Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=rwar20
Download by: [University of Hawaii at Manoa] Date: 02 October 2015, At: 11:30
World Archaeology
ISSN: 0043-8243 (Print) 1470-1375 (Online) Journal homepage: http://www.tandfonline.com/loi/rwar20
Dating human dispersal in Remote Oceania: a
Bayesian view from Hawai’i
Thomas S. Dye
To cite this article: Thomas S. Dye (2015) Dating human dispersal in Remote
Oceania: a Bayesian view from Hawai’i, World Archaeology, 47:4, 661-676, DOI:
10.1080/00438243.2015.1052845
To link to this article: http://dx.doi.org/10.1080/00438243.2015.1052845
Published online: 09 Jul 2015.
Submit your article to this journal
Article views: 149
View related articles
View Crossmark data
Dating human dispersal in Remote
Oceania: a Bayesian view from Hawaii
Thomas S. Dye
Abstract
Settlement date estimates for Hawaii and New Zealand are derived using Bayesian calibration of radio-
carbon dates on paleoenvironmental and archaeological samples to demonstrate that the Bayesian frame-
work provides the tools needed to resolve the order of settlement events in Remote Oceania, as well as the
time elapsed between them. It predicts that archaeologists will successfully rene the dating of human
dispersal elsewhere in Remote Oceania when they work collaboratively to build chronological models
within a Bayesian framework.
Keywords
Radiocarbon; Bayesian calibration; Pacic; Hawaii; New Zealand.
Introduction
The voyages of discovery that populated the far-ung islands of Remote Oceania (Fig. 1)
constitute one of mankinds great achievements. The magnitude of the enterprise, which saw
humans with a technology based on stone tools migrate through a region that covers about
one-thi rd of the globes surface, has red the Western imagination at least since the voyages
of Cook in th e eighteenth century (Howard 1967). Using radiocarbon dating, augmented
infrequently by U/Th (Burley, Weisler and Zhao 2012) and optically stimul ated lumines-
cence (Clark and Anderson 2009), archaeologists have established that human dispersal
through the Pacic was episodic (Anderson et al. 2006)orpulse-like (Rieth and
Cochrane forthcoming), or, more colorfully, was accomplished in a series of explosive
phases (Bellwood 2013, 197), with initial pushes out of Island Southeast Asia into Western
Micronesia perhaps as early as 1600
BC and by the Lapita peoples (Kirch 1997)ofNear
Oceania by about 1200
BC, expansion through central and eastern Micronesia about 200 BC
and, nally, settlement of the far-ung islands of Eastern Polynesia about AD 1000 (Kirch
World Archaeology Vol. 47(4): 661676 Prehistoric Bayesian Chronologies
© 2015 Taylor & Francis ISSN 0043-8243 print/1470-1375 online
http://dx.doi.org/10.1080/00438243.2015.1052845
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
2010). Although the dates of these explosive phases arelikelytochangeasmoreandbetter
data are collected, the focus of current archaeological debate concerns the tempo of dispersal
achieved during each phase. This article argues that a Bayesian framework is suited to this
task.
The central datum of a dispersal model is the relationship between the settlement date
estimates o f two islands or island groups. Tw o fundamental questions about the relationship
are of particular interest: (1) the order of settlement whichislandorislandgroupwas
settled before the othe r; and (2) the elapsed time between settlement of the rst island or
island group and the second. A Bayesian approach to these questions develops a chron-
ological model (Buck, Cavanagh and Litton 1996) suited both to the question at hand and
to the available data (e.g. Hamilt on and Kenney 2015). A review of the history of
settlement estimates in Hawaii and New Zealand, two island groups that have been
investigated intensively by Pacic archaeologists, indicates that the Bayesian model is
well-behaved, in contrast to alternatives that tend to m agnify discontinuities associated
with a disparity (Dean 1978). In New Zealand, where a sustained attempt to control for
disparity discontinuities and a concerted effort to collect suitable data forged a consensus
estimate more than twenty years ago, application of the Bayesian model yields a settlement
date estimate that closely matches the consensus. Comparison of the settlement date
estimates for Haw aii and New Zealand within a Bayesian framework yields a conclusive
answer to the question of order and a useful estimate of the elapsed time between settle-
ment events. The article concludes that debates over t he tempo of dispersal can be resolved
by archaeologists working collaboratively to develop chronological models within a
Bayesian framework.
Figure 1 A model of human dispersal to the islands of Remote Oceania, showing the islands and island
groups mentioned in the text.
662 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
Discovery and settlement of Hawaii
Estimates of the date Hawai i was discovered and settled by Polynesians have varied consider-
ably over the last century (Fig. 2). Prior to the radiocarbon revolution, tradition-based estimates
of the settlement date varied by about 500 years, from
AD 400500 (Fornander 1916 19)toAD
9001000 (Emory 1928). Subsequently, the trend over the rst four decades of the radiocarbon
era was towards increasingly older estimates, due primarily to the accumulating effect of dates
on materials with in-built age (Rieth and Stephen Athens 2013). Another contributing factor was
a move away from estimating the ages of early sites to investigations based on the age
distribution of large radiocarbon date corpora that resulted in an estimate that Hawaii was
settled in the rst century
AD (Hunt and Holsen 1991).
The rst challenge to this pattern of increasingly old settlement estimates was the application
of chronometric hygiene’–a protocol for accepting or rejecting radiocarbon dates that
yielded an estimate of the settlement date as late as
AD 1000 (Spriggs and Anderson 1993).
This estimate initiated a period of almost twenty years during which archaeologists split into
camps that favored either a long or a short chronology, despite the results of paleoenviron-
mental work that showed charcoal was absent through much of the Holocene and yielded
radiocarbon dates that supported a short chronology (Athens 1997; Athens et al. 2002). The
split was not resolved until recently, when three papers (Wilmshurst et al. 2011; Kirch 2011;
Dye 2011) independently estimated a chronology shorter than the short chronol ogy proposed
by Spriggs and Anderson (1993). This new consensus was strengthened recently by a fourth
settlement date estimate (Athens, Rieth and Dye 2014). Looking back on the rst six decades of
radiocarbon dating in Hawaii from the vantage point of todays consensus, it is striking that
most of the settlement date estimates proposed by archaeologists moved in the direction of
Figure 2 A history of settlement date estimates for Hawaii. The dashed vertical line marks the invention of
radiocarbon dating. The shaded band near the top shows the 68.2 per cent (dark) and 95 per cent (light)
highest posterior density regions for a recent Bayesian estimate of the settlement date (see Fig. 4, right).
Sources: Athens (1997), Bellwood (1979), Emory (1928), Fornander (191619), Graves and Addison
(1995), Hunt and Holsen (1991), Kirch (1985, 2000, 2011), Sinoto (1970), Spriggs and Anderson (1993),
Wilmshurst et al. (2011). Modied after https://commons.wikimedia.org/wiki/File:Settlement-estimates-
hawaii.svg.
Dating human dispersal in Oceania 663
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
greater error and not towards the current consensus value. The ad hoc methods used to estimate
the settlement date during this period behaved poorly in practice and proved incapable of
leading archaeologists to an accurate estimate.
Today, disagreements among settlement date estimates are narrower than previously, and the
basis for the disagreements can be specied. The disagreement between the esti mate by
Wilmshurst et al. (2011) and the Bayesian estimates of Dye (2011) and Athens, Rieth and
Dye (2014) can be attributed to difculties in estimating the settlement date with a disparity
(Dean 1978). The archaeologist dating with a disparity (Fig. 3 left), in which the reference event
is younger than the target event, must guard against the possibility that composition and
association errors yield an estimate of the dated event that is older than the target event,
which would confound their expected temporal relationship. Chronometric hygiene deals with
this situation by applying a protocol designed to eliminate dates with the potential to confound
the expected temporal relationship. A weakness of chronometric hygiene is that it is not possible
to know in advance whether a dated event is older or younger than a target event, leaving the
archaeologist to decide whether to accept or reject a date on the basis of indirect criteria more or
less plausibly related to chronology. In the case of Wilmshurst et al. (2011) the criteria employed
appear to have been too strict, resulting in an estimate of the settlement date that is too young.
The archaeologist has different concerns when dating with a disjunction, where the dated event
is older than the target event (Fig. 3 right). One advantage of dating with a disjunction is that
when discontinuities are large, either because suitable short-lived materials were not recovered
or because materials from the reference event could not be condently distinguished from
possibly residual materials with in-built age, the temporal relationship between the dated
event and the target event is not altered; the dated event is simply older than the target event
and estimates it imprecisely.
Figure 3 Directed graph representation of discontinuities in archaeological dating identied by Dean
(1978): left, dating with a disparity, in which the reference event is younger than the target event; right,
dating with a disjunction, in which the reference event is older than the target event. The arrows point from
a younger event to an older event. Labels on the arrows indicate potential sources of discontinuity.
Et = target event, the event to which the date is to be applied (Dean 1978, 228), Er = reference event,
the potentially dated event that is most closely related to the phenomenon to which the date is to be
applied (Dean 1978, 228), and E
d
= dated event, the event that is actually dated (Dean 1978, 226).
664 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
A Bayesian solution to the problem of estimating the settlement date of Hawai i does away
with the need to choose between a disparity and a disjunction. It establishes two abutting
phases, a pre-settlement phase, whose start and end dates can be represented as
α
pre and β
pre
,
respectively, and a post-settlement phase, whose start and end dates can be represented as α
post
and β
post
, respectively (Dye 2011). In outline, the chronological model is α
pre
>β
pre
= α
post
>β
post
,
where > means is older than and = indicates that the phases are abutting, i.e. that the pre-
settlement phase ended when the islands were discovered by Polynesians and the post-settle-
ment phase began.
Following the work of Athens and his colleagues (Athens 1997; Athens et al. 2002), Dye
(2011) argued that pre-settlement deposits could be identied by the absence of charcoal
particles in cores collected on the older, northern islands of the archipelago where volcanism,
the only plaus ible natural source of fores t res, has long been dormant. At Ordy Pond on the old
northern island of Oahu, where the paleoenvironmental record was especially well preserved,
charcoal rst appeared as tiny particles immediately before pollen from Polynesian introduced
plants (Athens 1997; Athens et al. 2002). This was interpreted as convincing proof that the
charcoal was anthropogenic, and it raised condence that charcoal-free sediments were reliable
indicators of pre-settlement deposition. Dye (2011) argued further that dates on plants and
animals introduced to the islands by Polynesians, including the human commensal Polynesian
rat, sweet potato, breadfruit, candlenut, bottle gourd and ti plant can be condently associated
with the post-settlement phase because the probability that archaeologists recovered and dated
materials brought from the homeland by the Polynesian discoverers is negligible. Bayesian
calibration of one pre-colonization phase radiocarbon date on a seed and six post-colonization
phase radiocarbon dates on Polynesian introductions yielded a settlement date estimate with a
95 per cent highest posterior density region of
AD 7801119 (Fig. 4, left).
Recently, this model was augmented to produce a more precise estimate of the settlement
date. Athens, Rieth and Dye (2014) argued that pre-settleme nt phase deposits on the geologi-
cally younger islands with active volcanism could be identied by a pollen spectrum that
included native taxa that had elsewhere been shown to decline precipitously following human
settlement. This resulted in the addition of a single date for the pre-settlement phase. In addition,
the criteria for recognizing post-settlement dates wer e relaxed to include short-lived plant taxa
Figure 4 Bayesian settlement date estimates for Hawaii: left, the 95 per cent highest posterior density
region is
AD 7801119; right , the 95 per cent highest posterior density region is AD 9401129.
Sources: Dye (2011, g. 1); Athens, Rieth and Dye (2014, g. 3).
Dating human dispersal in Oceania 665
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
and parts recovered from condently identied re-pit features. In all, tw enty-seven dates for
the post-settlement phase were added. Bayesian calibration yielded a settlement date estimate
with a 95 per cent highest posterior density region of
AD 9401129 (Fig. 4, right), a gain in
precision of 150 years, mostly at the early end of the estimate.
These results demonstrate that the Bayesian model is well-behaved; increasing the number of
valid observations increases the precision of the settlement date estimate.
Discovery and settlement of New Zealand
The history of archaeological estimates of the New Zealand settlement date differs from that of
Hawaii, in part because of differences in the archaeological records of the two island groups.
Unlike Hawaii, where settlement-era sites are difcult to identify and have yet to yield a
diagnostic artifact assemblage (Bayman and Dye 2013,345), the earliest sites in New Zealand
are highly visible and yield distinctive artifacts. They typically include rema ins of extinct fauna
endemic to New Zealand, especially the eleven species of moa bird, which, simulations and
direct evidence indicate, were driven to extinction within a century of human settlement
(Holdaway and Jacomb 2000; Holdaway et al. 2014). Early New Zealand sites also yield a
suite of artifacts directly comparable to artifacts found in the East Polynesian homeland and
unlike the artifacts produced by contact-era Maori (Davidson 1984, 1994). This so-called
Archaic assemblage, which appears to have developed in Eastern Polynesia, has not been
identied in Hawaii (Kirch 1986,201). The differences are also due to the contributions of
paleoenvironmental investigations in New Zealand, which have collected and analyzed abun-
dant data related to the settlement date question. A third contributing factor is that radio carbon
dating laboratories have been operating in New Zealand for many years and the collaboration of
archaeologists with radiocarbon scientists has yielded benets.
Prior to the radiocarbon revolution, S. Percy Smith synthesized regional Maori traditions and
brought them into line with nineteenth-century discoveries of a Moa Hunter culture antecedent
to the traditional Maori (Simmons 1969 ). Smith s synthesis posited three migrations to New
Zealand, the earliest of which, associated with Kupe , was dated genealogically to the early to
mid-tenth century
AD (Fig. 5). This version of the settlement chronology was popularized in
several inue ntial books (e.g. Buck 1925 ) and taught to New Zealand schoolchildren.
An attempt to interpret the earliest radiocarbon dates for New Zealand pushed back the
estimate of the settlement date to
AD 750850 (Groube 1968, 145), but over the next decade and
a half, as archaeologists gained some appreciation for the likely effects of old wood and
recognized some of the uncertainties of the radiocarbon method, the consensus estimate
broadened to include the traditional settlement date estimate within its range.
The much earlier settlement date estimate proposed by Sutto n (1987) was based primarily on
a review of palynological and geomorphological evidence and characterized by a willingness to
interpret changes in the paleoenvironmental record as anthropogenic rather than natural. The
response was immediate and vigorous (Sutton 1994a, 9), and it raised serious objections to
Suttons analysis and interpretation. Subsequent support for an early settlement date estimate
among archaeologists was rare (e.g. Bulmer 1989, 7001). A decade later, when dates on bones
of the human commensal Polynesian rat yielded unexpectedly early results (Holdaway 1996),
later shown to be erroneous (Wilmshurst et al. 2008), they were interpreted as implying an
666 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
early, transient, human contact with New Zealand more than 1,000 years before settlement
(Holdaway 1996, 225) and not as evidence for settlement per se.
Opposition to Suttons proposal appears to have galvanized New Zealand archaeologists in
the effort to minimize discontinuities associated with a disparity. Anderson (1991) applied
chronometric hygiene to the existing corpus of relevant archaeological dates, yielding an
estimate that ranged to the early fourteenth century and effectively established the modern
consensus. This was followed by the discovery that moa eggshell, which preserves well in
archaeological sites and lacks in-built age, could be made to yield reliable age determinations
(Higham 1994). Moa eggshells previously collected from an archaeological site with a rich
assemblage of Archaic artifacts at Wairau Bar yielded dates that were interpreted as indicating
the site was inhabited in
AD 12881300 (Higham, Anderson and Jacomb 1999). The dating of
the Wairau Bar site was subsequently conrmed by dates on nine moa eggshells collected from
a large oven pit at the site (Jacomb et al. 2014, 29). Most recently, Holdaway et al. (2014)
reported ninety-six dates on archaeological moa eggshell from seven sites, including Wairau
Bar, on the east coast of the South Island, and used ninety-three of them to estimate a 95 per
cent highest posterior density region of
AD 12941330 for the onset of Polynesian interaction
with moa. The posterior probability distribution for the start date of human interaction with moa
is constrained at its early end by the date of the Kaharoa tephra with the argument that there is
no denitive proof of human presence prior to deposition of the tephra, which covered
30,000km
2
of the northern and eastern North Island over a brief period in AD 1314±6 according
to a wi ggle-match date on wood encased by the tephra deposit (Hogg et al. 2003). This
argument has two potential aws. First, the Kaharoa tephra covers a large area and it is not
possible to look under it all for eviden ce of prior human activity. Absence of evidence is not
evidence of absence. Second, the Kaharoa tephra deposit, large as it is, covers only a portion of
the North Island. An argument that insists on direct stratigraphic evidence of pre-Kaharoa tephra
Figure 5 A history of settlement date estimates for New Zealand. The dashed vertical line marks the
invention of radiocarbon dating. The shaded band near the top shows the 68.2 per cent (dark) and 95 per
cent (light) highest posterior density regions for a Bayesian estimate of the settlement date (see Fig. 6).
Sources: Buck (1925), Groube (1968), Davidson (1984), Sutton (1987), Anderson (1991), McGlone and
Wilmshurst (1999). Higham, Anderson and Jacomb (1999), Hogg et al. (2003), Wilmshurst et al. (2008),
Holdaway et al. (2014).
Dating human dispersal in Oceania 667
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
human activity potentially ignores pertinent evidence elsewhere in New Zealand. The Kaharoa
tephra is precisely dated, but there appears to be no intrinsic connection between its eruption and
Polynesian settlement of New Zealand. The two events just happen to be pene-
contemporaneous.
Hand in hand with this archaeological progress, environmental scientists collected abundant
data on the timing of anthropogenic change. An analysis of pollen and charcoal evidence from
peat bogs, swamps, estua ries and lakes concluded that the rst evidence for the environmen tal
effects of human settlement dated to
AD 12001400 (McGlone and Wilmshurst 1999). Although
the Kaharoa tephra has not been found overlying an archaeological site, indicators of human
settlement are reported in paleoenvironmental cores immediately beneath the tephra (Lowe et al.
2000) and seeds gnawed by the human commensal Polynesian rat have been found encased in it
(Wilmshurst and Higham 2004). Rat-gnawed seeds and rat bones have both been dated
extensively (Wilmshurst and Higham 2004; Wilmshurst et al. 2008); materials associated with
rats are argued to be a good proxy for rst settlement because the animals reproduce quickly and
spread widely, a nd both seeds and bones are relatively easy to nd and can be dated condently
(e.g. Wilmshurst and Higham 2004, 8012).
A Bayesian chronological model similar to the one developed for Hawaii, which distin-
guishes pre-settlement and post-settlement phases, can be applied to the New Zealand situation
if criteria for distinguishing pre-settlement from post-settlement deposits can be developed. The
criteria used to identify pre-settlement phase deposits in Hawaiian paleoenvironmental cores do
not work in New Zealand, where pre-settlem ent res were widespread and the marked vegeta-
tion changes brought on by Polynesian settlement of Hawaii are not found (McGlone and
Wilmshurst 1999). A potential criterion for identifying pre-settlement deposits in New Zealand
builds on the argument that rat-gnawed seeds are a sensitive indicator of Polynesian settlement.
If this is so, then seed caches without evidence of rat gnawing probably represent deposits that
were not active into the post-settlement phase and the seeds within them must date to the pre-
settlement phase. Post-settlement phase dates on moa eggshell, Polynesian rat bones and rat-
gnawed seeds are abundant, as discussed above.
The chronological model can be expressed algebr aically as α
pre
>Θ
pre
>β
pre
= α
post
>Θ
post
>
β
post
, where Θ
pre
is the set of dated events θ
1
...θ
9
from the pre-settlement phase, represented by
the intact and bird-cracked seed cases from caches at Long Beach, Dunedin and Waitoetoe,
Taranaki (Wilmshurst et al. 2008, table S2); and Θ
post
is the set of dated events θ
10
...θ
108
from
the post-settlement phase, comprising eleven moa eggshells from Wairau Bar (Higham,
Anderson and Jacomb 1999), nine moa eggshells from an oven pit at Wairau Bar (Jacomb
et al. 2014), forty-eight rat-gnawed seed cases (Wilmshurst and Higham 2004; Wilmshurst et al.
2008) and thirty Polynesian rat bones (Wilmshurst et al. 2008).
Bayesian calibration yields a 95 per cent highest posterior density (HPD) region of
AD
12701309 (Fig. 6), which ts well with the current consensus estimate of the settlement
date. The 95 per cent HPD, which spans just four decades, overlaps the range of every
settlement date estimate since Anderson (1991).
As Denham, Ramsey and Specht (2012) point out, one advantage of estimating settlement
dates in a Bayesian framework is that the estimates can then be compared directly. It is possible
to answer the fundamental question of which island group was settled before the other with a
probability that expresses the condence that the answer is true, given the model and the data. In
this case, the probability Hawaii was settled before New Zealand is greater than 0.99, a virtual
668 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
certainty. A Bayesian answer to the second fundamental question, how many years elapsed
between the settlement of Hawaii and New Zealand, takes the form of a posterior probability
distribution with a modal value of 170 years and a 95 per cent HPD of 90259 years (Fig. 7).
Bayesian settlement date estimates in Remote Oceania
Other Bayesian settlement date estimates for islands and island groups in Remote Oceania use
chronological models based on disparities.
Green, Jones and Sheppard (2008) established a two-phase sequence for the SE-SZ-8 Lapita
site on Nendö Island in the Santa Cruz Group. The chronological phases were based on the
Figure 7 A Bayesian estimate of the time interval between Polynesian settlement of Hawaii and of New
Zealand. The 95 per cent highest posterior density region is 90259 years. The 68.2 per cent highest
posterior density region is 140209 years.
Figure 6 A Bayesian settlement date estimate for New Zealand. The 95 per cent highest posterior density
region is
AD 12701309.
Dating human dispersal in Oceania 669
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
cultural stratigraphy of the site, with the earlier phase corresponding to the Layer 2 Lapita
deposit and the late r phase to the modern garden soil of Layer 1. No dates were obtained from
pre-settlement Layer 3 deposits. Based on four marine shell dates from Layer 2 and a single date
on wood charcoal from Layer 1, Bayesian calibration estimated the onset of cultural activity at
17001050
BC. The wide range of this estimate is due to the few dates from Layer 2 and the lack
of dates from Layer 3, which might serve to constrain the early tail of the estimate.
Denham, Ramsey and Specht (2012) used a Bayesian framework to compare and contrast two
chronometric hygiene protocols applied to data on the settlement of Mussau Island and the rest
of the Bismarck Archipelago in Near Oceania and the subsequent dispersal of humans from the
Bismarck Archipelago to the island groups of Vanuatu and Fiji in Remote Oceania. The
chronological model established a single phase for each island or island group with no
constraints among the phases. The 95 per cent HPD of the estimated settlement date for
Vanuatu was 14841076
BC and for Fiji 13441024 BC. Depending upon which chronometric
hygiene protocol was used, the estimate of time elapsed between sett lement of the Bismarck
Archipelago and human dispersal into Remote Oceania varied between 99478 years and
36375 years. These variable and imprecise results are inherent in the attempt to estimate the
settlement date with a disparity. In addition, they highlight the central role of the subjectively
chosen rejection protocol in chronometric hygiene.
Discovery and excavation of a remarkable settlement-era burial ground at the Teouma site on
Efate Island in Vanuatu yielded dates on human bone, which eliminates potential context and
association problems, but the need to estimate the relative proportions of the diet provided by
terrestrial foods, which take up carbon from the atmospheric reservoir and marine foods, which
take up carbon from the older, marine reservoir, complicates the analysis (Petchey et al. 2014).
Based on measures of δ
13
C and δ
15
N in the human bone and comparison with published values
for various Pacic Island food sources, two potential diets were hypothesized that differed in the
contribution of marine foods. The dietary effects on bone dates from four burials were compared
with
14
C dates on associated marine shell artifacts, leading to the preference for a correction
routine based on the hypothesized diet with a greater contribution of marine foods. The dates on
human bone and marine shell artifacts were calibrated using a single-phase model with
trapezoidal priors for the phase boundaries, and compared to dates on materials from contem-
porary deposits adjacent to the cemetery, yielding a 68.2 per cent HPD estimate of 990930
BC
for the onset of regular cemetery use. This complex model yields precise results compared to
other analyses that also lack information from a pre-settlement phase.
Excavation of the Bourewa site in Fiji yielded a problematic suite of radiocarbon dates that
suffers from problems of context, association and sample composition (Nunn and Petchey 2013,
2730). Based on expectations about the age of the site, ten out of sixty-eight dates were
rejected as either unexpectedly old or too young given the archaeological context from which
they were collected. Approximately 20 per cent of the dated marine shell samples are on
medium to low-reliability species for radiocarbon dating (29); these samples are included in
the calibration because their radiocarbon ages conformed to expectations. Twelve dated charcoal
samples included in the calibration could be identied only as angiosperm wood, so it was not
possible to determine whether or not they might incorporate in-built age. Rather than reject age
determinations on unidentied wood charcoal, as is often the case in applications of chrono-
metric hygiene, a decision was made to weight samples according to how likely they are to be
correct based on assumptions outl ined within set parameters and rely on a model-averaging
670 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
approach (27). The 95 per cent HPD for the initial occupation of the Bourewa site yielded by
the calibration is 917822
BC, which is two to four centuries younger than the estimate of the
settlement date for Fiji calibrated by Denham, Ramsey and Specht (2012), also based on dates
from the Bourewa site and about three times as precise.
Kennett et al. (2012) used radiocarbon dating as a survey technique to yield a chronological
framework for landscape use on Rapa Island. Arc haeological contexts were categorized as
Initial Colonization, Coastal Expansion, Initial Fortication, First Fortication Expansion and
Final Fortication Expansion. OxCal software (Bronk Ramsey 2001) was used to implement an
island-wide chronological model with overlapping phases. Using seven radiocarbon dates from
the basal deposit of a coastal rock-shelter, Bayesian calibration estimated the start date of the
Initial Colonization phase at
AD 8001300 (Kennett et al. 2012, 197). The wide range of the
estimate is due to the small number of dates assigned to the Initial Colonization phase and the
fact that the chronological model lacks a constraint on the lower boundary of the phase.
Discussion
The precision of a settlement date estimate yielded by a Bayesian framework can be calculated
because the estimate is reported as a posterior probability distribution. The difference in
precision of the Hawaii settlement date estimate, which has a 95 per cent HPD of 190 years,
and the New Zealand estimate, which has a 95 per cent HPD of forty years, indicates the need
for additional pre-settlement and post-settlement phase data from Hawaii. Renewed analysis of
the Ordy Pond sequence that leverages technologi cal advances that make it possible to date
pollen (Piperno et al. 2007) has the potential to yield dates on pre-settlement phase material that
immediately precedes the introduction of charcoal to sediments. The utility of introduced
Polynesian rat bone as a post-settlement phase dating material is apparent in the experiments
carried out so far (Dye 2011; Athens, Rieth and Dye 2014), and Hawaiian archaeologists might
follow the productive lead of their New Zealand colleagues to expand the inventory of
Polynesian rat-bone dates.
Estimating the tempo of human dispersal through Remote Oceania using the two-phase
Bayesian model described here raises at least two practical problems. First, the model requires
dating of pre-settlement contexts, which archaeologists in Remote Oceania rarely attempt. As
the examples of Hawaii and New Zealand show, how best to identify and date pre-settlement
contexts varies from one place to another, a circumstance that highlights the potentially
important role of regional paleoenvironmental specialists. Second is that the comparison of
settlement date estimates at the heart of a dispersal model requires that Bayesian calibration
results for individual islands and island groups be distributed in a way that makes it possible for
other investigators to evaluate them fully and perhaps incorporate them into their own work.
Here, archaeologists nd themselves in the position of investigators in other elds, where an
increasingly important aspect of statistical practice is the dissemination of statistical methodol-
ogy, data analyses and statistical arguments. While statistical practice has evolved to encompass
more computation and larger and more complex datasets and models, the primary vehicle for
delivery has remained the static, printed page (Gentleman and Temple Lang 2007,45).
One solution to this general probl em is a digital compendium that includes the information
and instructions needed to reproduce a published analysis and can be conveniently distributed
Dating human dispersal in Oceania 671
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
over the Internet (Gentleman and Temple Lang 2007). Open access to Bayesian calibration
compendia would awaken the prospect of progress, where the community of archaeological and
paleoenvironmental researchers works to achieve the common goal of creating a detailed record
of human dispersal in Remote Oceania.
Conclusion
Bayesian calibration provides tools t hat archaeologists can use to establish the tempo of
human dispersal in Remote Oceania and move beyond the characterizatio n that it occurred in
aseriesofexplosive phases. One key to the Bayesi an project is to build models that control
for discontinuities associated with a disparity. In Hawaii, where archaeologists failed for
many years to control for the in-built ages of dating materials, settlement date estimates were
typically less accurate than their pred ecess ors, and it was only through applications of
chronometric hygiene and investigation of the paleoenvironmental record that progress was
made towards an accurate settlement date estim ate. Although chronometric hygiene played a
benecial role in Hawaii, its application there betrays one of its weakn esses , which is that
one cannot know whe ther the reje ction pro toco l is too strict or to o lax. A two-phas e
Bayesian model yields more desirable behavior; additional observations increase the preci-
sion of the settlement date estimate.
The history of New Zealand settlement date estimates contrasts strongly with Hawaii.
Archaeologists and environmental scientists in New Zea land have processed more than 100
high-quality dates from the post-settlement phase and this effort has yielded a consensus
estimate that has been relatively stable for more than two decades. Application of a two-
phase Bayesian model to a large collection of well-controlled archaeological and paleoen viron-
mental dates from New Zealand yields an estimate of the settlement date that ts well with the
modern consens us.
Two Bayesian tools are especially useful for dispersal studies. One calculates the probability
that one colonization event occurred before another, and the other estimates the elapsed time
between colonization events. Their application to the settlement date estimates from Hawaii
and New Zealand indicates that prehistorians can (1) be certain that New Zealand was settled
after Hawaii, and (2) begin to theorize about the causes and consequences of a 140209-year
lag between settlement events (Fig. 7).
Acknowledgements
The author thanks: Tim Rieth and Ethan Cochrane for sharing a draft manuscript on the
chronology of settlement in Remote Oceania; Chris Jacomb and Richard Walter for patiently
answering severa l queries about the stratigraphy and chronology of the Wairau Bar site; Janet
Wilmshurst for clarifyi ng the stratigraphic relationship of rat-gnawed seeds to the Kaharoa
tephra; Fiona Petchey for discussions about dating the Bourewa site in Fiji; and Steve Athens,
Tim Rieth, Magdalena Schmid and Chris Jacomb and his colleagues for reading various drafts
of the article and offering helpful comments. The author is solely responsible for errors of fact or
interpretation.
672 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
Disclosure statement
No potential conict of interest was reported by the author.
Supplemental data
The underlying research materials for this article can be accessed by sending an email message
to c.e.buck@shefeld.ac.uk with the subject line Dye WA Supplement. An account will be set
up on the BCal server, as necessary, and the projects will be copied to a subdirectory of the
account named WA Supplement.
Thomas S. Dye
University of Hawaii
tsd@tsdye.com
References
Anderson, A. 1991. The Chronology of Colonization in New Zealand. Antiquity 65: 76795.
Anderson, A., J. Chappell, M. Gagan, and R. Grove. 2006. Prehistoric Maritime Migration in the Pacic
Islands: An Hypothesis of ENSO Forcing. The Holocene 16 (1): 16. doi:10.1191/0959683606hl901ft.
Athens, J. S. 1997. Hawaiian Native Lowland Vegetation in Prehistory. Chap. 12 in Historical Ecology
in the Pacic Islands: Prehistoric Environmental and Landscape Change, edited by P. V. Kirch and T. L.
Hunt, 24870. New Haven, CT: Yale University Press.
Athens, J. S., T. M. Rieth, and T. S. Dye. 2014. A Paleoenvironmental and Archaeological Model-Based
Age Estimate for the Colonization of Hawaii. American Antiquity 79 (1): 14455. doi:10.7183/0002-
7316.79.1.144.
Athens, J. S., H. D. Toggle, J. V. Ward, and D. J. Welch. 2002. Avifaunal Extinctions, Vegetation Change,
and Polynesian Impacts in Prehistoric Hawaii. Archaeology in Oceania 37: 5778. doi:10.1002/
arco.2002.37.issue-2.
Bayman, J. M., and T. S. Dye. 2013. Hawaiis Past in a World of Pacic Islands. Washington, DC: SAA
Press.
Bellwood, P. S. 1979. Mans Conquest of the Pacic: The Prehistory of Southeast Asia and Oceania. New
York: Oxford University Press.
Bellwood, P. S. 2013. First Migrants: Ancient Migration in Global Perspective. Chichester: Wiley
Blackwell.
Bronk Ramsey, C. 2001. Development of the Radiocarbon Calibration Program Oxcal. Radiocarbon 43
(2A): 35563.
Buck, C. E., W. G. Cavanagh, and C. D. Litton. 1996. Bayesian Approach to Interpreting Archaeological
Data. Statistics in Practice. Chichester: John Wiley & Sons.
Buck, P. H. 1925. The Coming of the Maori. Nelson, NZ: Cawthron Institute of Scientic Research.
Bulmer, S. 1989. Gardens
in the South: Diversity and Change in Prehistoric Maaori Agriculture. Chap.
45 in Foraging and Farming: The Evolution of Plant Exploitation, edited by D. R. Harris and G. C.
Hillman, 688705. London: Unwin Hyman.
Dating human dispersal in Oceania 673
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
Burley, D., M. I. Weisler, and J.-X. Zhao. 2012. High Precision U/Th Dating of First Polynesian
Settlement. PLoS ONE 7 (11): e48769. doi: 10.1371/journal.pone.0048769.
Clark, G., and A. Anderson 2009. Site Chronology and a Review of Radiocarbon Dates from Fiji. Chap.
7inThe Early Prehistory of Fiji, Vol. 31, edited by G. Clark and A. Anderson, 15382. Terra Australis.
Canberra, AU: ANU E Press.
Davidson, J. M. 1984. The Prehistory of New Zealand. Auckland: Longman Paul.
Davidson, J. M. 1994. The Eastern Polynesian Origins of the New Zealand Archaic. Chap. 10 in The
Origins of the First New Zealanders, edited by D. G. Sutton. Auckland: Auckland University Press.
Dean, J. S. 1978. Independent Dating in Archaeological Analysis. In Advances in Archaeological
Method and Theory, Vol. 1, edited by M. B. Schiffer, 22365. New York: Academic Press.
Denham, T., C. B. Ramsey, and J. Specht. 2012. Dating the Appearance of Lapita Pottery in the Bismarck
Archipelago and Its Dispersal to Remote Oceania. Archaeology in Oceania 47: 3946. doi:10.1002/
j.1834-4453.2012.tb00113.x.
Dye, T. S. 2011. A Model-based Age Estimate for Polynesian Colonization of Hawaii. Archaeology in
Oceania 46: 1308. doi:10.1002/j.1834-4453.2011.tb00107.x.
Emory, K. P. 1928. Archaeology of Nihoa and Necker Islands. B. P. Bishop Museum Bulletin 53.
Honolulu: B. P. Bishop Museum.
Fornander, A. 191619. Fornander Collection of Hawaiian Antiquities and Folklore. Memoirs of the B. P.
Bishop Museum, 46. Translations revised and illustrated with notes by Thomas G. Thrum. Honolulu:
Bishop Museum Press.
Gentleman, R., and D. Temple Lang. 2007. Statistical Analyses and Reproducible Research. Journal of
Computational and Graphical Statistics 16 (1): 123. doi: 10.1198/106186007X178663.
Graves, M. W., and D. J. Addison. 1995. The Polynesian Settlement of the Hawaiian Archipelago:
Integrating Models and Methods in Archaeological Interpretation. World Archaeology 26 (3): 38099.
doi:10.1080/00438243.1995.9980283.
Green,
R. C., M. Jones, and P. Sheppard. 2008. The Reconstructed Environment and Absolute Dating of
SE-SZ-8 Lapita Site on Nendö, Santa Cruz, Solomon Islands. Archaeology in Oceania 43: 4961.
doi:10.1002/j.1834-4453.2008.tb00030.x.
Groube, L. M. 1968. Research in New Zealand since 1956. In Prehistoric Culture in Oceania: A
Symposium, edited by I. Yawata and Y. H. Sinoto, 1419. Honolulu: Bishop Museum Press.
Hamilton, W. D., and J. Kenney. 2015. Multiple Bayesian Modelling Approaches to a Suite of
Radiocarbon Dates from Ovens Excavated at Ysgol yr Hendre, Caernarfon, North Wales. Quaternary
Geochronology 25: 7282. doi:10.1016/j.quageo.2014.10.001.
Higham, T. F. G. 1994. Radiocarbon Dating New Zealand Prehistory with Moa Eggshell: Some
Preliminary Results. Quaternary Science Reviews 13: 1639. doi:10.1016/0277-3791(94)90043-4.
Higham, T. F. G., A. J. Anderson, and C. Jacomb. 1999. Dating the First New Zealanders: The
Chronology of Wairau Bar. Antiquity 73 (280): 4207.
Hogg, A. G., T. F. G. Higham, D. J. Lowe, J. G. Palmer, P. J. Reimer, and R. M. Newnham. 2003. A
Wiggle-Match Date for Polynesian Settlement of New Zealand. Antiquity 77 (295): 11625. doi:10.1017/
S0003598X00061408.
Holdaway, R. N., M. E. Allentoft, C. Jacomb, C. L. Oskam, N. R. Beavan, and M. Bunce. 2014. An
Extremely Low Density Human Population Exterminated New Zealand Moa. Nature Communications 5
(5436). doi:10.1038/ncomms6436.
Holdaway, R. N., and C. Jacomb. 2000. Rapid Extinction of the Moas (Aves: Dinornithiformes): Model,
Test, and Implications. Science 287: 22504. doi:10.1126/science.287.5461.2250.
Holdaway, R. N. 1996. Arrival of Rats in New Zealand. Nature 384: 2256. doi:10.1038/384225b0.
674 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
Howard, A. 1967. Polynesian Origins and Migrations. In Polynesian Culture History: Essays in Honor
of Kenneth P. Emory, edited by G. A. Highland, R. W. Force, A. Howard, M. Kelly, and Y. H. Sinoto, 45
101. B. P. Bishop Museum Special Publication 56. Honolulu: Bishop Museum Press.
Hunt, T. L., and R. M. Holsen. 1991. An Early Radiocarbon Chronology for the Hawaiian Islands: A
Preliminary Analysis. Asian Perspectives 30: 14761.
Jacomb, C., R. N. Holdaway, M. E. Allentoft, M. Bunce, C. L. Oskam, R. Walter, and E. Brooks. 2014.
High-Precision Dating and Ancient DNA Proling of Moa (Aves: Dinornithiformes) Eggshell Documents
a Complex Feature at Wairau Bar and Renes the Chronology of New Zealand Settlement by Polynesians.
Journal of Archaeological Science 50: 2430. doi:10.1016/j.jas.2014.05.023.
Kennett, D. J., B. J. Culleton, A. Anderson, and J. Southon. 2012. A Bayesian AMS
14
C Chronology for
the Colonisation and Fortication of Rapa Island. Chap. 11 in Taking the High Ground: The Archaeology
of Rapa, a Fortied Island in Remote East Polynesia, edited by A. Anderson and D. J. Kennett, 189202.
Terra Australis 37. Canberra, AU: ANU E Press.
Kirch, P. V. 1985. Feathered Gods and Fishhooks: An Introduction to Hawaiian Archaeology and
Prehistory. Honolulu: University of Hawaii Press.
Kirch, P. V. 1986. Rethinking East Polynesian Prehistory. Journal of the Polynesian Society 95: 940.
Kirch, P. V. 1997. The Lapita Peoples: Ancestors of the Oceanic World. The Peoples of South-East Asia
and the Pacic. Cambridge, MA: Blackwell.
Kirch, P. V. 2000. On the Road of the Winds: An Archaeological History of the Pacic Islands Before
European Contact. Berkeley, CA: University of California Press.
Kirch, P. V. 2010. Peopling of the Pacic: A Holistic Anthropological Perspective. Annual Review of
Anthropology 39: 13148. doi:10.1146/annurev.anthro.012809.104936.
Kirch, P. V. 2011. When Did the Polynesians Settle Hawaii? A Review of 150 Years of Scholarly Inquiry
and a Tentative Answer. Hawaiian Archaeology 12: 326.
Lowe, D. J., R. M. Newnham, B. G. McFadgen, and T. F. G. Higham. 2000. Tephras and New Zealand
Archaeology. Journal of Archaeological Science 27: 85970. doi:10.1006/jasc.1999.0614.
McGlone, M. S., and J. M. Wilmshurst. 1999. Dating Initial Maori Environmental Impact in New
Zealand. Quaternary International 59: 516. doi:10.1016/S1040-6182(98)00067-6.
Nunn, P. D., and F. Petchey. 2013. Bayesian Re-Evaluation of Lapita Settlement in Fiji: Radiocarbon
Analysis of the Lapita Occupation at Bourewa and Nearby Sites on the Rove Peninsula, Viti Levu Island.
Journal of Pacic Archaeology 4 (2): 2134.
Petchey, F., M. Spriggs, S. Bedford, F. Valentin, and H. Buckley. 2014. Radiocarbon Dating of Burials
from the Teouma Lapita Cemetery, Efate, Vanuatu. Journal of Archaeological Science 50: 22742.
doi:10.1016/j.jas.2014.07.002.
Piperno,
D. R., J. E. Moreno, J. Iriarte, I. Holst, M. Lachniet, J. G. Jones, A. J. Ranere, and R. Castanzo.
2007. Late Pleistocene and Holocene Environmental History of the Iguala Valley, Central Balsas
Watershed of Mexico. Proceedings of the National Academy of Sciences 104 (29): 1187481.
doi:10.1073/pnas.0703442104.
Rieth, T. M., and E. E. Cochrane. Forthcoming. The Chronology of Colonization i n Remote Oceania.
In The Oxford Handbook of Prehistoric Oceania, edited by T. L. Hunt. Oxford: Oxford University
Press.
Rieth, T., and J. Stephen Athens. 2013. Suggested Best Practices for the Application of Radiocarbon
Dating to Hawaiian Archaeology. Hawaiian Archaeology 13: 329.
Simmons, D. R. 1969. A New Zealand Myth: Kupe, Toi and the Fleet. New Zealand Journal of History
3: 1431.
Dating human dispersal in Oceania 675
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015
Sinoto, Y. H. 1970. An Archaeologically Based Assessment of the Marquesas Islands as a Dispersal
Center in East Polynesia. In Studies in Oceanic Culture History, edited by R. C. Green and M. Kelly,
10532. Pacic Anthropological Records 11. Honolulu: Anthropology Department, B. P. Bishop Museum.
Spriggs, M., and A. Anderson. 1993. Late Colonization of East Polynesia. Antiquity 67: 20017.
Sutton, D. G. 1987. A Paradigmatic Shift in Polynesian Prehistory: Implications for New Zealand. New
Zealand Journal of Archaeology 9: 13555.
Sutton, D. G. 1994a. Preface. In The Origins of the First New Zealanders, edited by D. G. Sutton, 118,
Auckland: Auckland University Press.
Sutton, D. G., ed. 1994b. The Origins of the First New Zealanders. Auckland: Auckland University Press.
Wilmshurst, J. M., A. J. Anderson, T. F. G. Higham, and T. H. Worthy. 2008. Dating the Late Prehistoric
Dispersal of Polynesians to New Zealand Using the Commensal Pacic Rat. Proceedings of the National
Academy of Sciences 105 (22): 767680. doi:10.1073/pnas.0801507105.
Wilmshurst, J. M., and T. F. G. Higham. 2004. Using Rat-Gnawed Seeds to Independently Date the
Arrival of Pacic Rats and Humans in New Zealand. Holocene 14: 80106. doi:10.1191/
0959683604hl760ft.
Wilmshurst, J. M., T. L. Hunt, C. P. Lipo, and A. J. Anderson. 2011. High- Precision Radiocarbon Dating
Shows Recent and Rapid Initial Human Colonization of East Polynesia. Proceedings of the National
Academy of Sciences 108 (5): 181520. doi:10.1073/pnas.1015876108.
Thomas S. Dye is President and Principal Archaeologist at T. S. Dye & Colleagues,
Archaeologists in Honolulu, Hawaii, and Afliate to the Graduate Faculty at the University
of Hawaii at Manoa. His research currently focuses on diachronic analysis of archaeological
materials, preparation of reproducible research documents and the use of directed graphs in
stratigraphic interpretation. He has an abiding interest in the structural analysis of eighteenth-
century records of contact between native Hawaiians and Westerners.
676 T. S. Dye
Downloaded by [University of Hawaii at Manoa] at 11:30 02 October 2015