr 1973: Fig. 114). However, these Philippine wares have a very large number
of features of form and decoration not paralleled at all at Paso, and they are
associated with imported ceramics. No imported ceramics were found with the Paso
Ware, although this does not automatically imply a date before the Late Period.
The Paso material also has parallels with the plain pottery recovered by Hutterer
(1973b) in association with porcelains in burials beneath Cebu City. The Cebu
material has Paso type 1 and 2, and possibly type 3 vessels, but slipping is much
rarer. The site also has rim notching like the Paso Ware and does seem to have more
parallels with it than do the Zamboanga and J 010 materials.
254 Asian Perspectives, XIX(2), 1976
At present I am unable to assess these two sets of parallels with finality, and the
Paso pottery badly needs independent absolute dates. Early Metal Period ceramics
of the Bagupantao and Novaliches types reported by Solheim (1964) also have
ring-feet and pedestals with cutout designs, but I feel these to be unrelated to Paso.
The choices seem to be late Neolithic or Late Period. I am inclined to choose the
former, since the Paso pottery has a stone adze in direct association, no porcelain or
clearly recognizable Early Metal Period pottery, and the similarities with Dimolit
are the closest.
Peterson is unable to find close parallels for his Dimolit material and feels it is
not closely related to any of the well-reported material from Taiwan. I have seen
similar red-slipped pottery from the Lal-Lo shell-mound in the Cagayan Valley of
northern Luzon, stored in the National Museum in Manila. This pottery is ap-
parently associated with radiocarbon dates of 3690 100 and 3580 ± 100 B.P.
(Ellen and Glover 1975: 376), but the exact nature of the site is unkno"vn to me.
Nevertheless, I feel that ceramic parallels of a specific nature may be beginning to
appear between the Philippines and northeastern Indonesia for the late Neolithic
Period, prior to the spread of jar-burial, which took place after about 1000 B.C.
ADDITIONAL INVESTIGATIONS IN MINAHASA
Although the Paso site accounted for most of the Minahasa research, a few other
sites were investigated as well, with minimal results. Several rock-shelters were
investigated about 4 km inland from Ratatotok on the southeastern coast of
Minahasa, but no important cultural deposits were located. We also spent some time
examining the well-known square stone burial jars (waruga) which are to be found
in many localities around Lake Tondano. These are often carved with figures in
Dutch costumes, and contain beads and Chinese ceramics of very modem appear-
ance, together with a well made local earthenware decorated by incision and dentate-
stamping. These waruga clearly need more detailed study. I would be inclined to
regard the observed examples as postdating A.D. 1500. However, stone burial jars
are found widely in adjacent regions, such as inland Borneo (Harrisson 1962a),
central Sulawesi (Kaudern 1938), and southern Mindanao (Maceda 1964, 1967;
Kurjack and Sheldon 1970, 1971), and since some of the Mindanao urns from the
Seminoho rock-shelter in the southern Cotobato Highlands may date back to about
A.D. 600 (Kurjack and Sheldon 1970), the Minahasa waruga jars may then have
a long and interesting prehistory as well.
EXCAVATED SITES IN THE TALAUD ISLANDS
The Talaud Islands are three, comprising Karakellang (60 km long) in the north
and the two smaller islands of Salebabu and Kabaruang in the south. We did not
visit Kabaruang. Karakellang is still mainly forested and has low mountains in its
northern and southern sectors rising to maximum heights of about 750 m. Across
the center of the island is a lower sector, followed by the only major track on the
island, which runs from Beo to Rainis. There are no roads, motor vehicles, or pack
animals anywhere in the group, and all movement is on foot or by boat. The
general topography of Salebabu is similar to that of Karakellang.
Neither the Talaud nor the Sangihe islands have barrier reefs and lagoons, but
coral does grow right against the shoreline in many areas. The lack of lagoons
reflects the tectonic instability of all the islands in this area. Today, the population
of the Talaud Islands is very small, and is confined to small coastal villages. No
modern population figures are available to me, but population density for Kara-
kellang prior to World War II was only 13 persons per km:!, compared to 141 on
Sangihe and 197 on Siau. The latter two islands have fertile volcanic deposits
which are lacking in Talaud.
Van Bemmelen (1949: 378-379) gives a brief description of the geology of the
Talaud group. The islands consist of varied rocks ranging from Cretaceous to
Pliocene in age, and the earlier Cretaceous series is of considerable importance
since it contains the cherts which were used on the islands for stone tools. Quite
large nodules of chert are very common on many beaches and along inland tracks
where soil is exposed, and the material would have been in plentiful supply through-
out prehistoric times. Many of the coastal regions of Karakellang and Salebabu are
occupied up to 50 m above sea level by raised reefs of coral limestone, which van
Bemmelen suggests are Pleistocene or Holocene in date. They grade into the living
reefs without interruption, a circumstance which suggests that the islands may have
been rising recently, and may still be doing so. This matter is discussed again below
in connection with the Leang Tuwo Mane'e site.
LEANG Tuwo MANE'E (LEANG-U-TuwoMANEE), KARAKELLANG ISLAND
The Leang Tuwo Mane'e site (henceforth LTM) is a wave-cut rock-shelter in

an uplifted coastal block of coral limestone. It lies about 150 m north of the northern
end of Arangka'a village and is 20 m inland from the head of the beach. The track
between the villages of Arangka'a and Geme passes just in front. The shelter is very
angular in shape (Fig. 4) and has a total length of about 25 m in three linked sectors.
Maximum breadth is about 6 m, and height at the time of first use would have been
everywhere adequate for standing. The southwestern sector of the site was partially
excavated by a cluster of trenches labelled P and Q, while the northeastern sector
was partially excavated by trenches labelled H, I, J, and K. These sectors are
referred to as the PIQ and HjK sectors below. Between them lies the middle part
of the shelter, now occupied by a walled tomb with a wooden lid, built in 1905.
This tomb is surrounded by a walled and gated enclosure, and it sits on a modern
sand terrace built up within the enclosure wall. There is every reason to assume
that prehistoric deposits still survive intact beneath the sand, but no excavation was
undertaken. The bones in the tomb, of 68 individuals, were collected from within
the shelter according to informants, but they are without goods except for two
Conus bracelets and are doubtless very recent. Christianity did not spread in these
islands until the end of the nineteenth century, and cave burial may have lasted
until this time.
Cultural deposits in both the P/Q and the H/K sectors extended down to a depth
of 1 to 1.5 m. These deposits were excavated in 10 cm levels in each 1 m 2 trench,
since it was virtually impossible to see the very slight changes in coloration in the
dry deposits during excavation. At the end of excavations, layers could be discerned
with some difficulty in the cleaned sections, and these were fully recorded for all
3M
--===--
\ K7
\.
1·0
Fig.4 Plan of Leang Tuwo Mane'e (contour interval 0.25 m).

trenches. Back in the laboratory at Australian National University, the ceramics
were analyzed both by 10 cm depth units and by natural layers, but no major
differences were observed between the results from the two methods. Ceramics and
stone tools are analyzed by 10 cm depth units in this report, since this simplifies
presentation without distorting the information content. I do not believe that all
the layers have cultural significance, since some are clearly products of differential
leaching and humus accumulation which would proceed regardless of human
activity.
The natural stratigraphy in the H/K sector is shown schematically in Figure SA ;
colors are described from the Munsell soil chart. Layer 1, 20-30 em thick, is a dark
gray soil containing many coral lumps from the roof and walls of the shelter. Such
lumps are found in all layers in the site. Beneath layer 1 is layer 2, a gray soil having
the same texture as layer 1. At present, I would regard layers 1 and 2 as closely
related in origin, with layer 2 being leached and layer 1 having more humus and
dampness. The division between layers 1 and 2 has no cultural significance in the
artifact analyses. Layer 3 is more clearly distinct from layers 1 and 2, and is basically
a pure beach sand of pale yellow color. In its top 20 cm or so it grades into a grayish
brown sand due to soil infiltration from layer 2. The H/K layers therefore consist
of a basal beach sand overlain by a two-part terrestrial deposit.
The natural layers in the PjQ sector are a little more complex (Figure SB), but
begin again with a dark gray topsoil 20-30 cm thick, best developed near the drip-
line in the front of the shelter. Beneath this is layer 2, up to 60 cm thick in some
trenches toward the back of the shelter. This layer owes its origin to two or possibly
three coral gravel floors with intervening occupation deposits. It rises almost to the
surface at the back of the shelter, where conditions were too dry for layer 1 to
develop fully. These floors have been disturbed beyond recognition in the P20t21
section illustrated, but they are visible elsewhere in the PjQ excavations, as well as
in a large hole dug by culprits unknown in the back of the cave behind Q22. They
are about 10 cm in individual thickness and are separated by layers of dark gray
earth with occupation material. Layer 2 grades at the drip-line into layer 3, which
seems simply to have been the muddy edge to the layer 2 floors. Traces of stone
paving were also found in layer 3. Layer 4, a gray soil similar to layer 2 in sector
HjK, gets darker toward its rocky base around and outside the drip-line, perhaps
due to humus accumulation. Layer 4 is about 50-60 cm thick, and it seems to form
the natural subsoil for layer 1, from which it is separated by the intrusive contents
of layers 2 and 3. It sits on the basal deposit of the shelter-layer S-which consists
of a light yellowish~brown clayey sand with many large coral blocks, grading down
into a light yellow sand similar to that at the base of the H/K sector.
The cultural stratigraphy of the site is fairly clear, even though there are rare
cases where sherds of one pot may be separated by up to 40 vertical cm. Disturbance
is fairly characteristic of the upper natural layers, and the amount of scuffing and
poking into the dry shelter floor must have been high at all times. The cultural levels
(not to be confused with the natural layers) are delineated in Figure 7 from the
data given in Figure 6, in which a number of cultural parameters are plotted by
depth. These cultural levels (see Bellwood 1976a) can be described as follows:
A
o 1M
LTM 1974
east J13 west
2·SY5/1
L _ _...r---=:='""-'-"""--"4 -2 :Sv5/2
2'5Y8 / 3
west east
B '21 P20
250±
410±
o 2'SY6/3
2'SY7/3
Fig. 5 Leang Tuwo Mane'e sections: A, through HjK sector (full section); B, through P/Q sector
(trenches P20 and P21). Carbon dates shown at left, Munsell soil colors at right.
1. Preceramic. H/K sector only, below 80 em (layer 3, upper). Absent in Pia
sector. Well-developed blade industry, using roughly equal proportions of
local gray and brown cherts.
2. Neolithic. The lower part of layer 2 in H/K (60-80 em) may be earliest
Neolithic, with a continuing blade industry and rare potsherds. The remainder
of the Neolithic in H/K is then represented by sparse deposits between 40
and 60 em in the middle part of layer 2. The Pia layer 4 (70-120 em) seems
also to be Neolithic, with prolific pottery and flake tools of brown chert.
Blades and the use of gray chert are phased out in this layer. The fairly high
proportion of gray chert in the upper layers of H/K is probably due to
disturbance of the lower layers.
3. Early Metal. Pia layers 2 (lower part) and 3, 40-70 em; H/K layer 2 (upper
part), 20-40 em (very approximately). Characterized by Metal Age pottery,
PtQ H/K P/Q
DEPTHS
eM
...
.:0:
0
z
0 10
10- 20 38
20-30
30-40
40 SO 58
50-60
60-70
67
70-80
80-90
90-100 64
0-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
90-100
100+
Fig. 6 Top, stone tool categories and chert colors at Leang Tuwo Mane'e, by 10 em level; bottom,
distribution of chert, pottery, and shell at Leang Tuwo Mane'e, by depth.
p/Q H/K
r--- r-- -
1 CM CM &U
r--- 20 1 ....c(
.....
1000-
r-- 20 .....
~
213 &U
~
A.D. ~
&:II::
0- c(
&U
B.C. r--- 70 40
2
1000-
4
-u....
::a::
....
o
&U
2000- 100 Z
60
3000- -80
~
I~
3 I~ &U
100 Iv
4000- - I:::
lAo.
Fig. 7 Proposed correlation between absolute dates, natural layers, depths,

and cultural levels at Leang Tuwo Mane' e.
copper or bronze, imported beads, and a very sharp decline in percentages

of stone tools, especially in P IQ.
4. Late. Layer 1 in both sectors and the upper part of layer 2 in P/Q. Imported
ceramics and Rarangunusa earthenwares (see below), possibly dating from
Sung or Yuan times to the present.
Dates for the L TM deposits are derived from archaeological considerations and
from five radiocarbon samples, four of charcoal and one of shell. These dates are
shown in stratigraphic location in Figure 5. A dating scheme for the site, summarized
in Figure 7, provides the essential framework for the following discussions.
The earliest date of 4860 ± 130 B.P. (AND 1717) comes from Turbo shell
near the top of layer 3 in trench H9, and it dates the Pre ceramic blade industry,
a little prior to the first appearance of pottery. The next date, 4030 ± 80 B.P.
(AND 1515), comes from charcoal near the base of layer 4 in the P/Q sector, and
dates the first appearance of pottery in the site. Allowing for carbon-14 calibration,
these dates suggest that the site was first occupied, in the H/K sector, at possibly
4000 B.C. Pottery then appears about 2500 B.C. or perhaps a little before. The date
of ca. 4000 B.C. for initial use of the shelter is of interest, since the surface of the
basal beach sand deposit in L TM is now between 6.5 and 7 m above high sea level.
I would suggest that the shelter was actually at sea level prior to first occupation.
Had it been high and dry it would surely have been used, and uplift of the surround-
ing coastline by about 7 m in the past 6000 years seems to me to be the only reason-
able explanation.
The three other dates from the site (all charcoal) are quite consistent, and come
from the Early Metal and Late Period layers in the P/Q sector. The date of
990 100 B.P. (AND 1715) comes from the lower part of layer 2 at 50-60 cm, and
falls within the Early Metal cultural level. The higher dates of 410 60 (AND
1514) and 250 ± 70 (AND 1513) are Late Period, and come from the upper part of
layer 2. There is also a sixth date of 530 ± 70 B.P. (AND 1716) from the 60-70
em level in trench H9, but this trench was clearly disturbed by a deep pit and the
date is therefore not shown on Figure 5.
Economically, there is little to report for L TM until many more analyses have
been completed. The shell percentages shown in Figure 6 might be noted here, for
these do not fall off in later levels in the way that Peacock (1971) has shown for
Malayan Hoabinhian sites. Clearly, shellfishing was always a favored activity for
L TM inhabitants, and remains are prolific throughout. The shells present are so
far unidentified. Other economic material, mainly from the recent pit dated to
530 ± 70 B.P. in trench H9, includes fishbone, plates from tidal invertebrates, and
Canarium nuts. No cultivated plant remains were found, despite use of saltwater
flotation, and no mammal bones were found.
The LTM Chert Industry (Plate Ii-m)

All stone tools from the L TM deposits are of chert, with the exception of one
obsidian nodule from the base of the Neolithic cultural level in the P /Q sector. The
source of this obsidian is unidentified, but it could come from Minahasa or the
Sangihe Islands. The cherts are of two main types: a reddish-brown, averaging
around Munsell color 2.5YR4/4, and a pinkish-gray, averaging around Munsell
color 7.5 YR7/2. Other colors, mostly intermediate between these two, are numeric-
ally insignificant.
The chart showing the distribution of stone tool characteristics by depth (Fig. 6)
reveals a number of very sharp differences between the P /Q and H/K sectors. The
P /Q industry. which is throughout associated with Neolithic ceramics, lacks blades
and is almost entirely made of brown chert. This brown chert, in nodules too small
for blade production, is scattered today all over the beaches and tracks around
Arangka'a. The P/Q inhabitants clearly used these nodules.
On the other hand, the H/K industry, from the HjK levels below 30 cm, is
characterized by up to SO % of blades and by a markedly higher percentage of gray
chert, although the brown chert was also in use throughout, particularly in the lowest
levels. The gray chert, according to my observations, does not lie around on the
ground in large quantities, and the H/K inhabitants may have gone to special
lengths to acquire it in nodules large enough for blade production. Flakes do
increase in percentages in the later Neolithic levels of H/K, and these flakes are on
brown chert as they are in P /Q.
My interpretation of the foregoing situation is that the Preceramic inhabitants
of the H/K sector used blade tools made of gray chert and also flake tools of the
more common local brown chert. Following this, the Neolithic inhabitants of both
sectors gradually stopped using the gray chert and simply made small flakes from
the brown chert. Although blades of gray chert are found as high as the Early
Metal levels of the H/K sector, I strongly suspect that this may indicate reuse of
tools already in the site, together with scuffing. Since there was no Preceramic
occupation in the P/Q sector, the factor of reuse would not apply here. My inter-
pretation is thus basically chronological, and I regard this as more likely than the
possibility of two contemporary but separate groups in the shelter, each using
different raw materials to make different types of tools.
For the purposes of analyzing the L TM assemblage, a blade is defined as an
elongated flake with length equal to or greater than twice the width, with parallel
sides, and with parallel ridges and scars down the dorsal surface (d. Crabtree
1972: 43). Blades tend to be produced from prepared cores which are struck in one
direction around the circumference of a flat striking platform. The resulting cores
often tend to have a cylindrical or conical shape. Such cores are present but rare
at LTM. In addition, the L TM blades (Plate Ii, k, m) are thin, have low edge-
angles, and occasionally present thin undamaged striking platforms which may
suggest the use of soft hammers, although stone hammers were certainly used as
well. Some of the cores show evidence of bipolar working on anvils (Plate Ii).
The L TM tools are similar in size to those from Paso, although percentages of
chunks are lower, and there are no flat-based and high-ridged tools. Blades fall
between 15 and SO mm in length, with modes between 20 and 30 mm. The flakes
have a similar size range, with dimensions under 10 and over 40 mm being rare.
Triangular coordinate plotting of length : breadth : thickness ratios adds an interest-
ing observation-in all levels throughout P/Q and H/K the ratio of thickness to
combined length breadth thickness is always between 10 and 20%. When
compared to the Paso obsidians, the L TM cherts show much greater control of
relative thickness.
Tool Function in the LTM Assemblage

The L TM and Paso assemblages clearly differ from each other in several crucial
ways, apart from the obvious difference in raw material. The Paso tools tend to
have high-angled edges, often retouched, that are always used with a scraping
BELL WOOD: Minabasa and tbe Talaud Islands 263
motion. The L TM tools tend to have lower edge-angles, negligible retouch, and
seem to have functioned as knives for cutting and perhaps whittling.
Use-wear on the L TM assemblage was much more difficult to define than for
Paso. Retouch was only present on six of the total of 655 tools, and two of these
examples appear to be tanged flakes, while the other four are of indeterminate form.
What may appear to be retouch on some cores has been interpreted as striking-
platform damage caused during manufacture. Otherwise, L TM use-wear consists
only of very slight edge-chipping. Although some of this could have taken place by
various means after the tools were discarded, most of the edge-chips did show what
are believed to be nonrandom concentrations caused by use.
Between 5 and 52 % of L TM tools in each 10 cm level show use-wear, and there
is no pattern behind the frequency variation. The average per level is between 10
and 20%. Lengths of edges showing use-wear are distributed fairly evenly between
5 and 40 mm at all levels, and angles range from 20 to 90 degrees, with a high
concentration between 30 and 60 degrees. Edge-angles and edge-lengths do not
correlate in any clear fashion.
The use-wear was observed to be either unifacial or bifacial. Bifacial wear is more
common (40% of total use-wear) in the H/K blade assemblage than it is on the
P/Q flakes (30% of total use-wear). In addition, bifacial edges tend to be slightly
longer than unifacial edges, and I would interpret them as knives used with a
to-and-fro motion. Bifacial edge-angles never exceed 60 degrees. Unifacial edges
tend to be shorter than bifacial edges, they very often have edge-angles up to 90
degrees, and they are more common in the P /Q flake assemblage. They are inter-
preted as whittling rather than scraping tools of the Paso type.
So far, I have not prepared detailed statistical tests to support the foregoing
observations, since the data base does not have sufficient rigidity. I know of no way
to resolve this problem at present. Peterson (1974a) has recently discussed some
of the many variables which must be assessed in analyzing use-wear on stone tools
from northern Luzon. The flake-blade industry of jasper described by Peterson for
the Dimolit site on Palanan Bay (discussed above in connection with the Paso
ceramics) has been divided into saw, scraper, burin, wedge, spokeshave, awl, and
whittIer categories, but I would hesitate to identify such a range of functions for
L TM. Peterson has provided a useful discussion of edge-gloss and notes that
bamboo-gloss is more widespread over the surface of the tool than the gloss caused
by cutting other grasses. Blades and flakes with edge-gloss do occur right to the
base of the L TM deposits in small numbers, and they seem to be of the nonbamboo
type. Similar flake-blade industries with some edge-gloss, roughly contemporary
with L TM, are also reported from islands in the southern Samar Sea (Scheans et al.
1973), from the Panhologan II Cave on Samar (Hutterer 1973a), and from eastern
Timor (Glover 1972) and southwestern Sulawesi (Glover 1975).
The L TM assemblage, together with the Luzon, Samar Sea, and Timor assem-
blages mentioned above, is one of a number of similar flake-blade industries
widespread through the Philippines, eastern Indonesia, and into Australia from
about 5000 B.C. onward. The Duyong Cave flake-blade industry from Palawan
(Fox 1970), dating to between about 5000 and 2000 B.C., is of similar type, as of
course is the Toalean of southwestern Sulawesi. In the cases of L TM and the
Toalean sites, the blade element seems to predate the flake element, with the latter
being specialized into the "microliths," backed-flakes, and Maros points of the
Middle and Late Toalean (van Heekeren 1972: Fig. 24; Glover 1975). Whether this
trend will be visible in other areas remains to be seen, but there may be a tendency
for blade industries to be Late Preceramic, while less elongated flake-tools become
more common during the Neolithic. The overall picture may indeed be complex,
for Hutterer (1973a) has reported a flake industry without blades from the Sohoton
Cave on Samar, which seems to be contemporary with the Late Preceramic flake
and blade industries elsewhere, and which also has parallels in manufacture with
the Paso assemblage. Hutterer suggests that the Sohoton Industry was made by
Negrito hunters and gatherers, and that the flake and blade industries are associated
with early horticulturalists. The latter is quite possible (except for Australia),
especially if the development of horticulture can be traced to Late Preceramic times.
The LTM Ceramics

Since LTM was purely a habitation shelter, the pottery does not reveal the
intricate decoration characteristic of Early Metal Period jar-burial sites, such as
Leang Buidane to be described below. Most of the pottery is plain and of simple
forms, mainly bowls with restricted orifices and everted rims with rounded lips.
Although sherds were uniformly small and no complete vessels could be recon-
structed, there is no evidence for any bases other than round, nor are there complex
elements such as ring feet, tripods, lids, handles, or spouts. These are, on the
contrary, much more common in Early Metal burial assemblages.
Since the L TM ceramic deposits span about 4500 years, it comes as no surprise
that there are many significant changes in attributes over time. These are discussed
according to the three cultural levels in L TM, which are attributed to the Neolithic,
Early Metal, and Late periods respectively. The rims illustrated in Figure 8 are all
from the P /Q sector.
The L TM sherds have been analyzed for all attributes apart from paste and
temper. The latter can be described only generally: the pottery is soft, bonfired, and
tempered with either river or beach sand. Women potters at Geme village near
Arangka' a use river sand today; they build pots either by modelling from lumps or
by coiling, and they finish with a paddle and anvil. Although local potting is
virtually extinct, we were able to see one demonstration during the L TM excava-
tions, and I have in my possession notes on all stages of manufacture prior to firing.
The prehistoric L TM pottery was almost certainly made with paddle and anvil,
and probably by coiling too; the latter method was certainly in use with the Leang
Buidane Metal Age pottery.
The pottery from the Neolithic cultural level at LTM is thin: body sherd thick-
nesses range between 2 and 7 mm, with a mode of 4 mm. No sherds are thicker than
7 mm, as may be seen from the graph for layer 4 in the P /Q sector, in Figure 8, top.
Rim profiles have rounded lips, low inclination angles generally approaching 90
degrees, and short stubby eversions. All vessels seem to be globular pots with
everted rims, apart from a very few bowls with direct rims and probably vertical
sides. About 10% of the sherdS are red-slipped on their exteriors, although some
vessels had red slip restricted to a band around the inside of the rim. No other form
BELLWOOD: Minahasa and the Talaud Islands
p"Q BODY St<ERD THICKNfSSfS
- LAYER 1 N.S07
i - - - LAYER 213 N.1625
..
:IE
--- -
- , - LAYER 4 N-U5
.
w
Z
~
"
.. -4 "-'-'-'- - " / ,. --- .::: ::--
Fig. 8 Pottery from Leang Tuwo Mane'e: tc;p, body sherd thickness distributions by natural layers;
bottom, rim profiles by 10 em level. Stippled halos denote red slip; rn = rim notched; rr = Raran~
gunusa decoration.
of decoration is present. Pot diameters range between 12 and 22 em, and there is
little change in this range throughout the L TM layers.
The pottery from the higher cultural level of Early Metal affinity trends away
from the Neolithic forms, but with no sharp breaks. Body thicknesses now range
from 2 to 13 mm, with a mode of 5 mm (Fig. 8, top: layer 2/3), and thick sherds
between 8 and 13 mm appear for the first time. Lips remain rounded, but inclination
angles increase and become much less angular. The profiles shown for P/Q 30-40
cm show this clearly, and this trend can be traced back to 70-80 cm, thus emphasiz-
ing the gradualness of the change, and the continuity. Red slip becomes very rare
in this cultural level, but it does continue with quite high percentages in H/K,
although this is because the H/K pieces are mostly from one large vessel broken on the
spot. The red-slipped sherds in the Neolithic cultural levels come from at least four
vessels. Other important ceramic appearances in the Early Metal levels are carina-
tions, thickened lips (e.g., the two right-hand profiles in the 50-60 cm level in
Fig. 8), and incision and cord-marking of Leang Buidane type (to be described).
The Late Period body thicknesses at L TM (Fig. 8, top: layer 1) range from 3 to
15 mm, and very thin sherds of 2 mm thickness now drop out of the distribution.
The skew caused by very thick sherds increases in length, and there is a definite
tendency at L TM for body sherds to get thicker through time. Late Period rims
are basically similar to those in the Early Metal levels, but very important new
occurrences comprise Rarangunusa incised decoration (to be described below) and
Chinese pottery. The latter ranges in date from T'ang onward, although I would
hesitate to claim that Chinese pottery was actually arriving in Talaud prior to
Sung times.
The L TM ceramic sequence thus presents a gentle gradation with no sign of any
sharp interruption. Sherd thicknesses increase gradually, as do inclination angles
and lengths of rims. Necks tend to become more rounded and less angular over
time. Carinations, incised decoration, fine cord-marking, and distinctive thickened
rims appear in tiny percentages in the Early Metal Period, but otherwise the pottery
remains plain and of simple household type. Most of the ceramic innovations which
characterize the Early Metal Period in Island Southeast Asia are confined to burial
assemblages. The L TM household pottery provides rather important evidence for
basic continuity in the region from Neolithic times onward.
The Early Metal and Late Period ceramic parallels for L TM will be considered
later, as more material remains to be presented from other sites. However, L TM
was the only site excavated in Talaud with Neolithic ceramics, and some comment
is necessary here. Pottery predating 1000 B.C. has rarely been reported from the
islands of Southeast Asia, apart from Taiwan, but recent work is altering the
situation very rapidly. From shelters in eastern Timor, Ian Glover (1972) has
excavated sherds of plain globular pots and open bowls, apparently not slipped.
dating to about 2500 B.C. After examining the L TM material in Canberra, Glover
informs me that it resembles quite closely this early Timor pottery, particularly in
form. Glover has also reported plain unslipped sherds from round-based pots with
everted and thickened rims from the Toalean site of Ulu Leang in the Maros region
(Glover 1975). and these may also date back to about 2500 B.C. Furthermore,
Peterson (1974a: 108) has reported sherds with some red-slipping, of unknown
form, from the Pintu rock-shelter in the Upper Cagayan basin of northern Luzon.
These may date back to 2000 B.C. Spoehr (1973) also reports red-slipped ware of
presumed Neolithic context from the Sangasanga rock-shelter in the Sulu Archipe-
lago. So we now have reported occurrences of plain pottery. usually globular in
form with everted rims, and sometimes red-slipped, extending through the greater
part of eastern Island Southeast Asia from 3000 B.C. to perhaps 1000 B.C. This, in
---=-
o
f -I
3CM
a-e
Plate II Items from the Talaud excavations: a, red-slipped dish with notched lip and flat ,base
from Leang Balangingi; b, section of hexagonal vessel foot from Leang Balangingi; c, typical
Rarangunusa decoration, from Tanjong Rarangunusa; d, incised tripod or tetrapod from Leang
Balangingi; e, carinated and lip-notched sherd from Leang Buidane, with herringbone paddle-
impression; I-h, cylindrical plugs of coral limestone, 1 and g from Leang Buidane, h from Leang
Balangingi; i, copper or bronze bell from Leang Tuwo Mane'e.
fact, is our first archaeological definition of the Neolithic in this area, and it is
beginning to provide a stimulating alternative to the adze migrations of former
scholars.
Nevertheless, the definition of this island Neolithic is still hazy, and ceramics
alone can tell us little. More details on the use of the red slip would be very helpful,
for this is clearly not confined to the earlier Neolithic and is in fact very common in
possibly late Neolithic assemblages such as those from Dimolit and Paso, and the
important late Neolithic assemblages from the Batungan caves 1 and 2 on Masbate
(Solheim 1968). The Batungan material has some incised decoration and a relatively
late carbon date of 750 ± 100 B.C. (Fox 1970: 98), but its rim form~ are similar
to those in the L TM Neolithic (Solheim 1968: Fig. 4). Red-slipping is also very
common in the Early Metal Period, and it is only at this time that cord-marking
makes its first appearance in the Philippines and eastern Indonesia. This is a point
of interest, given the overwhelming importance of cord-marking in Mainland
Southeast Asian Neolithic assemblages.
Other Artifacts from the LTM Excavations

Apart from pottery and stone tools, little was recovered from L TM, and the total
absence of stone adzes is disappointing. A copper or bronze heel-like plate of
unknown use was recovered from the Early Metal levels, as was a small copper or
bronze bell (Plate IIi) of shape similar to an example from Sa-Huynh in South
Vietnam (Parmentier 1924: Fig. 17), but lacking the spiral surface decoration
present on the latter. Since the Sa-Huynh Culture dates most probably to the later
first millennium B.C., this bell could then indicate the use of metal in the Talaud
Islands by as much as 2000 years ago. The only other metal object from L TM is a
copper or bronze finger ring, recovered from near the surface, that still awaits
cleaning.
Items of shell from L TM come from the Early Metal and Late levels, and
comprise 2 small perforated discs, a perforated whole bivalve and a perforated
cowrie shell, and a section of a thick triangular cross-sectioned Tridacna bracelet.
Bracelets like the latter are known from other Early Metal contexts in the Philip-
pines (Solheim 1964: 186-191). but are difficult to date. Because ofthe nature ofthe
L TM deposits, I am unwilling to suggest specific dates for any of the above objects.
Other Sites in the Vicinity of Leang Tuwo Mane'e

Thirty meters north-northeast of the main L TM shelter lies another small
shelter with a floor area of about 7 by 5 m. Excavations here revealed about 30 cm
of cultural deposits belonging entirely within the Late Period, with Rarangunusa
incised decoration. About 50 m southeast of the main shelter is a second small
shelter, with a floor area of 11 by 3 m. This had a liberal scattering of Rarangunusa
sherds on its surface, but was not excavated. Neither of these two small shelters is
thought to warrant further excavation.
LEANG BUIDANE, SALEBABU ISLAND

The Leang Buidane cave (henceforth LB) lies 2 km northwest of the town of
Lirung on the coast of Salebabu Island. The site is of great importance, although
entirely different in character from L TM. It contained an Early Metal Period jar-
burial assemblage dating from the later first millennium A.D., with ceramics,
imported beads, copper, bronze, and iron. The site was never used for habitation
and contains no economic refuse.
The plan and cross-section of Leang Buidane are shown in Figure 9. The cave is
18 m long, a maximum of 9 m wide, and roof height averages 2.5 m. Like LTM it is
formed in raised coral limestone; the site is now 24 m behind the beach and the
surface of its beach sand fill is about 5 m above high sea level. Like L TM, the cave
may have risen to its present position in Holocene times, although in this case there
is no Preceramic or Neolithic settlement within. Eight chert flakes and one tiny
blade-core found among the jar-burial deposits might attest to short visits by earlier
peoples, but this evidence is very €lim indeed.
The LB cave has a floor area of about 100 mll. The rear 35 m 2 , behind the excava-
tion trenches, contained no jar-burials, and only scattered Rarangunusa sherds were
-=-
o
LEANG BUIDANE
3M
LONGITUDINAL SECTION A-B
CULTURAL LAYER
--- --
Fig. 9 Leang Buidane: plan and section of site, with contour intervals of 10 em.
found here, the results of recent visits. The jar-burial deposit itself occupied the
front 65 m 2 of the cave, and 12.5 mil (about 20%) ofthis area was excavated. There
arises the problem of why the cave was not used before the Early Metal Period, for,
although the mouth is now very low due to the buildup of deposits around the
drip-line, the section shown in Figure 9 shows that these deposits postdate the
jar-burial interments, and they are formed of materials recently fallen from the
cliff face above the cave. Hence, prior to the jar-burials, the cave should have had
a good clear entrance eminently suitable for habitation. There is no obvious answer
to this problem. Today, the partial blockage of the entrance has made the cave quite
dark and airless inside, and all excavation had to be carried out with pressure lamps.
When the cave was first entered, we found an earthenware casting-mould for a
copper axe, a carnelian bead, and a glass bead on the surface, together with a number
of sherds. These had clearly been disturbed out of the buried cultural layer . During
excavation of the area shown in Figure 9, which was laid out and recorded in 1 m
squares, the cultural layer was found to lie between 10 and 20 cm beneath the surface,
covered by a layer of dry dusty soil. This ruled out any recent funerary use of the
cave. No Chinese imported sherds were found in the deposits, apart from 6 pieces
in this top layer.
Beneath the top sealing layer, the main cultural deposit occupied a thickness of
30 to 40 cm in the center of the cave, less near the sides. This deposit was absolutely
packed with fragmentary human bones, sherds, and other artifacts in remarkable
density. Some 15,200 sherds were recovered from the excavations, and the cave
could contain an estimated 50,000. The rather strange problem here is that no
complete pots were recovered at all, although there were a number of near-completes.
Furthermore, sherds of quite a number of individual vessels were found scattered
throughout the whole depth of the deposit over several square meters. Quite clearly,
none of this material was in situ when excavated, and ancient disturbance had taken
place long enough ago to allow for formation of the top sealing layer. Whether this
disturbance is to be related to the burial practices themselves (i.e., ritual smashing
of offerings) I cannot say, but there is also a possibility of treasure hunting from the
Spanish-Portuguese period, which in this area could have begun in the sixteenth
century. Recent visits for bat-hunting and other purposes could also have caused
damage, since the cave deposits are very soft and dry, although trampling alone
cannot account for the degree of disturbance. Saurin (1973) has suggested that
burial jars in the Hang-Gon site in South Vietnam were smashed purposefully after
placement, a kind of activity that is also evident in the Leang Balangingi site on
Karakellang, to be described below. In addition, Fox (1970: 53) records that
mediums in Palawan have been said to break pots in burial caves during times of
epidemics. Possible explanations for the LB situation are clearly numerous.
Because the ceramic material from the cave is so uniform in style, this disturbance
may pose few stratigraphic problems, since I suspect that the burials were placed
by a related group over a short period of time, perhaps a few centuries. However,
the scattering does make the archaeologist's job a little harder. The cave probably
contains sherds from a finite number of complete vessels, and so far some progress
has been made in piecing them together. However, many pieces of the jigsaw puzzle
are still in the cave, and laboratory sorting of the analyzed material is still far from
finished, due to the time~consuming nature of the job. Some of the reconstructed
vessels are shown in Figure 11, but otherwise, apart from recording attributes of
individual sherds, there is still a long way to go with reconstruction.
The majority of the human bones from the cave would appear to have been placed
in jars, as discussed below. However, at the north end of the excavated area five
separate skulls were found, in two cases with a few associated postcranial bones,
right at the base of the deposit beneath the jar-burial layer. The two skulls with
postcranial bones appear to have been placed between and under large lumps of
coral, and may thus have been placed in scoops in the cave floor and covered over.
These burials did not have any goods, and although some sherds were found among
them, I feel these could be derived from the jar-burials immediately above. There
is thus some evidence that the cave was used for burial prior to the period of the
jar-burials, but the details are very imprecise. Although skull fragments were
plentiful among the jar-burial remains, only the five skulls mentioned seem to have
been in situ when excavated.
The Leang Buidane Ceramics

The 15,200 sherds from the cave break down into 93 % plain, 4.4% red-slipped.
1.7% incised, 0.15% carved-paddle-impressed, 0.18% corded, and 0.15% Raran-
gunusa (top layer only). These percentages do not correspond to percentages of
whole vessels, since most decorated vessels have more than 50 % of their surfaces
plain, and the large burial jars, which produce the majority of sherds, are generally
entirely plain. In terms of whole vessels, these percentages would have to be
increased for all but plainware by an unknown amount.
The LB ceramics clearly relate most closely in form and decoration to the
Kalanay ceramics of the central Philippines (Solheim 1964), rather than to the
Tabon ceramics of Palawan (Fox 1970), where paddle impression is found on over
40% of the sherds from some sites. Unfortunately, no records survive of actual
methods of burial placement in Kalanay sites, but for the Tabon sites there is good
evidence that the large jars containing bones were placed upright on the floor of
the cave, while the smaller offering pots were placed around and perhaps sometimes
inside the larger jars. It is possible that the LB vessels were once placed similarly.
The many human bones found in the LB deposits were in unknown relationships
to the original positions of pots. None of this bone has yet been analyzed. However,
it seems very likely that all the jar-burials were secondary as in the Tabon caves,
and were not of complete skeletons.
The large burial jars at Leang Buidane were either round or of rectilinear shape,
and generally undecorated. Round vessels appear to have had cylindrical or globular
bodies with everted or thickened rims (Fig. 10a-e, g), and some may have had
ring-feet with small perforations (Fig. 101). Square or rectangular vessels have more
complex rim shapes, often with ledges or flanges, perhaps to support lids (Fig.
10h-h). No complete profiles of any burial jars can yet be reconstructed, and no
temper analyses have yet been undertaken on any LB material. Plain jars and
rectilinear forms such as these are unfortunately rather undiagnostic, particularly
when fragmentary; there are clearly many generalized parallels right through the
Tabon, Kalanay, and Sa-Huynh jar-burial cultures, although rectilinear vessels
appeal to be unknown from the latter.
perfora.tions
I)
__ ==-....3CM
o o
Fig. 10 Pottery and beads from Leang Buidane (stippling denotes red slip, numbers denote
diameters in em.): a-g, round vessels; h-k, rectilinear vessels; I, perforated ring foot; m, etched
barrel bead of black agate; n, identical specimen from Sirkap Mound, Taxila, dating from first century
A.D. (after Beck 1941: PI. II, 17); 0, flattened lozenge of carnelian: p, octagonal bicone of carnelian.
The smaller offering vessels from Leang Buidane are often much more complete,
and comprise round-bottomed carinated vessels, either plain or with decoration in
horizontal bands above the carination (see Fig. 11). Notching on lips and carina-
tions is common, especially on plain vessels. Rims are normally everted slightly,
and some show slight thickening and lip-squaring. The vessel shown in Figure l1a
is the only one known so far with a vertical direct rim. Also present are red-slipped
bottle necks (Figure 10f) and tripods or tetrapods (similar to Plate lId), although the
latter cannot yet be fitted to any actual pots. Lids are present too, perhaps for the
large burial jars.
The decoration on these small vessels is red-slipped, impressed, or incised.
Red-slipping and incision often occur together, with the former being on the inside
of the pot (Fig. lIb, f, g, i). The incision is nearly all done with a single point, and
comprises the normal spiral, rectangular meander, crossed diagonal, and triangular
motifs so well known from Southeast Asian sites. The pot shown in Figure I1b
has shallow cutout triangles beneath its lip, and that in Figure 11f has two-pronged
incision, but both these vessels are unusual. The "sloping-S" designs on Figure I1f
are very similar to those on pottery from the Sasak shelter in the Kalatagbak area
of central Palawan-a site which Fox (1970: 169-171) dates to between A.D. 400
and 600. These dates seem very reasonable. Some of the LB vessels also have an
upper zone of paddle-impression, as shown in Plate lIe.
Parallels for these offering vessels from Leang Buidane are fortunately very easy
to find, particularly from the Kalanay and Tabon sites- With the Kalanay sites of
the Visayas (central Philippines), the LB assemblage shares the general forms of the
burial jars, the carinated pots, bottles, and tripods or tetrapods, although it lacks
the complex cutout ring-feet of the Bagupantao vessels from Kalanay Cave.
Similarities to ceramics from the Tabon sites are about even, except that Tabon
lacks any tripod forms, as well as the cutouts in ring-feet.
In terms of decoration, LB shares with Kalanay the running scrolls and meanders,
but it lacks the paired diagonal and border motif which, Solheim (1964) states,
characterize Kalanay pottery, and it also lacks the long rolled facets on lips and
carinations which are distinctive of both Kalanay vessels and also the recently
discovered vessels from the Panhologan burial cave I on Samar (Hutterer 1973a).
LB also lacks the punctate fields between paired incised lines (e.g., Solheim 1964:
Pl. 17h), although this format is known from the Leang Balangingi site, to be
described below. Both Kalanay and LB are characterized by very low frequencies
of paddle-impression, both carved and cord-bound.
From Solheim's 1964 monograph on the Kalanay sites, some very close specific
similarities between Kalanay and LB material can be listed, and these more than
overcome the negatives listed above. Solheim's Plate lsf shows a ring-foot with
scroll-shaped cutouts from Siquijor, and LB has virtually an identical piece. I would
be almost inclined to suggest a single source of manufacture for them both. Solheim's
Plate 19 (Unknown Site, probably Visayas) shows incised scroll designs absolutely
identical to examples from LB, and the same applies to the sherds with notched lips
and rectangular meander designs in Solheim's Plate 21 (again from the Unknown
Site). Several more examples such as these could be listed, and the evidence clearly
suggests that LB is part of the same ceramic tradition as Kalanay, but with a few
obvious and expectable differences.
BELL WOOD: lV1inahasa and the Talaud Islands 273
o SCM
..... ......... ..
. . ... .... .... .
~ ~ ~
• • • • 4 ••• ~ .......... .
, ~ ~ , ,
.... . ............ .
,
0., ••• " ' •
: : . ; . : ::: ::: : :: ; :: ~
Fig. 11 Decorated vessels from Leang Buidane. Stippling denotes red slip
(note: designs not foreshortened for perspective).
Turning to Tabon, the picture is similar, but different in specifics from that
presented by the parallels with Kalanay. The Tabon vessels do not have such
closely related forms of scrolls and meanders as do LB and Kalanay, but a carinated
vessel with incised triangular motifs from Diwata Cave (Fox 1970: Fig. 27c) is very
close to the LB vessel shown in Figure 11d. However, Tabon does differ from LB
in the high frequency of paddle-impression, and my impression is that the latter
relates more closely to Kalanay.
The similarities to the other well-known jar-burial cultures of Niah and Sa-HuYnh
are more remote, despite the sharing of an obvious substratum in ceramic form and
decoration, and in the practice of jar-burial itself. The jar-burials at Niah and the
Sa-Huynh sites predate A.D. 1 on present evidence, and by the time LB was in use,
Sarawak was already dominated by the Bau-Malay pottery tradition (Solheim 1965),
while South Vietnam was the geographical focus of the Indianized Chamic civiliza-
tion. To the south, the only pottery jar-burials reported elsewhere in Sulawesi are
those from Sa'abang in the central part of the island (Willems 1940), but there is
no ceramic material here like that from LB. Neither does burial pottery recently
recovered from caves in the Maros region (Mulvaney and Soejono 1970; Glover
1975) have any close parallels. The undated Kalumpang material (van Heekeren
1972) relates more closely to the Maros material than to anything from northern
Sulawesi, and I suspect that relationships for the latter area in the Early Metal
Period were very strongly oriented toward the Philippines rather than toward the
south.
All the sites mentioned, '\\ith the possible exception of Galumpang, belong
in the Early Metal Period (Bellwood 1976a) and thus date somewhere between
500 B.C. and possibly A.D. 1200. The period of Chinese imports in the southern
Philippine region appears to have begun with the Sung dynasty (see below),
and the ornately decorated Early Metal Period pottery gradually ceases to be
produced after A.D. 1000. The LB assemblage, from evidence to be presented
below, dates to the second half of the first millennium A.D., and much of the
Kalanay and Tabon material is clearly of similar date, although Fox inclines to
date the latter to before A.D. 500. Since Solheim (1975) has recently reported
Kalanay-like pottery from Asin Cave on the coast of southeastern Mindanao, it is
easy to visualize the great extent of contact throughout the Philippines and north-
eastern Indonesia during the Early Metal Period. This extensive contact certainly
formed a suitable milieu for the later rapid development of the trade in Chinese
ceramics in the region.
Other Artifacts from Leang Buidane

Seventy-five non pottery artifacts were recovered from the Leang Buidane
deposits, as follows:
7 shell bracelet segments
3 beads of shell and bone
1 perforated mammal tooth
1 glass bracelet segment
4 glass beads
8 chert flakes and 1 core, possibly predating the burials
3 stone beads (jade, agate, and quartz)
23 carnelian beads
3 etched agate beads
2 coral bottle-stoppers or earplugs
11 artifacts and fragments of copper or bronze
4 earthenware casting valves, or fragments thereof
3 pieces of iron
1 baked clay penannular earring
Artifacts of Shell and Bone

The seven shell bracelet fragments are fairly crude, and of rectangular or irregular
cross-section. The shell beads are perforated subspherical discs, perhaps from the
bases of Conus shells. There is also a perforated fish vertebra, which may have
served as a bead. The perforated mammal tooth is 55 mm long, but is so far
unidentified. None of these items has diagnostic characters which might assist
dating, and none are figured.
Artifacts of Glass
The glass bracelet segment is of a semitransparent green glass, of triangular
cross-section, and about 50 mm in diameter. Solheim (1964) believes that glass
bracelets come quite late in the Kalanay sequence in the Visayas, but he has no
precisely dated examples. Harrisson (1962b) dates blue and green monochrome
glass bracelets from the Sarawak River Delta between A.D. 700 and 1300 on the
basis of Chinese ceramic associations, and in the Calatagan burials on Luzon they
may have been in use as late as the sixteenth century (Fox 1959: 41). On the other
hand, Fox (1970: 15) also mentions green glass bracelets with triangular cross-
sections from Manunggul Chamber Bon Palawan, associated with a carbon-14 date
of 190 100 B.C., so these artifacts may have had long popularity in Southeast
Asia, and I am unable to give a close date to the LB specimen.
The 4 glass beads from LB are rather restricted in variety. They are of an opaque
frit, roughly cylindrical in shape, and do not exceed 4 mm in any dimension. Three
are blue and one is dark red, the latter resembling the mutisalah beads of Indian
origin illustrated by Lamb (1965a : Fig. 5). The LB beads are therefore restricted
in range of color, and no poly chromes are present.
The literature on Southeast Asian glass beads is in a highly confused state, and
it is virtually impossible to give dated parallels for the LB beads. The sources of
manufacture remain virtually unknown, although Lamb (1965a, 1965b) has sug-
gested that manufacturing centers using scrap Western glass were in operation on
the Malay peninsula during the first millennium A.D.
Artt/acts of Stone
The chert flakes and cores are in the L TM tradition and require no comment.
The two items of coral limestone (Plate IIf, g) are either bottle stoppers or earplugs;
I suspect the latter.
The stone beads from LB are of particular importance, especially the carnelians.
There are twenty-three carnelians, three etched agates, one banded agate, one
quartz, and one possible jade or nephrite. Of the latter three, the banded agate is
of black and gray stone, and is drilled evenly with a metal bit, like the carnelians.
It was almost certainly manufactured in India, where such drilling techniques
were perfected. On the other hand, the quartz bead was made locally and split
longitudinally before being finished. It has a crude conical hole drilled partly into
one end, and clearly does not belong with the Indian group. The jade or nephrite
bead is a fragment of a broadly drilled and fluted cylinder, like the example from
Palawan illustrated by Fox (1970: color plate IA : g). It may be of Philippine
origin; the example shown by Fox is from either Duyong or Uyaw Cave, and is of
Early Metal Period date.
The carnelian beads are of two kinds: faceted and spherical. The faceted ones
range in length from 9 to 22 mm and are of a semitransparent red or orange. They
are drilled with admirable precision from both ends, a well-kno\VIl characteristic of
the class (Lamb 1965a). There are two groups in terms of shape (Fig. lOo,p): the
flattened lozenges with two broad flat sides and the octagonal or hexagonal faceted
bicones. There are seven beads in the former group, four in the latter (two octagonal
and two hexagonal). The spherical carnelians range from 4 to 16 mm in diameter,
but only three exceed 8 mm, and seven of the total of twelve spherical carnelians
have diameters of only 4-5 mm.
The majority, perhaps all, of these beads were probably made in India. However,
fragments of carnelian beads which may come from a workshop have been found
at Kuala Selinsing in Perak, West Malaysia (collections in University Museum,
Cambridge), and it may be that imported Indian carnelian was worked locally in
Malaya. Unless more definite proof is found to support this possibility I will accept
a more conventional Indian origin for the Leang Buidane beads.
The main characteristics of the Indian carnelian bead industry have recently
been summarized by Bridget Allchin (1977). Carnelian, which like agate is a variety
of chalcedony, has been used for beads since Harappan times, and in historical
times one of the main centers for manufacture has been Cambay in Gujerat, using
carnelian mined near Ratnapura (see also Arkell 1936; van der Hoop 1932: 137).
Carnelian and agate also occur in southern and eastern India, and were manufactured
into beads at the Roman trading station of Arikamedu (Lamb 1965a). Unfortunately
the sources of the Southeast Asian trade beads prior to the sixteenth century remain
unknown, and the Cambay industry is of rather uncertain antiquity. However, the
techniques of manufacture have changed little since Harappan times, and involve
roasting to deepen color, flaking and grinding, and drilling with a metal or perhaps
even a diamond drill. The Leang Buidane beads have certainly been drilled with
one of the latter. The faceted beads were ground individually on grindstones, while
the spherical ones may have been rotated in large quantities for long periods in
containers with sand and water.
The date of the export trade to Southeast Asia is hard to establish. Historically
the most extensive trade took place between A.D. 1300 and 1500 with the spread of
Islam (ArkellI936).. As Vlekke (1943: 52) has noted, the first Moslem traders to
penetrate Maluku during the thirteenth century were in fact Gujeratis from Cam-
bay, not Arabs, so reasons for a trade in carnelians after A.D. 1300 are not hard to
find. The carnelians wert: still available at this time; large numbers dating between
A.D. 850 and 1100 have been found at the north Indian site of Ahichchhatra (Dikshit
1952). They were being traded to East Africa by at least A.D. 800 (van der Sleen
1958).
However, carnelians were being traded into Southeast Asia long before the dates
given above, according to archaeological indications. They date back to the begin-
nings of the Early Metal Period and are found in faceted forms identical to those of
LB at Sa-Huynh (Parmentier 1924: Fig. 15), Hang-Gon (Saurin 1973), and Phu-
Hoa (Fontaine 1972) in Vietnam, in jar-burial contexts which probably date
between 700 B.C. and 1. They are probably shown carved on the Tanjungsirih
image in South Sumatra (van der Hoop 1932: PI. 10) in a context perhaps contem-
porary with the Dong-Son Culture of North Vietnam. They are also reported by
Fox (1970: 118) from Manunggul Chamber B on Palawan, in an assemblage dated
to 190 ± 100 B.C.
Several other Southeast Asian sites with faceted carnelians prior to A.D. 500 are
known (e.g., Oc-Eo; Malleret 1962: 205), and the list would be extended greatly
if one were to include the nonfaceted types, which are common as far east as the
Palau Islands, despite their absence at LB. Of more direct relevance to the latter,
carnelians of identical faceted form have been recovered from a pre-Sung burial in
Cebu City (Hutterer 1973b: Fig. 3g) and from several Early Metal sites on Palawan
(Fox 1970: color plate 1A " 0) and Java (I. M. Sutayasa, personal communication).
Harrisson reports them from the site of Jaong in Sarawak, dating between A.D. 900
and 1100 (Harrisson and O'Connor 1970: 143), and they were also being buried
with imported pottery of Sung and post-Sung date in Sulu (Spoehr 1973: 76, 243)
and Samar (Hutterer 1973a).
At the present time, these parallels can do no more than point out the importance
of carnelian beads in Southeast Asia. As far as the LB beads are concerned, date of
manufacture could perhaps be anywhere between 500 B.C. and A.D. 1500. As Dikshit
(1952; 41) states of the faceted carnelians in India, they "are shapes of such common
occurrence and are distributed over such a wide span of time that by themselves
they have little dating value." Since there is no historical evidence for Indian contact
with Indonesia prior to the first century A.D. (Wolters 1967), I would incline to date
the majority found in Island Southeast Asia to between 1 and A.D. 1500.
The three beads of etched black agate from LB (Fig. 10m) are also of Indian
origin. According to Mackay (1933) these beads were produced in Cambay and
Delhi, and the white designs were formed by painting on a mixture of soda and
vegetable paste, after which the bead was baked. The three LB beads are of sharply
contrasted black agate and white designs, and are paralleled with absolute precision
in a bead from Taxila dating from the first century A.D. (Fig. IOn). This form may
date back to about 400 B.C. in northern India (Dikshit 1952: 35). The LB examples
may thus be of first millennium B.C. manufacture, and may well have been in
circulation for several centuries before burial. I know of no exact parallels for the
LB beads from Southeast Asia, but spherical forms with similar etched patterns
are known from Early Metal sites in the Philippines (Fox 1970: color plate lA : j ;
Solheim 1964: PI. 30: I).
Metal Artifacts and Baked Clay Moulds

Bronze or copper objects found at Leang Buidane include many curved fragments
which appear to be from bracelets, a hollow cone-shaped item of bronze, and a
copper socketed axe 65 mm long. Since these objects have been considered in detail
and illustrated in a more recent paper (Bellwood 1976b), only a few comments are
necessary here. The axe has a splayed and curved cutting edge, similar in general
shape to the examples from the Tabon Caves (Fox 1970: Fig. 39). Three casting
valves of baked clay were also found, each presumably being one-half of a two-piece
mould. One of the valves, unfortunately fragmentary, appears to be for a socketed
axe similar but not identical to the one recovered. The other two valves are also
fragmentary; one appears to be for a parallel-sided axe, the other for a tubular
object.
Iron objects include what appears to be a tang for a knife or spearhead, and two
other small rodlike fragments. Places of manufacture of these objects are unclear for
the iron and bronze, although the copper items (including the axe) were clearly cast
locally. Copper is found commonly in Sulawesi and the Philippines. Tin, apparently,
is not, and it may be that scrap metal was imported and recast.
The Date of Leang Buidane

The jar-burial assemblage contains no imported Chinese or Siamese pottery, and
the artifacts in general (particularly beads and ceramics) suggest a date in the first
millennium A.D., perhaps the latter half. One radiocarbon date of A.D. 1440 80
(ANU 1516), obtained from charcoal collected from scatterings in the top sealing
layer in the cave, clearly postdates the assemblage. There is also a thermolumine-
scence date for one of the baked clay casting moulds of 960 years B.P. (courtesy of
Allan Mortlock, Australian National University). This date places the assemblage
at around A.D. 1000, and I accept it as valid. The date of 950 ± 130 B.P. for an
intermediate level of the assemblage of Buidane type from Leang Balangingi is also
directly relevant. Overall considerations would suggest that Leang Buidane, and the
Buidane Culture as a whole, dates to between A.D. 700 and 1200.
LEANG BALANGINGI, KARAKELLANG ISLAND

The rock-shelter of Leang Balangingi is situated about 1.5 km northwest of the
village of Rainis, on the central eastern coast of Karakellang. The shelter is about
! km inland from the sea, and, like Leang Buidane, was used entirely for burials
and offerings, mainly in the Early Metal Period. It contains no economic or habita-
tion remains (apart from some shellfish), despite its apparent suitability for settle-
ment, and this may be a reflection of its inland situation. Once sites such as Leang
Buidane and Leang Balangingi had been used for burial, future generations would
be unlikely to contemplate habitation in such caves, although this explanation cannot
account for the absence of Preceramic and Neolithic remains. This remains some-
thing of a mystery.
The Leang Balangingi shelter has a habitable area 10 m long and up to 5 m wide,
and is thus quite large. It is formed from coral limestone, and has a vertical back
wall and a roof height of over 4 m. Excavation trenches were laid out in the center
of the floor and were excavated to sterile deposits, which here consisted of stratified
clays (denoted layer 3). The shelter may have an ultimate base of coral sand, but
this was not reached.
Above layer 3 are layers 1 and 2, which contain the cultural materials. Layer 1
is a humic soil (Munsell lOYR 4/1) about 20 em thick, which has formed only in
the central and lowest portion of the shelter, and which owes its existence to the
wash of humic materials down the slopes on either side. Beneath layer 1 is layer 2,
more heavily leached and with a Munsell color of 10YR 5/3. Outside the boundaries
of layer 1, layer 2 rises to the surface of the shelter, and is up to 40 em thick in
places. Neither of these layers has any cultural significance, and the main Early
Metal assemblage runs through both. The maximum depth of cultural deposits is
70 cm.
The contents of layers 1 and 2 were kept separate during excavation, but the
material is more meaningfully presented in terms of 10 cm levels. As with LB.
fragments of individual vessels were found scattered from near the surface to a
depth of 30 cm in layers 1 and 2. Oddly enough, no bones (human or animal) were
found at all, but five separate human teeth were found among the sherds. Although
the shelter is rather damp, it is unlikely that bones have dissolved, since there are
shellfish in the deposits. Either the burials were ransacked in antiquity, or teeth
alone were buried in some unknown commemorative rituals; the latter would
appear to be the more likely situation.
We excavated 8.60 m 2 at Leang Balangingi, this being about 25% of the total
habitable area. A total of 2604 sherds was recovered; distributions by depth are
shown in Table 1. The columns for incised, impressed, and carinated sherds, square
vessels, and tripods refer to sherds which in many cases are almost identical to the
material from Leang Buidane. That the Balangingi assemblage is in the same tradi-
tion as Buidane is made even more evident by the rim forms, which have been
analyzed but which are not illustrated here. All the major Buidane types are present.
The Balangingi sherds have been graphed for thickness, and at all levels there are
two quite sharply separated groups-the small vessels between 4 and 7 mm in
thickness and the large burial jars between 10 and 13 mm. The 8 and 9 mm points
are very low on all curves. Clearly the Balangingi material comes from separate
groups of burial jars and offering pots, precisely as does that from Buidane.
TABLE 1. LEANG BALANGINGI: DISTRIBUTION OF POTTERY BY DEPTH

(in numbers of sherds)
TOTAL IM- CARI- RED- SQUARE TRIPODS RARAN- CHINESE

DEPTH POTTERY PLAIN INCISED PRESSED NATED SLIPPED VESSELS GUNUSA
0--10 532 466 13 9 15 5 5 17 2

10--20 1006 943 20 4 10 5 16 3 5
20-30 795 738 12 3 7 12 11 7 5
30-40 152 130 1 2 3 5 6 5
40--50 64 63 1
50-60 52 51
60-70 3 3
Totals 2604 2394 46 9 29 37 39 21 27 2
%sherdsl00 92 1.8 0.3 1.1 1.4 1.5 0.8 1.0

Asian Perspectives, XIX(Z), I976
One particularly interesting discovery at Balangingi comprised the deliberately
smashed remains of a large plain jar, placed under a small flat slab of stone near the
base of the cultural level. This pot was clearly broken prior to burial and the sherds
placed in a small hole scooped in the shelter floor and then sealed with the slab.
No bones were present, and the only possible association is a segment of a copper
or bronze bracelet. This interesting circumstance suggests that deliberate smashing
at placement could account for some of the disturbance in the Leang Buidane
assemblage.
Two other vessels from Balangingi are of interest. One is the flat-based dish
shown in Plate IIa, which is red-slipped inside and out, and has shallow rim notches.
It was again buried purposefully under a small stone slab near the base of the cultural
level, like the burial jar above, and the two vessels are presumably contemporary.
However, the flat-based dish is not evident as a Buidane form, and it does resemble
a common vessel form found in Lapita assemblages in Melanesia (e.g., Green 1973 :
Fig. 4). I am not sure how this can be explained, since the Balangingi vessel is
apparently of Early Metal date, and, like the burial jar, was broken before placement.
The other unusual piece is the large fragment of a vessel foot of apparent hexagonal
plan shown in Plate lIb. This is decorated with running scrolls enclosing rows of
punctate decoration, a type of decoration not found at Leang Buidane. However,
Solheim pictures a similar design on a lid of Kalanay affinity from Leyte (1964:
PI. 17h), and the Kalanay cave itself on Masbate has related material as well
(1964: PI. 7). This type of design is also paralleled on Late Neolithic material from
Palawan (Fox 1970: 113).
Chinese sherds (unidentified) are found only in the top 10 cm at Leang Balangingi,
although Rarangunusa sherds are found down to 30 em, a situation which may
reflect disturbance, since they are never found earlier than Chinese material
elsewhere. It is unlikely that the Rarangunusa ceramic style overlaps in time with
any Early Metal assemblages.
Other artifacts found with the jar burial material comprise one faceted carnelian
bead, three red glass and two green glass beads, a coral "earplug" (Plate IIh) similar
to the examples from LB, two pottery penannular earrings likewise similar to the
one example from LB, and three fragments of copper or bronze, including the
bracelet segment referred to above.
The Leang Balangingi assemblage is dated from charcoal to 950 ± 130 B.P.
(AND 1714) at the 20-30 cm level, and is thus satisfactorily contemporary with
Leang Buidane. Shellfish, so far unanalyzed, were found in all layers. These
could indicate that the shelter was used for occasional habitation. However, the
artifacts from the site are considered to be the results of burials and offerings
rather than habitation. Offerings to spirits are still made in secluded places on
Sangihe, and in 1974 we visited an offering place inland from Lapango village near
the southern tip of the island. Here a number of artifacts, including modern crockery
and coins, had been placed between the buttresses of a large tree. Informants
described it as a place of offerings to spirits. It is not hard to imagine that Leang
Balangingi may have served similar purposes-even the shellfish could be offerings
-and such a circumstance might account for the absence of skeletal material apart
from a few isolated teeth.
I thus interpret Leang Balangingi as a place of burials and offerings, in use

mainly in the latter half of the first millennium A.D. The site is an unquestionable
component of the Buidane Culture, a term which I would like to introduce for the
Early Metal Period in the Talaud Islands.
TAN}ONG RARANGUNUSA AND ITs CERAMICS
Tanjong Rarangunusa (Tanjong = peninsula) protrudes from the northeastern

shore of Karakellang and terminates in a raised coral knoll surrounded by a low
cliff. A very small burial cave is formed in this cliff; it has an entrance 3-4 m wide
by about 1.5 m high, but the inside of the cave rapidly dwindles to a chimney which
leads to the top of the knoll. Only a few hours were spent on the site and no plan
was made, so this description is unfortunately rather generalized.
The cave contained skulls and long-bones stacked in a niche at the back, and one
skull was observed to be sitting on a plate believed to be of Ch'ing date. Above the
cave, on the very rough coral surface of the knoll, we found very large numbers of
Chinese and local potsherds. These sherds clearly resulted from offerings, perhaps
connected with the burial cave, and were not associated with any habitation refuse.
The exact extent of sherdage was not recorded.
The imported pottery collected from the site was all Ch'ing or European; and
I would be inclined to date it to the eighteenth and nineteenth centuries. The local
earthenware, in the Rarangunusa style mentioned several times earlier, was found
in large sherds with clear details of decoration. For this reason, I have chosen Tan-
jong Rarangunusa as the type site for this style.
The decorative characteristics of Rarangunusa pottery are shown in Plate lIe.
Vessels, as far as can be determined, are round-based, globular, and have gently
everted rims with high inclination angles, a high degree of taper, and rounded lips
(see examples indicated rr in Fig. 8). Rarangunusa rims are in fact quite standardized
and can generally be recognized. The percentage of vessels decorated is also very
high (no exact figure can be given), and this clearly assists identification.
Rarangunusa decoration is on upper bodies and necks only, usually from a
carination or point of vertical tangency to the lip. All decoration is done with a
two-pronged tool and is generally vertically zoned. It is entirely unrelated to
anything in the Buidane range. Motifs include vertical paired lines and zigzags, and
the very distinctive trademark of paired notches, usually found in vertical rows.
Lips are frequently notched, and thickened or flattened lips of the Buidane type
are absent.
A point of great interest about the Rarangunusa style is its wide distribution. It
was found in the top layer or on the surface of virtually every Talaud site visited,
generally associated with Ch'ing ceramics. On Sangihe Island, sherds of precise
Rarangunusa affinity were found in sites around Tahuna and Manalu (see below),
and they have also been found at Beong on Siau Island. For Mindanao, Solheim
(1975: 177-178) described what appears to be almost identical decoration on late
pottery from Davao and Cotobato, in the latter case in association with stone burial
jars in caves (Kurjack et al. 1971). Some of the incision from the Panhologan II
cave on Samar could be related as well (Hutterer 1973a: 74). However, no pottery
related to Rarangunusa is known from Minahasa, or anywhere on Sulawesi to my

knowledge.
The Rarangunusa ceramic style may be interpreted as an expression of the high
degree of contact among Mindanao, Talaud, and Sangihe in early historic times.
Much of this contact took place under European and Islamic influence, in cultural
circumstances to be discussed in the concluding section. The style may have survived
until as recently as fifty years ago, although it is not produced today. Concerning
the origins of the Rarangunusa style, I regard it as a household ware lacking the
ceremonial vessel forms present in the Buidane assemblage. Such ceremonial vessels
were quite clearly replaced by Chinese ceramics. The Rarangunusa decoration is
unlike that of Buidane and might have developed in Mindanao rather than Sangihe-
Talaud. Wherever it first developed, the spread of the style is more likely to be a
result of widespread ties of intermarriage rather than of total population movement.
The style is pleasing when done well and may simply have "taken on" as an element
of fashion.
OTHER SITES ON KARAKELLANG AND SANGIHE
On Karakellang and Salebabu, a number of other caves and rock-shelters which

may have cultural deposits were visited. None were excavated, but all produced a
few surface sherds. Leang Totonbatu was the most interesting (Fig. 1, no. 4), being
a burial cave near the top of an outcrop of raised coral limestone. The outcrop is on
the end of the small peninsula called Tanjong Totonbatu. The cave, open at either
end (and thus strictly a tunnel), is about 8 m long, 4 m wide, and 1 m high. It
contains one stack of skulls and another separate stack of long-bones, from approxi-
mately thirty-five individuals. Separate from the bones is another stack of Ch'ing
dishes, many polychrome. Local earthenware sherds were found only on the cave
floor, and some of the material is of Buidane affinity, so the cave might contain
earlier remains below the obvious surface material, which is not more than 300 years
old. The cave also contains one Ch'ing brown glazed storage vessel, which has a
large piece broken out of its side. Informants stated that this vessel once contained
a skull. My impression was that the long-bones and skulls had been stacked recently,
although I cannot be certain. A number of sides of wooden coffins or bone boxes
were lying on the cave floor as well. Apart from a few small sherds, none of the
material in this cave was disturbed or removed.
On Sangihe Island, very little archaeological material was recovered. Two shelters
called Leang Mangki and Leang Kaluhaghi, near Tahuna, were excavated but found
sterile, apart from Rarangunusa sherds on their surfaces. Near Manalu, two shelters
called Leang Bowonleba and Leang Timpalo were visited, and the latter produced
surface sherds of possible T'ang and Yuan pottery, as well as Rarangunusa sherds.
Sangihe was neglected because so much material was found in Minahasa and
Talaud, and because time was short.
A Possible Pebble and Flake Industry in Talaud

While walking along the track which leads across Karakellang from Beo to
Rainis, we collected a number of chert tools. Most finds were made on the western
half of this track, for about 3 km inland from Beo. Several of the specimens are
flakes, and there is a single blade, which would not be out of place in the levels at
Leang Tuwo Mane'e. However, there are also a number of chert pebble tools, as
shown in Plate In. These have varying amounts of cortex, and are both unifacially
and bifacially worked. Nothing like them is known from the excavated LTM and
Paso stone-tool assemblages, and they could be older than 8000 years on these
grounds.
The question arises whether these tools are of Pleistocene date. Certainly, they
are smaller than the Pacitanian tools of Java I have seen in Jakarta, but they appear
to be in a similar tradition. Pebble tools are present in Upper Pleistocene assem-
blages from Niah and Tabon caves, in the Cagayan Valley of northern Luzon
(von Koenigswald 1958), and also in the industries of presumed Pleistocene date
from Flores and Timor (Maringer and Verhoeven 1970a, 1970b, 1972). I hesitate
to suggest Pleistocene settlement in the Talaud Islands without better evidence, but
the material is at least suggestive.
THE MINAHASA AND TALAUD RESULTS IN WIDER PERSPECTIVE
Apart from the possible Pleistocene tools from Karakellang, the earliest site
excavated by the 1974 expedition was the shell-mound at Paso on Lake Tondano,
which dates from about 6000 B.C. The obsidian tools from this site are related to
other, nonblade industries from the Philippines, eastern Indonesia, and Australia,
and they are part of a hunting and gathering way of life that was widespread through
these islands in the earlier Holocene. No definite evidence for horticulture was
recovered from Paso, and the site is basically a shell-mound of a type known from
Hoabinhian contexts on the mainland of Southeast Asia and on Sumatra. The later
red-slipped ceramics from Paso are of unknown date, but have both Neolithic and
Late Period affinities in the Philippines.
The Leang Tuwo Mane'e site on Karakellang belongs to the period after 5000 B.C.
when blade industries had spread through the eastern islands of Southeast Asia, and
its levels also record the arrival of plain Neolithic ceramics at about 2500-3000 B.C.
By this date, linguistic evidence for the reconstruction of early Austronesian
languages would suggest that horticulture was present, as well as a canoe technology
and voyaging ability. Neolithic cultures of the LTM type developed soon after
3000 B.C. from the Philippines down to Timor.
Following the Neolithic, which at present is best known from ceramics, we have
a much richer record for the Early Metal Period, from 500 B.C. to about A.D.
1000-1200. The internal developments of Early Metal cultures are still poorly
known, but Fox (1970) has presented a good sequence for Palawan. He divides the
period into two: an Early Metal Age beginning about 5-700 B.C. with bronze and
glass beads, and a Developed Metal Age from about 200 B.C. with iron, glass
bracelets, and an increasing use of carnelian beads. The Buidane Culture of the
Talaud Islands has very strong Philippine affinities, but is more closely related to
the Kalanay Culture of the central Philippines than to Palawan, and dates to the
latter half of the first millennium A.D. Earlier material may be present at Leang
Balangingi, but I am not certain of this.
Broader relationships in the Preceramic, Neolithic, and Early Metal periods in the
Philippines and eastern Indonesia have been discussed in more detail in two other
works (Bellwood 1976a, 1977: chaps. 3 and 8), and I cannot repeat all of the
details here. The Late Period, as I have defined it elsewhere (Bellwood 1976a),
is characterized by the widespread use of imported Chinese and Siamese pottery,
and it witnesses the demise of the more flamboyant local ceramics of the Early
Metal Period. In the western Philippines and Sarawak, this import trade goes back
as far as the T'ang dynasty (Fox 1967; Zainie and Harrisson 1967), but the greater
part of the trade in the Philippines took place from Sung times (A.D. 960-1279)
onward (Beyer 1948; Cole and Laufer 1912; Reynolds 1967). Historical Chinese
records of this trade go back to A.D. 982 (Scott 1968: 67), and it seems likely from
the findings of the 1974 expedition that Sangihe and Talaud were being reached by
at least this date. For Minahasa I have no definite data.
Identification of social and political changes in Minahasa and Sangihe-Talaud in
the Late Period is much more difficult, and impossible from present archaeological
data alone. There is no good evidence for any direct extension of fourteenth-century
Majapahit control into these regions (Vlekke 1943; Rausa-Gomez 1967), and they
do not appear to have been Islamicized during the fifteenth century. All the evidence
suggests that they remained as small-scale pagan tribal societies until well into the
nineteenth century. This is how Wallace (1962: 186) described the people of Mina-
hasa for the period prior to 1822, when the Dutch introduced coffee cultivation.
Although the four local chiefs of Sangihe were called "Raia" (Rajah) in 1521
(Pigafetta 1969: 64), it is not clear that this term indicates any high degree of
political integration. since its use was so widespread. Archaeologically, there is no
evidence to contradict a view of small-scale tribal societies prior to the nineteenth
century, and the widespread Rarangunusa style may well postdate 1800 and thus be
contemporary with the period of active Dutch government.
In conclusion, the importance of the sites described in this report is that they
give a continuous chronological sequence from 6000 B.C. to the present, with only
a short break between Paso and the basal layers of Leang Tuwo Mane' e. The four
periods represented are Preceramic (8000-3000 B.C.), Neolithic (3000-500 B.C.),
Early Metal (500 B.C. to A.D. 1000), and Late (A.D. 1000 to the ethnographic present).
These periods are taken from my two previous discussions of prehistory in the
eastern islands of Southeast Asia (Bellwood 1976a, 1977: chap. 8), and I have tried
to be systematic without rejecting too much of the terminology previously in use.
Southeast Asia is now far from being archaeologically unknown, and some sort of
order needs to be brought into the data. A long and detailed sequence, such as that
presented here, provides a good excuse for synthesis.
ACKNOWLEDGMENTS
The archaeological work described was conducted in Sulawesi Utara between

24 May and 25 July 1974, with a grant from the Australian Research Grants
Committee. Joint leaders were P. S. Bellwood (Australian National University) and
I. M. Sutayasa (Pusat Penelitian Purbakala dan Peninggalan Nasional, Jakarta).
Other expedition staff included P. F. Sondakh (Kepala Kantor Pembinaan Permu-
seuman at Manado) and his staff members M. L. Roring, C. Sentinuuo, and C. G.
Siwi. In the Sangihe and Talaud Islands we were joined by E. Tatimu (Kepala
Binkab Kebudayaan Kabupaten Kepulauan Sangihe-Talaud), by his staff member
BELLWOOD: Minahasa and the Talaud Islands 28 5
S. Makaluase, and by two members of the Sangihe-Talaud Police Force-H. N.
Dotulong and P. Dalenoh. In Minahasa the team included J. A. Makal (Kepala
Binkab Kebudayaan Kabupaten Minahasa). Other local individuals joined the
expedition staff from time to time, to make an average team of about ten persons.
I would like to thank all of the persons mentioned above, without whom the
expedition could never have taken place. R. P. Soejono of the Pusat Penelitian
Purbakala dan Peninggalan Nasional in Jakarta also gave his support to the expedi-
tion and first introduced the two leaders in 1973. Technical assistance has been
provided by the Department of Prehistory and Anthropology and by the Radiocar-
bon and Thermoluminescence laboratories, at the Australian National University.
I would like also to acknowledge the help of Amtje Clason of the Biologisch-
Archaeologisch Instituut in Groningen (animal bones), and of Robert Griffing of
Honolulu (Chinese ceramics).
Artifacts will be deposited with the Pusat Penelitian Purbakala dan Peninggalan
Nasional in Jakarta by mid-1979.
REFERENCES
ALLCHIN, B.
1977 The agate and carnelian industry of western India and Pakistan. South Asian Archaeology
3 (in press).
ARKELL, A. J.
1936 Cambay and the bead trade. Antiquity 10: 292-305.
AUDLEy-CHARLIlS, M. G., and D. A. HOOIJER
1973 Relations of Pleistocene migrations of pygmy stegodonts to island arc tectonics in eastern
Indonesia. Nature 241: 197-198.
BANDI, H. G.
1951 Die obsidianindustrie der Umgebung von Bandung in Westjava. In Sudseestudien,
pp. 127-161. Basel: Museum fUr Volkerkunde.
BECK, H. C.
1941 The Beads from Taxila. Archaeological Survey of India, Memoir 65.
BELLWOOD, P. S.
1975 Report on archaeological research in Sulawesi Utara. Far Eastern Prehistory Assockltitm
Newsletter 4: 30-36.
1976a Indonesia, the Philippines and Oceanic prehistory. In La Prehistoire Oceanienne, ed. by
J. Garanger, pp. 7-26. Paris: Centre National de la Recherche Scientifique.
1976b The significance of excavated bronze objects from the Talaud Islands, northeastern
Indonesia. In Ancient Chinese Bronzes and Southeast Asian Metal and Other Archaeological
Artefacts, ed. by N. Barnard, pp. 413-420. Melbourne: National Gallery of Victoria.
1977 Man's Conquest of the Pacific. London and Auckland: Collins.
VAN BEMMELEN, R. W.
1949 The Geology of Indonesia, vol. tA. The Hague: Nijhoff.
BEYER, H. O.
1947 Outline review of Philippine archaeology by islands and provinces. Philippine Journal of
Science 77: 205-390.
1948 Early history of Philippine relations with foreign countries, especially China. In Chinese
Elements in the Tagalog Language, by E. A. Manuel, pp. ix-xxv. Manila: Filipiniana
Publications.
BULMER, S.
1975 Settlement and economy in prehistory Papua New Guinea. Journal de la Societe des
Oceanistes 31: 7-76.
COLE, F. C., and B. LAUFER
1912 Chinese Pottery in the Philippines. Anthropological Series, Field Museum of Natural
History, vol. 12, no. 1. Chicago.
CRABTREE, D. E.
1972 An Introduction to Flintworking. Occasional Papers of the Idaho State University Museum
no. 28.
DIKSHIT, M. G.
1952 The beads from Ahichchhatri, U.P. Ancient India 8: 33-63.
ELLEN, R. F., and 1. C. GLOVER
1975 Pottery manufacture and trade in the central Moluccas, Indonesia: the modern situation
and the historical implications. Man (n.s.) 9: 353-379.
FERNANDEZ, C. A. II, and F. LYNCH, S. J.
1972 The Tasaday: cave-dwelling food gatherers of South Cotobato, Mindanao. Philippine
Sociological Review 20: 277-330.
FONTAINE, H.
1972 Nouveau champ de jarres dans Ie Province de Long-Khanh. BSEI 47: 397-486.
Fox, R. B.
1959 The Calatagan Excavations. Reprinted from Philippine Studies 7(3). Manila.
t967 The archaeological record of Chinese influences in the Philippines. Philippine Studies
15: 41-62.
1970 The Tabon Caves. Monograph of the National Museum no. 1. Manila.
GLOVER, I. C.
1972 Excavations in Timor. 2 vols. Ph.D. thesis, Australian National University.
1975 Uiu Leang Cave, Maros: a preliminary sequence of post-Pleistocene cultural develop-
ment in South Sulawesi. Archipelll: 113-t54.
GOULD, R. A., D. A. KOSTER, and A. H. L. SONTZ
1971 The lithic assemblage of the Western Desert Aborigines of Australia. American Antiquity
36: 149-169.
GOULD, R. A., and J. QUILTER
1972 Flat adzes-a class of flaked stone tools from southwestern Australia. American Museum
Novitates no. 2502.
J
GROVllS, C. P.
1976 The origins of the mammalian fauna of Sulawesi (Celebes). Zeitschriftfur Siiugertierkunde
41: 201-216.
GREEN, R. C.
1973 Lapita pottery and the origins of Polynesian culture. Australian Natural History 17(10):
332-337.
HARRISSON. T.
1962a Borneo death. BTLV 118: 1-41.
1962b Translucent glass rings from Borneo. AP 6: 236-238.
1970 The prehistory of Borneo. AP 13: 17-46.
HARRISSON, T., and S. J. O'CONNOR
1970 Gold and Megalithic Activity in Prehistoric and Recent West Borneo. Cornell University
Southeast Asia Program, Data Paper no. 77.
VAN HBBKBREN, H. R.
1972 The Stone Age of Indonesia. 2nd ed. The Hague: Nijhoff.
HOOI}BR, D. A.
1958 The Pleistocene vertebrate fauna of Celebes. AP 2: 71-76.
1969 Pleistocene vertebrates from Celebes. XIII. Sus celebensis Milller and Schlegel, 1845.
Beaufortia 16, no. 222: 215-218.
VAN DER Hoop, A. N.
1932 Megalithic Remains in South-Sumatra. Zutphen: Thieme.
BELLWOOD: Minahasa and the Talaud Islands
HUTTERBR, K. L.
1973a Basey archaeology: prehistoric trade and social evolution in the Philippines. Ph.D.
dissertation, University of Hawaii.
1973b An Archaeological Picture of a Pre-Spanish Cebuano Community. San Carlos Publications,
Series A: Humanities, no. 9. Cebu City, Philippines.
KAUDERN, W.
1938 Megalithic Finds in Central Celebes. Ethnographic Studies in Celebes, vol. 5. GOteborg:
privately printed by the author.
KBBLBY, L. H.
1974 The methodology of microwear analysis: a comment on Nance. American Antiquity
39: 126-128.
VON KOBNIGSWALD, G. H. R.
1958 Preliminary report on a newly-discovered stone age culture from northern Luzon. AP
2: 69-70.
KURJACK, E. B., and C. T. SHELDON
1970 The archaeology of Seminoho Cave in Lebak, Cotobato. Silliman journal 17: 5-18.
1971 The urn burial caves of Southern Cotobato, Mindanao, Philippines. Silliman journal
18: 127-153.
LAMB, A.
1965a Some observations of stone and glass beads in early Southeast Asia. jMBRAS 38: 87-124.
1965b Some glass beads from the Malay Peninsula. Man 65: arrlcle 30.
MACEDA, M. N.
1964 Preliminary report on ethnographic and archaeological fieldwork in the Kulaman Plateau,
Island of Mindanao, Philippines. Anthropo$ 59: 75-82.
1967 Preliminary studies of the figures and ornamentation of some selected jar covers from
Kulaman Plateau (southwestern Cotobato), Island of Mindanao, Philippines. Anthropo$
62: 509-532.
MACKAY, E.
1933 Decorated carnelian beads. Man 33: article 150.
MALLERBT, L.
1962 L'Archilologie du Delta du Mekong. Tome 3: La Culture du Fan-Nan. PEFEO, vol. 43.
MARINGER, J., and TH. VERHOBVEN
1970a Die oberftachenfunde aus dem Fossilgebiet von Mengeruda und Olab~'la auf Flores,
Indonesien. Anthropos 65: 530-546.
1970b Die steinartefacte aus der Stegodon-Fossilschicht von Mengeruda auf Flores, Indonesien.
Anthropos 65: 229-247.
1972 Steingerate aus dem Waiklau-Trockenbett bei Maumere auf Flores, Indonesien.
Anthropos 67: 129-137.
MULVANBY, D. J.
1975 The Prehistory of Australia. Rev. ed. Harmondsworth. Middlesex: Penguin Books.
MULVANEY, D. J.• and R. P. SOE]ONO
1970 The Australian-Indonesian Archaeological Expedition to Sulawesi. AP 13: 163-177.
PARMENTIER, H.
1924 Dep8ts de jarres a Sa Huynh (Quang-ngai, Annam). BEFEO 24: 325-343.
PEACOCK, B. A. V.
1971 Early cultural development in South-East Asia with special reference to the Malay
Peninsula. APAO 6: 107-123.
PBTBRsON, W.
1974a Anomalous archaeology sites of northern Luzon and models of Southeast Asian prehistory.
Ph.D. thesis, University of Hawaii.
197% Summary report of two archaeological sites from northeastern Luzon. APAO 9: 26-35.
PIGAPETTA. A.
1969 First Voyage Around the World. Manila: Filipiniana Book Guild.
RAUSA-GOMIlZ, L.
1967 Sri Vijaya and Madjapahit. Philippine Studies 15: 63-107.
REYNOLDS, H.
1967 Why Chinese traders approached the Philippines late-and from the south. Studies in
Philippine Anthropology, ed. by C. P. Zamora, pp. 463-479. Quezon City: Alemar
Fhoenix.
SAURIN, E.
1973 Le champ de jarres de Hang Gon, pres Xuan Loc (Sud Viet-Nam). BEFEO 60: 329-358.
SCOTT, W. H.
1968 Prehispanic Source Materials for the Study of Philippine Prehistory. Manila: University
of Santo Tomas Press.
SCHEANS, J., K. L. HUTTERER, and R. L. CHERRY
1973 A newly discovered blade tool industry from the central Philippines. Leyte-Samar
Studies 7: 8-14.
VAN DER SLEEN, W. G. N.
1958 Ancient glass beads with special reference to the beads of east and central Africa and the
Indian Ocean. JRAI 88: 203-216.
SOLHEIM, W. G. II
1964 The Archaeology of CentTal Philippines. Manila: Bureau of Printing.
1965 The prehistoric earthenware of Tanjong Kubor, Santubong. SM] 12: 1-62.
1968 The Batungan Cave sites, Masbate, Philippines. Anthropology at the Eighth Pacific Science
Congress, ed. by Wilhelm G. Solheim II. Asian and Pacific Archaeology Series 2: 21-62.
1975 Manuscript on southern Mindanao research. Department of Anthropology, University
of Hawaii.
SPOEHR, A.
1973 Zamboanga and Sulu. University of Pittsburgh, Ethnology Monographs no. 1.
TRINGHAM, R., G. COOPER, G. ODELL, B. VOYTEK, and A. WHITMAN
1974 Experimentation in the formation of edge damage: a new approach to lithic analysis.
Journal of Field Archaeology 1: 171-196.
TuGGLE, H. D., and K. L. HUTTERER, eds.
1972 Archaeology of the Sohoton Area, Southwestern Samar. Divine Word University of
Tacloban, Leyte-Samar Studies no. 6(2).
URDANETA, A.
1911 Narrative of the Loaysa Expedition. Early Spanish Voyages to the Strait of Magellan,
ed. by Sir C. Markham, pp. 41-89. London: Hakluyt Society.
VLEKKE, B. H. M.
1943 Nusantara: A History of Indonesia. Brussels: Les £ditions a Manteau S. A.
WALLACE, A. R.
1962 The Malay Archipelago. New York: Dover.
WHITE, J. P.
1969 Typologies for some prehistoric flaked stone artefacts from the Australian New Guinea
Highlands. APAO 4: 18-46.
WHITE, J. P., and D. H. THOMAS
1972 'What mean these stones? Ethno-taxonomic models and archaeological interpretations in
the New Guinea Highlands. In Models in Archaeology, ed. by D. L. Clarke, pp. 275-308.
London: Methuen.
WILLEMS, W. J. A.
1940 Preliminary report on the excavation of an urn-burial ground at Sa'abang, central Celebes.
In Proceedings of the Third Congress of Prehistorians of the Far East, ed. by F. N. Chasen
and M. W. F. Tweedie, pp. 207-208. Singapore: Government Printer.
WILMSEN, E. N.
1970 Lithic Analysis and Cultural Inference: a Palaeo-Indian Case. Tucson: University of
Arizona Press.
WOLTERS, O. W.
1967 EaT~V Indonesian Commerce: a Study of the Origins of Srivijaya. Ithaca: Cornell University
Press.
ZAINIE. C., and T. HARRISSON
1967 Early Chinese stonewares excavated in Sarawak 1947-67. SMJ 15: 30-90.

Archaeological Research in Minahasa and The Talaud Islands, Northeastern Indonesia

Uploaded by

Copyright:

Available Formats

Archaeological Research in Minahasa and The Talaud Islands, Northeastern Indonesia

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Archaeological Research in Minahasa and The Talaud Islands, Northeastern Indonesia

Uploaded by

Copyright:

Available Formats

Archaeological Research in

Minahasa and the Talaud Islands,

HE archaeological work described in this paper was undertaken by the author

T with the Indonesian archaeologist 1. M. Sutayasa in the Indonesian province

Fauna and the Question of Land Bridges

MINAHASA-THE PASO SITE

church ard terrace wall

C15 Cl6 Cli

ARTIFACTS FROM THE PASO SITE

The Obsidian Industry (Plate la-e)

Tool Function in the Paso Assemblage

o '.' ~ .... '

·..... ~. ... " ..

...... ·.·.".0 .....

slipped, often on both surfaces. Vessels of type 2A account for 31 % of

ADDITIONAL INVESTIGATIONS IN MINAHASA

EXCAVATED SITES IN THE TALAUD ISLANDS

LEANG Tuwo MANE'E (LEANG-U-TuwoMANEE), KARAKELLANG ISLAND

The Leang Tuwo Mane'e site (henceforth LTM) is a wave-cut rock-shelter in

Fig.4 Plan of Leang Tuwo Mane'e (contour interval 0.25 m).

east J13 west

PtQ H/K P/Q

Fig. 7 Proposed correlation between absolute dates, natural layers, depths,

copper or bronze, imported beads, and a very sharp decline in percentages

The LTM Chert Industry (Plate Ii-m)

Tool Function in the LTM Assemblage

The LTM Ceramics

p"Q BODY St<ERD THICKNfSSfS

Other Artifacts from the LTM Excavations

Other Sites in the Vicinity of Leang Tuwo Mane'e

LEANG BUIDANE, SALEBABU ISLAND

LONGITUDINAL SECTION A-B

The Leang Buidane Ceramics

Other Artifacts from Leang Buidane

Artifacts of Shell and Bone

Metal Artifacts and Baked Clay Moulds

The Date of Leang Buidane

LEANG BALANGINGI, KARAKELLANG ISLAND

TABLE 1. LEANG BALANGINGI: DISTRIBUTION OF POTTERY BY DEPTH

TOTAL IM- CARI- RED- SQUARE TRIPODS RARAN- CHINESE

0--10 532 466 13 9 15 5 5 17 2

Totals 2604 2394 46 9 29 37 39 21 27 2

%sherdsl00 92 1.8 0.3 1.1 1.4 1.5 0.8 1.0

I thus interpret Leang Balangingi as a place of burials and offerings, in use

TAN}ONG RARANGUNUSA AND ITs CERAMICS

Tanjong Rarangunusa (Tanjong = peninsula) protrudes from the northeastern

related to Rarangunusa is known from Minahasa, or anywhere on Sulawesi to my

On Karakellang and Salebabu, a number of other caves and rock-shelters which

A Possible Pebble and Flake Industry in Talaud

THE MINAHASA AND TALAUD RESULTS IN WIDER PERSPECTIVE

The archaeological work described was conducted in Sulawesi Utara between

You might also like