You are on page 1of 20

Journal of Archaeological Science (1999) 26, 527–546

Article No. jasc.1998.0332, available online at http://www.idealibrary.com on

Integrating Phytoliths within Use-Wear/Residue Studies of


Stone Tools
Lisa Kealhofer
Department of Anthropology, William and Mary College, P.O. Box 8795, Williamsburg, Virginia 23187–8795,
U.S.A.

Robin Torrence and Richard Fullagar


Division of Anthropology, Australian Museum, 6 College Street, N.S.W. 2000, Australia

(Received 14 January 1998, revised manuscript accepted 3 July 1998)

Analyses of phytolith assemblages extracted from residues on obsidian artefacts and from the surrounding soil matrix
of two sites in Papua New Guinea address two important methodological issues. First, multivariate analysis of
phytolith assemblages extracted from the edges of stone artefacts, from soil directly in contact with the artefact, and
from soil derived from the same archaeological context demonstrates the integrity of phytolith residues. Second,
reconstructions of tool use based on the analysis of phytoliths are compared with results from an independent
use-wear/residue study. The interpretation of phytolith assemblage variability enhances and strengthens the use-wear/
residue results. Finally, integrating the phytolith data within the broader use-wear/residue study produces significant
new findings about changes in obsidian tool-use in prehistoric Papua New Guinea.  1999 Academic Press

Keywords: PHYTOLITHS, STONE TOOLS, USE-WEAR, RESIDUES, HAFTING, OBSIDIAN,


PAPUA NEW GUINEA.

Introduction sents a systematic attempt to evaluate the relationship


between phytolith residues and stone tool use.

P
lant residues have been noted on chipped Two methodological issues were investigated: (1) the
stone artefacts by archaeologists for some time integrity of phytolith residues and (2) interpretations of
(Bruier, 1976; Shafer & Holloway, 1979; artefact use based on phytolith assemblage compo-
Anderson, 1980; Hurcombe, 1992), however, few sys- sition and use-wear/residue analysis. First, in any cul-
tematic attempts have been made to use this potentially tural context it is highly likely that there will be
rich source of information for evaluating artefact use incidental contact between stone artefacts and other
(cf. Sobolik, 1996). A notable exception is the ongoing materials creating residues which may transfer phyto-
work by a group of Australian archaeologists who liths or other materials onto artefacts (cf. Bruier, 1976;
have been developing various experimental, labora- Barton, Torrence & Fullager, 1998). We tested for the
tory, and archaeological techniques for plant residue unique association of residue assemblages with stone
and use-wear analyses (e.g., Hall, Higgins & Fullagar, artefacts, by comparing phytolith assemblages from
1989; Fullagar, 1986, 1988, 1992, 1993a, b; Barton, residues with those from soil samples directly associ-
1991; Fullagar, Meehan & Jones, 1992; Loy, Wickler & ated with the individual artefact and from soil samples
Spriggs, 1992; Barton & White, 1993; Loy, 1994; derived from the same archaeological context.
Fullager, Furby & Hardy, 1996; Fullagar & Furby, Second, interpretations of tool use based on the
1997; Atchison & Fullagar, 1998; Fullagar, Loy & analysis of phytolith assemblages extracted from resi-
Cox, 1998, 1992). Their residue work has primarily dues on artefacts were compared with interpretations
focused on starch grains, although the potential for of use-wear and residues using direct observation on
interpreting phytolith residues was demonstrated by the same artefact. The two studies were conducted by
Fullagar (1993a, b; Bowdery, n.d.) in a study independent analysts as ‘‘blind tests’’. First, Fullagar
of one obsidian artefact from Papua New Guinea. identified and recorded use-wear and residues. Next the
Although Sobolik (1996) noted phytoliths on stone artefacts were sonically cleaned and the residue re-
artefacts, an analysis of assemblage composition was moved was processed for phytoliths by Kealhofer. The
not undertaken. The pilot study reported here repre- cleaned artefacts were later reassessed by Fullagar.
527
0305–4403/99/050527+20 $30.00/0  1999 Academic Press
528 L. Kealhofer et al.

Willaumez
Peninsula

NEW
BRITAIN
PAPUA NEW GUINEA

0 200 400 km

0 2 4 km FAO

FRL
N

GARUA IS.

Figure 1. Location of sites used in this study.

Analysis of phytolith residues and direct observation of ian artefacts derived from highly acidic, volcanic soils
use-wear/residues provide distinct but complementary in a tropical region with high rainfall. Not surprisingly,
types of information. no macroscopic organic materials were recovered from
Use-wear analysis of stone artefacts contributes the sites, so the interpretation of past subsistence and
a wide range of information about artefact use at a settlement systems relies heavily on determining uses
rather general level. Mode of use (e.g., cutting, scrap- for the abundant chipped stone artefacts. As phytolith
ing, drilling) and broad categories of contact material analysis considerably enhanced the use-wear residue
are typical outcomes. The analysis of residues can analyses in this worst case scenario, we argue that
substantially enhance this information if integrated phytoliths have a very important role to play in the
within a larger study based on use-wear. Residues, study of chipped stone tool use.
however, should not be interpreted in the absence of
use-wear analysis since contamination may have oc-
curred (Fullagar, 1988; Fullagar, Furby & Hardy,
1996; Barton, Fullagar & Torrence, 1998).
Archaeological Context
Residues can accumulate on stone tools in a number Obsidian artefacts from two sites, FRL and FAO,
of ways. The most commonly considered residues are located in West New Britain province, Papua New
those resulting from use. They can be also formed, Guinea (Figure 1), were chosen for the study. The sites
however, when a material is used as a haft or sheath to are notable for their well-defined stratigraphy and
protect hands or aid prehension or as decoration of the relatively long time depth. The Bitokara Mission Site
tool: for example, paint, feathers, binding substances. (FRL) was excavated by Specht and co-workers
Finally, a residue may be created when an artefact (Specht et al., 1988), while the FAO excavations are
comes into accidental contact with a material. Since part of the larger Garua Island project (Torrence et al.,
residues result from a number of processes, it is in press). Both sites have a sequence of diagnostic
extremely important that analyses be undertaken to volcanic tephras with a well-developed soil in the upper
discriminate between these and other potential forma- horizon of each. The volcanic events which deposited
tion processes. Our interpretation of phytoliths as the tephras have been dated by radiocarbon, and
residues attempted to account for as many sources the tephras have been sourced and cross correlated
as possible. by both macroscopic and geochemical techniques
Whereas residue analysis can be an important aid for (Machida et al., 1996; Torrence et al., in press) thereby
many archaeological studies, our research illustrates providing a relative chronology for the region.
the particular importance of phytolith residues for Stratigraphic units at the sites have been divided into
archaeological research in regions where organic pres- four chronological periods: (1) older than W-K1
ervation is generally poor. In the case study reported tephra, 10,000–6000 ; (2) between W-K1 and W-K2
here, phytoliths were found to be abundant on obsid- tephras, 5900–3600 ; (3) between W-K2 and Dk
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 529

tephras, 3600–1100 ; and (4) post-Dk tephra, 1100 bag, air-dried, sealed, and stored. The FAO artefacts
–present. Radiocarbon dates from FRL and were selected from a single 1 m1 m test pit, 970/
FAO provide additional confirmation for the tephra 1000. At FRL the test pit was initially a 2 m2 m
chronology (Specht et al., 1988; Torrence et al., in unit, stepped down to 1 m1 m at the base. Soil
press). samples were also taken from each site. At FRL
Although organic preservation of macroscopic plant samples were collected within each stratigraphic unit
material is very poor at these sites due to the high from one wall of the unit. A column sample was taken
acidity of the soils and the high rainfall, the presence of on the wall of the FAO test unit, with samples divided
organic residues on stone artefacts was not unexpected into 5-cm thick units.
because phytoliths are extremely resistant to decay.
For example, Anderson (1980) and Hurcombe (1992)
have also identified phytoliths adhering to and par-
Sampling
tially incorporated into polishes on stone artefacts even
after treatment in acid and alkaline washes. Fullagar Seventeen obsidian artefacts were selected: three from
(1993a, b) has reported the presence of phytoliths FRL which had previously been analysed by Fullagar
extracted from soil adhering to a stone artefact (1992) and a grab sample of 14 from samples obtained
recovered from FRL, one of the sites in our study. from each period in trench 970/1000 at site FAO. This
These artefacts are particularly appropriate for phy- latter sample was selected for a ‘‘blind test’’ comparing
tolith studies because Fullagar’s (1992, 1993a, b) use- direct observation on the artefacts with phytolith
wear/residue analyses have shown that prehistoric analysis of extracted residues. As shown in Table 1 and
obsidian artefacts at six sites in West New Britain, Figures 2 and 3, the majority of the artefacts are simple
including FRL, were mainly used to process plant flakes and the amount of retouch decreases through
materials. He also identified a trend through time such time.
that more recent site assemblages showed evidence of a Assemblage composition of phytoliths in sediment
wider range of tasks. At the same time multi-purpose, derived from three contexts was examined: (1) soils
retouched, curated artefacts were gradually replaced by near the artefacts, site soil samples (Table 2); (2) soil
expedient unretouched flakes which were used for only adhering loosely to the artefact, artefact soil samples
one task and then discarded immediately (Torrence, (one from each artefact at FAO only); and (3) residues
1992). Fullagar’s (1992) residue and use-wear data on the surface of the artefacts, artefact residues (Table
support Torrence’s (1992) hypothesis that changes in 1). To assess the general background of phytolith
FRL technology are related to increasing sedentism assemblages through time at each site, three site soil
and an intensification in gardening. Based on these samples from FAO and two from FRL were chosen
studies, we expected that the residues on individual from the same stratigraphic contexts as the stone
tools should become less diverse through time, artefacts. The artefact soil sample, which comprises
reflecting the increase in one-off, expedient use. In sediment in contact with the artefact when it was
contrast, the total assemblage from a site should excavated, provides a further check against contamina-
appear more diverse, since increasing sedentism tion. As the artefacts dried, much of the soil originally
would lead to a wider range of tasks taking place at adhering to them became detached from them. The
individual sites. artefacts were also gently rubbed inside the bag to
remove additional loose sediment. One half of the
resulting artefact soil sample was analysed for phyto-
Methodology liths. All 14 artefact soil samples in this study are from
This pilot study was undertaken primarily to assess the the FAO artefacts. Artefact residues were obtained
value of analysing phytoliths extracted from artefact from sonic cleaning each artefact for 5 min in pure
residues. Since phytolith analysis is very time consum- water.
ing, we began by studying three artefacts from FRL for As noted above, analyses by Fullagar (1992, 1993a,
which Fullagar had identified use-wear and residues. b) at FRL have shown that through time there is a shift
After obtaining good results from these cases, we from multiple to single purpose artefacts and some of
expanded the study to include soils and artefacts from the early, larger artefacts with a relatively long use-life
the FAO site, which had not been the subject of were probably hafted. In order to examine whether
previous use-wear/residue analyses. these patterns occur at FAO, and to look for other
types of variability within individual artefacts, some of
the artefacts were divided into various sections which
Excavation were sonic cleaned separately. For example, on several
During the excavations at FRL and FAO, a sample of artefacts the retouched stem was distinguished from
10 artefacts was taken haphazardly from each arbitrary the unretouched blade portion of the artefact, and the
unit (usually 10-cm thick) of a 1 m1 m unit. Using a material from one blade was divided longitudinally
metal trowel (the artefacts were not handled), the to investigate differential use on the two opposing
artefacts plus adhering soil were placed into a plastic lateral margins. A total of 22 artefact residue samples
530 L. Kealhofer et al.

Table 1. Summary of artefacts and artefact residue samples

Macroscopic Tool residue


Artefact Period Type Segment Cortex Retouch use-wear sample

FAO
285 4 Flake Complete Partial None None Whole artefact
288 4 Flake Complete Partial None None Whole artefact
303 4 Flake Complete Partial None None Whole artefact
306 4 Flake Distal Partial None None Whole artefact
310 3 Flake Complete None None Micro-scarring on distal edge Whole artefact
313 3 Flake Complete Total None None Whole artefact
317 3 Blade Complete None None Micro-scarring on one lateral Whole artefact
margin
325 3 Flake Distal Partial Present on left lateral Micro-scarring and rounding Stem (325s), Blade (325b)
margin on one lateral margin
336 3 Flake Complete None None None Lateral left (336a), Lateral
right (336b)
339 2 Flake Complete Partial None None Whole artefact
340 2 Flake Complete Partial None None Whole artefact
359 2 Flake Complete None Stemmed tool, retouch on None Stem (359s), Blade (359b)
two lateral margins of stem
364 2 Flake Complete None Present on dorsal side at Micro-scarring on both lateral Proximal (364p), Distal
proximal end of flake; edges of blade (364d)
differentiates proximal
from distal
367 2 Blade Complete None Retouch on two lateral Micro-scarring on left lateral Stem (367s), Blade (367b)
margins of blade to form side of blade
stem; retouch on distal end
of blade
FRL
021 4 Blade Distal None None None Distal (021d), Proximal
(021p)
281 2 Flake Proximal Partial None None Whole artefact
352 1 Blade Medial Partial Present proximal end of Extensive micro-scarring and Stem (352s), Blade (352b)
blade to form notches rounding on both lateral edges
differentiating stem and
blade

from the 17 artefacts were included in the study which in this case was recorded as follows: (1) ‘‘not
(Tables 1 & 3). pitted to slightly pitted/etched’’ which means the use-
wear is unaffected and 80% of the surface is clearly
visible with fresh fracture scars and stress lines; (2)
Direct analysis of use-wear and residues ‘‘pitted/etched’’ has pitting to the extent that the pres-
Analysis began with the microscopic examination of ence of use-wear is not in doubt but the interpretations
the artefacts. Use-wear and residues were assessed of polish have some uncertainty; and (3) ‘‘heavily
concurrently using both low and high power magnifi- pitted/etched’’ surfaces still have use scarring but
cations. The reconstruction of artefact use(s) was low reliability for the interpretation of polishes.
based on experimental results (Fullagar, 1991), Sometimes flakes with heavily etched surfaces have
whereas the contact material was identified using a edges which are well preserved and maintain a distinc-
reference collection from New Britain (Fullagar, 1992). tive use polish. The process which permits polished
The results are reported in Table 4. A variety of edges to be better preserved than other surfaces is not
residues, including starch and plant fibres, were ob- understood.
served on most tools and most artefact edges showed As a first measure of whether or not an artefact had
some degree of scarring. been used, the edges were examined under a stereo-
To distinguish use from post-depositional contami- scopic microscope with oblique incident light (6,
nation and to obtain reliable assessments of artefact 12, 25 and 50), specifically noting scarring
function, an integrated analysis of residues with the characteristics. Subsequently, the artefacts were exam-
four main kinds of use-wear is required (Fullagar, ined under vertical incident light with bright- and
Furby & Hardy 1996). These are (1) scarring; (2) darkfield illumination (between 200 and 500).
rounding; (3) polish; and (4) striations (Figures 4 & 5). Edge rounding, or the loss of a sharp edge due to use,
An important factor to consider particularly with is visible in cross-section and measured by degree of
obsidian is surface preservation (Ambrose, 1996), loss: none; very slight; slight; slight–medium; and
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 531

Figure 2. Artefacts from FAO used in this study.

medium. Polish provides a third measure of artefact (generally not diagnostic of material worked) are
use. Four stages of polish formation have been defined observed; during the third stage polished areas are
for obsidian (Fullagar, 1991): the first stage involves linked and the pattern becomes diagnostic of materials
abrasive smoothing and loss of features of fresh frac- worked (e.g., Figure 4); and the fourth stage leads to
tured surfaces; in the second stage, patches of an extensively polished, featureless surface. Stage 4
smoothed polished surfaces within abrasive surfaces polish can be achieved very rapidly when moist,
532 L. Kealhofer et al.

Figure 3. Artefacts from FAO and FRL used in this study.

siliceous materials are processed. The nature of polish these. Particular contact materials and modes of use
often precludes defining what the artefact was used on, can cause unique patterns of striations.
without consideration of other use-wear and residue In addition to use-wear, residues were recorded
characteristics (Fullagar, 1991). Striations, the fourth (Table 4). These included traces of resin, plant fibres,
form of evidence (e.g., Figure 5), are defined by their cracked films of blood, starch grains, and phytoliths
regularity or randomness, their directionality (parallel (cf. Fullagar, 1986, 1988; Fullagar, Furby & Hardy,
or perpendicular to the edge), and by combinations of 1996). Starch grains were observed with cross polarized
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 533

Table 2. Site soil samples Phytolith analysis


Site Unit Sample Period Two different methods were used to extract phytoliths.
First, site soil samples (80 g) and artefact soil samples
FAO 970/1000 30–35 cm 4 (varied but generally less than 0·1 g) were processed
80–85 cm 3 according to procedures outlined by Piperno (1988),
150–155 cm 2
FRL 11 4
with a few modifications in chemicals (substitution of
8 2 zinc bromide) and processing times. Soils were defloc-
culated for a week, sieved to remove >250 ìm material,
and fractionated into 50–250 ìm and 5–50 ìm frac-
tions. Gravity sedimentation procedures were applied
Table 3. Abundance of phytoliths and starch extracted from artefact to the fine fraction (5–50 ìm) to remove clays. Both
soils and residues fractions were then wet ashed in Schulze solution
(nitric acid and potassium chlorate) to remove
Phytolith abundance
Starch abundance organic materials. Samples were rinsed in water,
Sample Period Artefact soil residue residue dried, and the phytoliths were floated on a heavy
liquid, specific gravity 2·35, to separate them from the
FAO remaining heavier minerals. Soil processing required
285 4 + ++ + approximately 1 month.
288 4 ++ — ++
303 4 ++ ++ +
Artefact residue samples (generally less than 0·001 g)
306 4 ++ ++ + were processed in a similar way to modern reference
310 3 ++ ++ ++ material. The liquid resulting from the sonic cleaning
313 3 ++ ++ + of the artefacts was deflocculated overnight, and sieved
317 3 ++ ++ — through 250-ìm mesh. No clays were removed as
325 3 ++
325b 3 — — the fraction was very small. The artefact residues were
336 3 ++ then wet ashed in Schulze solution to remove organic
336a 3 ++ + materials, and as above, rinsed and floated on a heavy
336b 3 ++ + liquid solution.
339 2 + + ++
340 2 ++ + +
After removal from the sediment matrix, the
359 2 ++ — — phytoliths were mounted on slides, viewed, and photo-
364 2 — graphed at 400 magnification. Phytolith types were
364p 2 + — identified when possible, assigned a descriptor if not,
364d 2 ++ — and counted and photographed. For most slides a
367 2 +
367d 2 + + minimum of 100 diagnostic phytoliths were counted,
367p 2 — — but where there was sufficient material, counts were
FRL extended to c. 200.
21 4 ++ Phytoliths were identified on the basis of comparison
21p 4 ++ +
21d 4 ++ + with more general South-East Asian reference collec-
281 2 ++ ++ ++ tions, and a reference collection of grasses from around
352 1 ++ the world (Brown, 1984; Piperno, 1988; Piperno &
352b 1 — — Pearsall, 1990a, b; Kealhofer & Piperno, in press;
352s 1 ++ ++
Piperno & Pearsall, in press; Kealhofer, unpubl.). The
phytolith assemblages included a wide variety of
Sample number: p, d are proximal and distal ends of the artefact;
b, s are blade and stem; a, b are lateral edges. grasses falling mainly within the Bambusoideae and
Phytolith abundance: —, <20 phytoliths; +, 21–75 phytoliths; + +, Panicoideae subfamilies, herbaceous types (Composi-
>75 phytoliths. tae, Marantaceae, Cyperaceae), a diverse selection
Starch abundance: —, Absent; +, present; + +, abundant. of arboreal types (Palm, Moraceae, Musaceae,
Burseraceae, Annonaceae, and others), and unknown
light. They were predominantly less than 5 ìm in Dicotyledon types (Figure 6). Possibly economically
diameter and so identification to species was not useful species including arboreal phytoliths from
attempted (cf. Therin, Torrence & Fullager, 1997). Burseraceae species, potentially Canarium, as well as
A summary assessment of use is presented in scleroid from Annonaceae, and hair cells from Mora-
Table 4 using the following categories: 0=not used; ceae taxa, were present in the residues from both sites.
1=possible use; 2=probable use; 3=definite use. For a Specific identifications were made where possible but
positive attribution of use, residues had to be associ- the focus was on establishing assemblage patterns
ated with an edge, rather than randomly distributed for each sample. More detailed identification would
over the surface of the artefact. A more precise defini- require extensive analysis of modern reference material
tion of function was based on the most likely material which has not as yet been carried out for West New
worked. Britain.
Table 4. Results of use-wear/residue analysis by direct observation of artefacts under reflected light microscopy
534

Scars Scar Scars Edge Obsidian Contact


Tools <5 mm types >5 mm rounding Polish Striations Alignment Other surface Residues material Use

FAO 1
285 Present b,f None None 0 None Not used 1 0
288 Present b,f,s None Slight–medium 2 Present Perpendicular Probably cf. collagen Animal 2
& parallel used
L. Kealhofer et al.

289 Present f None None 0 None Not used 1 0


303 Present f,s None Slight 1–2 None 1 Fibres, Uncertain 2
spicules
306 Present f None Very slight 0–1 None 1 Starch grains, Uncertain 1
fibres
310 Present f,s None Slight–medium 2–3 Parallel 1 Starch grains, Plant 3
phytoliths
313 None None Very slight 1–2 Present Perpendicular 2 Plant 1–2
317 Present b,f None Slight–medium 2–3 Present Perpendicular Multiple use 1 Plant 2–3
& parallel
325d Present b,f,c Present Slight 1–2 Present 45 deg. 3 Non-sliceous wood 2
325p Present Present Slight 1–2 Present Perpendicular 3 Non-siliceous soft plant 2
& 45 deg.
336a, b Present b,f,s None Medium 1–2 Present Perpendicular Bevel on 2 Resin? Plant 2
& parallel proximal end
339 Present Very slight 1–2 Present Parallel & 45 3 Starch Uncertain 2
deg.
340 Present b,f,s,c None None None 3 Uncertain 0–1
359 Present b,f Present None None None 2 Starch grains Uncertain 0–1
on stem
364p Present None Very slight 0–1 Present Perpendicular 2 Starch grains, Soft plant 0–1
fibres
367d Present f,s Present Slight 2–3 Present Parallel Haft, pitted 1 Starch grains cf. plam 3
polish on
high spots
367p None f,s Present Slight 1 Present Perpendicular 1 Starch grains Non-siliceous starchy plant 1–2
FRL
M021d Present f,s None Slight 3 Present Parallel & 1 Starch grains, Siliceous plant: palm or reed 2
perpendicular fibres,
cracked film
(cf. blood)
M021p Present f,s None Slight 2 Present Parallel & Pitted polish 1 Starch, fibres Siliceous plant: palm or reed 2
perpendicular
M281 Present b,f None Slight 2–3 None Soft siliceous plant 2
M352b Present f,s Present Medium 2–3 Present Perpendicular Scraping 2 Starch Wood 3
& 45 deg.
M352s Present f,s Present Very slight 2 Present None Haft 2 Starch, fibres Siliceous plant 3

Key: ‘‘Scar types’’ refers to presence of distinct termination types of use-wear scars as follows: f, feather; s, step; h, hinge; b, bending.
‘‘Polish’’ refers to polish development from intial to later stages 1–3. See text for discussion.
‘‘Obsidian surface’’ refers to the degree of cortex development, erosion, or other weathering ranked from lower to higher impact on use-wear interpretations 1–3. See text for discussion.
‘‘Use’’ refers to the interpretation of tool use from lower to higher levels of confidence 1–3. See text for discussion.
‘‘Tools’’ refers to edge samples as follows (cf. Table 1): d, distal; p, proximal; b, blade; s, stem; a,b, lateral edges.
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 535

Figure 4. Examples of use-wear polishes. (a) FRL. M021. Polish (cf. palm) on high points near edge with small pits. Negative taken at 200.
Width of field approximately 0·3 mm; (b) FRL. M021. Polish (cf reeds/bamboo) and rounding on distal edge. Negative taken at 200. Width
of field approximately 0·3 mm.

The analysis of starch grains, often from plants grains surrounding gelatinized starch were observed.
that do not make phytoliths (e.g., root crops such Starch grains were relatively common in the artefact
as taro and yams) expands our knowledge about residues (Figure 7) but were rare in the artefact soil
plant use (cf. Loy, Wickler & Spriggs, 1992; Loy, or site soil phytolith assemblages, a pattern seen in
1994). Given the use of aggressive chemicals in other analyses of stone artefacts (Barton, Torrence &
phytolith extraction techniques (e.g., Schulze sol- Fullagar, 1998). In general, the starch grains were
ution), we were surprised to find that starch grains spherical and ranged in size from 10 to 25 ìm. It seems
survived processing. Currently, the only explanation likely that smaller grains have not been preserved
for their appearance in the phytolith samples is or did not survive the processing. A summary of
that the starch grains were silicified in some way. In starch abundance extracted from artefact residues is
some cases what appeared to be silicified casts of presented in Table 3.
536 L. Kealhofer et al.

Figure 5. Examples of use-wear polishes and striations. (a) FRL, M352. A thin band of polish (stage 2–3) with striations at 45deg. to the edge.
Negative taken at 500. Width of field approximately 0·1 mm. (b) FRL, M021. Slight rounding and perpendicular striations along distal
edge. Negative taken at 200. Width of field approximately 0·3 mm.

Test for Integrity of Residues both the phytolith assemblage from each artefact resi-
due with a site soil sample from the associated exca-
We began by testing whether the residues extracted vation unit and the phytolith assemblage from the
from the artefacts could be assumed to be associated artefact residue sample with that from the artefact soil
with tool use. We argue that if phytoliths were sample adhering to the artefact.
deposited on artefacts in the context of use, then the Forty-four samples were analysed for phytoliths: 22
phytolith assemblage in a residue sample should be artefact residues, 17 artefact soil samples; and five site
significantly different from that in the adjacent soil. soil samples (Tables 2 & 3). Phytoliths were abundant
Alternately, similarity between artefact residues and in the site soils but mixed results were obtained from
both site and artefact soil samples would suggest the the artefact soil and residue samples (Table 3). Counts
artefact came in contact with the phytoliths after of less than 20 were interpreted to mean real absence.
deposition. This hypothesis was tested by comparing The mid range of 21–75 was considered to be too small
Figure 6. Examples of phytoliths found in the samples, centre of photograph. (a) Panicoid bilobate c. 20 ìm (grass); (b) sclereid (arboreal
form); (c) possible palm c. 18 ìm (Caryota?-type); (d) UNID arboreal form c. 25 ìm.
538 L. Kealhofer et al.

Figure 7. Starch grains found in the phytolith preparation.

Figure 8. Graph of phytolith types.

to be meaningful and, in the case of residues, open to (56% and 67% variability explained, respectively) show
an interpretation of contamination. Samples with a clear separation of the artefact residue samples and
greater than 75 phytoliths were included in the corre- the soil samples (Figures 9 & 10) confirming that the
spondence analyses discussed below. It is worth stress- phytoliths are derived from different processes. For
ing that given the extremely small size of the artefact Period 3 (3600–1100 ) the three residues on the upper
residue samples, less than 0·001 g, the presence of left of Figure 9 are associated with a specific assem-
quantifiable numbers of phytoliths lends much support blage of Palm and Bambuseae phytolith types. In con-
to their interpretation as residues rather than contami- trast, the two residues on the upper right are
nation from surrounding soils, for which 80 g of characterized by a wide variety of unknown Dicotyle-
material often produced only 100 diagnostic phytoliths. don and arboreal phytolith types. As would be
The question of the integrity of the residues is best expected of samples representing the general environ-
examined where there were sufficiently high phytolith ment at the site, rather than artefact use, the artefact
counts for quantitative analysis: 11 artefact residues soil and site soil assemblages include a more diverse
(eight FAO artefacts, and three FRL artefacts), eight assemblage of Bambusoid, arboreal, Panicoid, and her-
artefact soil samples, and all five soil samples (FRL baceous types than the residue samples. The artefact
and FAO). The basic data are presented in Figure 8 soil sample from FAO317 includes an unusual herba-
and the results of correspondence analyses using these ceous component, skewing the assemblage somewhat.
counts are presented in Figures 9–13. A multivariate In Figure 10, showing Period 4 (1100 –present), the
approach helps overcome potential contamination be- two residue samples cluster on the bottom right of the
cause it focuses on the overall structure of the data (cf. graph, while the soil and artefact soil samples are
Powers, Padmore & Gilbertson, 1989 for a similar further left. The residues have a more diverse assem-
approach). blage of Bambusoid, herbaceous, and arboreal types
Due to chronological changes in vegetation at FAO than the previous period and the soils include a wider
(cf. Figure 8), each period was analysed separately. spectrum of arboreal types, Palms, and Panicoid
Plots of the first two eigenvalues for Periods 3 and 4 grasses.
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 539

Soils versus residues: FAO Period 3 Phytoliths


Correspondence analysis
1.5
Residues
1 310R
313R
336RBA 317R
0.5 336RBB

313S
0
Eigenvalue 2

336S
–0.5 80 cm
310S

–1

–1.5 Bag soils and soil sample

–2

–2.5
317S

–3
–2 –1.5 –1 –0.5 0 0.5 1 1.5
Eigenvalue 1
Figure 9. Correspondence analysis of phytolith assemblages extracted from residue and soils samples from FAO Period 3. Plot of cases on the
first two eigenvalues.

Soils versus residues: FAO Period 4 Phytoliths


Correspondence analysis

30 cm
1.5

1 Bag soils and soil sample

Residues
Eigenvalue 2

0.5 306S

0
303R
–0.5 303S

–1

–1.5 306R

–2
–2.5 –2 –1.5 –1 –0.5 0 0.5 1 1.5
Eigenvalue 1
Figure 10. Correspondence analysis of phytolith assemblages extracted from residue and soils samples from FAO Period 4. Plot of cases on
the first two eigenvalues.

Due to the small sample size, all periods at FRL Among the samples with low counts, the much
were combined leading to a greater range of artefact larger counts for the FAO285 artefact residue suggests
and soil variability. Nevertheless, Figure 11 shows that the phytoliths are not due to contamination but
there are significant differences in phytolith assemblage result from tool use. The cases where there are virtually
composition between the two soil samples and the no phytoliths in the residue samples (FAO288,
artefact residues. FAO325b, FAO359s, FAO367p) contrast with the
540 L. Kealhofer et al.

FRL residues and soils: Phytoliths


Correspondence analysis
2

1.5 352S

1 Soil 1 Period 4
Eigenvalue 2

Residues
0.5
Soils
0
Soil 8 Period 2

–0.5 021D

021P
–1
281

–1.5 –1 –0.5 0 0.5 1 1.5 2 2.5


Eigenvalue 1
Figure 11. Correspondence analysis of phytolith assemblages extracted from soils and residues at site FRL. Plot of cases on the first two
eigenvalues.

Phytolith group residues: FAO and FRL stone tools


Correspondence analysis
2
Bamboo group
FAO306
1.5
FAO303
1 FAO285
FAO310
Eigenvalue 2

0.5
Grass and mixed group FRL021D
0 FAO3645
FRL021P
FRL281 FRL3525
–0.5
FAO313

–1 FAO317
Palm group
–1.5 FAO336BIB Arboreal group
FAO336BIA

–2 –1.5 –1 –0.5 0 0.5 1 1.5 2 2.5


Eigenvalue 1
Figure 12. Correspondence analysis of phytolith assemblages extracted from artefact residue samples. Plot of cases on the first two eigenvalues.

presence of phytoliths in the artefact soil samples, use rather than contamination from the surrounding
therefore supporting the proposition that the informa- soil.
tion from the residue samples is independent of the In summary, the phytolith assemblage compo-
soils. For the only two examples where there is a match sition of the artefact residue samples at both FRL
(FAO339 and FAO367), the quantity of phytoliths is in and FAO are distinct from the soil matrix. We
the mid range for which we feel interpretation is not therefore accept the proposition that the assem-
warranted. Although these results are not quantifiable blages of phytoliths extracted from the artefact resi-
due to small sample sizes, the data do suggest that due samples are related in a meaningful manner to
the residues, where present, accumulated as a result of artefact use.
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 541

Phytolith residue groups: FAO and FRL stone tools


Correspondence analysis
3 Bamboo group
Wide B Saddle
TallSaddle Tall Narrow Saddle
T2tpC UNID7Oryza UND2fuzz
SpinSph
2 UNID13elongnodul Tc-c C-cUNID12ribbedelong
2ptBil
Caryota-type UNID11irregelong
Saddle
TabSph
1 INIDmixed Chloridoid
Grass and mixed group Sedge T2ptBil
Eigenvalue 2

4ptBil SRS Arboreal group


T4ptBil Cross BulbSph
0 Bilobate
Multilobate LRS
Musa LnkdBodies UNID6flatsp
SS
Palm UNID3spky
–1 MFP Sclereid
SmMFP
UNID14nodular
EMFP Canarium-type
UNID1disk UNID8acorn
UNID10projn

–2 Palm group UNID9trd 2-cell Hair UNID5wrvol UNID16bulbbod


UNID15spic/cone

FaintMFP
–3 KBPalm

–4 –3 –2 –1 0 1 2 3
Eigenvalue 1
Figure 13. Correspondence analysis of phytolith assemblages extracted from artefact residue samples. Plot of variables on the first two
eigenvalues.

Analyses of Artefact Use contact (such as casual wrapping of a tool edge for use,
storage or transport) and also inadvertent contact (e.g.,
The next stage of the analysis was to compare the trampling) could have taken place. Only one tool
inferences about use provided by the direct observation (FAO288) has evidence of contact with soft animal
of use-wear and residues with that derived from an tissue, most likely skin. It is possible that this pattern
interpretation of the phytolith assemblages from of use-wear was due to handling of the artefact during
the artefact residues. The aim was two-fold. First, very short term use.
since direct observation of use-wear and residues is As shown in Table 5, artefacts with reliable
now well established in archaeological methodology, evidence of plant processing can be divided into two
it could serve as a check on the validity of the groups according to the presence of residues, devel-
reconstructions derived from phytolith analysis. oped polish types and other forms of use-wear: (1)
Second, we examined whether phytolith analysis, woody plants and (2) soft, non-siliceous, starchy
which is quite time-consuming and requires a great plants (tubers). Woody plants can be broken down
deal of specialist knowledge, could significantly into (1a) hard non-siliceous wood, (1b) soft siliceous
enhance the understanding of artefact use. plants (e.g., woody grasses like reeds or bamboo),
and (1c) very starchy and siliceous woody plants
(e.g., palm).
Direct observation of artefacts The most distinctive polish was found on artefacts
Use-wear and residue analysis by direct observation FRL352 (Figure 5), FRL021 (Figures 4 & 5), FRL281,
identifies which artefacts were used. In many cases the FAO325 and FAO367s which form an easily recogniz-
broad category of contact material, for example, skin able group relating to woody plants. In contrast, the
or plant tissue, can also be identified. For some arte- processing of non-siliceous, soft starchy plants (e.g.,
facts fairly specific information about contact material tubers) is difficult to identify because use-wear alone is
is possible, especially when diagnostic residues are not diagnostic. Reliable determination depends upon
present. Table 4 summarizes the data. A category of further study of residues in terms of the frequency, size
‘‘uncertain’’ was included for cases where use-wear was and other characteristics of starch grains (cf. Loy,
insufficiently diagnostic to specify contact material 1994; Therin, Torrence & Fullagar, 1997). With these
with a high level of confidence. The almost exclusive reservations, artefacts likely to have been in dynamic
presence of plant residues in the assemblage, however, contact with soft, starchy, non-siliceous plants
suggests that contact with plant tissue was the most include FAO339, FAO340, FAO364a, FAO364b and
likely for those artefacts having an uncertain function. FAO367b and possibly FAO325s.
Dynamic contact with plants includes processing of In terms of the data necessary to test hypotheses
particular materials, but other kinds of deliberate about chronological patterning in lithic technology in
542 L. Kealhofer et al.

Table 5. Summary of artefact uses based on direct observation of use-wear and residues

Use-wear groups Artefacts

(1) Woody plants


(1a) Hard, non-siliceous FRL21d, FRL021p, FAO325d, FRL352b
(1b) Soft, siliceous FRL352s, FRL021d, FRL021p, FRL281, FAO310, FAO313
(1c) High starch, siliceous FRL021d, FRL021p, FAO367s
(2) Soft, non-siliceous, starchy plants FAO339, FAO340, FAO359, FAO364p, FAO367b, FAO325p
(3) Animal FAO288
(4) Plant, uncertain FAO303, FAO306, FAO317, FAO336a, FAO336b, FAO364d
(5) Unused FAO285
Starch residues present FAO306, FAO310, FAO313, FAO339, FAO359s, FAO364p, FAO364d,
FAO367d, FAO367s, FRL021d, FRL021p, FRL281, FRL352b, FRL352s

West New Britain, two important forms of information Table 6. Summary of artefact uses based on extracted phytoliths and
can be derived from the data: multi-functionality and starch grains
presence of hafting. The tool with the largest number
of contact materials (3) was FRL021 from the most Residue
groups Artefacts
recent period. Two materials were found on FRL352
(Period 1), FAO367 (Period 2) and FAO325 (Period 3).
(1) Bamboos FAO303, FAO306, FAO310
Whereas multiple uses were found on both parts of (2) Palms FAO336A, FAO336B, FAO285,
FRLO21, the contact materials were separated on the (3) Arboreal FAO313, FAO317, FAO364d
other three tools, suggesting the presence of a used end (4) Multiple FRL021p; FRL021d, FRL281, FRL352b, FRL352s
and a hafted end. In the case of FAO367, palm may (5) Uncertain FAO285, FAO339, FAO340, FAO367b, FAO364p
have been used as a handle as in the case of recent (6) Absent FAO288, FAO325b, FAO359s, FAO367s
obsidian spears and daggers from Manus (cf. Torrence,
1993). The handle for FAO325 may have been
wooden, whereas for FRL352 some form of wrapping The results of correspondence analysis of the phyto-
involving grass, bamboo or a reedy plant is indicated. lith assemblage structure from the artefact residue
Two further Period 2 artefacts with retouch suggestive samples with high phytolith counts is presented as a
of hafting, FAO359 and FAO364, had very little plot of the first two eigenvalues which explain over
evidence of use, although fibres were observed on 60% of the variability in the data. Four slightly over-
FAO364. However, FAO359 had a very etched surface lapping groups of tools can be distinguished in Figure
prohibiting a reliable interpretation for use. 12 and the phytolith types that define them can be
Only the FAO data are large enough to be analysed identified by examining the same region of the variable
for chronological trends. Comparing Tables 1 and 4, a plot in Figure 13. The tool use groups as discriminated
very general trend occurs in the data from multi- by phytoliths include the following (Table 6):
functional or hafted tools used to process soft, starchy (1) primarily bamboos;
plants in Period 2–Period 4 which is characterized by (2) defined by a specific type of Palm phytolith, and
hand-held tools used for short periods of time on an unidentified Dicotyledon type;
uncertain plants. Period 3 is transitional with fewer (3) strongly dominated by arboreal and Dicotyledon
hafted tools and more uncertain plants than Period 2, phytolith forms, including potential Burseraceae
whereas all the samples from Period 4 are unused or taxa such as Canarium (an important economic
slightly used. In addition the importance of soft, non- species in this region) and no grasses or herba-
siliceous starchy plants appears to decline through ceous types were found in this group;
time. (4) a wide mixture of Palms, Panicoid grasses,
banana, and diverse arboreal, herbaceous and
Dicotyledon forms defines a group of tools for
Phytoliths extracted from artefact residues which multiple uses can be inferred.
Surprisingly, the relative abundance of phytoliths dif-
fered between the two sites. Phytoliths were abundant Not surprisingly the artefacts from FRL, which were
in both the soils and residues of all periods at FRL, but specifically chosen for this pilot study because Fullagar
were relatively rare in FAO Period 2 artefact soil and (1992) had previously identified distinctive use-wear
residue samples. Given that most of the FAO Period 2 and residues on them, have the most diverse group of
surfaces are also poorly preserved (Table 4, FAO339– phytoliths, suggesting multi-purpose uses. Examining
367), it seems likely that residues had been removed differences within individual tools relating to function
through soil diagenesis and/or that phytoliths are not also produced good results. FRL021 (Figure 3) pro-
particularly well preserved in Period 2 contexts. duced the most striking results. The distal end is
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 543

characterized by nearly 50% Palms, but also has a In summary, the study of phytoliths and starch
significant percentage of Panicoid grasses and bam- grains extracted from artefact residues provides rela-
boos. In contrast, residue from the proximal end of the tively specific interpretations for contact material. Cor-
same artefact contains a much lower percentage of respondence analyses identified three functional classes
Palms, fairly comparable amounts of grasses, and a based on the predominance of bamboos, palms, and
higher arboreal count. It seems likely that the distal arboreal taxa, as well as a fourth group of multi-
end was used for processing palms, whereas the proxi- functional tools. Starch grains are primarily associated
mal end was hafted, possibly with a wooden handle with the multi-functional and palm groupings but a
secured by grass binding. Similarly, FRL352 had phy- fifth functional class may be justified by the very high
toliths on the stem, potential handle region, and not on concentrations of starch on two other artefacts. Differ-
the blade. FRL281 may also have been hafted since ential use on several artefacts is interpreted as the
grasses are present in the residue. Separate parts of result of hafting.
three FAO artefacts were analysed separately. FAO367
produced too few phytoliths for interpretation and
may have not been used, whereas the presence of
primarily arboreal phytoliths on the stem of FAO364
An Integrated Approach to Tool Use
may be associated with a wooden haft. There is no When comparing the two methodologies it is import-
evidence from the phytolith analysis that the two edges ant to stress that each detects particular forms of
of FAO336 were used on different materials. behaviour and one method does not necessarily dupli-
The distribution of starch grains provides further cate the other. For instance, use-wear can determine if
support for the phytolith interpretations, although the an artefact has been used, but phytolith analysis can-
data must be interpreted cautiously since we are unsure not. The presence of phytoliths indicates contact (not
what proportion of the original material survived the necessarily use) with only certain plants since phyto-
phytolith extraction procedures. Abundant quantities liths are only produced by some species. So, for
were recorded for five artefacts (Table 3). FRL352 and example, use-wear analysis can detect the processing of
FRL281 are probably multi-functional tools involving non-siliceous plants, but phytolith analysis cannot. In
starchy and other plants. The absence of phytoliths on contrast, the two techniques will agree on some uses,
FAO288 combined with abundant starch strongly sup- for example, soft siliceous or high starch siliceous
ports the interpretation of a hand-held tool used to plants, but the presence of characteristic phytoliths
process a starchy, non-phytolith producing plant. The permits identification of contact material to plant taxa
only definite indication of root processing is from which is not possible through use-wear analysis. In
FAO310 which had masses of large starch grains other words, the data obtained by the two techniques
tentatively identified as Dioscorea (yam). Since bam- should either complement or add a different type of
boo leaf phytoliths were also common on this artefact, information to that provided by the other.
it seems very likely that it was partially wrapped to aid If the findings of both techniques are combined,
prehension while used to process a starchy tuber. The every artefact examined had use-wear or phytoliths
last example with a high abundance of starch is and/or starch residues. In only two cases, discussed
FAO339 which was classified by the correspondence further below, did interpretations from the two tech-
analysis into the group characterized by palms. Since niques disagree. Total agreement was reached in terms
palms are quite starchy plants, this finding is not of the ascription of plant as the contact material. All
surprising. Starch was also recovered in lower quan- the tools in the class of soft, non-siliceous plants
tities from all the other artefacts in the phytolith palm established in the use-wear analysis either lacked phy-
group. toliths or were placed in the uncertain class because the
Although sample sizes are small, some interesting phytolith counts were too low to give meaningful data
chronological trends in contact materials can be seen in in terms of use. The only example with evidence of skin
the correspondence analysis of the phytolith assem- working, FAO288, lacked phytoliths, although starch
blages from FAO (Figures 12 & 13). The only Period 2 was recovered from the phytolith extraction. Since
artefact, FAO364, forms an outlier because it contains starch was not found through direct observation of the
a mix of herbaceous and arboreal phytoliths. The artefact nor was there evidence of use on starchy
arboreal and palm groups are comprised solely of plants, we view the starch as possible contamination
Period 3 artefacts, whereas the bamboo group is domi- (cf. Barton, Torrence & Fullagar, 1998).
nated by Period 4 artefacts. These data may be moni- These cases demonstrate that the absence or low
toring a shift through time from sturdier wooden hafts frequency of phytoliths is generally significant. A sig-
to more expedient handles using leaves. The mixed nificant exception is FAO367 which lacked quantifiable
group in the middle of Figure 12 is comprised of the phytoliths and yet had well-preserved polishes and
multi-functional tools from FRL. These tools show residues that allowed direct observation of the artefact
little chronological patterning, although there is some to suggest that it was hafted. This discrepancy may
separation within the cluster with the more recent be explained by the relatively poor preservation of
examples located closest to the bamboo group. phytoliths for Period 2 samples at FAO.
544 L. Kealhofer et al.

The second case where there was substantial dis- ment. The majority of the tools in this category could
agreement between the two techniques is FAO285. only be assigned an uncertain plant function by use-
This artefact was determined as unused by use-wear wear residue analysis, but the phytolith study has
analysis but had abundant quantities of phytoliths provided a new perspective on these artefacts by iden-
placing it into the group characterized by palms. Once tifying differences in the nature of the hafting material
again we feel that the use-wear is the best judge of use, used to grasp them.
although the phytoliths might indicate contact ma-
terial, which in this case could be either some form of
deliberate wrapping of the artefact or non-intentional,
post-depositional contact. The former seems more
Chronological Change
likely since only a small number of phytoliths were Integrating the results from the use-wear/residue and
extracted from the corresponding artefact soil sample. phytolith analyses allows us to address predictions
In a number of examples the phytolith data provided concerning chronological changes in tool use on the
additional information on the types of plants which north coast of West New Britain. Following on from
produced the observed use-wear. FAO313 and Fullagar’s (1992) and Torrence’s (1992) previous work,
FAO336 which were in the use-wear uncertain plant we expected to find a change from curated, multi-
category were assigned to a phytolith group and functional to single use tools combined with a change
specific plant types could be given to others. It is from less to more diverse range of tasks at individual
interesting that the phytolith bamboo group is largely sites. The very small sample size prohibits an adequate
comprised of artefacts characterized by use-wear test of these hypotheses, but the integrated data do
analysis as uncertain plant. It is possible that Bambu- contribute significantly to an understanding of the
soid residues came from the expedient use of bamboo changes in obsidian artefacts.
leaves to grasp sharp obsidian flakes during very The earliest Periods 1 and 2 appear to have been
short-term use that left the undiagnostic traces characterized by both multi-functional and single pur-
observed by Fullagar. Until further study of modern pose tools primarily for processing starch. Hafting was
reference material from the area has been undertaken, common for many of the artefacts but unretouched
a specific identification of the leaves is not possible. flakes were also used on soft, starchy plants. Since
FAO310 for which bamboo use-wear was observed phytolith preservation is poor for Period 2 at FAO, we
may also have been used in the same way but used for cannot determine whether the unretouched flakes were
a slightly longer period of time. FAO336 which had hafted or had handles of some form. The fact that the
uncertain plant use-wear and was placed in the phyto- most highly curated artefacts in our sample were
lith palm group may also have been an expedient discarded at FRL and not at FAO is significant
tool held in palm leaves and used only for a brief because the former is a quarry site (Specht et al., 1988).
time. It is significant that the only tool which had It seems likely that these tools were abandoned at FRL
evidence of contact with skin, FAO285, did not have after fresh replacements had been manufactured using
phytoliths, thereby supporting the proposition that the plentiful supplies of raw material. Similar tools
other expediently used tools were grasped directly in may have been used at FAO but were carried away and
the hand. discarded elsewhere.
The FAO364s sample provides additional evidence Period 3 is characterized by a change to more
for the use of plants as expedient holding materials. expedient tool use. Only one artefact has been
The retouched stem had a few herbaceous phytoliths, retouched so as to create an area appropriate for
possibly suggesting a material used to wrap or haft this hafting, although the unretouched flakes appear to
stem. When combined with use-wear identified as have had informal handles consisting of wrapped plant
relating to soft, starchy plant for the blade end, it materials. Use-wear analysis did not specifically iden-
seems this tool may have been used to process tubers. tify artefacts used to process starchy plants, although
The interpretation of FAO364 demonstrates how the starch was noted as a residue by both techniques for
two techniques can be used in concert with data on two cases.
manufacturing technique to infer tool function. No By Period 4 the trend to expedient tool use had
single approach definitively reconstructed tool use, but progressed further. In the majority of cases the very
together they provide a fuller picture. sharp, unretouched, obsidian flakes were grasped with
These examples demonstrate the need for an inte- bamboo leaves—presumably as much to reduce dam-
grated approach to the study of tool function. Rather age to fingers as well as to increase the grip. Only one
than rely solely on either the analysis of residues artefact was held directly in the hand, but it was used
extracted from artefacts or direct observation of use- only very briefly. With the exception of the starch
wear/residues, a combination of approaches produces a extracted from FAO288 as part of the phytolith analy-
more comprehensive reconstruction. The case study sis, which for reasons stated above we are hesitant to
also highlights the very important role phytoliths can accept uncritically, there is no evidence for starch
play in identifying the function of unretouched, expe- processing in Period 4. This result is extremely im-
dient tools with very low levels of use-wear develop- portant because as Specht (1981) has noted, recent
Integrating Phytoliths within Use-Wear/Residue Studies of Stone Tools 545

ethnographic accounts only mention uses associated about chronological variability in the way the tools
with the human body (e.g., shaving). In the recent past were handled.
processing of starchy tubers involved shell not stone While contributing further taxonomic information
artefacts. Barton & White (1993) have also detected a on plant processing and craft production in West New
change from stone to shell tool use at Balof Cave in Britain, phytolith studies also provided quantitative
New Ireland. The plant contact material which information for testing hypotheses about functional
Fullagar has noted in the use-wear analysis may simply variability. This new information extends the category
be hafting of flakes as suggested by the phytolith study of siliceous plant processing, available from use-wear/
for uses similar to the ethnographic accounts. The residue studies, to the presence and quantification of
absence of cracked protein films interpreted as animal specific plants, particularly bamboo, palm, grasses
blood residues is not surprising for this tropical, high and a range of Dicotyledons. New insights were also
rainfall environment. obtained about the nature of tool use such as handling,
In summary, our results provide substantial confir- hafting, single versus multi-purpose, and expediency.
mation for a change from curated to expedient tool use The extra information concerning expediently used
at sites FRL and FAO. The new data from the tools, for which use-wear was poorly developed, was
integrated study of direct observation and phytolith particularly valuable since the change from curated to
analysis also provides important new insights into expedient tools is a major focus for our research. In
the associated changes in the nature of handles. In future work each research project will need to assess
addition, the recognition of a possible change from the need for a phytolith study given the particular
primarily plant processing to cutting of animal skin or questions asked, the likely functions of the tools,
tissue not only helps explain the previous conflict and the specific types of information which phytolith
between Fullagar’s (1992) use-wear/residue study and analysis provides.
recent ethnography but also provides a much richer It should be stressed that on its own, phytolith
understanding of chronological changes in obsidian analysis is not an adequate means for understanding
tool use in West New Britain. tool function. Not only is there potential for non-
intentional contact with materials and/or post-
depositional contamination, but also phytoliths can
only provide information about a subset of plants.
Conclusions Analyses of phytolith assemblages from residues
At the beginning of this paper we framed two ques- should always be integrated into a wider study which
tions. The first was answered affirmatively for the incorporates use-wear/residue and technological analy-
artefacts used in this study: phytolith assemblages sis of the stone artefacts. When carried out in conjunc-
represent residues resulting from pre-depositional con- tion with adequate soil testing and artefact analyses,
tact with materials. Since contamination from the the study of phytolith assemblages extracted from
surrounding soil is likely to occur in most archaeologi- residues on artefacts will play a very important role in
cal settings, each case will need evaluation comparable the understanding of tool use.
to the methodology used here. We recommend that
systematic testing of site soil samples from archaeologi-
cal contexts and random testing of artefact soil samples
be carried out in conjunction with residue analyses.
Acknowledgements
When phytoliths are absent or present in low numbers Much of the research was carried out while L.
in residue samples, preference should be given to the Kealhofer held Australian Museum Visiting Fellow-
interpretation from direct observation of use-wear/ ships in 1995 and 1997. R. Fullagar and R. Torrence
residues. This prescription applies to all forms of held an Australian Research Council Research Fellow-
residue analysis (Fullagar, Furby & Hardy, 1996; ship and a Senior Research Fellowship, respectively.
Barton, Torrence & Fullagar, 1998). The excavation of FRL was supported by an
The second question was how does the analysis of Australian Research Council grant to Jim Specht and
phytolith assemblage data complement and/or add to Chris Gosden. Fieldwork on Garua Island was funded
information derived from use-wear/residue analyses. by the Australian Research Council and the Papua
Since the analysis of phytoliths involves costly and New Guinea Biological Foundation with assistance
time-consuming procedures, it is important to assess from Garua Plantation Pty, Kimbe Bay Shipping
whether these are justified. In terms of the obsidian Agencies, Walindi Plantation, the West New Britain
artefacts analysed in our work, we conclude that Provincial Cultural Centre and the support of local
phytolith analysis makes a very important contribution communities. Both projects were affiliated to the
to the understanding of chronological changes in tool National Museum and Art Gallery, PNG. Permission
use. First, the analysis of phytolith assemblage com- for research was granted by the West New Britain
position using correspondence analysis categorized Provincial Government and visas were obtained with
contact material by particular plant types. Second, the assistance of the National Research Institute.
phytolith analysis provided important information The authors are also grateful to Anita van der
546 L. Kealhofer et al.

Meer and Kelly Ann Sullivan for technical assistance, Fullagar, R., Loy, T. & Cox, S. (1998). Starch grains, sediments and
Glenn Summerhayes for Figure 1, Peter White and stone tool function: archaeological evidence for plant foods from
Bitokara, Papua New Guinea. In (R. Fullagar, Ed.). Accessing
Peter Graves for comments, and Annie Clarke for Stones: Recent Australian Studies of Stone Tools. Sydney: Sydney
lunches. University, Archaeological Methods Series.
Hall, J., Higgins, S. & Fullagar, R. (1989). Plant residues on stone
tools. In (W. Beck, A. Clarke & L. Head, Eds) Plants in Australian
References Archaeology. St. Lucia: Tempus Monograph Series, pp. 136–160.
Hurcombe, L. M. (1992). Use Wear Analysis and Obsidian: Theory,
Ambrose, W. (1996). Obsidian hydration dating of the Reef/Santa Experiments and Results. Sheffield: Sheffield Archaeological
Cruz Lapita sites. In (J. Davidson, G. Irwin, F. Leach, A. Pawley Monographs 4.
& D. Brown, Eds) Oceanic Culture History. Dunedin North: New Kealhofer, L. & Piperno, D. (in press). Phytolith Distribution in
Zealand Journal of Archaeology Special Publication, pp. 245–255. Modern Southeast Asian Flora. Washington D.C.: Smithsonian
Anderson, P. (1980). A testimony of prehistoric tasks: diagnostic Contributions to Botany.
residues on stone tool working edges. World Archaeology 12, Loy, T. H. (1994). Methods in the analysis of starch residues on
181–193. prehistoric stone tools. In (J. Hather, Ed.) Tropical Archaeo-
Atchison, J. & Fullagar, R. (1998). Starch residues on pounding botany. London: Routledge, pp. 86–114.
implements from Jinmium rock-shelter. In (R. Fullagar, Ed.)
Loy, T., Wickler, S. & Spriggs, M. (1992). Direct evidence for human
Accessing Stones: Recent Australian Studies of Stone Tools.
use of plants 28,000 years ago: starch residues on stone artifacts
Sydney: Sydney University, Archaeological Methods Series.
from the northern Solomon Islands. Antiquity 66, 898–912.
Barton, H. (1991). Raw material and the tool function: a residue and
use-wear analysis of artefacts from a Melanesian rock shelter. B. A. Machida, H., Blong, R., Moriwaki, H., Hayakawa, Y., Talai, B.,
Hons. Thesis, University of Sydney, Sydney. Lolock, D., Specht, J., Torrence, R. & Pain, C. (1996). Holocene
Barton, H., Fullagar, R. & Torrence, R. (1998). Clues to stone tool explosive eruptions of Witori and Dakataua volcanoes in West
function re-examined: comparing starch grain frequencies on used New Britain, Papua New Guinea and their possible impact on
and unused obsidian artefacts. Journal of Archaeological Science human environment. Quaternary International 35/36, 65–78.
25, 1231–1238. Piperno, D. R. (1988). Phytolith Analysis: An Archaeological and
Barton, H. & White, J. P. (1993). Use of stone and shell artifacts at Geological Perspective. San Diego, CA: Academic Press.
Balof 2, New Ireland, PNG. Asian Perspectives 32, 169–181. Piperno, D. R. & Pearsall, D. M. (1990a). Phytolith morphology in
Bowdery, D. (n.d). Report on phytoliths from a tool at FRL. tropical Bambusoideae and Arundinoideae: taxonomy, new forms
Unpublished report for the Australian Museum. and implications for archaeological maize identification. Ms on
Brown, D. (1984). Prospects and limits of a phytolith key for grasses file. Philadelphia: MASCA, University of Pennsylvania.
in the Central United States. Journal of Archaeological Science 11, Piperno, D. R. & Pearsall, D. M. (1990b). Phytoliths in the repro-
345–368. ductive structures on Zea and Tripsacum: implications for the
Bruier, F. L. (1976). New clues to stone tool function: plant and study of maize evolution. Ms on file. Philadelphia: MASCA,
animal residues. American Antiquity 41, 478–484. University of Pennsylvania.
Fullagar, R. (1986). Use-wear and residues on stone tools: functional Piperno, D. R. & Pearsall, D. M. (in press). The Silica Bodies of
analysis and its application to two southeastern Australian ar- Tropical American Grasses: Morphology, Taxonomy, and Implica-
chaeological assemblages Ph.D. Thesis, La Trobe University, tions for Grass Systematics and Fossil Phytolith Identification.
Melbourne. Washington D.C.: Smithsonian Contributions to Botany.
Fullagar, R. (1988). Recent developments in Australian use-wear Powers, A. H., Padmore, J. & Gilbertson, D. D. (1989). Studies of
and residue studies. In (S. Beyries, Ed). Industries Lithiques, late prehistoric and modern opal phytoliths from coastal sand
Tracéologie et Technologie. Oxford: British Archaeological dunes and machair in northwest Britain. Journal of Archaeological
Reports International Series 411(ii), pp 133–145. Science 16, 27–46.
Fullagar, R. (1991). The role of silica in polish formation. Journal of Shafer, H. J. & Holloway, R. J. (1979). Organic residue analysis in
Archaeological Science 18, 1–25. determining stone tool function. In (B. Hayden, Ed.) Lithic
Fullagar, R. (1992). Lithically Lapita. Functional analysis of flaked Use-Wear Analysis. New York: Academic Press, pp. 385–399.
stone assemblages from West New Britain Province, Papua New Sobolik, K. D. (1996). Lithic organic residue analysis: an example
Guinea. In (J.-C. Galipaud, Ed.). Poterie Lapita et Peuplement. from the Southwestern Archaic. Journal of Field Archaeology 23,
Nouméa: ORSTOM, pp. 135–143. 461–70.
Fullagar, R. (1993a). Flaked stone tools and plant food production:
Specht, J. (1981). Obsidian sources at Talasea, West New Britain,
a preliminary report on obsidian tools from Talasea, West New
Papua New Guinea. Journal of the Polynesian Society 90, 337–56.
Britain, PNG. In (P. Anderson, S. Beyries, M. Otte & H. Plisson,
Eds). Traces et Fonction: Les Gestes Retrouvés. Liegè: ERAUL, Specht, J., Fullagar, R., Torrence, R. & Baker, N. (1988). Prehistoric
Vol. 50, pp. 331–337. obsidian exchange in Melanesia: a perspective from the Talasea
Fullagar, R. (1993b). Taphonomy and tool use: a role for phytoliths sources. Australian Archaeology 27, 3–16.
in usewear and residue analysis. In (B. Fankhauser & J. R. Bird, Therin, M. J., Torrence, R. & Fullagar, R. (1997). Australian
Eds) Archaeometry: Current Australasian Research. Canberra: Museum starch reference collection. Australian Archaeology 44,
Australian National University. Occasional Papers in Prehistory 52.
22, 740–745. Torrence, R. (1992). What is Lapita about obsidian? A view from
Fullagar, R. & Furby, J. (1997). Pleistocene seed-grinding imple- the Talasea source. In (J.-C. Galipaud, Ed.) Poterie Lapita et
ments from the Australian arid zone. Antiquity 71, 300–307. Peuplement. Nouméa: ORSTOM, pp. 111–26.
Fullagar, R. B., Meehan, R. & Jones, R. (1992). Residue analysis Torrence, R. (1993). Ethnoarchaeology, museum collections and
of ethnographic plant-working and other tools from northern prehistoric exchange: obsidian-tipped artifacts from the Admiralty
Australia. In (P. Anderson-Gerfaud, Ed.) Préhistoire de Islands. World Archaeology 24, 467–81.
l’agriculture: nouvelles approches experimentales et ethno- Torrence, R., Pavlides, C., Jackson, P. & Webb, J. (in press)
graphiques. Monographie du CRA. 6, Paris: CNRS, pp. 39–53. Volcanic disasters and cultural discontinuities in the Holocene of
Fullagar, R., Furby, J. & Hardy, B. (1996). Residues on stone West New Britain, Papua New Guinea. In (C. Firth & W.
artifacts: state of a scientific art. Antiquity 70, 740–745. McGuire, Eds) Volcanoes in the Quaternary. Oxford: Blackwell.

You might also like