Alternative models of volcanic glass quarrying and exchange in Hawai'i

Jeffrey L. Putzi

, Nathaniel J. DiVito

, Carl E. Sholin

, Peter R. Mills

, Steven Lundblad

Bobby Camara

,ThomasS.Dye

⁎

Scientiﬁc Consulting Services, Inc., 1347 Kapiolani Blvd., Suite 408, Honolulu, HI 96814, United States

T. S. Dye & Colleagues, Archaeologists, Inc., 735 Bishop St., Suite 315, Honolulu, HI 96813, United States

Department of Anthropology, Arntzen 315, Western Washington University, 516 High Street, Bellingham, WA 98225, United States

Department of Anthropology, University of Hawai'i at Hilo, 200 Kawili St., Hilo, HI 96720, United States

Department of Geology, University of Hawai'i at Hilo, 200 Kawili St., Hilo, HI 96720, United States

P.O. Box 485, Volcano, HI 96785, United States

abstractarticle info

Article history:

Received 18 October 2014

Received in revised form 21 February 2015

Accepted 15 March 2015

Available online 28 March 2015

Keywords:

Exchange

Common pooled resource

Volcanic glass

EDXRF

Hawai'i

Social complexity

138 volcanic glass artifacts recovered from Site 50–10–19–30173 at Ka'ūpūlehu, Hawai'i Island were sourced to

Pu'uwa'awa'a using EDXRF. Site 50–10–19–30173 is a beach sand deposit with volcanic glass and other tradition-

al Hawaiian artifacts that was sealed by an AD 1800–1801 lava ﬂow. The proportion of Pu'uwa'awa'a volcanic

glass in the assemblage is consistent with a cost surface model proposed recently. It is shown that the fall off

in Pu'uwa'awa'a volcanic glass is exponential for the cost surface for Hawai'i Island, as it is for two alternative dis-

tance decay models, which also yield good ﬁts to the volcanic glass data. A straight line distance overland model

provides an easy way to generate predictions. A depot model, where Pu'uwa'awa'a volcanic glass is brought to

Kahuwai Bay at Ka'ūpūlehu and distributed by canoe, ﬁts the existing data somewhat better than the two over-

land transport models. It has been argued on the basis of distributional data and technological analyses that

Pu'uwa'awa'a volcanic glass was a common pooled resource. The analysis presented here supports this idea by

noting the lack of evidence for directional trade in the residuals of the ﬁt to the exponential curve. Recommen-

dations for future research are offered.

1. Introduction

At the time of Western contact in AD 17 78, traditional Hawaiian

society was among the most highly stratiﬁed and complex in Polynesia

(Sahlins, 1958; Goldman, 1970; Kirch, 2010; Hommon, 2013). Compar-

ison of Contact-era Hawaiian social organization with other Polynesian

archipelagos indicates that, at some point in its history, Hawai'i

underwent a transformation that ruptured the genealogical connection

between elite ali'i and common maka'āinana (Sahlins, 1992; Hommon,

1976, 1986). One spatial expression of this process was the establish-

ment of ahupua'a land divisions where communities of maka'āinana

were managed by an appointed ofﬁcial of the state government

known as a konohiki (Hommon, 1976, 1986). A typical ahupua'a con-

tains the full range of re sources needed for subsistence and, at the

time of European contact, functioned as a unit for the payment of trib-

ute in kind and labor to a multi-tiered hierarchy of elite ali'i.These

circumstances have led to a characterization of ahupua'a that highlights

the managemen t role of konohiki as collectors of tribute in isolated

and sel f-sustained c ommunities (e.g., Hommon, 2013; Kirch, 2010).

However, within this br oad characterization of the Contact-era

ahupua'a there remains many questions potentially addressed by ar-

chaeological information. How deeply did this Hawaiian social transfor-

mation affect the lives of maka'āinana? Were tribute production and

konohiki management a central part of maka'āinana life? Or, did they re-

quire maka'āinana to make relatively minor adjustments to a social life

with deep roots in the past?

An important contribution to the study of these questions was made

recently by McCoy et al. (2011), who modeled the distribution of volca-

nic glass from the Pu'uwa'awa'a source as overland transport across a

cost surface and investigated technological attributes of the transported

glass to argue that there was “unfettered access” (McCoy et al., 2011,

2547) to Pu'uwa'awa'a. In this view, Pu'uwa'awa'a volcanic glass was a

common-pooled resource that was exchanged amon g maka'āinana

free of control by ali'i.Thisﬁnding contrasts with a preliminary, small-

scale investigation of adze rock distrib ution that posits ali'i control

over access to this material that becomes a source of economic power

(Kirch et al., 2012

). According to McCoy et al. (2011), it is possible to dis-

cover a mix of maka'āinana behaviors responsible for the distribution of

Pu'uwa'awa'a vol canic glass, including direct acces s for people who

lived close to the source, down-the-line exchange to a distance of a sin-

gle day's travel overland, and direct access from more distant locations,

perhaps involving transport of volcanic glass by canoe.

Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

⁎ Corresponding author at: 735 Bishop St., Suite 315, Honolulu, HI 96813, United States.

E-mail address: tsd@tsdye.com (T.S. Dye).

JASREP-00054; No of Pages 12

http://dx.doi.org/10.1016/j.jasrep.2015.03.006

Contents lists available at ScienceDirect

Journal of Archaeological Science: Reports

journal homepage: http://ees.elsevier.com/jasrep

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

This paper reports a new collection of volcanic glass recovered from

Site 50–10–19–30173 located close to the Pu'uwa'awa'a source (Fig. 1)

and attempts to build upon and strengthen the analysis carried out by

McCoy et al. (2011). It does this by suggesting certain reﬁnements to

the technological analysis, by proposing alternative distribution routes

to the overland cost surface, and by showing that the three component

model, with separate components for varying distances from the source,

can be replaced with a single distance decay model based on an expo-

nential function. Exponential models have a strong theoretical and

empirical foundation in geography (e.g., Haggett et al., 1977; Taylor,

1971) and their utility in the archaeological situation was established

many years ago (Hodder, 1974; Renfrew, 1975, 1977). It is argued that

systematic deviations from an exponential model indicative of direc-

tional trade are absent in the distribution of Pu'uwa'awa'a volcanic

glass and that this provides strong support for the hypothesis of unfet-

tered access. Finally, differences in the chaînes opératoire of volcanic

glass and adze rock from the Mauna Kea quarry are brieﬂy noted and

the exponential model is recommended for investigating the distribu-

tion of adze rock so that a direct comparison with the distribution of vol-

canic glass can be achieved.

2. Description of Site 50–10–19–30173

Site 50–10–19–30173 was discovered during archaeological moni-

toring at the Kona Village Resort for a project to repair facilities dam-

aged during the tsunami of March 11, 2011. Current building codes

require utilities be placed underground; at the resort this entailed exca-

vating a trench with a hoe ram through low portions of a lava ﬂow. Al-

though a large-scale geologic map of lava ﬂows on Hawai'i Island (Wolfe

and Morris, 1996)showsa1500yearoldpāhoehoe ﬂow here and most

of the interface between this older ﬂow and an AD 1800–1801 'a'ā ﬂow

that partially overlies it has been obscured by development of the resort,

inspection of sections exposed by the hoe ram conﬁrms that this lava is

a pāhoehoe member of the 1800–1801 ﬂow and not part of the older

pāhoehoe ﬂow. Lava extended beneath the water table in most excava-

tions, but at the inland end of the resort, about 400 m from the shore-

line, the hoe ram exposed a calcareous coarse sand beach deposit that

had been covered by the AD 1800–1801 lava

ﬂow (Fig. 2).

Close insp ection of the trench face revealed unusual stratiﬁcation

within the beach deposit beneath the lava ﬂow (Fig. 3). The basic stra-

tigraphy consisted of four coarse calcareous sand layers, Contexts

104–107, underlying the AD 1800–1801 lava ﬂow, which was designat-

ed Context 103 (Fig. 4). Context 104, immediately beneath the lava ﬂow

is a pale yellow layer of contact metamorphosed sandstone. Context 105

is a dark gray layer of thermally altered sand that contained a few tradi-

tional Hawaiian artifacts and marine shell whose colors lacked chroma

due to exposure to extreme heat. Context 106 is a light brownish gray

sand with numerous thin dark lenses that might represent organic-

rich deposits carbonized by the lava ﬂow. It contained the bulk of the

traditional Hawaiian artifacts, none of which show any obvious effect

of exposure to extreme heat. Context 107, at the base of excavation, is

culturally-sterile, very pale brown sand.

Also present were very dark grayish brown irregularly-shaped

lense s, two of which, Contexts 116 and 117, are shown on Fig. 4.

These lenses are located at the base of Context 106 and extend into

the basal sand, Context 107. It was not possible to determine whether

these contexts were cultural features or by-products of heat from the

lava ﬂow, perhaps the burned root systems of vegetation. The cultural

content of the lenses was similar to Context 106.

Remnant cultural deposits located in the base of the trench and be-

neath the lava ﬂowinthesidesofthetrenchwereexcavatedwithatrow-

el and the excavated sediments were passed through 0.125 in. mesh

screen to facilitate identiﬁcation and collection of small items of cultural

material. These excavations were made difﬁcult by the degree to which

the hoe ram had disturbed deposits within the trench and safety issues as-

sociated with undermining the lava ﬂow. These challenging conditions

Fig. 1. Project location on a portion of a USGS quadrangle map.

2 J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

and the limited extent of the excavations made it impossible to character-

ize conﬁdently the archaeological events responsible for the cultural de-

posits. A general impression based on ﬁeld observation of poorly sorted

coarse sand is that the deposits were laid down close to the shoreline,

likely seaward of more substantial cultural deposits along the back beach.

Nevertheless, 166 traditional Hawaiian artifacts were collected and

cataloged during the excavations, including 138 pieces of volcanic

glass, eight pieces of modiﬁed mammal bone, ﬁ ve echinoid spine

abraders, three coral abraders, three basalt abraders, three mammal

bone ﬁshhooks, two adzes, two pieces of modiﬁed shell, one piece of ba-

salt debitage, and a cowry shell octopus lure (Fig. 5). No historic period

artifacts were recovered from sediments beneath the lava ﬂow.

Charcoal samples collected from Contexts 105, 106, and 116 beneath

the lava ﬂow were identiﬁed to the lowest possible taxonomic level by

Gail Murakami of the International Archa eological Research Institute

Wood Identiﬁcation Laboratory. Thirteen native and Polynesian intro-

duced plants were identiﬁed (Table 1). Historically introduced plants

were not identiﬁed in the charcoal.

Two pieces of charcoal identiﬁed as the Polynesian introductions

kī and kukui were sent to Beta Analytic, Inc. for dating with AMS.

After pre-treatment at the laboratory, the sample of kukui nutshell

was found to be incompletely carbonized; it returned a modern date.

The charcoal identiﬁed as kī wood was analyzed as Beta-376428 and

returned a conventional radiocarbon age of 250±30 bp. This age deter-

mination was calibrated with the BCal software (Buck et al., 1999) using

a recent atmospheric calibration curve (Reimer et al., 2013) and a single

phase model with a terminus ante quem of AD 1800, which corresponds

to the known age of the lava ﬂow that sealed the site: α

N θ

N β

=AD

Fig. 2. Photograph of the utility trench showing the beach sand deposit capped by the 1800–1801 lava ﬂow. The scale, which sits on top of the lava ﬂow, is marked in 10 cm increments.

Fig. 3. Photograph of the stratiﬁcation showing the 1800–1801 lava ﬂow (top) and the discoloration of the sand due to the heat of the lava ﬂow. The scale is marked in 10 cm increments.

3J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

1800, where α

and β

are the start and end dates of the single phase,

respectively, and θ

represents th e true calendar age of the kī wood

charcoal. The calibrated estimate of θ

indicates that the kī plant grew

for a relatively short time in AD 1530–1799 (95% probability), probably

at the later end of this range, AD 1640–1799 (68% probability).

3. Technological analysis of volcanic glass assemblages

The 138 pieces of volcanic glass recovered from the beach sand de-

posit under the lava ﬂow include 132 complete and partial ﬂakes and

six pieces of angular debris. No cores were recovere d. There a re 105

complete ﬂakes, the distal ends of 23 ﬂakes, the proximal ends of two

ﬂakes, and two medial sections. Two ﬂakes each exhibit two ﬂake

bulbs indicative of bipolar reduction, but the rest have a single bulb of

percussion. Cortex is present on 68 ﬂakes, eleven of which have primary

cortex over the entire dorsal surfac e and 57 have secondary cortex

mixed with ﬂake scars. 64 ﬂakes have no, or tertiary, cortex. The lengths

of complete ﬂakes range from 6 to 31 mm; the median complete ﬂake

length is 15.3 mm. Flakes with secondary cortex tend to be longer

than ﬂakes with either primary or tertiary cortex (Fig. 6).

McCoy et al. (2011) carried out technological analyses designed to

investigate the use lives of volcanic glass cores in an effort to identify

down-the-line exchange. They reason that, in the presence of down-

the-line exchange, sites far from the source will yield assemblages of

Fig. 4. Stratigraphic proﬁle of the east face of the utility trench showing identiﬁed contexts. Contexts 104–107, 116, and 117 represent a sand beach deposit capped by a lava ﬂow, Context

103, in AD 1800–1801. Context 102 is ﬁll topsoil laid down for resort development; its surface, Context 101, supports various resort facilities today.

Fig. 5. Traditional Hawaiian artifacts recovered from beneath the lava ﬂow: a–c, Pu'uwa'awa'a volcanic glass ﬂakes; d, e, echinoid spine abraders; f,boneone-pieceﬁshhook fragments; g,

mammal bone manufacturing waste; h, vesicular lava abrader; i,basaltadze;j,boneﬁshhook fragment with barb; k,boneﬁshhook tab; l, cowry shell octopus lure. m,highly-polished

basalt adze fragment. The scale bar is 1 cm.

4 J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

volcanic glass that reﬂect the use of well-used cores as raw material. In

this situation, it is expected that ﬂake size will be small, the frequency of

cortex will be low, and wasted cores will be common.

Based on six assemblages ranging from 7.41 to 18.79 h round trip

duration to and from Pu'uwa'awa'a, they found an exponential rise in

the proportion by weight of wasted cores ( McCoy et al., 2011, 2555).

The absence of cores in the assemblage from Site 50–10–19–30173,

which is close r to Pu'uwa'awa'a than any of the ot her assem blages,

generally supports this ﬁndin g. However, there are two concerns.

First, the exponential rise in the proportion by weight of wasted cores

was established with a relatively small number of sites located near

Pu'uwa'awa'a and excludes sites more distant from the source. The as-

semblage from Site 50–Ha–B21–20, with a nearly 40-hour round trip

duration, yielded a high proportion of wasted cores, consistent with

the general pattern but less than predicted by the exponential function.

The other assemblage from Sites 50–Ha–B22–65 and –84 with a 38.39-

hour round trip duration, yielded an unexpectedly low proportion of

wasted cores, consistent with a round trip travel time less than 11 h.

Perhaps more important, however, most of the variability in the propor-

tion of wasted cores comes from assemblages with 12 or fewer pieces of

volcanic glass (Fig. 7). The small size of these samples might have intro-

duced variability that is not present in the underlying populations of

volcanic glass at these sites.

McCoy et al. (2011, 2555) found that volcanic glass artifacts with

cortex were relatively common at three sites near Pu'uwa'awa'a and

nearly absent at three sites farther than 11 h of round trip travel time.

The relatively high proportion of ﬂakes with cortex from Site 50–10–

19–30173, which is close to Pu'uwa'awa'a, lends support to this obser-

vation. However, this pattern breaks down if two more distant sites

are included. If all of the assemblages for which cortex information is

available are plotted, rather than just the close ones, then there appears

to be no association between the proportion of artifa cts with cortex

and round trip duration (Fig. 8). The small sample sizes from several

sites yield a regressio n line with a large s tandard error. The lack of

sites with round trip travel times between 19 and 38 h means the

small assemblages distant from Pu'uwa'awa'a wield undue inﬂuence

on the slope of the regression line. When these distant sites are includ-

ed, no simple relationship between distance from Pu'uwa'awa'a and in-

cidence of cortex is apparent.

Using assemblages from four sites close to Pu'uwa'awa'a, McCoy

et al. (2011, 2556) found a regular decline in the “average” (presumably

mean) length of ﬂakes with distance from Pu'uwa'awa'a. The mean ﬂake

length at Site 50–10–19–30173 is somewhat shorter than predicted by

the regression line formula (McCoy et al., 2011, Fig. 9), and this has the

effect of lessening the slope of the regression line. If distant sites are

included, then the slope of the regression line is positive (Fig. 9).

Again, the small number of assemblages, the small sizes of several of

them, and the lack of sites with round trip tra vel times between 19

and 38 h all have an effect on the analysis. In any event, there is

no clear relationship between mean ﬂake length and distance from

Pu'uwa'awa'a. The small number of assemblages and the small sizes of

many of them make conﬁdent inferences impossible.

4. Distribution of Pu'uwa'awa'a volcanic glass on Hawai'i Island

In hand sample the volcanic glass ﬂakes from Site 50–10–19–30173

exhibit the dark olive green color indicative of a Pu'uwa'awa'a source.

The volcanic glass pieces were analyzed with EDXRF at the University

of Hawai'i at Hilo (Lundblad et al., 2008, 2011). The elemental composi-

tion of the volcanic glass pieces when compared using trace element ra-

tios of Y, Sr, and Zr indicates that the volcanic glass pieces recovered

from Site 50–10–19–30173 all derived from the trachytic Pu'uwa'awa'a

source (Fig. 10).

The distribution of Pu'uwa'awa'a volcanic glass away from the

source was ﬁrst investigated by McCoy et al. (2011), who computed

with ArcGIS 9.3.1 software cost surface overland routes and round trip

travel times between Pu'uwa'awa'a and 19 archaeological sites with col-

lections of 15–489 volcanic glass pieces (McCoy et al., 2011, Table 4)

(Fig. 11). Round trip travel times varied from 7.41 h to 41.23 h. When

the percentage frequency of Pu'uwa'awa'a volcanic glass in 15 assem-

blages from sites close to the source was plotted against round trip trav-

el time, a strong linear relationship was found (McCoy et al., 2011,

2553). This linear relationship does a good job of predicting the propor-

tion of Pu'uwa'awa'a volcanic glass in the Kona Village assemblage. Four

sites with round trip travel times greater than 22 h were considered

outliers and were excluded from the spatial analysis.

Table 1

Taxa identiﬁed from charcoal.

Family Taxon Name Habit Origin

Agavaceae cf. Cordyline fruticosa

Shrub Poly. intro.

Apocynaceae cf. Rauvolﬁa sandwicensis hao Tree Native

Asteraceae cf. Bidens sp. ko'oko'olau Shrub Native

Ebenaceae Diospyros sandwicensis lama Tree Native

Euphorbiaceae Aleurites moluccana kukui Tree Poly. intro.

Euphorbiaceae Chamaesyce sp. 'akoko Shrub–tree Native

Malvaceae Hibiscus tiliaceus hau Shrub–tree Native

Myrtaceae Metrosideros polymorpha 'ōhi'a lehua Tree Native

Pandanaceae Pandanus tectorius hala Tree Native

Pittosporaceae cf. Pittosporum sp. hō'awa Tree Native

Rubiaceae Canthium odoratum alahe'e Shrub–tree Native

Rubiaceae cf. Coprosma sp. pilo Shrub–tree Native

Sapindaceae Dodonaea viscosa 'a'ali'i Shrub-tree Native

Sapindaceae cf. Dodonaea viscosa 'a'ali'i Shrub–tree Native

primary secondary tertiary

Cortex

Length (mm)

Fig. 6. Boxplot of complete volcanic glass ﬂake lengths for different types of cortex.

5J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

The linear relationship between percentage frequency of

Pu'uwa'awa'a volcanic glass and round trip travel time breaks down

when distant sites are included (Fig. 12). It underestimates the propor-

tion of Pu'uwa'awa'a glass close to the source, and overestimates it for

round trips longer than about 15 h. An exponentia l curve, where a

straight line is ﬁtted to the relation ship between round trip travel

time and the logarithm of percentage frequency of Pu'uwa'awa'a volca-

nic glass, does a better job, although it too tends to underestimate the

proportion of Pu'uwa'awa'a glass close to the source (Fig. 12). Renfrew

(1977, 78) identiﬁed positive residuals close to the source as evidence

of a “supply zone”, however, Hodder (1974, 182) noted that this pattern

is characteristic of situations where the movement of materials away

from a source result from a random walk process.

Although a cost surface provides an intuitively attractive way to de-

termine effective distances between two points, in practice it relies on

proprietary software that requires considerable training to use properly,

and which is likely to be unavailable to many ﬁeld archaeologists. An al-

ternative model that is simpler to use connects the Pu'uwa'awa'a source

to sites with volcanic glass assemblages with straight lines that can be

measured by hand on a US GS quad ma p (Fig. 13). The straight line

model yields an exponential decay curve with distance from the source,

but the ﬁt is not as good as the cost surface (Fig. 14).

McCoy et al. (2011, 2547) note the possibility that transport of volca-

nic glass to sites distant from Pu'uwa'awa'a might have been by canoe

rather than overland. Accordingly, a model of volcanic glass distribution

by canoe transport was investigated using a model that posits use of Site

50–10–19–30173 at Kahuwai Bay as a depot. Either re sidents of the

village at Kahuwai Bay collected volcanic glass cores from Pu'uwa'awa'a

and brought them to the coast to exchange with canoe travelers from

other parts of Hawai'i Island, or canoe travelers put ashore at Kahuwai

Bay and accessed Pu'uwa'awa'a from there with a hike that took the

better part of a day. Coastal distances from Site 50–10–19–3 0173 to

the coastal locations closest to other sites with volcanic glass assem-

blages were derived by measuring the length of coastline with GIS soft-

ware (Fig. 15). An exponential distance decay curve ﬁts this model

better than the two overland transport models, which suggests that

transport of volcanic glass by canoe might have been important for peo-

ple living relatively close to Pu 'uwa'a wa'a, as well as those living far

away (Fig. 16).

5. Discussion

McCoy et al. (2011) concluded that Pu'uwa'awa'a volcanic glass was

a common pooled resource to which there was “unfettered access” in

traditional Hawaiian times, both through the exercise of a right not to

be excluded from direct access to the source and through down-the-

line exchange of volcanic glass cores outside the purview of ali'i control.

They support this conclusion with a linear regression on the relationship

between the contribution of the Pu'uwa'awa'a source to site assem-

blages of volcanic glass and the site to source round trip travel time as

determined by cost surface analysis using GIS software for a subset of

sites close to Pu'uwa'awa'a. The linear regression calculated on the

050100

Number of volcanic glass pieces

Proportion of assemblage weight

as cores (%)

Fig. 7. The proportion of assemblage weight as cores varies with the size of the assemblage. Note that most of the variability in this measure is introduced by small assemblages.

−25

100

10 20 30 40

Round trip (hr)

Proportion of assemblage

with cortex (%)

Collection

size

100

Fig. 8. Proportion of assemblage with cortex as a function of round trip duration from Pu'uwa'awa'a. The regression line was computed using the rlm method of Venables and Ripley

(1994). The shaded area is the 95% conﬁdence interval for the regression.

6 J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

untransformed variables ﬁts well for sites located close to Pu'uwa'awa'a,

but it intersects the x-axis at a round trip travel time of about 24 h

McCoy et al. (2011, 2554) and sites with Pu'uwa'awa'a volcanic glass

more distant from the source, where the model predicts Pu'uwa'awa'a

volcanic glass is absent, are excluded from the spatial and technological

analyses as outliers.

McCoy et al. (2011) do not discuss the decision to carry out the re-

gression on untransformed variables, which assumes that the fall-off

in the contribution of Pu'uwa'awa'a volcanic glass to site assemblages

is dependent only on change in distance (Renfrew, 1977; Hodder,

1974). Based on numerous empirical studies, archaeologists and geog-

raphers have found that variabi lity in interaction is dependent not

only on dista nce, but also on the transferability and relative value of

the materials being transported (Hodder, 1974; Haggett et al., 1977).

When these other factors are taken into account, the best ﬁt to the un-

transformed variables is a curved line that drops rapidly close to the

source and then levels out as distance increases (Haggett et al., 1977;

Renfrew, 1977; Taylor, 1971). The demonstration that the distribution

of Pu'uwa'awa'a volcanic glass can be described with an exponential

curve using alternative measures of distance, including s traight line

overland, cost surface overland, and coastal transport from a depot at

Kahuwai Bay, indicates that identiﬁcation by McCoy et al. (2011) of dis-

tant site outliers is likely an artifact of model selection, rather than a re-

ﬂection of the factors underlying the observed distribution of volcanic

glass. Factors other than distance are also responsible for the distribu-

tion of volcanic glass away from the Pu'uwa'awa'a source.

What are these other factors? Transferability plays a major role with

bulky or heavy items; the canonical archaeological examples are the

distribution of Romano-British rooﬁng tiles around Cirencester and sev-

eral vari eties of coarse pottery for which distributional studies have

been co mpleted (Hodder, 1974, 179 –182). In contrast, Pu'uwa'awa'a

volcanic glass nodules are small and light; transferability should not

constrain their distribution away from the source. Rather, it seems likely

that the relative value given to Pu'uwa'awa'a volcanic glass in tradition-

al Hawai'i was a factor in its distribution away from the source. Geogra-

phers and others have noted that the value of a material is related to the

shape of the drop-off curve. Materials that are highly valued are used far

from the source and are distributed widely. They yield drop-off curves

with low slopes and long tails . In contrast, low-valued materials are

used near to the source and are not distributed widely. They yield

drop-off curves with high slopes and short tails. The exponential drop-

off of Pu'uwa'awa'a volcanic glass indicates that it was highly valued

in traditional Hawai'i.

Because the exponential curve does not intersect the x-axis, there is

no upper bound to the distribution and no way to predict the full size of

the distribution ﬁeld. Thus, geographers use the concept of mean ﬁeld to

measure and compare the sizes of distribution ﬁelds (Haggett et al.,

1977,48–50). The mean ﬁeld is calculated here as the point along the

exponential curve where the proportional change of distance exceeds

the proportional change in the percentage of Pu'uwa'wa'a volcanic

glass in an assemblage. Calculation of the mean ﬁeld indicates that

the value of Pu'uwa'awa'a volcanic glass in traditional Hawai'i was

10 20 30 40

Round trip (hr)

Mean length (mm)

Collection

size

100

Fig. 9. Mean ﬂake length as a function of round trip duration to and from Pu'uwa'awa'a. The linear regression line was computed using the rlm method of Venables and Ripley (1994).The

shaded area is the 95% conﬁdence interval for the regression.

0 100 200 300

Sr/Zr * 100

Y/Zr * 100

Provenience

Pu‘uwa‘awa‘a

Mauna Loa

Kilauea

50−10−19−30173

Fig. 10. Trace element ratios of Kona Village volcanic glass artifacts compared to volcanic glass from Pu'uwa'awa'a, Klauea, and Mauna Loa (after Lundblad et al., 2013, Fig. 2).

7J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

such that it was transported regularly to sites within a radius of about

21 h round trip travel time or 46 km straight line overland from

Pu'uwa'awa'a, or a distance of about 63 km along the coast from

Kahuwai Bay. Past the mean ﬁeld, a small quantity of Pu'uwa'awa'a vol-

canic glass is expected at sites as far away as 41 h round trip travel time

or 98 km straight line overland from Pu'uwa'awa'a, or 127 km along the

coast. Future work on Hawai'i Island, especially in the windward dis-

tricts of Hamakua, Hilo, and Puna, is likely to expand the range of sites

where Pu'uwa'a wa'a volcanic glass is known to have be en used and

deposited.

Fitting an exponential curve to the full data set makes it possible

to distinguish the effects of directional trade, in which certain sites

function as centra l places that receive materials directly from the

source, regardless of dista nce, and from which the materials a re

redistributed to nearby sites that don't receive them directly from the

source (Renfrew, 1977,85–87). Directional trade is typically associated

with redistribution of materials under the direction of some central con-

trol (Renfrew, 1975, 48). An investigation into the distribution of obsid-

ian within the Maya sphere lends empirical support to the theoretical

expectation that one effect of directional trade is to decrease the ﬁtof

Fig. 11. Cost surface overland routes to and from Pu'uwa'awa'a. Adapted from McCoy et al. (2011, ﬁg. 6). Note that the routes to and from Kahuwai Bay and Site 50–10–19–30173 are not

shown.

100

10 20 30 40

Round trip duration (hr)

Percent Pu‘u Wa‘awa‘a glass

Collection

size

100

200

300

400

Fig. 12. Exponential and straight line ﬁts for the percentage frequency of Pu'uwa'awa'a volcanic glass and overland round trip travel times computed with a cost surface model. The ex-

ponential function is I = exp(5.20 − 0.09*D) and the linear function is I =81.80− 1.83*D,whereI = percent Pu'uwa'awa'a glass and D = round trip duration in hours. The value of R

for

the exponential function is 0.56.

8 J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

an exponential curve to the distributional data (Sidrys, 1977). So far, de-

viations from the exponential curve that might indicate directional

trade are not evident in the Pu'uwa'awa'a volcanic glass data. The distri-

bution of Pu'uwa'awa'a volcanic glass appears to have taken place in the

context of traditional social relationships independent of the political

hierarchy.

The important observation that Pu'uwa 'awa'a volcanic glass was

not d istributed outside Hawai'i Island (McCoy et al., 2011, 2552)

indicates that Pu'uwa'awa'a volcanic glass wasn't valued so highly that

it was transported to other islands. In this respect, the distribution of

Pu'uwa'awa'a volcanic glass contrasts with Mauna Kea adze basalt,

which has been identiﬁed recently from six sites on Maui Island, includ-

ing an elite residence and three religious structures (Kirch et al., 2012),

and perhaps as far aﬁeld as O'ahu Island (Mills and Lundblad, 2014,34).

The distribution of non-local basalt in the Maui Island sites was

interpreted as evidence for directional trade where “access to and distri-

bution of these stone resources was controlled by elites” (Kirch et al.,

2012, 1060). Given this difference, it is useful to distinguish

the chaînes opératoire fo r volcanic glass and adze rock by comparing

evidence for the organization of production at Pu'uwa'awa'a and

the Mauna Kea adze quarry. The area around Pu'uwa'awa 'a has

been described as barren of t raditional Hawaiian habitation and

Fig. 13. Straight line routes from Pu'uwa'awa'a to archaeological sites with volcanic glass assemblages.

100

25 50 75 100

Distance (km)

Percent Pu‘u Wa‘awa‘a glass

Collection

size

100

200

300

400

Fig. 14. Exponential distance decay function for the straight line overland model: I = exp(5.10 − 0.04*D), where I = percent Pu'uwa'awa'a glass and D =distanceinkm.ThevalueofR

for

the exponential function is 0.53.

9J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

agriculture, and the source of volcanic glass nodules on the north face

of Pu'uwa'awa'a hill lacks a formal quarry (McCoy et al., 2011,

2548–2549). In contrast, the Mauna Kea adze quarry has a multitude

of shrines; in fact, shrines are the most common type of architectural

feature at the quarry. It has been argued that ritualization of production

at the quarry, including rites of passage sites used to initiate apprentices

into an adze-makers' guild (McCoy, 1999), is strong evidence for adze

makers having been specialists attached to ali'i (McCoy et al., 2010).

These preliminary observations on the differences between the ritu-

alized production and c onsumption of Mauna Kea adze roc k, which

appears to have been exchanged through directional trade, and the

exchange of Pu'uwa'awa'a volcanic glass, the raw material for an expe-

dient tool that was exchanged through traditional social relations, are

based on the current state of distributional data (Mills and Lundblad,

2014), which a re deﬁcient in several respects, including (i) sample

sizes that are too small to support conﬁdent inferences, (ii) patchy spa-

tial coverage with large gaps between some sites and uncertain distri-

butional boundaries, and (iii) an almost complete lack of chronological

control.

The low cost and non-destructive nature of EDXRF and the conﬁ-

dence with which the technique is able to identify Pu'uwa'awa'a volca-

nic glass make it possible to model traditional Hawaiian exchange in an

Fig. 15. Coastal routes from Pu'uwa'awa'a with a depot at Kahuwai Bay.

100

50 100

Distance (km)

Percent Pu‘u Wa‘awa‘a glass

Collection

size

100

200

300

400

Fig. 16. Exponential distance decay function for the coastal model: I = exp(4.92 − 0.02*D), where I = percent Pu'uwa'awa'a glass and D = coastal distance in km. The value of R

for the

exponential function is 0.65.

10 J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

analytically usef ul wa y. However, at this early sta ge of investigation

the small sizes of ma ny volcanic glass collections restricts the infor-

mation they have yielded. Using a ruler and a suitabl e map of Hawai'i

Island, the straight-line method descri bed abov e convenie ntly pro-

duces an expected valu e, p, for the percentage of Pu' uwa'awa'a vol -

canic glass at a site. Given this information, a problem-oriented

excavation strategy can be designed to yield a sample size, n ,that

establishes the observed percentage within a speciﬁed precision

using the formula for the standard deviation of a binomial distribu-

tion,

ﬃﬃﬃﬃﬃﬃﬃﬃ

npq

,whereq =1− p.

Tec hnological analyses play a potentially important rol e in deter-

miningthenatureofvolcanicglassexchange,buttheirpromisehas

yet to be ful ﬁlled, in part because relatively few sites have been ana-

lyzed in this way an d sa mple sizes at the sites that ha ve been ana-

lyzed are small. In addition, the technological attributes chosen for

study might be reﬁned. It is recommended that observations of cor-

tex on whole ﬂakes distinguish primary, secondary, and tertiary cor-

tex. Using the proporti on of each cortex type at Site 50 –10–

19–30173 as a baseline for a si te with direct access to the source , it

is expected that down-the-line exchange will yield drop off curves

that are s teepest for primary cortex and ﬂattest for tertiary cortex.

This pattern should be relatively easy to distinguish from the expect-

ed pattern yielded by direct access, where the wide distribution of

fresh cores should yield drop off curves that are consistent across

cortex types. Alte rnatively, a more labo r intensive method might

be used (Ditchﬁeld et al., 2014).

Equall y pres sing is the need to track change over time (Mills and

Lundblad, 2014, 36). The volcanic glass assemblages analyzed by

McCoy et al. (2011) were collected in the 1960's through the

1980's before Hawaiian archaeologists were able to distinguish suit-

able d ating materials and their ages cannot be estimated wi th conﬁ -

dence with the information at hand. Similarly, Lass (1994)

attempted to chart change over time in the distribution of Mauna

Kea basalt with assembla ges that were insecurely dated. In addition,

quarrying activitie s at both sources are difﬁcult to date. The lack of a

formal quarry at Pu'uwa'awa'a depri ves arc haeologists of a site to

date and dates on unidentiﬁed charcoal with a potential for in-built

age is the basis for the c hronology of the Mauna Kea adze quarry

(McCoy et al., 2009). Thus, the histori es of the Pu'uwa'awa'a volcanic

glass quarry and the Mauna Kea adze quarry—when they were dis-

covered and how their use and the distribution of their pro ducts

grew and perhaps declined—re main to be investigated with well-

dated assemblages.

6. Conclusions

The exponential fall-off of Pu'uwa'awa'a volcanic glass with distance

from source at arch aeological sites on Hawai'i Island indicates a

resource that was highly valued in traditional Hawai'i. The lack of

evidence for directional trade supports th e inference developed by

McCoy et al. (2011) that Pu'uwa'awa'a volcanic glass was a common

pooled resource distributed through traditional social relations outside

the purview of elite control. The distribution of Pu'uwa'awa'a volcanic

glass contrasts with preliminary indications that Mauna Kea adze rock

was distributed by directi onal trade. The cultural elaboration of adze

production on Mauna Kea contrasts strongly with the lack of evidence

for similar elaboration at Pu'uwa'awa'a.

Alternative models of Pu'uwa'awa'a volcanic glass distribution

cannot distinguish overla nd transport from transport by cano es,

given the data at hand.

Future work should ﬁll gaps in the known distribution o f

Pu'uwa'awa'a volcanic glass and de ﬁne the limits of its dist ribution

on Hawai'i Island. In addition, more detailed technological analyses

are needed to dist inguish do wn-the-line exchange from direct

access.

Acknowledgments

The aut hors th ank Mark D. McCoy, Tim Rieth, Marshall Weisler,

Patrick C. McCoy, Andy Howard, and an anonymous re viewer for

many useful comments on drafts of the paper. Mike Vitousek of the

Hawai'i Historic Preservation Division required the archaeological mon-

itoring that led to the discovery of Site 50–10–19–30173. Frank Trusdell,

USGS, spent considerable time and effort identifying the AD 1800–1801

lava ﬂow. Fieldwork was supported by Hualālai Resort and especially its

staff members Jay Uyeda and Leina'ala Lightner. Any errors that remain

are the responsibility of the authors.

References

Buck, C.E., Christen, J.A., James, G.N., 1999. Bcal: an on-line Bayesian radiocarbon calibra-

tion tool. Internet Archaeology 7 (URL: http://intarch.ac.uk/journal/issue7/buck/).

Ditchﬁeld, K., Holdaway, S., Allen, M.S., McAlister , A., 2014. Meas uring stone artefact

transport: the experimental demonstration and pilot application of a new method

to a prehistoric adze workshop, southern Cook Islands. J. Archaeol. Sci. 50, 512–523.

Goldman, I., 1970. Ancient Polynesian Society. University of Chicago Press, Chicago.

Haggett, P., Cliff, A.D., Frey, A., 1977. Second ed. Locational Analysis in Human Geography:

Locational Models vol. 1. John Wiley & Sons, New York.

Hodder, I., 1974. Regression analysis of some trade and m arketing patterns. World

Archeol. 6, 172–189.

Hommon, R.J., 1976. The Formation of Primitive States in Pre-contact Hawaii. (Ph.D. the-

sis). University of Arizona, Tucson, AZ.

Hommon, R.J., 1986. Social evolution in ancient Hawaii. In: Kirch, P.V. (Ed.), Island Socie-

ties: Archaeological Approaches to Evolution and Transformation. Cambridge Univer-

sity Press, Cambridge, pp. 55–68.

Hommon, R.J., 2013. The Ancient Hawaiian State: Origins of a Political Society. Oxford Uni-

versity Press, Oxford.

Kirch, P.V., 2010. How Chiefs Became Kings: Divine Kingship and the Rise of Archaic States in

Ancient Hawai'i. University of California Press, Berkeley, CA.

Kirch, P.V., Mills, P.R., Lundblad, S.P., Sinton, J., Kahn, J.G., 2012. Interpolity exchange of ba-

salt tools facilitated via elite control in Hawaiian archaic states. Proc. Natl. Acad. Sci.

109, 1056–1061.

Lass, B., 1994. Hawaiian Adze Production and Distribution: Implications for the Develop-

ment of Chiefdoms. Number 37 in Institute of Archaeology Monograph. University of

California, Los Angeles, Los Angeles, CA.

Lundblad, S.P., Mills, P.R., Hon, K., 2008. Analyzing archaeological basalt using non-

destructive energy dispersive X-ray ﬂuorescence (EDXRF): effects of post-depositional

chemical weathering and sample size on analytical precision. Archaeometry 50, 1–11.

Lundblad, S.P., Mills, P.R., Drake-Raue, A., Kikiloi, S.K., 2011. Non-destructive EDXRF anal-

yses of archaeological basalts. In: Schackley, M.S. (Ed.), X-Ray Fluorescence Spec-

trometry in Geoarchaeology. Natural Sciences in Archaeology. Springer, New York,

pp. 65–80.

Lundblad, S.P., Mills, P.R., McCoy, M.D., Kahn, J.G., Mulliken, K., Kaylor, D., 2013. Identiﬁca-

tion of volcanic glass sources inferred from geochemical analysis of artifacts on lee-

ward Hawai'i Island. In: Summerhayes, G.R., Buckley, H. (Eds.), PaciﬁcArchaeology:

Documenting the Past 50,000 Years. Papers from the 2011 Lapita Paciﬁc Archaeology

Conference. Dunedin, NZ. number 25 in University of Otago Studies in, Archaeology,

pp. 67–75.

McCoy, P.C., 1999. Neither here nor there: a rites of passage site on the eastern fringes of

the Mauna Kea adze quarry, Hawai'i. Hawaii. Archaeol. 7, 11–34.

McCoy, P.C., Weisler, M.I., Xin Zhao, J., Feng, Y.X., 2009.

230

Th dates for dedicatory corals

from a remote alpine desert quarry on Mauna Kea, Hawai'i. Antiquity 83, 445–457.

McCoy, P.C., Nees, R.C., Clark, S.D., 2010. Final report, archaeological inventory survey of

the astronomy precinct in the Mauna Kea Science Reserve, Ka'ohe A hupua'a,

Hāmākua District, Island of Hawai'i, TMK: (3) 4–4–015:09 (por.), Prepared for Ofﬁce

of Mauna Kea Management, University of Hawai'i at Hilo, Honolulu.

McCoy, M.D., Mills, P.R., Lundblad, S.P., Rieth, T.M., Kahn, J.G., Gard, R., 2011. A, cost surface

model of volcanic glass quarrying and exchange in Hawai'i. J. Archaeol. Sci. 38,

2547–2560.

Mills, P.R., Lundblad, S.P., 2014. Current perspectives on Hawai'i's stone tool economies.

J. Pac. Archaeol. 5, 30–39.

Reimer, P.J., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., Ramsey, C.B., Buck, C.E., Cheng,

H., Edwards, R.L., Friedrich, M., Grootes, P.M., Guilderson, T.P., Haﬂidason, H., Hajdas,

I., Hatté, C., Heaton, T.J., Hoffmann, D.L., Hogg, A.G., Hughen, K.A., Kaiser, K.F., Kromer,

B., Manning, S.W., Niu, M., Reimer, R.W., Richards, D.A., Scott, E.M., Southon, J.R., Staff,

R.A., Turney, C.S.M., van der Plicht, J., 2013. Intcal13 and Marine13 radiocarbon age

calibration curves 0–50,000 years cal BP. Radiocarbon 55, 1869–1887.

Renfrew, C., 1975. Trade as action at a distance: questions of integration and communica-

tion. In: Sabloff, J.A., Lamberg-Karlovsky, C.C. (Eds.), Ancient Civilization and Trade.

School of American Research Book. University of New Mexico Press, Albuquerque,

NM, pp. 3–59 (chapter 1).

Renfrew, C., 1997. Alternative models for exchange and spatial distribution. In: Earle,

Erikson (Eds.), pp. 71–90.

Sahlins, M., 1958. Social Stratiﬁcation in Polynesia. American Ethnological Society Mono-

graph. University of Washington Press, Seattle.

Sahlins, M., 1992. Anahulu: the Anthropology of History in the Kingdom of Hawaii. Historical

Ethnography vol. 1. University of Chicago Press, Chicago.

11J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006

Sidrys, R., 1977. Mass-distance measures for the Maya obsidian trade. In: Earle, Ericson

(Eds.), pp. 91–107 (chapter 5).

Taylor, P.J., 1971. Distance transformation and distance decay functions. Geogr. Anal. 3,

221–238.

Venables, W.N., Ripley, B.D., 1994. Modern Applied Statistics with S-plus. Statistics and

Computing. Springer-Verlag, New York.

Wolfe, E.W., Morris, J., 1996. Geologic map of the Island of Hawaii. Number I-2524 in

Miscellaneous Investigations SeriesU.S. Geological Survey, Denver, CO.

12 J.L. Putzi et al. / Journal of Archaeological Science: Reports 2 (2015) xxx–xxx

Please cite this article as: Putzi, J.L., et al., Alternative models of volcanic glass quarrying a nd exchange in Hawai'i, Journal of Archaeological

Science: Reports (2015), http://dx.doi.org/10.1016/j.jasrep.2015.03.006