Tracking Emergent Spatial and Social Patterns across Terraced Landscapes in Polynesia

Abstract

Successful settlement on Polynesian islands required the alteration of environments, and such alteration produced extensive cultural landscapes. While some of the characteristics of these landscapes are well-established, what drives the spatial and temporal structure of these settlements is not clear across the entire region. Here, we present data on the nature and structure of settlement along one geological substrate in the interior of Ta‘ū Island, Manu‘a Group, American Sāmoa. Our results suggest that variability in slope and soil fertility were key drivers of archaeological patterns. Early use of the area seems to meet expectations of an ideal free distribution wherein the community was dispersed and located in relatively optimal locations for settlement. Characteristics of the settlement in the 15th century a.d. and later are consistent with landscape packing and community integration, signaling a shift to an ideal despotic distribution.

Full text

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=yjfa20
Journal of Field Archaeology
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/yjfa20
Tracking Emergent Spatial and Social Patterns
across Terraced Landscapes in Polynesia
Seth Quintus, Dolly Autufuga, Stephanie Day, Jennifer Huebert, Noa
Kekuewa Lincoln, Nolita Motu & Kyungsoo Yoo
To cite this article: Seth Quintus, Dolly Autufuga, Stephanie Day, Jennifer Huebert, Noa
Kekuewa Lincoln, Nolita Motu & Kyungsoo Yoo (2021): Tracking Emergent Spatial and Social
Patterns across Terraced Landscapes in Polynesia, Journal of Field Archaeology, DOI:
10.1080/00934690.2021.2018259
To link to this article: https://doi.org/10.1080/00934690.2021.2018259
View supplementary material
Published online: 27 Dec 2021.
Submit your article to this journal
View related articles
View Crossmark data

Tracking Emergent Spatial and Social Patterns across Terraced Landscapes in
Polynesia
Seth Quintus
a
, Dolly Autufuga
a
, Stephanie Day
b
, Jennifer Huebert
c
, Noa Kekuewa Lincoln
a
, Nolita Motu
d
, and
Kyungsoo Yoo
e
a
University of Hawai‘iatMānoa, Honolulu, HI;
b
North Dakota State University, Fargo, ND;
c
Sunrise Archaeology, Mangonui, New Zealand;
d
National Park of American Samoa, Pago Pago, AS;
e
University of Minnesota-Twin Cities, Falcon Heights, MN
ABSTRACT
Successful settlement on Polynesian islands required the alteration of environments, and such
alteration produced extensive cultural landscapes. While some of the characteristics of these
landscapes are well-established, what drives the spatial and temporal structure of these
settlements is not clear across the entire region. Here, we present data on the nature and
structure of settlement along one geological substrate in the interior of Ta‘ūIsland, Manu‘a Group,
American Sāmoa. Our results suggest that variability in slope and soil fertility were key drivers of
archaeological patterns. Early use of the area seems to meet expectations of an ideal free
distribution wherein the community was dispersed and located in relatively optimal locations for
settlement. Characteristics of the settlement in the 15th century A.D. and later are consistent with
landscape packing and community integration, signaling a shift to an ideal despotic distribution.
KEYWORDS
Luatele Site; Ta‘ūIsland;
ideal-free distribution;
landscape archaeology;
Samoa
Communities distribute themselves across space in response
to a number of drivers, both culturally and environmentally
defined. Cultural meaning is constructed and social relations
manifested through activities undertaken across landscapes
(Knapp and Ashmore 1999). At times, normative cultural
perceptions of social space affect the orientation of commu-
nities and notions of social centrality (Ashmore and Sabloff
2002). These processes of community formation occur across
variable environmental conditions, and patterning of land-
scape alteration, especially at a regional scale, is often driven
by environmental variability (Winterhalder et al. 2010). The
environment offers a set of opportunities and constraints
(Ladefoged et al. 2009) that contributes to patterns of archi-
tectural construction in so far as these opportunities and
constraints are associated with costs and benefits of using
those locations for different activities.
The intersecting impact of cultural and environmental
drivers of settlement are visible across Polynesia (e.g., DiNa-
poli et al. 2019; Kahn and Kirch 2013; Ladefoged, McCoy,
and Graves 2020; Ladefoged et al. 2009; Lepofsky and
Kahn 2011; Wesiler and Kirch 1985), and the Sāmoan archi-
pelago exemplifies this situation (Morrison and O’Connor
2018). Sāmoa offered a diverse set of environments for
human settlement, which drove variation in the structure
of settlement across the island group (Green 2002). Villages
were largely coastal in the historic period, but extensive
interior settlement was the norm during the 2nd millennium
A.D. (Davidson 1969,1974; Glover, Ladefoged, and Cochrane
2020; Green 2002; Holmer 1980; Jackmond, Fonoti, and Tau-
tunu 2018; Jennings, Holmer, and Jackmond 1982; Martins-
son-Wallin 2007). Extensive built landscapes are well
documented across the archipelago (Davidson 1974; Glover,
Ladefoged, and Cochrane 2020; Jennings, Holmer, and Jack-
mond 1982; Morrison and O’Connor 2018), including in the
interior uplands of the small eastern islands that collectively
constitute the Manu‘a Group (Quintus 2020; Quintus, Day,
and Smith 2017). The small size of these latter islands and
the bounded nature of their interior landscapes make them
ideal case studies for examining the nature and development
of built landscapes in the region. While past research has
highlighted recurrent archaeological patterns (Quintus and
Clark 2016), little research has investigated the spatial drivers
and temporal development of these settlements.
We build on this research by describing, analyzing, and
interpreting synchronic and diachronic patterns of a built
landscape on Ta‘ūIsland in the Manu‘a Group of American
Sāmoa (Figure 1). First, archaeological, soil fertility, and land
morphology investigations provide a dataset from which to
examine the spatial structure of settlement and the effects
of environmental variability at the scale of an individual dis-
persed village. Second, an extensive set of radiocarbon dates
(Quintus et al. 2020) allows an examination of how settle-
ment structure changed through time. Finally, the combi-
nation of these datasets presents an opportunity to test
expectations of ideal free and ideal despotic distribution
models in this small island environment to assess the roles
of soil fertility, slope, and social status in the development
of a village-scale settlement.
Sāmoan Settlement Units, Patterns, and
Processes
Sāmoan societies were historically organized as “house
societies.”The household, defined by a social title, included
an extended family with land, a set of domestic architecture,
and other property (see Mead 1969). Multiple interacting
households constituted pitonu‘u(subvillages), and multiple
pitonu‘u were grouped to form nu‘u(roughly, villages).
© Trustees of Boston University 2021
CONTACT Seth Quintus squintus@hawaii.edu Department of Anthropology, University of Hawai‘iatMānoa, 2424 Maile Way, Honolulu, HI 96822.
JOURNAL OF FIELD ARCHAEOLOGY
https://doi.org/10.1080/00934690.2021.2018259

These are relational social units (Shore 1982)defined on the
basis of interaction, but each had a spatial component (Jen-
nings, Holmer, and Jackmond 1982). Titles, land, and
resources were held by descent groups (Mead 1969, 18, 71–
73), and rights to cultivate land or build a house were
based on participation in familial activities and with the per-
mission of the title holder. Titled individuals from each des-
cent group were ranked, reflecting the social position of the
descent group to which the family title belonged, and these
different ranks reflected the social level of authority of each
individual (e.g., household, pitonu‘u, or nu‘u). Ranking
was dynamic and dependent on actions. In particular, acts
of generosity and fertility were closely tied to status, ideologi-
cally linked to the concept of mana, or supernaturally-
derived power and efficacy (Shore 1989). Both in Samoa
and elsewhere, the failure of chiefs to provide materially
for themselves and the rest of the population was interpreted
as a loss of divine power and met with political instability
(Allen 2010; Howard 1985; Shore 1989). Thus, access to fer-
tile land to produce materials for exchange and redistribu-
tion was of high importance.
House societies are proposed to have a long history in the
Sāmoan archipelago, extending back to the 1st millennium
A.D. (Green 2002), and are thought to be identifiable archae-
ologically. Holmer (1976,1980; see also Jennings, Holmer,
and Jackmond 1982) has argued that material records of
these social relations are visible in some areas of the western
islands of the archipelago through a nested set of archaeolo-
gical remains termed household units, wards, and villages
(see also Morrison and O’Connor 2018). Household units
were the basic level of organization, represented by a set of
domestic architecture and associated land, bounded by
walls (Holmer 1976,1980). The size of house foundations
was sensitive to variable status, documented both archaeolo-
gically and historically (Holmer 1976, 48; see also Jennings,
Holmer, and Jackmond 1982), with those of high status pos-
sessing higher and/or larger house foundations or house lots.
Importantly, the houses of elites, built of perishable material
with a limited archaeological signature, were no larger than
commoner houses (see Davidson 1974, 214); it was the size
—height and area—of the house foundation (e.g., platform
or terrace of stone and earth) that was the important marker
of difference. Distinct clusters of large residential structures
have been further interpreted as elite architecture and argued
to represent areas of chiefly authority (Holmer 1980; Jen-
nings, Holmer, and Jackmond 1982). Each cluster, around
which is scattered architecture of those of lower status, was
argued to represent authority over a group of extended
families, or wards. These are interpreted as equivalent to eth-
nographic pitonu‘u (Jennings, Holmer, and Jackmond 1982;
Martinsson-Wallin 2007), but the term “ward”has been pre-
ferred because pitonu‘u are relationally rather than spatially
defined. Groups of wards are interpreted to reflect villages
roughly equivalent to nu‘u (Jennings, Holmer, and Jack-
mond 1982). These nested settlement units were integrated
by a diverse set of stone and earthen walkways and paths
that connected household units and wards. Primary walk-
ways, defined by their large size, were situated especially
close to clusters of high-status architecture (Holmer 1980;
Jennings, Holmer, and Jackmond 1982), and Green (2002,
129) opined that the configuration of these paths may have
helped define or reaffirm political networks.
While broadly similar to settlement zones documented in
the western islands, archaeological communities in the
interiors of Ofu and Olosega in the Manu‘a group are also
defined by a set of patterns that reflect the topography of
the islands. Terraces are the dominant element of these cul-
tural landscapes, and many of these terraces exhibit evidence
of residential use (e.g., pavings and curbing alignments). This
is consistent with ethnohistoric and ethnographic references
to former villages in these locations (La Perouse 1798; Mead
1969), and the distribution of terraces suggests the presence
of distinct communities (Quintus 2020). These settlements
are best characterized as low-density agro-residential villages
Figure 1. The Manu‘a Group and Ta‘ūisland. The shaded areas with the inset image of Ta‘ūare areas where contiguous terracing is visible in the lidar dataset.
2S. QUINTUS ET AL.

within which was located a large set of households and, at
times, sub-village groups. Individual or small groups of ter-
races are inferred to represent centers of domestic activity
equivalent to household units (Quintus 2020; Quintus and
Clark 2016). Mean terrace size generally decreases with
increasing slope, and clusters of large terraces interpreted
as elite residences are positioned in central or seaward
locations (Quintus and Clark 2016). These clusters of large
terraces are thought to represent sub-village- and village-
level authority (Quintus 2015; Quintus and Clark 2016).
Agricultural infrastructure (i.e., drainage ditches) has been
found interspersed among residential terracing (Quintus
2012; Quintus, Allen, and Ladefoged 2016), and at least
some of the terraces, especially those that are too narrow
to support structures, may have functioned solely as garden
spaces. The distribution of vegetation types across the
interiors of these islands hints that agroforestry was practiced
at lower elevations, amongst residential features, while forms
of extensive shifting cultivation were practiced at higher
elevations (Quintus 2015). These archaeological patterns
are consistent with observations of French explorer La Per-
ouse (1798,55–56), who observed from his vessel an interior
settlement on Ta‘ūwherein houses were situated about mid-
way up the mountain with cultivated plots around them.
Ethnographic sources document the symbolic significance
of the seaward and center sections of villages in Sāmoa
(Mead 1969; Shore 2014). Chiefly residences and associated
features located in these central areas were focal points of vil-
lages (Lehman and Herdrich 2002; Mead 1969; Shore 2014),
around which commoner households and other village
activities were oriented. These spatial relations served to pro-
duce and reproduce social relations. As Shore (2014, 386)
noted in the context of chiefs serving as focal points of com-
munities in West Polynesia: “This, then, is not simply a cul-
tural model of space. It is a cognitively grounded form of
political socialization affording those who use it an intrinsi-
cally hierarchical perception of the world. This model lit-
erally forces the self, in a range of ordinary orientational
tasks, to view the world from the perspective of a social
superior.”This spatial configuration of villages was histori-
cally a mechanism of political legitimization in this cultural
context. The construction of architecture in the geographical
core of a larger system also had practical implications in that
such a location would make management or oversight of the
larger area easier (see Weisler and Kirch 1985).
These ethnographic patterns seem to extend at least into
the last few hundred years prior to European contact, as
the symbolic significance of the center of villages helps
explain some archaeological patterns (Quintus and Clark
2016). What drove the selection of specific areas for focal
points or as centers in the past is unclear, but previous
research has called attention to the role of slope in generating
some residential patterns in these interior settlements (Quin-
tus and Clark 2016). Slope is a key technological constraint to
terracing in these environments in that labor costs increased
drastically with increasing slope, as more fill material was
needed to produce each unit of terrace width. This mani-
fested in a reduction in terrace width and variation with
increasing slope as the construction of wide terraces became
too costly.
Still, these focal points could be established anywhere in
these interior landscapes where slope is suitable for the con-
struction of terracing, which covers a broad area. Given the
well documented relationship between the fertility of land
and Polynesian chieftainship (Howard 1985; Shore 1989),
we hypothesize that these focal points were established in
agriculturally productive spaces. Indeed, variation in agricul-
tural production is a well documented driver of patterning in
archaeological landscapes across multiple spatial scales in
Polynesia. Variation in agricultural potential frequently con-
strained the distribution of cultivation (Vitousek et al. 2014).
Even within cultivated zones, however, differences in agricul-
tural potential had substantial effects on the distribution and
nature of the archaeological record, as various factions
attempted to gain access to productive locations, legitimize
authority, or move away from the demands of elites. In the
‘Opunohu Valley of Moʿorea, Kahn and colleagues (Kahn
and Kirch 2013; Lepofsky and Kahn 2011) have found that
elite residences were positioned in agriculturally productive
locations, with lower status individuals occupying more mar-
ginal spaces. Kahn and Kirch (2013) relate these archaeologi-
cal patterns to competition between social houses as different
corporate groups attempted to invest in landed house estates
in order to maintain or accumulate status through time.
Similar trends are visible in Hawai‘i. In both Kaupo and
Kahikinui on Maui, the spatial distribution of temple sites
tied to those in power is correlated with the boundaries of
productive locations, serving both to mark territory and sig-
nal authority over these locations (Baer 2016; Kirch et al.
2004). The location of temples in the Leeward Kohala Field
System (LKFS) has similarly been tied to managerial oversite
and the formation of authority (McCoy et al. 2011; Phillips
et al. 2015), while there is a lower density of agricultural
infrastructure and habitation, as well as little evidence of
elite presence, in less optimal areas of the LKFS relative to
its core (Ladefoged, McCoy, and Graves 2020).
Documentation of the impact of environmental variabil-
ity on synchronic patterns of settlement elsewhere hint that
such variation may have been important on Ta‘ūin produ-
cing both spatial and temporal patterns of the built land-
scape. Pacific archaeological landscapes are palimpsests,
and the growth of these landscapes is structured by previous
land use, as well as environmental conditions (Ladefoged
et al. 2011). Past land use, often in the form of constructed
environments, intersects with environmental variation to
influence where factions of a population can live, how they
can exploit an environment, and how they should interact
with other groups in that same environment. While previous
temporal data is limited (but see Quintus 2018), ideal free
and ideal despotic distribution models provide testable
expectations of these phenomena. These models assume
that individuals will attempt to maximize their evolutionary
fitness, often by improving their economic well-being and
maximizing their net-energetic returns (Weitzel and Cod-
ding 2020). The ideal free distribution (IFD) predicts that
when access to land is unconstrained, individuals will settle
first the most suitable locations until the density of settle-
ment increases to such an extent that those most suitable
areas are no longer an option or are less suitable (Kennett,
Anderson, and Winterhalder 2006; see also Fretwell and
Lucas 1969). After, individuals will choose to occupy pre-
viously more marginal zones. Suitability is defined by acces-
sibility, the availability of livable space, and appropriate
resources (Winterhalder et al. 2010), which, in agricultural
societies, is often based on soil fertility. However, access to
land is often not unconstrained, especially in ranked
JOURNAL OF FIELD ARCHAEOLOGY 3

societies, where some individuals are better able to compete
for optimal land when resources are defensible (Mattison
et al. 2016). In these cases, which form an ideal despotic dis-
tribution (IDD), territoriality may develop that interferes
with the ability of individuals to settle optimal zones, push-
ing individuals to settle lower ranked habitats before it
would be expected in an IFD (Bell and Winterhalder 2014;
Prufer et al. 2017). Because highly suitable areas are defended
by those of higher rank, less suitable areas are settled by those
with limited social authority and opportunities. Thus, spatial
sorting by rank and markers of resource control should be
visible across resource-heterogeneous landscapes. These pat-
terns may be emergent in archaeological sequences. As both
Kennett and colleagues (2009) and Prufer and colleagues
(2017) demonstrate, initial unconstrained access to suitable
land, as in an IFD settlement model, can give rise over
time to control of land associated with an IDD settlement
model as social power is differentially accumulated, land-
scape suitability changes, and knowledge of landscape varia-
bility increases.
Based on these models, we expect that individuals in the
interior uplands of the Manu‘a group initially occupied
areas of gentle slope and high soil fertility relative to the sur-
rounding areas. As settlement density increased, we expect
settlement occurred in more marginal areas, here defined
as areas of steeper slope and lower soil fertility. Landscape
packing may also have increased competition, especially as
internal social ranking became more marked, which could
necessitate investment in boundary walls that more clearly
marked land ownership or elaboration of status symbols.
Intensive survey and excavation across a terraced land-
scape on the Luatele substrate of the island of Ta‘ūprovide
an opportunity to evaluate these patterns and expectations.
In particular, we aim to compare Luatele to settlement
zones on Ofu and Olosega to evaluate whether recurrent pat-
terns documented on those islands extend to Ta‘ū. We then
assess the role of slope and soil fertility in driving those pat-
terns before testing expectations of IFD and IDD models
against the development of this interior settlement.
Environmental Setting and Methods of
Investigation
All the islands of Manu‘a are small, with Ta‘ūthe largest at 36
km
2
. Settlement of the group occurred around 2650–2750
CAL B.P. (Clark et al. 2016; Petchey and Kirch 2019), though
sites dating to this period are not yet documented on Ta‘ū.
Occupation throughout the group was largely coastal
through the 1st millennium B.C. and much of the 1st millen-
nium A.D. (Quintus 2015), with only limited evidence of the
use of the interior uplands at this time (Quintus et al. 2020).
Intensive use of island interiors occurred by the end of the
1st millennium A.D. and continued through the 2nd millen-
nium A.D. (Quintus 2015; Quintus et al. 2020).
The Luatele site (AS-11-123) was identified in the north-
eastern quadrant of the Ta‘ūinterior using lidar imagery
and first investigated in 2015. The name of the site derives
from the geological substrate on which the site primarily
sits, though some archaeological remains are found on the
adjacent Lata substrate as well (Figure 2). The area is heavily
vegetated with a mix of secondary (e.g., Rhus taitensis and
Myristica spp.), economic (e.g., Cocos nucifera,Artocarpus
altilis,andHibiscus tiliaceus), and invasive (e.g., Adenanthera
pavonina) forest species. The ages of most surface substrates
on the island are unknown, but all are younger than 100,000
years (McDougall 2010) and, on stratigraphic grounds, the
Luatele substrate is younger than the adjacent and more
spatially extensive Lata substrate. It is generally hot, humid,
and rainy across the island. Lower elevations of the island’s
interior receive roughly 4,000 mm annually, while the higher
elevations receive some 7,500 mm. Slope values generally
range from 10° up to 30° in the project area, though slopes
under 10° and over 30° are locally present.
Fieldwork was undertaken across three field seasons
(Motu 2018; Quintus, Day, and Smith 2017). These efforts
focused on four transects and an intensive recording of fea-
tures near the center of the Luatele substrate (see Figure 2).
Additional spot checks were undertaken between these
transects to confirm the continuous distribution of features
and to further map key linear features that extended outside
of transects. Field-recorded features were supplemented by a
dataset of terraces documented between transects previously
identified through analysis of a lidar dataset. Methods of
digital feature identification and a discussion of character-
istics of those features can be found in Quintus, Day, and
Smith (2017). The lidar dataset also allows confirmation of
the continuation of linear features that were partially
mapped in the field.
We focus here on two major categories of documented
features: terraces and linear mounds. Terraces are defined
as artificially flattened surfaces with at least three free stand-
ing sides. The length and width of each feature was
measured, and the presence of secondary features was
noted for each terrace. The area of field-recorded terraces
was calculated using a simple length × width equation. We
use this simple measure because the exact boundaries of indi-
vidual terraces were often ambiguous given the density of the
vegetation, the way that the features graded into surrounding
slopes, and post-construction processes of erosion and depo-
sition that have impacted the form of each feature. While
there is some uncertainty, these measurements are useful
for relative comparison across the site. Linear mounds are
mounded features built of cobbles, boulders, and soil that
are longer than they are wide. The configuration of these fea-
tures falls into three broad types: a single linear mound
(single wall), two mounds running parallel to each other
(double wall), and mounds that are continuous but change
in morphology from single to double walls across their extent
(mixed form). The single walls are morphologically similar to
low walls across the western islands of Sāmoa that form
boundaries around sets of features and open spaces (David-
son 1974; Holmer 1980), while the double walls are morpho-
logically equivalent to walled walkways (Holmer 1980). As
such, they are interpreted in these terms here. The height
and width of each feature were measured, though these
measures vary along the length of each feature. All features
were plotted using the internal GPS of Apple iPads and the
ESRI Collector App with 5–10 m accuracy, with the shape
of these features later modified to reflect field-recorded
dimensions if errors were identified. Each archaeological fea-
ture was then described and photographed.
A small percentage of field-identified features were cho-
sen for excavation. Two methods of excavation were used:
small test pits to acquire datable material under constructed
features and larger test units to examine the internal struc-
ture of features. Twenty-six test pits and 10 controlled
4S. QUINTUS ET AL.

units were dug through 20 terraces (ca. 6% of total terraces)
and 13 linear mound segments (ca. 14% of total linear
mound segments). Two units were dug into a single terrace,
and three test pits were dug into a single linear mound seg-
ment. Additional methods of excavation, charcoal analysis,
and radiocarbon dating can be found in Quintus and col-
leagues (2020).
Slope and elevation data were extracted for each terrace
identified in the field and digitally using ArcDesktop 10.7.
A 10 m DEM produced by the USGS was downloaded (cat-
alog.data.gov) from which elevation in meters and slope in
degrees was derived. Soil fertility across the project area
was measured through soil chemistry. Our analysis focused
on three measures that correlate with boundaries of intensive
agriculture in Hawai‘i and Rapa Nui (Vitousek et al. 2014):
pH, % base saturation, and exchangeable Ca. pH is an impor-
tant variable that, among other things, impacts nutrient
availability to plants and Al toxicity, with values between 6
and 7 considered optimal. Percent base saturation is linked
to pH and is a good predictor of plant-available nutrients.
Exchangeable Ca is not a soil fertility indicator, per se, but
it is an important base cation and one factor that impacts
the availability of Ca to plants. We include it here because
it was considered the best predictor of field boundaries in
Hawai‘i (Vitousek et al. 2014). We principally sampled
soils along three transects across the Luatele substrate
(Figure 3)asdefined by previous geological mapping (Stice
and McCoy 1968): one vertical transect (n = 13) and two
Figure 2. A) The distribution of field-recorded archaeological features (terraces and linear mounds) in relation to geological substrates. Field transects are labeled
with Roman numerals. B) The distribution of terrace sizes within the four intensive transects from northwest (top) to southeast (bottom). Geological map and
shapefiles were produced by the National Park Service (NPS) Geologic Resources Inventory (GRI) program.
JOURNAL OF FIELD ARCHAEOLOGY 5

lateral transects (upper, n = 17; lower, n = 22). Four samples
at the center of the lateral transects were included as part of
the vertical transect. Six additional samples were taken from
below and between these two lateral transects nearer the east-
ern and western boundaries of the Luatele substrate. Soil pH
was measured for 54 samples, while exchangeable Ca and %
base saturation were measured for 53. At each location, we
composited three integrated samples collected from unmo-
dified slopes to 30 cm depth, following methods in Vitousek
and colleagues (2004). All samples were air-dried and passed
through a 2 mm sieve. Measurement of pH was accom-
plished with air-dried soils mixed with deionized water in
a 1:2 ratio at the Indigenous Cropping Laboratory at the Uni-
versity of Hawai‘iatMānoa. Exchangeable cations and base
saturation were analyzed following procedures in Soil Survey
Laboratory Staff(1992) using the ammonium acetate (NH
4-
OAc) method buffered at pH 7 at the University of Hawai‘iat
Hilo Analytical Laboratory.
The segmented package (Muggeo 2008) was used in R (R
Core Team 2021) to carry out a small set of segmented or pie-
cewise regressions. Soil characteristics, slope, and terrace size
were each evaluated against elevation. While patterning of
soil fertility on the island is likely to be driven by variable pre-
cipitation (see Vitousek et al. 2014), we explore the relation-
ship between our soil fertility indicators and elevation to
maintain consistency with other datasets and because fine-
grained precipitation datasets are lacking. We also assume
that elevation is a good proxy for rainfall in this high island
environment. Segmented regression assesses whether the
relationship between two variables is better explained with
more than a single line and highlights at what elevation the
nature of the relationship begins to differ. Breakpoints are
defined as the position where the relationship between
variables begins to differ. The Davies’test (Davies 1987),
which tests for a change in the slope of a regression parameter,
was used to test the statistical significance of breakpoints using
a 0.05 alpha level. Moran’s I (Spatial Autocorrelation tool) and
Anselin Local Moran’s I (Optimized Cluster and Outlier tool)
tests were performed in ArcDesktop 10.7 to examine point
patterns in a combined field and digital dataset of terraces
in order to identify clusters of terraces of different sizes at
different scales. Chi-squared tests and confidence intervals
were calculated in Minitab. Analyzed data are provided in
Supplemental Materials 1, 3, and 4.
The Spatial Structure of the Luatele Settlement
Three hundred twelve terraces and 93 linear mound seg-
ments were recorded across a ca. 130 ha area during ped-
estrian survey. Terracing and linear mounds occur between
these transects and have been recorded in digital surveys
(Motu 2018; Quintus, Day, and Smith 2017). The vast
majority of terraces we recorded are located on the Luatele
substrate, with a small number located on the Lata substrate
at the far eastern end of the project area that were documen-
ted digitally and field confirmed. Given some uncertainty of
substrate boundaries, it is possible that the Luatele substrate
does extend further east to an intermittent stream that
defined the southeastern side of our project area. The area
of contiguous terracing is roughly bounded on the northwes-
tern side of the project area by an intermittent stream and by
steep cliffs to the north. The inland boundary is more
difficult to define, as archaeological remains continue slightly
past the crater (Klenck 2016). However, a series of large
cross-slope walls does mark an elevation at which the nature
of the archaeological record changes from a dense
Figure 3. The location of soil samples across slopes of the Luatele substrate.
6S. QUINTUS ET AL.

concentration of terracing to more specialized features and
more sporadic linear mounds.
The size of field-recorded terraces varies considerably
across Luatele, ranging from 8–714 m
2
(see Figure 2, Sup-
plemental Material 1). Features in the southeastern transect
(Transect III [95% confidence interval for mean, 46–65
m
2
]) are smaller than those in the center (Transects II
[121–165 m
2
] and IV [103–163 m
2
]) and northwest (Trans-
ect I [100–140 m
2
]). Terrace size also varied by elevation,
though the relationship is not linear. Mean terrace size
increases with increasing elevation in the lower half of the
site and then decreases with increasing elevation moving
further inland (Figure 4A). While the relationship is weak,
the point at which the relationship between elevation and
terrace size changes, the breakpoint, is estimated by segmen-
ted regression at 198 m (r
2
= 0.08; Davies’p< 0.001; see Sup-
plemental Material 5). These spatial patterns are even more
apparent when terraces are broken into four size classes
(Supplemental Material 2); in particular, very large terraces
(over 300 m
2
) are distributed unevenly. Seventy-seven per-
cent of these features recorded in the field are in a roughly
8 ha area near the center of the Luatele substrate (seaward
of the absolute center), including the two largest in the data-
set located next to each other.
These field data can be supplemented with 457 digitally
recorded terraces (the analytical dataset from Quintus, Day,
and Smith 2017)tocreateacombinedfield and digital dataset
(Figures 5,6). Identified patterns are largely consistent with
those documented in the field dataset, with larger terraces situ-
ated in the middle elevations of the site (see Figures 4A, 6;
Supplemental Materials 2, 3). Thirty-one terraces over 300
m
2
in size are included in the combined dataset, with 61%
located in the roughly 8 ha area in the center of the Luatele
substrate. The breakpoint between terrace size and elevation
in this dataset is also at 198 m (r
2
= 0.05; Davies’p<0.001;
see Supplemental Material 5), the same as that identified in
the field dataset. More generally, terraces are spatially autocor-
related (Moran’s I; z-score = 14.56; p<0.001;seeSupplemen-
tal Material 5). An Optimized Cluster and Outlier Analysis
(Anselin Local Moran’s I; see Supplemental Materials 5, 6)
using terrace area as the value parameter identifies the group-
ing of large terracing near the center of the project area as a
significant high value cluster (False Discovery Rate [FDR] cor-
rection, 95% confidence,p<0.026) and terracing at the upper
elevations and near the southeastern field transect as clusters
of low values. In addition to these clusters that are consistent
with patterns in the field data, a significant high value cluster
was highlighted at the southeastern end of the project area (see
also Figure 6). This area was not intensively surveyed in the
field, but spot checks did confirm the presence of a few
large features in the area. Several individual outliers were
identified in all of these clusters, while some terraces at the
northwestern edge of the substrate were classified as a low
value cluster. The presence of outliers indicates that terrace
sizes are poorly sorted across the site; small and large terraces
are intermixed, even though there are areas where large or
small terraces are clustered.
The majority of linear mounds (boundary walls) are
oriented parallel to the slope, with segments running from
near the bluffto near Luatele crater. The distance between
adjacent boundaries within surveyed transects is reasonably
consistent, ranging from ca. 30–80 m with no clear spatial
patterning in the field data available vertically or laterally
across the project area. The longest linear mound in the
site, which is of mixed form but is a double wall for much
of its extent, is interpreted as a path that stretches 1.6 km
across the settlement (Wall 2). This feature was only partially
mapped in the field, with connecting segments visible using
the lidar dataset. This feature is interpreted to have served a
function comparable to primary walkways in the western
islands of the archipelago, as it effectively integrates the site
from one side to the other. It does so at an elevation around
200 masl for most of its extent, though it does extend to an
elevation of 240 masl toward the southeastern boundary of
the site. Sixty-four percent of very large terraces in the field
dataset are located within 50 m of this path, compared to
19% of terraces smaller than 300 m
2
(n = 307; excludes ter-
races beyond extent of feature; χ
2
= 24.28; p< .001). Other
cross-slope walls are interspersed amongst the terraces,
some of which form retaining walls for the terraces. How-
ever, a series of these walls, which are far larger than others,
occur at and seem to mark the upper elevation boundary of
the settlement. These wall segments do not connect across
the entire breadth of the site, but they do define a boundary
at a relatively consistent elevation: at ca. 310 masl in the east-
ern transect (Wall 78), at ca. 315–330 masl in the center, and
ca. 340 masl in the center-west transect. It is above this point
that there seems to be a shift in the nature of the archaeolo-
gical record wherein terracing is more sporadic, linear
mounds parallel to the slope are rare, and ritual sites are pre-
sent (see Klenck 2016).
Features are generally distributed in slopes less than 20°
(Figure 4B, C), with some exceptions along the seaward
boundary of the site. Extensive gentle sloping land is one
of the characteristics that seems to delineate the Luatele sub-
strate in the eastern half of the island. The size of terracing is
generally more variable in gentler or moderate slopes, where
mean terrace size is higher, though this pattern is not as
apparent in the combined dataset. These differences relate
to variation in the width of terraces rather than length.
Long terraces were still constructed in steeper slopes but
wide terraces were not, as the labor requirements of wider
terraces outweighed benefits of their construction. There is
a breakpoint in the relationship between slope values associ-
ated with terraces and elevation at 183 m in the field dataset
(r
2
= 0.23; Davies’p< 0.001; see Supplemental Material 5)
and 218 m in the combined dataset (r
2
= 0.14; Davies’p<
0.001; see Supplemental Material 5). From the cliffto this
breakpoint, slope values associated with terraces decrease
substantially with increasing elevation, while mean terrace
size increases (Figure 4B, C). The decrease in mean terrace
size above the 200 m contour, however, is not related to
changes in slope, as slope values associated with terraces gen-
erally stabilize below 20° above the 200 m contour.
Soil fertility varies by elevation, with lower elevations
being more fertile than higher elevations (Figure 7, Sup-
plemental Material 4; Autufuga 2021). This suggests that pre-
cipitation is a key driver of soil fertility across the Luatele
substrate, as demonstrated elsewhere (Vitousek et al. 2014),
with areas of higher rainfall on Luatele being less fertile.
The location of cross-slope boundary walls between 310
and 340 masl coincides with low values of soil parameters.
It is above this elevation that some soils begin to fall below
30% base saturation and the majority of soils exhibit pH
values of less than 5.7, both values of soil fertility that seem
to constrain agriculture elsewhere in Polynesia (Vitousek
JOURNAL OF FIELD ARCHAEOLOGY 7

Figure 4. Relationship between key variables across Luatele with the field dataset on the left and the combined dataset on the right. A) The relationship between
terrace size and elevation. Each dot is a terrace, and dotted lines track changes in mean size. B) The relationship between terrace size and slope. Lines track
changes in mean size, and asterisks identify outliers. Small boxes represent the IQR, whiskers are the range without outliers, and large boxes mark the range
within each group. C) The relationship between slope and elevation associated with each terrace. Each dot represents a terrace, and the dotted lines are
LOESS smoothers with 0.5 degrees of smoothing and two steps.
Figure 5. The distribution of terraces in the combined dataset compared to geological substrate.
8S. QUINTUS ET AL.

et al. 2014). Measured exchangeable Ca values are low
throughout the project area, at least relative to agricultural
landscapes in Hawai’i and Rapa Nui, and these values also
decline with increased elevation. There is some evidence
that soil fertility is higher in the northwestern half of the pro-
ject area than in the southeastern half in the lower elevations
(Table 1), which correlates with increased mean terrace size
in the northwest relative to the southeast, but two of these
differences are not statistically significant, and all of the
differences are relatively small.
Segmented regression provides further insights into the
relationship between elevation and soil characteristics (see
Supplemental Material 5). The vertical soil transect was used
to assess breakpoint patterns, as it is not affected by the lateral
variation in soil fertility measures across the project area. No
statistically significant breakpoint was present in the relation-
ship between pH and elevation, as the relationship is relatively
linear. The breakpoint in exchangeable Ca occurs between the
second and third sample point (Breakpoint Est. = 84 m; see
Supplemental Material 5), after which exchangeable Ca
declines with increased elevation in a relatively linear fashion.
In contrast, a breakpoint is present between elevation and base
saturation at 204 masl (r
2
= 0.80; Davies’p=0.03). This is
similar to the breakpoint between terrace size and elevation,
wherein terrace size began to decrease above ca. 200 masl,
and between slope and elevation, wherein slope values stabil-
ize above ca. 185–215 masl.
Subsurface Investigations
The excavation of terraces allowed an opportunity to evalu-
ate their function and internal structure. This is important
since no excavation has been conducted through the surfaces
of terraces on Ofu or Olosega. All but one terrace through
which controlled excavation was undertaken on Ta‘ū(n =
8) displayed a similar stratigraphic sequence. These terraces
were largely built on sloping ground onto which a fill was
transferred to flatten the surface. In those units dug from
the terrace front to back, this layer of fill increased in thick-
ness the closer one was to the front of the terrace (Figure 8A).
The lone exception to this general pattern was a feature built
near the edge of the bluffleading down to the coastal plain
(Terrace 98, Figure 8B). Here, the individuals that con-
structed the terrace used large angular boulders to help cre-
ate a flat surface onto which sediment was dumped to
complete the terrace. A rough cobble pavement was then
constructed, which seems to have provided a base to the liv-
ing floor. This style of construction is a unique solution in a
steep location where construction of a terrace with earthen
fill alone would be difficult.
All terraces excavated with controlled methods exhibited
evidence of residential use. In three cases, surface curbing
alignments evince such a function. These outlines are all rela-
tively small, being between 20 and 30 m
2
, but fall within the
range of house sizes documented elsewhere in the archipe-
lago (Davidson 1974). Subsurface layers of loosely placed
sub-angular to sub-rounded boulders and cobbles create
floor pavement in three units (Figure 8C, 8D). These pave-
ments, even though they are quite rough, are interpreted to
mark locations of previous superstructures. Postholes,
which also document the presence of superstructures, were
identified in two units, and basin-shaped features interpreted
as trash pits were recorded in two units. Small basalt flakes
indicative of tool maintenance and use were recovered in
Figure 6. Mean terrace size in 10,000 m
2
cells across Luatele. Terraces with centroid points within each cell were included in the calculation of that cell’s mean
terrace size. Note cells with higher mean terrace size in the center and southeast.
JOURNAL OF FIELD ARCHAEOLOGY 9

all but one controlled excavation, while waterworn basalt
gravel and small cobbles (ili‘ili) of former pavings were
identified in all controlled units. Locally available angular
cobbles were found on the surface of most terraces in the
project area, similar to those found scattered on residential
platforms in the Mt. Olo tract on ‘Upolu. Holmer (1976)
argued that these cobbles were an important addition to
foundation surfaces, as they increased drainage and limited
rot of superstructures. It is plausible that scatters of angular
stone and subsurface foundations would have effectively
served this function in Luatele. Coral gravel, a common
house paving material on the adjacent island of Ofu and Olo-
sega, was noticeably rare in Luatele. It was found during
excavation in only one unit, while a small coral boulder
was found in the retaining wall of another feature.
The construction ages of 15 terraces in Luatele have been
modelled previously (Quintus et al. 2020). Each terrace can
be placed into an early (n = 6; pre-A.D. 1400) or late (n = 9;
post- A.D. 1400) temporal group (Table 2). We include in
the early group all terraces for which a pre- A.D. 1400 age
is plausible, which includes two terraces with terminus
post quem (TPQ) but no terminus ante quem (TAQ)
dates. With no upper constraint, these terraces have
modelled construction ages that range into the historic
period, even though their associated TPQ are in the 12th
and 13th century A.D., respectively. Thus, we use an assump-
tion in this analysis that there is little temporal lag between
these TPQ dates and the event of terrace construction
when a terrace is lacking a TAQ.
The spatial distribution of dated terraces can be found in
Figure 9. All terraces dating to the early period are larger
than 100 m
2
and smaller than 300 m
2
, with both smaller
and larger terraces present in the later period. Terraces
dated to the early period are all located below 221 masl,
and all but one is within a 30 m elevation envelope that cor-
responds well with various breakpoints between slope, soil
measures, and elevation. The four dated terraces in the high-
est associated elevations date to the late period. There is little
difference between slope values associated with terraces of
different time periods, and both early and late terraces are
located across the lateral extent of the Luatele substrate.
The two dated terraces that may fall on the older Lata sub-
strate were both built in the later temporal period.
One controlled unit was dug through a linear mound to
better understand the internal morphology of these features.
The internal structure of the feature is similar to that
Figure 7. The relationship between three soil fertility indicators and elevation: A) pH, B) exchangeable Ca, and C) % Base Saturation. The figures in the right-hand
column include all soil samples. The figures in the left-hand column include samples from the vertical transect. Reference lines represent the breakpoint in the
relationship between terrace size and elevation (200 masl), the elevation of the cross-slope walls on the southeastern side of the project area (310 masl), and the
elevation of the cross-slope walls on the northwestern side of the project area (340 masl). The LOESS curves were calculated using one degree of smoothing and
two steps.
Table 1. Mean and standard error of soil measurements in the northwestern and southeastern halves of the project area using samples from the lower lateral
transect combined with a small number of samples from the vertical transect in proximity. Bottom row shows the results of Mann-Whitney U Tests.
Area pH (SE) Exchangeable Ca (SE) % Base Saturation (SE)
Southeast (n = 13) 5.97 (0.06) 7.74 (0.55) 63.4 (3.83)
Northwest (n = 17) 6.21 (0.07) 8.15 (0.59) 67.7 (3.24)
Mann-Whitney U Test U = 53; z = 2.39; p= 0.02 U = 91.5; z = 0.77; p= 0.44 U = 83; z = 1.13; p= 0.26
10 S. QUINTUS ET AL.

documented for some linear mounds in Hawai‘i (Quintus
and Lincoln 2020), with a mounded profile and construction
fill of earth, cobbles, and small boulders. There appears to
have been little deposition around the basal stones of these
features, implying that they were relatively late additions to
the site. Radiocarbon dating supports this conclusion. Mod-
elled construction ages are available for 10 linear mound seg-
ments (Quintus et al. 2020). Almost all linear mound
segments are dated to the 17th century A.D. and later
(Table 3). Of the two examples that potentially date before
this time, one is found in the center and one at the southeast-
ern end of the project area. These two features have the low-
est associated elevations of the dated features, but there may
be substantial lag between the TPQ date and the construction
of each feature. Both the large pathway that stretches across
nearly the length of the site (Wall 2) and the only dated large
cross-slope feature (Wall 78) that forms part of the upper
boundary of the settlement were constructed in or after the
17th century A.D.
Discussion
Large size is a characteristic of high-status architecture in
Sāmoa (Davidson 1974; Holmer 1976,1980; Jennings, Hol-
mer, and Jackmond 1982), and clusters of large architecture
are interpreted as centers of authority (Jennings, Holmer,
and Jackmond 1982, 89). These clusters in the Manu‘a
group have spatial characteristics of community focal points
documented in ethnohistoric and ethnographic literature
(Quintus and Clark 2016). In Luatele, this architecture is lar-
gely, but not solely, concentrated in a roughly 8 ha area near
the center of the Luatele substrate and near the center of the
distribution of terraces across this landscape, a pattern
confirmed by spatial autocorrelation analyses. We interpret
this cluster of large terraces to be the focal point of the settle-
ment, comparable to those found in communities on the
adjacent islands of Ofu and Olosega. Outside this focal
zone, mean terrace size is smaller in the southeast than in
the center and northwest. Mean terrace size also decreases
above the 200 m contour.
Slope played a role as an important technological con-
straint to terrace size in Luatele. That slope was a technologi-
cal constraint is demonstrated by the fact that construction
of a terrace in steep slopes necessitated the use of a different
construction method relative to terraces in gentler slopes.
The presence of terraces, some large, in steep slopes below
the 200 m contour highlights a tradeoffbetween slope and
soil fertility. Given that soil fertility declined with increasing
elevation, it seems that at least some individuals or groups
were willing to invest additional labor to construct terraces
in steeper slopes to gain access to more fertile land.
Slope partially contributes to the location of the focal
point within Luatele, but other areas of gentle slope would
also be suitable for the construction of large terraces. In
fact, a larger area of contiguous slope values below 10° is
located to the southwest of the recorded focal point, but
comparable terraces are not found in the area. Instead, the
focal point is situated at an intersection of stabilizing slope
and declining soil fertility. The position of the focal point
takes advantage of gentler slopes relative to areas downslope
and higher soil fertility relative to areas upslope.
Construction in this location, without the limitation of
slope and with still relatively high soil fertility, is important
in a social context where fertility was often taken as a sign
of the efficacy of leadership and proximity to the divine
(Howard 1985; Shore 1989). Access to productive land that
Figure 8. Characteristics of terraces documented in excavation. A) The eastern wall of Terrace 117 showing the wedge-shaped Layer II interpreted as fill to level
the terrace. B) The southern wall of Terrace 98, illustrating the bolder fill (Layer IIb) and basalt cobble pavement (Layer Ib) of the terrace. The boulder fill was
excavated down two courses without an identifiable base. C) The top of a rough pavement/foundation in a 1 ×1 m unit excavated into Terrace 282 (Transect
I). D) The base of a rough pavement/foundation in a 1 ×1 m unit excavated into Terrace 120 (Transect III).
JOURNAL OF FIELD ARCHAEOLOGY 11

made one’s group more capable of meeting social obligations
was a key component of status maintenance and accumu-
lation. Furthermore, based on Sāmoan cultural models of
space in the ethnographic and ethnohistoric data (Shore
2014), the position of this focal point at the center of the
archaeological distribution served to legitimize the inhabi-
tants of these terraces. The position also affords more
efficient access across the settlement to those living in the
focal point, perhaps increasing some level of community
oversight. The orientation of the settlement across Luatele
and the position of the focal point at the center of the sub-
strate may speak to attempts to control the Luatele land-
scape, specifically. The youthful age of the substrate and
expansive gentle slopes, relative to adjacent areas, may
have made the area especially attractive. In general, smaller
residential structures in Luatele seem to be situated in less
suitable locations based on slope and soil fertility. While
those in the center of the settlement had access to both gentle
slope and relatively high soil fertility, individuals elsewhere
had to choose between the two. The other cluster of large ter-
races at the southeastern edge of the project area may rep-
resent more local authority at the sub-village level, akin to
those documented on ‘Upolu and Savai‘i (Jennings, Holmer,
and Jackmond 1982). This pattern, wherein architecture
associated with those in power was built in agriculturally
productive landscapes, is well documented broadly across
the region (Glover, Ladefoged, and Cochrane 2020; Kahn
and Kirch 2013; Kirch et al. 2004; Lepofsky and Kahn
2011; McCoy et al. 2011).
This is not to say that no small terraces are located in the
focal point, as several are, and there is no evidence of con-
centric circles of decreasing status with increased distance
from the settlement center, as would be expected from the
ethnographic model proposed by Shore (2014). Rather,
Table 2. Modelled ages and characteristic of 15 dated terraces in Luatele (from Quintus et al. 2020). TPQ = Terminus Post Quem; TAQ = Terminus Ante Quem.
Name Modelled Age (95.4% HPD) Associated Constraint(s) Temporal Period Area (m²) Slope (in Degrees) Elevation (m) Location in Luatele
Terrace 117 A.D.86–1052 TPQ and TAQ Early 108 15.1 213 Southeast
Terrace 120 A.D. 725–1329 TPQ and TAQ Early 112 17.6 202 Southeast
Terrace 238 A.D. 1182–1896 TPQ Early 286 14.9 192 Center
Terrace 282 A.D. 901–1336 TPQ and TAQ Early 231 12.9 221 Northwest
Terrace 320 A.D. 1272–1801 TPQ Early 140 13.9 215 Southeast
Terrace 98 A.D. 1310–1411 Construction Early 270 26.5 104 Center
Terrace 110 A.D. 1458–1864 TPQ Late 360 11.2 178 Center
Terrace 163 A.D. 1708–1905 TPQ Late 48 20.9 186 Southeast
Terrace 210 A.D. 1708–1904 TPQ Late 119 17.6 190 Southeast
Terrace 252 A.D. 1670–1810 TPQ and TAQ Late 171 13.0 230 Center
Terrace 311 A.D. 1490–1895 TPQ Late 42 8.5 299 Southeast
Terrace 321 A.D. 1544–1903 TPQ Late 128 18.9 188 Southeast
Terrace 48 A.D. 1470–1895 TPQ Late 390 14.0 281 Northwest
Terrace 68 A.D. 1707–1904 TPQ Late 56 12.3 260 Center
Terrace 8 A.D. 1415–1586 TPQ and TAQ Late 319 16.4 210 Northwest
Figure 9. The distribution of early (squares) and late (triangles) terraces.
12 S. QUINTUS ET AL.

there is a general centering pattern to the status architecture
(see Hutson 2016, 160–165), with a few large terraces found
outside the center and a number of smaller terraces inter-
mixed among the larger features in the focal point. This is
consistent with what might be expected if descent groups
lived near each other and if these descent groups were intern-
ally ordered hierarchically. Thus, the distribution of terrace
sizes may represent ordering at two scales: the community
as a whole and within each descent group.
The most notable difference between Luatele and settle-
ments zones on Ofu and Olosega is the presence of linear
mounds in the former. This is partially a result of raw
material availability; there is more stone available across
and around Luatele than on either Ofu or Olosega because
of the youthful age of Ta‘ū. However, the presence of these
linear mounds also indicates the perceived need to bound
and control land at multiple spatial scales. At the site-scale,
the large cross-slope walls (e.g., Wall 78) that form the
upslope boundary of the site demarcate a point on the land-
scape wherein soil fertility thresholds occur that correlate
with boundaries of intensive agriculture elsewhere in Polyne-
sia (Vitousek et al. 2014). The long boundaries parallel to the
slope highlight localized efforts to partition the land around
smaller social groups. The mixed form pathway (Wall 2)
stretching across the settlement is more integrative. Proxi-
mity to the long pathway may have been an important mar-
ker of status, as traffic along this pathway would have
exposed inhabitants in Luatele to the large complexes located
along the 200 m contour. This is consistent with the relation-
ship between primary walkways and elite structures docu-
mented in the western islands of the archipelago (Davidson
1974; Green 2002; Holmer 1980; Jennings, Holmer, and Jack-
mond 1982).
The available modelled construction ages of terraces and
linear mounds allow for an examination of the formation
of these patterns. Apart from one exception, the early ter-
races were built in a relatively narrow elevation and slope
band dispersed laterally across the Luatele substrate. This
slope and elevation band is similar to that of the focal
point. The one exception is the uniquely constructed terrace
on the side of the bluff(Terrace 98). Terrace construction in
and after the 15th century A.D. occurred around previously
built terraces and began to extend into higher elevations,
though the average slope values targeted seem to have stayed
consistent. The construction of linear mounds was largely
late, indicating that settlement density did not reach a
threshold where physical demarcation of land was necessary
until late in the cultural sequence. It further suggests that
these features were not merely the result of the piling of exca-
vated rocks from garden activities but, rather, were inten-
tional boundaries. The construction of these features,
especially the pathway across the settlement, evince
increased settlement integration in the 17th century A.D.or
later.
These data are largely consistent with expectations
derived from an IFD settlement model (Weitzel and Codding
2020). Early occupation across Luatele occurred in relatively
optimal locations in terms of soil fertility (elevations around
200 masl) and, generally, within a narrow band of slope
values. It is surprising that earlier terraces were not built
on gentler slopes relative to later terraces, but this may relate
to sampling. Alternatively, it may imply that differences in
slope below a certain threshold were not practically impor-
tant for inhabitants. Later construction marks population
growth and movement into more marginal locations, as
defined by soil fertility. Three of the later terraces are
below 100 m
2
in size and are relatively narrow, with widths
between 4 and 5 m. It is possible that these functioned as gar-
den spaces, but their size could still support small residential
structures. The two dated terraces that may lie on the Lata
substrate also date to the 16th century A.D. or later. Older
substrates tend to be less fertile (Vitousek et al. 2014), and
our results may document later movement into less optimal
areas, though the Lata substrate is still relatively young (less
than 100 kya), and soil fertility on the substrate is poorly
understood (Autufuga 2021).
The current configuration of the landscape, specifically
the correlation between terrace size, slope, and soil fertility,
in addition to the presence of boundary walls, meets expec-
tations of an IDD settlement model (see Bell and Winter-
halder 2014). However, the formation of this pattern is
difficult to unravel with our current data. The construction
of most linear mounds late in the cultural sequence following
increased landscape packing highlights an increased concern
with land boundaries. These features might mark a late tran-
sition to an IDD. If the two relatively early radiocarbon
determinations under linear mounds accurately date the
construction of those linear mounds, it would indicate that
boundaries were constructed first at the more fertile lower
elevations, as expected by an IDD model. That the three lar-
gest terraces from the early period were in the central and
northwestern sections of the Luatele substrate, areas of
slightly higher soil fertility relative to the southeast, could
imply early control of more suitable areas by higher-ranked
individuals. However, size differences were more muted at
this time, lateral soil fertility differences across Luatele are
limited, and there is little other evidence of competition or
attempts at control. Certainly, the equivocal evidence of sta-
tus distinctions during this early period contrasts with indi-
cations of differential rank visible in the early ‘Opunohu
Valley sequence in the Society Islands (Kahn and Kirch
2013).
Table 3. Modelled ages and characteristics of 10 linear mound segments (from Quintus et al. 2020). Two segments of Wall 2, the mixed form feature that runs
across much of the project area, were dated. Wall 78 is one of several cross-slope linear mounds at the upper elevation boundary of the site.
Name Modelled Age (95.4% HPD) Elevation (masl) Location Type
Wall 89 A.D. 1050–1897 186 Southeast Single Parallel to Slope
Wall 36 A.D. 1490–1895 180 Center Single Parallel to Slope
Wall 78 A.D. 1674–1902 311 Southeast Cross-Slope Boundary
Wall 92 A.D. 1685–1903 232 Center Double Parallel to Slope
Wall 48 A.D. 1698–1904 228 West-Center Single Parallel to Slope
Wall 2 Double Segment A.D. 1707–1904 213 Northwest Cross-Slope Double (Part of Larger Mixed Form Feature)
Wall 86 A.D. 1708–1903 214 East-Center Single Parallel to Slope
Wall 93 A.D. 1708–1905 270 West-Center Cross-Slope Single
Wall 2 Single Segment A.D. 1709–1905 200 West-Center Cross-Slope Single (Part of Larger Mixed Form Feature)
Wall 46 A.D. 1711–1905 261 Center Single Parallel to Slope
JOURNAL OF FIELD ARCHAEOLOGY 13

At present, we favor a model of gradual focal point emer-
gence. We hypothesize that initial use of the area established
the boundaries of the settlement. The creation of settlement
boundaries defined a central point after which time materi-
alizations of rank developed as a mechanism of political
legitimization and as a product of access to high suitability
land. Additional dating of the large terraces in the focal
point of the settlement would test this interpretation, and
we would expect those large terraces to date in later time
periods if our favored model is correct.
Conclusions
Terraced landscapes are an important component of the
archaeological record across the Pacific and, indeed, the
world (Treacy and Denevan 1994). In the Manu‘a group,
these cultural landscapes share a set of recurrent patterns,
indicating a shared settlement structure and, likely, the
impacts of shared environmental characteristics. In Luatele,
consideration of slope and soil fertility contributed to the
production of these patterns. Social status was a key variable
that intersected with slope and soil fertility to produce the
archaeological patterns that we documented. It does seem
that soil fertility was more influential than slope, as the archi-
tectural focal point was located in a position that privileged
soil fertility over low slope and because it appears that
slope was only a constraint when it reached values above
roughly 15°.
Material inequality seems to have been a product of these
environmental conditions and was emergent across Luatele,
which then was maintained by these same environmental
conditions. The architectural materialization of rank distinc-
tions appears to have been relatively muted when terracing
began to be constructed in Luatele, at least in comparison
to later times. This is consistent with the general lack of
boundary walls in this early time period, which suggests a
lower level of competition for land relative to later periods.
Terraces were constructed in more marginal locations over
time and between already established features. While smaller
features were built in more marginal locations, more elabo-
rated terraces were constructed in previously occupied
spaces. This landscape packing and eventual territorial
behavior appear to have been late phenomena in Luatele.
Evidence of paths and boundaries that cross-cut the settle-
ment suggest the presence of larger social units at this
time, as well. The late emergence of larger social units and
small-scale institutionalized inequality in Luatele is consist-
ent with evidence of the development of suprahousehold
authority after the 15th century A.D. elsewhere in the
Manu‘a group (Quintus, Allen, and Ladefoged 2016). Within
a social structure where extended families were land-holding
entities and rank was perceived through abundance and fer-
tility, access to and control of productive land was an impor-
tant mechanism of status maintenance and accumulation.
Our analysis of Luatele adds to a growing body of data
examining the growth of interior settlements across Polyne-
sia. Communities across the region targeted optimal
locations for initial settlement that reflected a concern with
agricultural production. Later settlement occurred in more
marginal locations, with settlement in optimal locations
becoming more elaborated to reflect power and status
accumulation. The latter is especially well documented in
the Society Islands and was likely a key driver of increased
institutionalized social inequality (Kahn and Kirch 2013).
While social inequality is far more muted in Manu‘a than
in the Society Islands, the position of the settlement focal
point in Luatele speaks to a concern of elites with access to
and control over fertile land. As noted globally, the transge-
nerational transmission of fertile lands through corporate
descent groups is an important mechanism that gives rise
to social inequality (Prufer et al. 2017; Shennan 2011). This
is notably true in Polynesia, where political practice was con-
cerned with demonstrations of mana. In this context, access
to and transgenerational transmission of fertile landscapes
provided foundations for stable hierarchical political organ-
izations in environmentally predictable locations.
Acknowledgements
We thank the people of Fitiuta, especially Eseta Kese and Pastor Fred
Scanlan, for hosting us during our research. We thank Logoleo Feagai
Logoleo for permission to work in Luatele. We wish also to recognize
the contributions of Tiffany Lee, Darby Filimoehala, Malone Ieti, Prin-
cecharles Faleagafulu, Christina Fu‘afu‘a, Tafa Fu‘afu‘a, Paulo Paulo,
Oceana Te‘i, Arthur Sega, Fafeta‘i Lauofo, Joshua Fu‘afu‘a, Falani
Masunu, Visa Vaivai Tiapusua, Brian Vivao, Fa‘afutai Lauofo, Fauato
Aukuso, Taumakai Atautia, Jonathon Mauga, Leonard Vivao, Lawrence
Fautua, Robert Mauga, J. J. Tanielu, and Achilles Tevasea to the success
of this research. We appreciate the helpful comments of JeffClark, Tom
Dye, Robert DiNapoli, Christian Peterson, and Jim Bayman on a pre-
vious draft of this manuscript. Finally, we thank the American Samoa
Historic Preservation Office, specifically Letitia Peau-Folau, Teleai
Christian Ausage, and Lancelot Leutu‘utuofiti Te‘i, for archaeological
and logistical support. Logistical assistance was provided by the
National Park of American Samoa under permit NPSA-2019-SCI-0001.
Funding
This research is based upon work supported by the National Science
Foundation under Grant No. NSF BCS-1732360.
Conflicts of Interest
We have no potential conflicts of interest to report.
Notes on Contributors
Seth Quintus (Ph.D. 2015, University of Auckland) is an Associate Pro-
fessor at the University of Hawai’iatMānoa. His research examines the
intersections of communities and environments in the past, especially
through food production practices. https://orcid.org/0000-0003-4388-
3862.
Dolly Autufuga (M.Sc. 2021, University of Hawai’iatMānoa) is a Ph.D.
student at the University of Hawai’iatMānoa. Her research interests
explore plant and soil sciences as they pertain to indigenous crops.
Stephanie Day (Ph.D. 2012, University of Minnesota, Twin Cities) is an
Associate Professor at North Dakota State University. Her research
makes use of terrestrial laser scanning and lidar to explore landscape
change in a range of geomorphological settings.
Jennifer Huebert (Ph.D. 2014, University of Auckland) is an archaeol-
ogist and archaeobotanist at Sunrise Archaeology (NZ). She specializes
in the study of wood and charred plant materials and has research inter-
ests in anthropogenic landscape change and food production systems in
Oceania.
Noa Kekuewa Lincoln (Ph.D. 2013, Stanford University) is an Associate
Researcher at the University of Hawai’iatMānoa. His research explores
the interactions between soil biogeochemistry and indigenous agroecol-
ogy using multi- and inter-disciplinary approaches. https://https://
orcid.org/0000-0003-1070-4290.
14 S. QUINTUS ET AL.

Nolita Motu (M.A. 2018, North Dakota State University) is an Archae-
ologist at the National Park of American Samoa. Her interests include
Samoan archaeology, GIS, and lidar applications.
Kyungsoo Yoo (Ph.D. 2003, University of California, Berkeley) is a Pro-
fessor at the University of Minnesota, Twin Cities. His research inter-
ests include soil genesis, landscape evolution, bioturbation, and
human impacts on soils. https://orcid.org/0000-0002-4677-5093.
References
Allen, M. S. 2010.“Oscillating Climate and Socio-Political Process: The
Case of the Marquesan Chiefdom, Polynesia.”Antiquity 84: 86–102.
Ashmore, W., and J. A. Sabloff.2002.“Spatial Orders in Maya Civic
Plans.”Latin American Antiquity 13: 201–215.
Autufuga, D. 2021.“Soil Nutrients and Traditional Agriculture on
Young Volcanic Soils of Ta’ū, American Samoa.”MA Thesis,
University of Hawai’iatMānoa, Honolulu.
Baer, A. 2016.“Ceremonial Architecture and the Spatial Proscription of
Community: Location Versus Form and Function in Kaupōm Maui,
Hawaiian Islands.”Journal of the Polynesian Society 125: 289–305.
Bell, A. V., and B. Winterhalder. 2014.“The Population Ecology of
Despotism: Concessions and Migration Under the Ideal Despotic
Distribution.”Human Nature 25: 121–135.
Clark, J. T., S. Quintus, M. Weisler, E. St. Pierre, L. Nothdurft, and Y.
Feng. 2016.“Refining the Chronology for West Polynesian
Colonization: New Data from the Samoan Archipelago.”Journal of
Archaeological Science: Reports 6: 266–274.
Davidson, J. M. 1969.“Settlement Patterns in Samoa Before 1840.”
Journal of the Polynesian Society 78: 44–88.
Davidson, J. M. 1974.“Samoan Structural Remains and Settlement
Patterns.”In Archaeology in Western Samoa, Vol. II, edited by R.
C. Green and J. M. Davidson, 225–244. Auckland: Auckland
Institute and Museum Bulletin 7.
Davies, R. B. 1987.“Hypothesis Testing When a Nuisance Parameter is
Present Only Under the Alternative.”Biometrika 74: 33–43.
DiNapoli, R. J., C. P. Lipo, T. Brosnan, T. L. Hunt, S. Hixon, A. E.
Morrison, and M. Becker. 2019.“Rapa Nui (Easter Island)
Monument (Ahu) Locations Explained by Freshwater Sources.”
PLoS One,https://doi.org/10.1371/journal.pone.0210409.
Fretwell, S. D., and H. L. Lucas. 1969.“On Territorial Behavior and
Other Factors Influencing Habitat Distribution in Birds.”Acta
Biotheoretica 19: 16–36.
Glover, H. J., T. N. Ladefoged, and E. E. Cochrane. 2020.“Costly
Signaling and the Distribution of Monumental Mounds in Savai’i
and ‘Upolu, Sāmoa.”Archaeology in Oceania 55: 141–152.
Green, R. C. 2002.“A Retrospective View of Settlement Pattern Studies
in Samoa.”In Pacific Landscapes: Archaeological Approaches, edited
by T. N. Ladefoged and M. Graves, 125–152. Los Osos: Easter Island
Foundation.
Holmer, R. N. 1976.“Mt. Olo Settlement Pattern Interpretation.”In
Excavations on ‘Upolu, Western Samoa, edited by J. D. Jennings, R.
N. Holmer, J. Janetski and H. L. Smith, 41–56. Honolulu: Pacific
Anthropological Records No. 25.
Holmer, R. N. 1980.“Mt. Olo Settlement Pattern Interpretation.”In
Archaeological Excavations in Western Samoa,edited by J. D.
Jennings and R. N. Holmer, 93–102. Honolulu: Pacific
Anthropological Records, No. 32.
Howard, A. 1985.“History, Myth, and Polynesian Chieftainship: The
Case of Rotuman Kings.”In Transformations of Polynesian
Culture, edited by A. Hooper and J. Huntsman, 39–77. Auckland:
Polynesian Society.
Hutson, S. R. 2016.The Ancient Urban Maya: Neighborhoods,
Inequality, and Built Form. Gainesville: University Press of Florida.
Jackmond, G., D. Fonoti, and M. M. Tautunu. 2018.“Samoa’s Hidden
Past: LiDAR Confirms Inland Settlement and Suggests Larger
Populations in Pre-Contact Samoa.”Journal of the Polynesian
Society 127: 73–90.
Jennings, J. D., R. N. Holmer, and G. Jackmond. 1982.“Samoan Village
Patterns: Four Examples.”Journal of the Polynesian Society 91: 81–
102.
Kahn, J. G., and P. V. Kirch. 2013.“Residential Landscapes and House
Societies of the Late Prehistoric Society Islands.”Journal of Pacific
Archaeology 4: 50–72.
Kennett, D. J., A. J. Anderson, and B. Winterhalder. 2006.“The Ideal
Free Distribution, Food Production, and the Colonization of
Oceania.”In Human Behavioral Ecology and the Origins of
Agriculture, edited by D. J. Kennett and B. Winterhalder, 265–288.
Berkeley: University of California Press.
Kennett, D., B. Winterhalder, J. Bartruff, and J. M. Erlandson. 2009.“An
Ecological Model for the Emergence of Institutionalized Social
Hierarchies on California’s Northern Channel Islands.”In Pattern
and Process in Cultural Evolution, edited by S. Shennan, 297–314.
Berkeley: University of California Press.
Kirch, P. V., A. S. Hartshorn, O. A. Chadwick, P. M. Vitousek, D. R.
Sherrod, J. Coil, L. Holm, and W. D. Sharp. 2004.“Environment,
Agriculture, and Settlement Patterns in a Marginal Polynesian
Landscape.”Proceedings of the National Academy of Science 101:
9936–9941.
Klenck, J. D. 2016.“Report for the Manu‘a Archaeological Survey,
Luatele or Judds Crater, Island of Ta‘u, American Samoa.”
Unpublished report on file at the American Samoa Historic
Preservation Office. Pago Pago: American Samoa.
Knapp, A. B., and W. Ashmore. 1999.“Archaeological Landscapes:
Constructed, Conceptualized, Ideational.”In Archaeologies of
Landscapes: Contemporary Perspectives, edited by W. Ashmore and
A. B. Knapp, 1–30. Oxford: Blackwell.
La Perouse, J. F. G. 1798.Voyage Round the World in the Years 1785,
1786, 1787, and 1788. London: J. Johnson.
Ladefoged, T. N., P. V. Kirch, S. M. Gon III, O. A. Chadwick, A. S.
Hartshorn, and P. M. Vitousek. 2009.“Opportunities and
Constraints for Intensive Agriculture in the Hawaiian Archipelago
Prior to European Contact.”Journal of Archaeological Science 36:
2374-2383.
Ladefoged, T. N., M. D. McCoy, G. P. Asner, and P. V. Kirch. 2011.
“Agricultural Potential and Actualized Development in Hawai’i:
An Airborne LiDAR Survey of the Leeward Kohala Field System
(Hawai’i Island).”Journal of Archaeological Science 38: 3605–3619.
Ladefoged, T. N., M. D. McCoy, and M. W. Graves. 2020.“The
Dynamics of Collective Action and Political Agency in Leeward
Kohala Hinterlands, Hawai’i Island.”Journal of Pacific Archaeology
11: 10–20.
Lehman, F. K., and D. J. Herdrich. 2002.“On the Relevance of Point
Fields for Spatiality in Oceania.”In Representing Space in Oceania:
Culture in Language and Mind. Pacific Linguistics 523, edited by G.
Bennardo, 179–197. Canberra: The Australian National University.
Lepofsky, D. S., and J. Kahn. 2011.“Cultivating an Ecological and Social
Balance: Elite Demands and Commoner Knowledge in Ancient
Ma’ohi Agriculture, Society Islands.”American Anthropologist 113:
319–335.
Martinsson-Wallin, H. 2007.“Samoan Archaeology: A Review of
Research History.”Archaeology in Oceania 42 (Supplement): 11–27.
Mattison, S. M., E. A. Smith, M. K. Shenk, and E. E. Cochrane. 2016.
“The Evolution of Inequality.”Evolutionary Anthropology 25: 184–
199.
McCoy, M. D., T. N. Ladefoged, M. W. Graves, and J. W. Stephen. 2011.
“Strategies for Constructing Religious Authority in Ancient
Hawai’i.”Antiquity 85: 927–941.
McDougall, I. 2010.“Age of Volcanism and Its Migration in the Samoa
Islands.”Geological Magazine 147: 705–717.
Mead, M. 1969.Social Organization of Manua. Honolulu: B.P. Bishop
Museum Bulletin No. 76.
Morrison, A. E., and J. T. O’Connor. 2018.“Settlement Pattern Studies
in Polynesia: Past Projects, Current Progress, and Future Prospects.”
In The Oxford Handbook of Prehistoric Oceania, edited by E.
Cochrane and T. Hunt. New York: Oxford University Press. DOI:
10.1093/oxfordhb/9780199925070.013.024.
Motu, N. 2018.“Light Detection and Ranging (Lidar) Technology in
Archaeology and the Human-Environment Interaction: The Case
of Ta‘u Island, Manu‘a, American Samoa.”MA Thesis. North
Dakota State University, Fargo.
Muggeo, V. M. R. 2008.“Segmented: An R Package to Fit Regression
Models with Broken-Line Relationships.”R News 8 (1): 20–25.
Petchey, F. J., and P. V. Kirch. 2019.“The Importance of Shell: Redating
of the To‘aga Site (Ofu Island, Manu‘a) and a Revised Chronology
for the Lapita to Polynesian Plainware Transition in Tonga and
Sāmoa.”PLoS One 14: e0211990.
Phillips, N., T. N. Ladefoged, B. W. McPhee, and G. P. Asner. 2015.
“Location, Location, Location: A Viewshed Analysis of Heiau
JOURNAL OF FIELD ARCHAEOLOGY 15

Spatial and Temporal Relationships in Leeward Kohala, Hawai’i.”
Journal of Pacific Archaeology 6: 21–40.
Prufer, K. M., A. E. Thompson, C. R. Meredith, B. J. Culleton, J. M.
Jordan, C. E. Ebert, B. Winterhalder, and D. J. Kennett. 2017.“The
Classic Period Maya Transition from an Ideal Free to an Ideal
Despotic Settlement System at the Polity of Uxbenkãi.”Journal of
Anthropological Archaeology 45: 53–68.
Quintus, S. 2012.“Terrestrial Food Production and Land Use in
Prehistoric Samoa: An Example from Olosega Island,
Manu’a, American Samoa.”Archaeology in Oceania 47: 133–
140.
Quintus, S. 2015.“Dynamics of Agricultural Development in
Prehistoric Samoa: The Case of Ofu Island.”PhD Thesis, The
University of Auckland, Auckland.
Quintus, S. 2018.“Historicizing Food Production in Polynesia: A Case
Study of 2700 Years of Land Use on Ofu Island, American Samoa.”
Journal of Field Archaeology 43: 222–235.
Quintus, S. 2020.“Community Relationships and Integration at a
Small-Island Scale.”Journal of Pacific Archaeology 11: 61–70.
Quintus, S., M. S. Allen, and T. N. Ladefoged. 2016.“In Surplus and in
Scarcity: Agricultural Development, Risk Management, and Political
Economy, Ofu Island, American Samoa.”American Antiquity 81:
273–293.
Quintus, S., and J. T. Clark. 2016.“Space and Structure in Polynesia:
Instantiated Spatial Logic in American Samoa.”World Archaeology
48: 395–410.
Quintus, S., S. S. Day, and N. J. Smith. 2017.“The
Efficacy and Analytical Importance of Manual Feature Extraction
Using Lidar Datasets.”Advances in Archaeological Practice 5:
351–364.
Quintus, S., J. Huebert, S. Day, N. Lincoln, K. Yoo, T. Lee, D.
Filimoehala, and D. Autufuga. 2020.“Tempo and Trajectory of the
Built Landscape on Ta’u Island, Manua’Group, American Samoa:
Integrating Extensive Radiocarbon Dating with Joint Posterior
Modeling.”Radiocarbon 62: 1317–1337.
Quintus, S., and N. K. Lincoln. 2020.“Integrating Local and Regional in
Pre-Contact Hawaiian Agriculture at Kahuku, Hawai’i Island.”
Environmental Archaeology 25: 53–68.
RCoreTeam.2021.R: A Language and Environment for Statistical
Computing. Vienna, Austria: R Foundation for Statistical Computing.
Shennan, S. 2011.“Property and Wealth Inequality as Cultural Niche
Construction.”Philosophical Transactions of the Royal Society B
366: 918–926.
Shore, B. 1982.Sala’ilua: A Samoan Mystery. New York: Columbia
University Press.
Shore, B. 1989.“Mana and Tapu.”In Developments in Polynesian
Ethnology, edited by A. Howard and R. Borofsky, 137–173.
Honolulu: University of Hawaii Press.
Shore, B. 2014.“A View from the Islands: Spatial Cognition in the
Western Pacific.”Ethos 42: 376–397.
Soil Survey Laboratory Staff..1992.Soil Survey Laboratory Methods
Manual. Washington D.C.: Soil Survey Investigations Report No.
42, USDA-SCS.
Stice, G. D., and F. W. McCoy. 1968.“The Geology of the Manu’a
Island, Samoa.”Pacific Science 22: 427–457.
Treacy, J. M., and D. M. Denevan. 1994.“The Creation of Cultivatable
Land Through Terracing.”In The Archaeology of Garden and Field,
edited by N. F. Miller and K. L. Gleason, 91–110. Philadelphia:
University of Pennsylvania Press.
Vitousek, P. M., O. A. Chadwick, S. C. Hotchkiss, T. N. Ladefoged, and
C. M. Stevenson. 2014.“Farming the Rock: A Biogeochemical
Perspective on Intensive Agriculture in Polynesia.”Journal of
Pacific Archaeology 5: 51–61.
Vitousek, P. M., T. N. Ladefoged, P. V. Kirch, A. S. Hartshorn, M. W.
Graves, S. C. Hotchkiss, S. Tuljapurkar, and O. A. Chadwick. 2004.
“Soils, Agriculture, and Society in Precontact Hawai’i.”Science
304: 1665–1669.
Weitzel, E. M., and B. F. Codding. 2020.“The Ideal Free Distribution
Model and Archaeological Settlement Patterning.”Environmental
Archaeology, doi:10.1080/14614103.2020.1803015.
Wesiler, M., and P. V. Kirch. 1985.“The Structure of Settlement Space
in a Polynesian Chiefdom: Kawela, Molokai, Hawaiian Islands.
Hawaiian Islands. New Zealand Journal of Archaeology 7: 129–158.
Winterhalder, B., D. J. Kennett, M. N. Grote, and J. Bartruff.2010.
“Ideal Free Settlement of California’s Northern Channel Islands.”
Journal of Anthropological Archaeology 29: 469–490.
16 S. QUINTUS ET AL.