ChapterPDF Available

A comparative study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens

Authors:

Abstract and Figures

An experimental program was designed to compare and contrast the stone tool-making skills of modern African apes (bonobos or Pan paniscus), of prehistoric toolmakinghominins from the earliest known Palaeolithic sites at Gona, Ethiopia (sites EG 10 and EG 12) dating to approximately 2.6 million years ago (possibly Australopithecusgarhi), and of modern humans or Homo sapiens. All three species used the same range of raw materials, unmodified water-rounded cobbles of volcanic rock from the 2.6 Ma river conglomerates in the Gona study area. A detailed attribute analysis of the three samples was conducted, examining flaking patterns and artifact products and, from these products, inferring relative levels of stone tool-making skill in the three species. Results of this comparative analysis indicate that, inthe majority of artifact attributes that appear to be linked with skill, the Gona hominin tool-makers grouped either with the modern human sample or were intermediate between the bonobos and modern humans. This indicates that the biomechanical and cognitive skills required for efficient stone tool-making were already present at Gona by 2.6 Ma. Although some of the individual stone artifacts generated by the bonobos [with cranial capacitiesand probable EQ (encephalization quotient) values similar to estimates for prehistoric hominins contemporary with the Gona sites] would clearly be recognized as artifactualby palaeolithic archaeologists, the label “Pre-Oldowan” might be more appropriate for the overall assemblage of artifacts they produce. The level of flakingskill seen in the bonobo assemblage may represent an earlier stage of lithic technology not yet discovered inthe prehistoric record. Or, alternatively, if the Gona sites indeed represent the earliest phases of stone tool-making with no precursors to be found, it may be that by the time of the Gona sites early hominins were already “preadapted” to more skilled flaking of stone. This study also highlights interesting aspects of the stone tool-making behaviors of the early tool-makers at Gona by 2.6 Ma: 1) the Gona hominins were selective in choosing raw materials, sometimes selecting excellent quality raw materials from the river gravels available within the Gona region; 2) the Gona hominins conducted earlier stages of the reduction of cobbles off-site, prior to the transport of cores to the floodplain sites; and 3) the Gona hominins likely transported numerous larger, more usable flakes away from the floodplain sites. This suggests a higher level of early tool-making complexity, and presumably subsistence complexity, than many prehistorians have appreciated.
Content may be subject to copyright.
stone age institute
publication series
Series Editors Kathy Schick and Nicholas Toth
Number 1.
THE OLDOWAN: Case Studies into the Earliest Stone Age
Nicholas Toth and Kathy Schick, editors
Number 2.
BREATHING LIFE INTO FOSSILS:
Taphonomic Studies in Honor of C.K. (Bob) Brain
Travis Rayne Pickering, Kathy Schick, and Nicholas Toth, editors
Number 3.
THE CUTTING EDGE:
New Approaches to the Archaeology of Human Origins
Kathy Schick, and Nicholas Toth, editors
Number 4.
THE HUMAN BRAIN EVOLVING:
Paleoneurological Studies in Honor of Ralph L. Holloway
Douglas Broadeld, Michael Yuan, Kathy Schick and Nicholas Toth, editors
Stone Age Institute
Gosport, Indiana
and
Indiana University,
Bloomington, Indiana
Stone Age Institute Press · www.stoneageinstitute.org
1392 W. Dittemore Road · Gosport, IN 47433
STONE AGE INSTITUTE PUBLICATION SERIES
NUMBER 1
Edited by Nicholas Toth and Kathy Schick
THE OLDOWAN:
Case Studies Into the Earliest Stone Age
Published by the Stone Age Institute.
ISBN-10: 0-9792-2760-7
ISBN-13: 978-0-9792-2760-8
Copyright © 2006, Stone Age Institute Press.
All rights reserved under International and Pan-American Copyright Conventions. No part of this
book may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopying, without permission in writing from the publisher.
COVER PHOTOS
Front, clockwise from upper left:
1) Excavation at Ain Hanech, Algeria (courtesy of Mohamed Sahnouni).
2) Kanzi, a bonobo (‘pygmy chimpanzee’) fl akes a chopper-core by hard-hammer percussion (courtesy Great Ape
Trust).
3) Experimental Oldowan fl aking (Kathy Schick and Nicholas Toth).
4) Scanning electron micrograph of prehistoric cut-marks from a stone tool on a mammal limb shaft fragment (Kathy
Schick and Nicholas Toth).
5) Kinesiological data from Oldowan fl aking (courtesy of Jesus Dapena).
6) Positron emission tomography of brain activity during Oldowan fl aking (courtesy of Dietrich Stout).
7) Experimental processing of elephant carcass with Oldowan fl akes (the animal died of natural causes). (Kathy
Schick and Nicholas Toth).
8) Reconstructed cranium of Australopithecus garhi. (A. garhi, BOU-VP-12/130, Bouri, cranial parts, cranium recon-
struction; original housed in National Museum of Ethiopia, Addis Ababa. ©1999 David L. Brill).
9) A 2.6 million-year-old trachyte bifacial chopper from site EG 10, Gona, Ethiopia (courtesy of Sileshi Semaw).
Back:
Photographs of the Stone Age Institute. Aerial photograph courtesy of Bill Oliver.
ABSTRACT
An experimental program was designed to compare
and contrast the stone tool-making skills of modern Afri-
can apes (bonobos or Pan paniscus), of prehistoric tool-
making hominins from the earliest known Palaeolithic
sites at Gona, Ethiopia (sites EG 10 and EG 12) dating to
approximately 2.6 million years ago (possibly Australo-
pithecus garhi), and of modern humans or Homo sapi-
ens. All three species used the same range of raw materi-
als, unmodifi ed water-rounded cobbles of volcanic rock
from the 2.6 Ma river conglomerates in the Gona study
area. A detailed attribute analysis of the three samples
was conducted, examining fl aking patterns and artifact
products and, from these products, inferring relative lev-
els of stone tool-making skill in the three species.
Results of this comparative analysis indicate that, in
the majority of artifact attributes that appear to be linked
with skill, the Gona hominin tool-makers grouped either
with the modern human sample or were intermediate be-
tween the bonobos and modern humans. This indicates
that the biomechanical and cognitive skills required for
effi cient stone tool-making were already present at Gona
by 2.6 Ma. Although some of the individual stone arti-
facts generated by the bonobos [with cranial capacities
and probable EQ (encephalization quotient) values simi-
lar to estimates for prehistoric hominins contemporary
with the Gona sites] would clearly be recognized as ar-
tifactual by palaeolithic archaeologists, the label “Pre-
Oldowan” might be more appropriate for the overall as-
semblage of artifacts they produce. The level of fl aking
skill seen in the bonobo assemblage may represent an
earlier stage of lithic technology not yet discovered in
the prehistoric record. Or, alternatively, if the Gona sites
indeed represent the earliest phases of stone tool-mak-
ing with no precursors to be found, it may be that by the
time of the Gona sites early hominins were already “pre-
adapted” to more skilled fl aking of stone.
This study also highlights interesting aspects of the
stone tool-making behaviors of the early tool-makers at
Gona by 2.6 Ma: 1) the Gona hominins were selective
in choosing raw materials, sometimes selecting excellent
quality raw materials from the river gravels available
within the Gona region; 2) the Gona hominins conduct-
ed earlier stages of the reduction of cobbles off-site, pri-
or to the transport of cores to the fl oodplain sites; and 3)
the Gona hominins likely transported numerous larger,
more usable fl akes away from the fl oodplain sites. This
suggests a higher level of early tool-making complexity,
and presumably subsistence complexity, than many pre-
historians have appreciated.
INTRODUCTION
The evolution of technology has occurred in tandem
with human biological evolution during at least the past
2.6 million years. This unique co-evolution ultimately
has led to the modern human condition, and it is likely
that major cognitive and biomechanical changes oc-
curred during this time through selection for more ef-
cient tool-related activities. Different stages or grades
of human evolution tend to be associated with different
levels of technology, with a general trend of increasing
technological complexity and sophistication through
time.
A persistent challenge in paleoanthropology is how
to determine levels of cognitive and biomechanical skill
based upon the archaeological record, which consists
primarily of fl aked stone artifacts and, sometimes, as-
sociated faunal remains. It would be interesting to be
able to observe and compare different levels of fl aked
stone technology between prehistoric hominins and ex-
A COMPARATIVE STUDY OF THE STONE
TOOL-MAKING SKILLS OF PAN,
AUSTRALOPITHECUS, AND HOMO SAPIENS
BY NICHOLAS TOTH, KATHY SCHICK, AND SILESHI SEMAW
CHAPTER 6
156 The Oldowan: Case Studies Into the Earliest Stone Age
tant apes and humans. Unfortunately, Early Stone Age
hominins are extinct, and modern apes are not known to
ake stone intentionally in the wild. However, the last-
ing products of early tool-making persists in the form of
stone artifacts at early archaeological sites at Gona in
Ethiopia, and, beginning in 1990, captive bonobos have
been producing fl aked stone artifacts in an experimen-
tal setting (Toth et al., 1993; Schick et al., 1999). This
provides a unique opportunity to conduct a three-way
comparison of tool-making patterns evident in the earli-
est known tool-makers, in stone tool-making apes, and in
modern human knappers, in order to investigate techno-
logical patterns and abilities evident in each group.
This study, part of a long-term investigation of stone
tool-making and tool-using abilities in captive African
apes, is an attempt to approach this problem through rig-
orous comparisons of the artifacts produced by the earli-
est stone tool-makers and those produced by bonobos and
modern humans in controlled experiments. In this study,
the bonobos and modern humans used volcanic cobbles
from the same river gravels that had served as the source
of raw materials for the early Gona stone tool-makers.
Thus, all three samples, the prehistoric tool-makers and
the two groups of experimental tool-makers, were effec-
tively using the same raw material source.
This study provides a valuable three-species com-
parison (probably in three different genera) of stone
tool-making and spanning a time period of 2.6 million
years. This makes it possible to make detailed com-
parisons of the stone technologies of the earliest known
stone tool-makers, those produced by modern humans,
and those made by modern apes. Inferences can then be
made regarding discrete attributes, and combinations of
attributes, that emerge as sensitive indicators of relative
levels of skill. Further insights can thereby be gained
regarding behavioral implications of the early archaeo-
logical sites at Gona, as well as regarding ape stone tool-
making abilities and a possible evolutionary ‘substrate’
for the development of technological skills in human
evolution.
THE COMPARATIVE SAMPLES
In effect, then, this study is a comparison of techno-
logical skill between three species over some 2.6 million
years, with good control over raw materials. These three
different samples whose artifacts will be compared and
contrasted are: 1) an experimental sample produced by
African apes who are practiced in stone tool manufac-
ture (bonobos); 2) an archaeological sample produced
by early hominins (at Gona, Ethiopia); and 3) an experi-
mental sample produced by modern humans who are ex-
perienced stone tool-makers.
African Apes
The African ape sample consisted of two bonobos
(“pygmy chimpanzees”), Kanzi and his half-sister Pan-
banisha, both born and raised in captivity with daily
human contact. At the time of these experiments Kanzi
was twenty years old and Panbanisha fi fteen (Figures 1
through 12). Kanzi had been knapping stone for ten years
by this time and Panbanisha for four years. The average
cranial capacity of Pan about is 380 cc, and the Homocen-
tric EQ (a human-centered encephalization quotient, or
ratio of brain size to body size, in which human EQ=1.0)
is 0.38 (Holloway, 2000). The bonobos have learned to
ake stone to produce large, usable fl akes as cutting tools
(cutting through a rope or membrane to access a food re-
source). They were encouraged to reduce cobbles as far
as possible with a stone hammer to produce a range of
Figure 1
1. Portrait of Kanzi.
Figure 2
2. Portrait of Panbanisha.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 157
akes and fragments from which they could select a tool
to use for the cutting activity. Since this study, two of
Panbanisha’s sons (Nyota, now eight years old and Na-
than, now six years old) have started fl aking stone, pri-
marily from observing Kanzi and Panbanisha. In effect,
we now have set forth a fl aked lithic cultural tradition in
this bonobo group which is now long-term and transgen-
erational.
Biomechanical studies by Harlacker (2006), study-
ing the kinesiology of Oldowan tool-making in bonobos
and in unskilled and skilled modern humans have dem-
onstrated that the bonobos are only accelerating their
hammerstones to about one-half of the impact velocities
of the expert human sample (3.67 meters/second versus
7.12 meters/second, respectively). This lower impact ve-
locity of the bonobos will almost certainly affect their
lithic assemblage, which will be discussed below.
The history of the bonobo acquisition of stone tool-
making is presented in Savage-Rumbaugh and Fields
(this volume). (For more details on bonobo acquisition
and development of stone tool-making skills, see Toth et
al., 1993; Schick et al., 1999; and Savage-Rumbaugh et
al., 2006). The study of bonobo stone tool-making is part
of a long-term, ongoing research program that will soon
also include chimpanzees (Pan troglodytes) and orang-
utans (Pongo pongo) as subjects.
Gona Hominins
The archaeological sample was from two contem-
poraneous sites in the Afar Rift at Gona, Ethiopia named
EG (East Gona) 10 and EG 12, located a few hundred
meters from each other. These two sites are almost iden-
tical in their lithic technology (Semaw, 1997), so for
this study the two sites were combined for a statistically
larger sample size. These sites have been dated to ap-
Figure 3
3. Kanzi fl aking a cobble core.
Figure 4
4. Kanzi fl aking another cobble core.
Figure 5
5. Kanzi using a fl ake to cut through rope to
access a food resource.
158 The Oldowan: Case Studies Into the Earliest Stone Age
Figure 6
6. Sample of bonobo cores. Note the predominance of end choppers. (Small squares on scale represent one cm).
Figure 7
7. Bonobo unifacial end chopper.
Figure 8
8. Large bonobo end chopper with some bifacial
aking.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 159
Figure 9
9. Large bonobo unifacial side chopper. 10. Bonobo unifacial chopper.
Figure 11
11. Bonobo end chopper exhibiting very
heavy hammerstone battering.
Figure 12
12. Sample of bonobo fl akes. Note the abundance of side-struck fl akes.
Flakes oriented with striking platforms at top.
Figure 10
160 The Oldowan: Case Studies Into the Earliest Stone Age
proximately 2.6 Ma. Although the species of tool-maker
is not known, the only species known in the Afar Rift
during this time period is Australopithecus garhi, known
from the Middle Awash some 60 km from Gona (Asfaw
et al., 1999; White et al., 2005). The cranial capacity of
this taxon, based on one skull from Bouri, Middle Awash
(the holotype, BOU-VP-12/130), is 450 cc, about 70cc
larger than the Pan mean. The EQ value for A. garhi is
not known. Fossils representing a non-robust Australo-
pithecus that may also be attributable to A. garhi include
a mandible fragment (GAM-VP-1/1) and parietal frag-
ments (GAM-VP-1/2) from Gamedah, Middle Awash
(White et al., 2005) and isolated teeth from the Omo
Shungura Formation (Suwa et al.,
1996). (See Figures 13, A. garhi,
and 14 through 23, setting of and
artifacts from the Gona EG sites).
Geoarchaeological evidence
suggests that there has not been
signifi cant hydrological action
at theses sites: they are found in
ne-grained slickenside clays and
contain not only heavier cores but
also a debitage sample including
very small fl akes and fragments
(Semaw, 1997, 2000; Semaw et
al., 1997, 2003). Thus far, fossil
bone has been found only on the
surface at these archaeological
localities and may have derived
from higher deposits, but cut-
marked bones have been found in
the same stratigraphic level else-
where in the study area (Domin-
guez-Rodrigo et al., 2005).
Modern Humans
The modern human sample
consisted of two experienced stone tool-makers (NT and
KS) that had, at the time of these experiments, fl aked
stone for over two decades. The mean cranial capacity
for modern humans is about 1350 cc, with a Homocen-
tric EQ value of 1.0. Gona cobbles were fl aked unifacial-
ly and bifacially to reduce them by roughly half of their
original cobble mass. As the aim was to produce a con-
trol sample of cores and resultant debitage representing
approximately 50% cobble reduction, rather than to rep-
licate precisely the Gona assemblage, there was some-
what more bifacial fl aking in the human sample than was
represented at the Gona sites (roughly 68% of the human
akes are from unifacial fl aking, versus ~79% of Gona
akes). Direct, freehand, hard-hammer percussion was
employed using Gona cobble percussors, and no spe-
cial attempt was made to prepare platforms or remove
especially long or thin fl akes: the goal was to produce
serviceable fl akes that could be used for cutting activi-
ties. The human experimental sample thus provided an
important baseline that could be compared to the bonobo
and archaeological samples (Figures 21 and 22) to exam-
ine for similarities and differences with the bonobo and
Gona stone technologies.
As analysis proceeded, it became clear that prefer-
entially later stages of cobble reduction were typical of
the Gona EG sites, whereas the experimental samples of
bonobo and human reduction differed in two major as-
pects: they contained all stages of fl aking, from initiation
of cobble reduction to cessation of fl aking, and, overall,
their cores were not as extensively reduced. In the case
of the human experiments, this had been a deliberate de-
sign to produce a control sample of cores and debitage
with reduction of the cobble mass held at approximately
50%, in order to effectively compare the products to the
other two samples. Comparisons with the experimental
human sample highlighted the much heavier reduction
of the Gona cores. Thus, in an attempt to more closely
match the Gona pattern, ten of the experimental human
cores (fi ve unifacial choppers, fi ve bifacial choppers)
were subsequently reduced further in order to examine
assemblage characteristics in such later stages of core
reduction. These “later stage” experiments provided a
database that were then used in more direct compari-
sons with the Gona sites, and especially salient results
of these comparisons are highlighted in special sections
in this analysis.
A major aim of the human experimental sample
was to produce data to generate models to understand
how the Gona archaeological assemblages could have
formed. As will be discussed below, this experimentation
has suggested that the Gona sites represent preferentially
later stages of core reduction with subsequent selection
of certain artifacts that were transported off-site. Another
Figure 13
13. Reconstructed cranium of Australopithecus garhi. Original housed in National
Museum of Ethiopia, Addis Ababa. © 1999 David L. Brill
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 161
Figure 14
14. Gona site EG 10, dated to approximately 2.6 mya.
Figure 15
15. The fossil river gravel conglomerate at East Gona: the source of raw material for the Gona hominins and the
experimental sample.
162 The Oldowan: Case Studies Into the Earliest Stone Age
Figure 16
16. Gona artifacts from EG 10: cores (below) and fl akes (above).
Figure 17
17. Gona unifacial side choppers from EG 10.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 163
Figure 18
18. A set of six Gona EG cores. (Photo courtesy of Tim White).
Figure 19
19. The other opposite face of the six Gona cores in Figure 18. (Photo courtesy of Tim White)
164 The Oldowan: Case Studies Into the Earliest Stone Age
important aim of this study was to compare the human
sample with the bonobo sample (each having all stages
of reduction represented) to evaluate levels of skill and
gain insight into possible stages of development of stone
technology in the course of hominin evolution.
METHODOLOGY
This study was designed as a detailed examination of
the lithic assemblages produced by three samples: mod-
ern African apes (Pan paniscus, called bonobos or “pyg-
my chimpanzees”), prehistoric Gona hominins (possibly
Australopithecus garhi), and modern humans or Homo
sapiens. This analysis would also address basic episte-
mological questions in Palaeolithic archaeology: why do
we measure and record the attributes that we do? Beyond
pure description, what do the attributes that we study tell
us about levels of cognitive and/or biomechanical skills,
and what stages of core reduction (e.g. early, late) are
represented in an assemblage, and what does this imply
regarding transport and land-use patterning?
Analysis and Statistics
It was decided that this analysis would include the
detailed statistical descriptions and testing that was em-
ployed. For more qualitative attributes (e.g. fl ake scar
types, fl ake shapes) chi-square tests were employed,
while for metrical, quantitative attributes (e.g. linear
measurements, weights, and angles) the Mann-Whit-
ney/Wilcoxon test (and sometimes additionally the
Kolmogorov-Smirnov test) was used. In both cases, the
threshold for assessing statistical signifi cance was at the
.05 confi dence level. A summary of statistical tests and
overview of results are presented in appendices at the
end of this chapter.
Raw Materials
The raw materials for the experiments were selected
from the 2.6 Ma river gravels at East Gona, the source
for the Gona hominins as well. With the permission of
the Ethiopian government, unmodifi ed cobbles (i.e. geo-
logical samples) were collected from the surface scatter
and in situ gravels. In practice, only one of perhaps every
fty cobbles in the Gona conglomerates was considered
suitable for fl aking. Cobbles were chosen for 1) their
smooth cortex, suggesting a fi ne-grained rock type; 2)
shapes that would be suitable for fl aking: not too spheri-
cal and not too thin; 3) a range of sizes (from about 10
to 20 cm); and 4) absence of heavy cortical pitting or
hairline fractures, which would suggest a poor-quality
rock with unwanted vesicles or weathering fl aws. All
raw materials in the experimental sample (and it appears
almost all from the archaeological sites) started as water-
worn river pebbles composed of a range of volcanic raw
materials.
The experimental sample of raw material was ran-
domly divided into two groups, one for the bonobo sub-
jects and one for the human sample. As will be discussed,
the experimental raw material sample was very similar
to that exploited by the Gona hominins, so that there
was an excellent control of raw materials between the
experimental and archaeological samples. The hammer-
Figure 20
20. More fl akes from Gona EG 10.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 165
stones used to fl ake cores were also Gona cobbles. Sev-
eral Gona cores show battering on their cortical surfaces
showing that they were also used as percussors, and it is
likely that other hammerstones were transported off-site.
For this analysis, only stone artifacts greater or equal to
20 mm in maximum dimension were examined.
ATTRIBUTES ANALYZED
The following lithic attributes were used in this study:
1. Assemblage composition (lithic class: cores/fl akes/
fragments)
2a. Debitage breakdown [whole fl akes/broken fl akes
(splits & snaps)/angular fragments & chunks]
2b. Ratio of split fl akes to whole fl akes
3. Raw material breakdown (cores)
Figure 21
21. A selection of cores from the human experimental sample. Note the predominance of side choppers.
Figure 22
22. A selection of fl akes from the human experimental sample. Note the abundance of end-struck fl akes.
166 The Oldowan: Case Studies Into the Earliest Stone Age
4. Quality of raw material (cores)
5. Cores: Original form (blank)
6. Cores: Flaking mode (Unifacial, bifacial, etc.)
7. Cores: Invasiveness of fl aking
8. Cores: Percentage of circumference fl aked
9. Cores: Quality of fl aking
10. Cores: Platform edge battering/microstepping/
crushing
11. Cores: Modifi ed Leakey typology
12. Cores: Weight
13. Cores: Maximum dimension (length)
14. Cores: Breadth
15. Cores: Thickness
16. Cores: Ratio of breadth to length
17. Cores: Modifi ed ratio of breadth to length
18. Cores: Ratio of thickness to breadth
19. Cores: Maximum dimension of largest scar
20. Cores: Ratio of largest scar to core maximum
dimension
21. Cores: Percentage of original cobble remaining
(estimate)
22. Cores: Number of fl ake scars
23. Cores: Ratio of step & hinge scars to total scars
24. Cores: Edge angle
25. Cores: Percentage of surface cortex remaining
26a. Flakes: Flake types
26b. Flakes: Simulated fl ake type population after
selection and the removal of the best fl akes
26c. Flakes: Simulated fl ake type population: later
stages of fl aking
26d. Flakes: Simulated fl ake type population: later
stages after selection and removal
27. Flakes: Location of Cortex
28. Flakes: Dorsal scar pattern
29. Flakes: Flake shape
30. Flakes: Platform battering/microstepping, crushing
31. Flakes: Weight
32. Flakes: Maximum dimension
33. Flakes: Thickness
34. Flakes: Ratio of fl ake breadth to length
35. Flakes: Ratio of fl ake thickness to breadth
36. Flakes: Ratio of platform thickness to platform
breadth
37. Flakes: Number of platform scars
38. Flakes: Number of dorsal scars
39. Flakes: Ratio of step & hinge scars to all scars
40. Flakes: Exterior platform angle (“core angle”)
41. Flakes: Interior platform angle (“bulb angle”)
42. Flakes: Percentage of dorsal cortex (“cortex
index”)
RESULTS
In this study, these 42 separate attributes, both qual-
itative and quantitative, were examined and compared/
contrasted among the three lithic assemblages, and then
interpreted with regard to similarities and differences.
When appropriate, statistical tests were conducted to
examine for signifi cant differences among the three tool-
maker samples in their attribute characteristics.
1. Assemblage: Composition (Lithic Class)
Rationale
The major classes of fl aked stone artifacts consist
of cores, fl akes, and fragments. Fragments comprise
a variety of broken portions of fl akes, including split
akes (broken more or less perpendicular to the plat-
form), snapped fl akes (broken more or less parallel to
the platform, into proximal, distal, and sometimes also
mid-section snaps), angular fragments (fairly fl at frag-
ments, apparent portions of conchoidally fl aked pieces
but without clear indication of which part is represent-
ed), and chunks (miscellaneous, usually thicker, nonde-
script fragments of conchoidally fl aked materials). Here
we examine whether there are signifi cant differences in
the major fl aked artifact categories among the three tool-
maker samples (bonobos, Gona knappers, humans).
Figure 23
23. Flake types. Flake types 1-3: cortical platforms with, in order, all dorsal cortex, partial dorsal cortex, and no dor-
sal cortex. Flake types 4-6: non-cortical platforms with, in order, all dorsal cortex, partial dorsal cortex, and no
dorsal cortex. Examination of these fl ake types in an artifact assemblage can yield critical technological informa-
tion regarding core reduction.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 167
Results
The overall breakdown of the assemblages into the
major lithic classes (cores, fl akes, and fragments) by
tool-maker is shown in the table and graph below:
Assemblage Composition by Tool-maker
Core ake frag
core ake frag Total
Tool-maker bonobo Count 33 158 130 321
% within
Tool-maker 10.3% 49.2% 40.5% 100.0%
gona Count 23 178 294 495
% within
Tool-maker 4.6% 36.0% 59.4% 100.0%
human Count 31 244 370 645
% within
Tool-maker 4.8% 37.8% 57.4% 100.0%
Total Count 87 580 794 1461
% within
Tool-maker 6.0% 39.7% 54.3% 100.0%
humanGonabonobo
Tool-maker
60.0
50.0
40.0
30.0
20.0
10.0
0.0
Percent
Frag %
Flake %
Core %
Assemblage Composition by Tool-maker
Chi-Square tests revealed that the Gona and human
tool-makers group together in terms of their general as-
semblage composition (.788 level of signifi cance), while
the artifact assemblages produced by the bonobo tool-
makers were signifi cantly different from both the Gona
and the human samples (at the .000 level of signifi cance).
The bonobo assemblage differed from both the Gona
and human samples in its much higher core proportion,
higher fl ake proportion, and much lower fragment pro-
portion.
Chi-Square Test: Assemblage Composition, Bonobo v. Gona
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 30.703a2 .000
Likelihood Ratio 30.716 2 .000
N of Valid Cases 816
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 22.03.
Chi-Square Test: Assemblage Composition, Bonobo v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 28.157a2 .000
Likelihood Ratio 27.888 2 .000
N of Valid Cases 966
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 21.27.
Chi-Square Test: Assemblage Composition, Gona v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square .478a2 .788
Likelihood Ratio .478 2 .787
N of Valid Cases 1140
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 23.45.
Discussion
While the Gona and human proportions of cores,
akes, and fragments are similar, the bonobo sample
has a higher core percentage (more than double the as-
semblage percent in the Gona and human samples), indi-
cating the bonobos are removing less debitage per core.
Interestingly, they are producing more whole fl akes than
fragments, while in both the Gona and human samples
there are proportionally more fragments than whole
akes. While some lithic analysts might argue that a
higher fl ake-to-fragment ratio may be an indication of
greater skill, it is unlikely in the case of the bonobos.
Their hammerstone velocities were much lower than the
modern human subjects, and the bonobos would often re-
peatedly strike the core at a location until fracture fi nally
occurred and proceed to “chew” down the cobble with
further fl aking. The majority of spalls so produced by
the bonobos tend to be whole fl akes. In contrast, modern
humans (and probably Gona hominins) appear to have
exploited cores more effi ciently, reducing raw material
in fl ake production more readily and with much higher
hammerstone velocities, in the process producing more
shatter (fragments) relative to whole fl akes.
2a. Assemblage: Debitage Breakdown
Rationale
The more subtle differences between debitage cat-
egories (whole fl akes, split and snapped fl akes, and an-
gular fragments/chunks) can be a useful means of com-
paring different lithic assemblages. Although we had no
clearly-defi ned expectations, we were interested to see if
there were any major differences between the tool-mak-
ing groups in the types of debitage they produced.
Results
The overall breakdown of the debitage type by tool-
maker is shown in the table and graph below:
168 The Oldowan: Case Studies Into the Earliest Stone Age
Debitage Breakdown by Tool-maker
(Whole Flakes, Broken Flakes, Angular Fragments)
Deb type
Flake Flk
brok Frag-
ment Total
Tool-maker bonobo Count 158 34 96 288
% within
Tool-maker 54.9% 11.8% 33.3% 100.0%
gona Count 178 95 199 472
% within
Tool-maker 37.7% 20.1% 42.2% 100.0%
human Count 244 226 144 614
% within
Tool-maker 39.7% 36.8% 23.5% 100.0%
Total Count 580 355 439 1374
% within
Tool-maker 42.2% 25.8% 32.0% 100.0%
humanGonabonobo
Tool-maker
60.0
50.0
40.0
30.0
20.0
10.0
0.0
Debitage %
Angular
fragment
Broken flake
Whole flake
Debitage Breakdown by Tool-maker
Chi-Square tests revealed that each of the three tool-
making groups produced debitage breakdowns that were
statistically different from the other two groups (at the
.000 level of signifi cance).
Chi-Square Test: Debitage Breakdown, Bonobo v. Gona
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 22.786a2 .000
Likelihood Ratio 22.959 2 .000
N of Valid Cases 760
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 48.88.
Chi-Square Test: Debitage Breakdown, Bonobo v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 59.767a2 .000
Likelihood Ratio 66.431 2 .000
N of Valid Cases 902
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 76.63.
Chi-Square Test: Debitage Breakdown, Gona v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 54.975a2 .000
Likelihood Ratio 55.655 2 .000
N of Valid Cases 1086
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 139.51.
Discussion
Bonobos produced higher percentages of whole
akes and lower percentages of broken fl akes than the
other two samples. As mentioned above, it is likely that
this pattern results from knapping with lower hammer-
stone velocities, so that less spontaneous fl ake fragmen-
tation occurs during the manufacturing process. To test
this, as a follow-up experiment, we reduced cobbles with
high-velocity/high-impact hammerstone percussion as
well as low-velocity/low-impact hammerstone percus-
sion. The low-velocity/low-impact fl aking yielded deb-
itage with an appreciably a greater proportion of whole
akes (31% higher) than did high-velocity/high-impact
aking. This supports our contention that the higher per-
centage of whole fl akes in the bonobo sample and the
smaller mean fl ake size are primarily due to their lower-
velocity impacts.
In addition, the Gona sample has appreciably more
angular fragments and fewer broken fl akes than the hu-
man sample. It is possible that the elevated proportion
of angular fragments may partially be a result of post-
depositional breakage of debitage (Hovers, 2003). Also,
some subtle signs of conchoidal fracture are more dif-
cult to identify on volcanic archaeological specimens,
as their fractured surfaces tend to be slightly weathered;
therefore it is likely that some pieces that would be as-
signed to splits or snaps on fresh specimens would be
demoted to angular fragments on archaeological speci-
mens. Finally, as will be discussed below, it appears that
a portion of the larger, sharper whole fl akes and broken
akes at Gona may have been subsequently transported
off-site, which would increase the percentage of angular
fragments in the debitage population.
2b. Ratio of Split to Whole Flakes
Rationale
A higher ratio of split fl akes to whole fl akes could be
an indication of the average impact velocity of hammer-
stones when fl aking cores. As noted above, experiments
have shown that for a given raw material, higher percus-
sion forces tend to produce higher ratios of split to whole
akes (but also larger whole fl akes).
Results
The bonobos had the lowest number of split to
whole fl akes (ratio of .095, 8.7% split fl akes), with Gona
hominins intermediate (ratio of .331, 24.9% split fl akes)
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 169
and the human sample the highest (ratio of .643, 39.2%
split fl akes). Each of these samples showed signifi cant
difference from the other two in their whole and split
ake proportions at the .05 confi dence level.
Whole and Split Flakes by Tool-maker
Whole or
split ake
split whole Total
Tool-maker bonobo Count 15 158 173
% within Tool-maker 8.7% 91.3% 100.0%
gona Count 59 178 237
% within Tool-maker 24.9% 75.1% 100.0%
human Count 157 244 401
% within Tool-maker 39.2% 60.8% 100.0%
Total Count 231 580 811
% within Tool-maker 28.5% 71.5% 100.0%
Chi-Square Test: Whole and Split Flakes, All Tool-makers
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 57.243a2 .000
Likelihood Ratio 64.190 2 .000
N of Valid Cases 811
a. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 49.28.
Chi-Square Test: Whole and Split Flake Proportions,
Bonobo v. Gona
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 17.796b1 .000
Continuity Correctiona16.716 1 .000
Likelihood Ratio 19.135 1 .000
Fisher’s Exact Test .000 .000
N of Valid Cases 410
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 31.22.
Chi-Square Test: Whole and Split Flake Proportions,
Bonobo v. Human
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 53.509b1 .000
Continuity Correctiona52.066 1 .000
Likelihood Ratio 62.038 1 .000
Fisher’s Exact Test .000 .000
N of Valid Cases 574
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 51.84.
Chi-Square Test: Whole and Split Flake Proportions,
Gona v. Human
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 13.522b1 .000
Continuity Correctiona12.893 1 .000
Likelihood Ratio 13.861 1 .000
Fisher’s Exact Test .000 .000
N of Valid Cases 638
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 80.24.
Discussion
It would appear that the bonobos has the lowest
hammerstone velocity, the Gona hominins intermedi-
ate, and the humans the highest velocities. As previ-
ously mentioned, experiments and analysis by Harlacker
(2006) have demonstrated that the mean hammerstone
velocities just prior to impact of the bonobos was 3.67
meters per second, and the hammerstone velocities of
the experienced human sample was 7.12 meters per sec-
ond. If the split to whole fl ake ratio is strongly correlated
with hammerstone velocities, then the extrapolated mean
hammerstone velocity of the Gona hominins could be es-
timated to have been a minimum of 5 meters per second
(and perhaps more if, as we believe, the Gona hominins
were selectively removing numbers of larger fl akes, as
discussed in section 26 below).
3. Assemblage: Raw Material Type
Rationale
Examining differential use of raw materials can
give insights into possible selectivity of early hominin
tool-makers. Raw materials in the Gona archaeological
sample were divided three major types: lighter volcanics
(especially trachyte), darker volcanics (especially rhyo-
lite), and vitreous volcanics (very fi ne-grained). Subse-
quent geological classifi cation by Jay Quade (reported
in Stout et al. (2005), has subsequently separated Gona
raw materials into a number of more discrete types. The
distribution of raw material types was examined in the
archaeological sample and the experimental sample
(knapped by humans and bonobos) for possible differ-
ences among the tool-makers in the fl aked cores (as a
proxy for cobble selections).
The experimental sample of cobbles was selected in
proximity to the EG sites from the 2.6 million year old
conglomerate that lies stratigraphically just below the
sites. At the time of occupation of these sites, this con-
glomerate would have been a readily accessible cobble
source in the overall region as exposed gravel bars along
stream courses, and the raw material types at the EG
sites are found within this conglomerate. Selection was
based upon sizes and shapes of cobbles deemed suitable
170 The Oldowan: Case Studies Into the Earliest Stone Age
for fl aking (excluding very large or very small cobbles,
spherical clasts, etc.), as well as the likelihood or expec-
tation of good, fi ne-grained stone (normally indicated by
a cobble with smooth cortex but without extensive pit-
ting or incipient cracks). Selection was based on these
criteria, and not upon observable indications of the rock
type itself. No initial testing (fl aking) was done prior to
the experiments. In practice, only approximately one in
fty cobbles from these conglomerates was selected via
this procedure. We were interested to see whether the
Gona EG sites showed more selectivity than this sam-
pling procedure.
Results
The experimental and archaeological samples can
be placed in placed in one of three general raw material
categories. The types observable among cores include,
in order of prevalence, lighter volcanics (approximately
two-thirds to three-quarters of each sample), darker vol-
canics (approximately one-quarter of each sample), and
a small proportion (approximately 9%) of vitreous vol-
canic material in the Gona archaeological sample.
Raw Material Types for Cores: Experimental v. Archaeological
Raw material
lighter
volcanics darker
volcanics vitreous
volcanics Total
Exp or
arch arch Count 15 6 2 23
% within
Exp or arch 65.2% 26.1% 8.7% 100.0%
exp Count 47 17 0 64
% within
Exp or arch 73.4% 26.6% .0% 100.0%
Total Count 62 23 2 87
% within
Exp or arch 71.3% 26.4% 2.3% 100.0%
exparch
80
60
40
20
0
Percent
Vitreous
volcanics
Darker
volcanics
Lighter
volcanics
Raw Materials for Cores: Archaeological and
Experimental
Chi-Square Test: Raw Material for Cores,
Archaeological v. Experimental
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 5.727a2 .057
Likelihood Ratio 5.487 2 .064
N of Valid Cases 87
a. 2 cells (33.3%) have expected count less than 5.
The minimum expected count is .53.
Discussion
Although the two samples show some small differ-
ences in their raw material types, the assemblage differ-
ences are not signifi cant at the .05 level. One difference
is the presence at Gona of two cores in a vitreous volca-
nic material that was absent in the experimental sample.
In other words, the experimental sample (already highly
selected for “fl akability”) is largely similar to the Gona
sample with regard to the major rock types fl aked. As
Stout et al. (2005) have shown, Gona hominins select-
ed higher-quality cobbles compared to the proportions
found in the stream conglomerates, especially the light
trachytes. This pattern is also seen in the sample selected
for experiments. More interesting, however, is the as-
sessment of raw material quality from a knapper’s per-
spective, which will be discussed in the next section.
4. Assemblage: Raw Material Quality
Rationale
The cores from the assemblages were assigned a
quality-of-fl aking value (excellent/good/fair) based on
how fi ne-grained the rock type and how homogeneous
the fracture surface. This was done from a stone-knap-
per’s perspective to examine possible selectivity by early
hominin tool-makers and to examine possible similarities
or differences between the Gona archaeological sample
and the experimental samples (bonobo and human) in
terms of raw material quality. Again, the experimental
sample was highly selected in the fi eld based on size,
shape, and cortex appearance but without any testing of
the cobble (fl aking) or examination of the interior rock
quality.
Results
The archaeological and experimental samples do
not signifi cantly differ at the .05 confi dence level. Nev-
ertheless, there are some differences, for example, the
Gona archaeological material has more “excellent” raw
material.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 171
Raw Material Quality for Cores: Archaeological and Experimental
Raw Material Quality
excellent good fair Total
Exp or
arch arch Count 3 16 4 23
% within
Exp or arch 13.0% 69.6% 17.4% 100.0%
exp Count 2 47 15 64
% within
Exp or arch 3.1% 73.4% 23.4% 100.0%
Total Count 5 63 19 87
% within
Exp or arch 5.7% 72.4% 21.8% 100.0%
Chi-Square Tests: Raw Material Quality for Cores,
Archaeological and Experimental
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 3.214a2 .200
Likelihood Ratio 2.813 2 .245
N of Valid Cases 87
a. 2 cells (33.3%) have expected count less than 5.
The minimum expected count is 1.32.
ExpArch
80.0
60.0
40.0
20.0
0.0
Percent
Fair %
Good %
Excellent %
Raw Material Quality for Cores: Experimental and Archaeological
Discussion
The archaeological and experimental populations
are similar and not signifi cantly different, although there
is a higher proportion of “excellent” raw material ob-
served in the archaeological sample from Gona than in
the experimental samples. This is probably due to the
Gona hominins testing the raw materials at the conglom-
erate sources before transporting cores to the fl oodplain
sites for further reduction (which was deliberately not
done for the experimental sample). It should be noted
that selecting raw materials merely on their exterior
characteristics (size, shape, cortex appearance), as was
done for the experimental sample, yielded a sample with
superior fl aking characteristics overall (over 75% of the
experimental sample assessed to have “good” to “excel-
lent” fl aking quality).
5. Cores: Original Form
Rationale
Cores were examined to identify what the original
(blank) form was. In the early Oldowan, most cores were
made on natural water-rounded river cobbles and this can
readily be observed in the archaeological assemblage.
However, sometimes heavily-reduced cores retain little
or no evidence of their original form, and are assigned to
an indeterminate category. Other, miscellaneous catego-
ries for original core form include fl akes or fl ake frag-
ments, tabular chunks, or broken cobbles.
Results
In all three samples, the dominant original forms for
cores were cobbles. In addition at Gona, six cores were
placed in the indeterminate category, with one more core
made on a fl ake and another on a tabular chunk. The
bonobos also primarily reduced whole cobbles but also
made four simple cores from broken cobble fragments
(broken in half during the knapping process).
Core Original Form by Tool-maker
Original form
cobble
ake,
ake
frag-
ment tabular
chunk indet broken
cobble Total
Tool-maker bonobo Count 29 0 0 0 4 33
% within
Tool-maker 87.9% .0% .0% .0% 12.1% 100.0%
gona Count 15 1 1 6 0 23
% within
Tool-maker 65.2% 4.3% 4.3% 26.1% .0% 100.0%
human Count 30 0 0 0 1 31
% within
Tool-maker 96.8% .0% .0% .0% 3.2% 100.0%
Total Count 74 1 1 6 5 87
% within
Tool-maker 85.1% 1.1% 1.1% 6.9% 5.7% 100.0%
humanGonabonobo
Tool-maker
100.0
80.0
60.0
40.0
20.0
0.0
Percent
Cobble frag
Indet
Tabular chunk
Flake
Cobble
Core Original Form by Tool-maker
172 The Oldowan: Case Studies Into the Earliest Stone Age
As the frequencies of original forms other than cob-
bles were small to absent, comparisons were also made
between the tool-makers looking at cobbles v. “other”
original forms (lumping fl akes, tabular chunks, indeter-
minate, and cobble fragments). Chi-square tests indicate
a difference between the Gona sample and the two ex-
perimental samples, bonobo and human, in the original
forms of their cores (.042 and .002 levels of signifi cance,
respectively). The difference between the two experi-
mental samples, bonobos and humans, was not signifi -
cant.
Core Original Form by Tool-maker (Cobble v. Other)
Original form
lumped
Cobble Other Total
Tool-maker bonobo Count 29 4 33
% within Tool-maker 87.9% 12.1% 100.0%
gona Count 15 8 23
% within Tool-maker 65.2% 34.8% 100.0%
human Count 30 1 31
% within Tool-maker 96.8% 3.2% 100.0%
Total Count 74 13 87
% within Tool-maker 85.1% 14.9% 100.0%
Chi-Square Tests: Core Original Form (Cobble v. Other),
Bonobo v. Gona
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 4.134b1 .042
Continuity Correctiona2.898 1 .089
Likelihood Ratio 4.097 1 .043
Fisher’s Exact Test .054 .045
N of Valid Cases 56
a. Computed only for a 2x2 table.
b. 1 cells (25.0%) have expected count less than 5.
The minimum expected count is 4.93.
Chi-Square Tests: Core Original Form (Cobble v. Other),
Bonobo v. Human
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 1.756b1 .185
Continuity Correctiona.738 1 .390
Likelihood Ratio 1.882 1 .170
Fisher’s Exact Test .356 .198
N of Valid Cases 64
a. Computed only for a 2x2 table.
b. 2 cells (50.0%) have expected count less than 5.
The minimum expected count is 2.42.
Chi-Square Tests: Core Original Form (Cobble v. Other),
Gona v. Human
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 9.467b1 .002
Continuity Correctiona7.331 1 .007
Likelihood Ratio 10.105 1 .001
Fisher’s Exact Test .003 .003
N of Valid Cases 54
a. Computed only for a 2x2 table.
b. 1 cells (25.0%) have expected count less than 5.
The minimum expected count is 3.83.
Discussion
With regard to the original form of cores, the pres-
ence at Gona of a number of cores (n=6) whose origi-
nal blank form was indeterminate constitutes the major
difference between the experimental samples and the
archaeological sample. This proportion of Gona cores
(26.1%) assigned to the “indeterminate” category with
regard to original form strongly suggests that these cores
were more heavily-reduced, obliterating most or all signs
of the original blank type used. Most likely these were
all cobbles, but in the absence of refi tting this cannot be
demonstrated conclusively. That Gona cores overall were
more heavily reduced than the experimental samples will
also become evident in further attributes examined be-
low. Nevertheless, cobbles were by far the predominant
original form in all samples in this study, with the Gona
indeterminate core forms likely representing an interest-
ing indication of higher intensity of core reduction.
6. Cores: Flaking Mode
Rationale
The dominant fl aking mode was recorded to exam-
ine the major patterns of cobble reduction for the Gona
and bonobo samples, thus focusing on the nonhuman
subjects. The major fl aking modes were:
a) unifacial (normal), or fl aking on one face along a
core edge (unidirectional fl aking)
b) bifacial, or fl aking on two faces along a core edge
(bidirectional fl aking)
c) unifacial plus bifacial, or fl aking unifacially along
one core edge and bifacially along another edge
d) unifacial alternate, or fl aking unifacially on one face
along one core edge and then unifacially on the op-
posite face from another edge
Results
Overall, both the bonobo and Gona samples showed
a preponderance of unifacial fl aking of cobbles (from
63.6 to 78.3% respectively), with relatively low inci-
dence of the other fl aking modes. These populations
did not differ signifi cantly in fl aking mode at the .05
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 173
confi dence level [nor was there a signifi cant difference
when all fl aking modes (b) through (d) were combined
as “other,” resulting in a chi-square value of .253 with 0
cells having expected value of less than 5]. The table of
results and chi-square test for all four fl aking modes are
presented below:
Cores: Mode of Flaking by Tool-maker
Mode of aking
unifacial
(normal) bifacial uni -+
bifacial
uni-
facial
alter-
nate Total
Tool-maker bonobo Count 21 2 5 5 33
% within
Tool-maker 63.6% 6.1% 15.2% 15.2% 100.0%
gona Count 18 2 3 0 23
% within
Tool-maker 78.3% 8.7% 13.0% .0% 100.0%
human Count 17 13 1 0 31
% within
Tool-maker 54.8% 41.9% 3.2% .0% 100.0%
Total Count 56 17 9 5 87
% within
Tool-maker 64.4% 19.5% 10.3% 5.7% 100.0%
Chi-Square Tests: Mode of Flaking, Bonobo v. Gona
Val ue df
Asymp.
Sig.
(2-sided)
Pearson Chi-Square 4.075a3 .253
Likelihood Ratio 5.872 3 .118
Linear-by-Linear
Association 2.789 1 .095
N of Valid Cases 56
a. 6 cells (75.0%) have expected count less than 5.
The minimum expected count is 1.64.
Gonabonobo
Tool-maker
80.0
60.0
40.0
20.0
0.0
Percent
Unifacial
alternate
Uni+Bifacial
Bifacial
Unifacial
Cores, Mode of Flaking: Bonobo v. Gona
Discussion
Both the bonobo and Gona samples have a prepon-
derance of unifacial fl aking of cobbles. It can be argued
that this is the simplest, easiest approach to reducing
cobbles and producing sharp fl akes. This tendency to-
wards unifacial fl aking could also have been a habitual
norm among the hominins who produced these EG sites
at Gona, or, at the very least, their tendency at these par-
ticular sites but perhaps variable at other sites and other
times.
7. Cores: Invasiveness.
Rationale
This attribute is an assessment of how invasive fl ake
scars are relative to core morphology (heavily, moder-
ate, light). This assessment is a qualitative one, based on
the distance fl ake scars have traveled across the core sur-
faces. Heavy invasiveness suggests that tool-makers are
either driving larger fl akes off of cores or heavily reduc-
ing cores, or both.
Results
The bonobo cores are characterized by higher per-
centages of light and moderate invasiveness, and low
percentage of heavy invasiveness. The Gona cores have
a preponderance of heavy invasiveness. Differences be-
tween the bonobo, Gona, and human samples were sig-
nifi cant at the .05 level.
Invasiveness of Flaking by Tool-maker
Degree of invasiveness
light,
shallow moderate heavy,
invasive Total
Tool-maker bonobo Count 4 27 2 33
% within
Tool-maker 12.1% 81.8% 6.1% 100.0%
gona Count 0 7 16 23
% within
Tool-maker .0% 30.4% 69.6% 100.0%
human Count 1 19 11 31
% within
Tool-maker 3.2% 61.3% 35.5% 100.0%
Total Count 5 53 29 87
% within
Tool-maker 5.7% 60.9% 33.3% 100.0%
174 The Oldowan: Case Studies Into the Earliest Stone Age
humanGonabonobo
Toolmaker
100.0
80.0
60.0
40.0
20.0
0.0
Percent
Heavy
Moderate
Light
Invasiveness of Flaking by Tool-maker
Chi-Square Tests: Degree of Invasiveness, Bonobo v. Gona
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 25.687a2 .000
Likelihood Ratio 28.705 2 .000
N of Valid Cases 56
a. 2 cells (33.3%) have expected count less than 5.
The minimum expected count is 1.64.
Chi-Square Tests: Degree of Invasiveness, Bonobo v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 9.369a2 .009
Likelihood Ratio 10.123 2 .006
N of Valid Cases 64
a. 2 cells (33.3%) have expected count less than 5.
The minimum expected count is 2.42.
Chi-Square Tests: Degree of Invasiveness, Gona v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 6.420a2 .040
Likelihood Ratio 6.882 2 .032
N of Valid Cases 54
a. 2 cells (33.3%) have expected count less than 5.
The minimum expected count is .43.
Discussion
The bonobo cores have a low and moderate invasive-
ness because they are lightly reduced with small fl akes
being detached for the most part, whereas the Gona cores
have a heavy invasiveness, indicating that they have been
heavily and effectively reduced with larger fl akes being
detached overall. As mentioned previously, the experi-
mental human sample intentionally tried to reduce most
cores by about 50% of their original cobble mass in order
to provide a controlled sample for comparisons with the
bonobo and Gona samples. In this attribute, the human
sample is intermediate between the more lightly fl aked
bonobo cores and the more heavily fl aked Gona cores.
8. Cores: Percentage of
Circumference Flaked
Rationale
The percentage of cobble circumference fl aked is
yet another criterion that can be used to determine how
extensively an Oldowan core has been exploited. Al-
though a very long, roller-shaped cobble can be exten-
sively reduced and still retain a relatively low percentage
of circumference fl aked, most of the cobbles available to
Gona hominins tend to fall into the disc and sphere cat-
egories of geological clast shape, so that their percentage
of circumference fl aked is reasonably indicative of the
extent of fl aking.
Results
The bonobo cores showed the lowest percentage of
circumference fl aked, with the human sample intermedi-
ate and the Gona sample having a very high percentage
of circumference fl aked. The Gona cores were fl aked
along nearly twice as much of the cobble circumference
overall than the bonobo cores. The differences between
the bonobos and both the Gona and human core samples
were signifi cant at the .05 level, but the difference be-
tween the Gona and human samples was not signifi cant
at this level.
Extent of Core Reduction by Tool-maker
Extent of core reduction
Tool-maker Mean N Std. Deviation
bonobo 30.91 33 9.475
gona 58.67 23 23.799
human 45.81 31 9.924
Total 43.56 87 18.362
humanGonabonobo
toolmaker
60.0
50.0
40.0
30.0
20.0
10.0
0.0
% Circumference
Percentage Circumference by Tool-maker
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 175
Test Statistics: Extent of Core
Reduction, Bonobo v. Gonaa
Extent of core
reduction
Mann-Whitney U 96.000
Wilcoxon W 657.000
Z -4.855
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Extent of Core
Reduction, Bonobo v. Humana
Extent of core
reduction
Mann-Whitney U 152.000
Wilcoxon W 713.000
Z -4.982
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Extent of Core
Reduction, Gona v. Humana
Extent of core
reduction
Mann-Whitney U 263.000
Wilcoxon W 759.000
Z -1.684
Asymp. Sig. (2-tailed) .092
a. Grouping Variable: Toolmkrcod
Discussion
The bonobos relatively low percentage of circum-
ference fl aked (~31%) is an indication of minimal reduc-
tion of cobbles, while the high percentage of circumfer-
ence fl aked in the archaeological sample (~59%) is an
indication of much heavier reduction of cobble cores by
the Gona hominin tool-makers. Again, the experimen-
tal human sample, which intentionally reduced most
cores by about half their original mass, was intermediate
(~46%).
9. Cores: Quality of Flaking
Rationale
From a fl intknapper’s perspective, cores were ranked
by the quality of fl aking. Higher quality of fl aking denot-
ed well-reduced cores, clean fl ake detachments, and crisp
non-battered edges. Although a very subjective category,
experience has shown us that this is nonetheless a useful
category in assessing different levels of skill. Such cores
are typically illustrated in Oldowan site reports. Quality
categories were excellent, good, fair, and poor.
Results
The bonobo core sample was characterized by a
modal value of fair-quality fl aking, the Gona sample by
good-quality fl aking, and the human sample by excel-
lent-quality fl aking. Using all four quality categories,
statistical testing showed differences between the human
sample and the other two samples but not between the
bonobo and Gona samples at the .05 level of signifi cance,
although 4 cells had expected counts less than 5. Testing
after combining categories (good to excellent as “high”
and poor to fair as “low”), on the other hand, points to
a signifi cant difference between the bonobo sample and
the other two samples in this attribute, while the Gona
and human samples show a much higher and relatively
equivalent quality of fl aking.
Quality of Flaking by Tool-maker
Quality of aking (skill)
excellent good fair poor Total
Tool-maker bonobo Count 1 15 16 1 33
% within
Tool-maker 3.0% 45.5% 48.5% 3.0% 100.0%
gona Count 1 18 4 0 23
% within
Tool-maker 4.3% 78.3% 17.4% .0% 100.0%
human Count 14 12 5 0 31
% within
Tool-maker 45.2% 38.7% 16.1% .0% 100.0%
Total Count 16 45 25 1 87
% within
Tool-maker 18.4% 51.7% 28.7% 1.1% 100.0%
humanGonabonobo
80.0
60.0
40.0
20.0
0.0
Percent
Poor
Fair
Good
Excellent
Quality of Flaking by Tool-maker
176 The Oldowan: Case Studies Into the Earliest Stone Age
Chi-Square Tests: Quality of Flaking, Bonobo v. Gona,
4 categories
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 6.907a3 .075
Likelihood Ratio 7.574 3 .056
N of Valid Cases 56
a. 4 cells (50.0%) have expected count less than 5.
The minimum expected count is .41.
Chi-Square Tests: Quality of Flaking, Bonobo v. Human,
4 categories
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 18.317a3 .000
Likelihood Ratio 21.164 3 .000
N of Valid Cases 64
a. 2 cells (25.0%) have expected count less than 5.
The minimum expected count is .48.
Chi-Square Tests: Quality of Flaking, Gona v. Human,
4 categories
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 11.648a2 .003
Likelihood Ratio 13.576 2 .001
N of Valid Cases 54
a. 1 cells (16.7%) have expected count less than 5.
The minimum expected count is 3.83.
Quality of Flaking (2 categories) by Tool-maker
Quality of aking
(2 categories)
excellent
or good fair or
poor Total
Tool-maker bonobo Count 16 17 33
% within
Tool-maker 48.5% 51.5% 100.0%
gona Count 19 4 23
% within
Tool-maker 82.6% 17.4% 100.0%
human Count 26 5 31
% within
Tool-maker 83.9% 16.1% 100.0%
Total Count 61 26 87
% within
Tool-maker 70.1% 29.9% 100.0%
Chi-Square Tests: Quality of Flaking, Bonobo v. Gona,
2 categories
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 6.734b1 .009
Continuity Correctiona5.357 1 .021
Likelihood Ratio 7.124 1 .008
Fisher’s Exact Test .012 .009
N of Valid Cases 56
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 8.63.
Chi-Square Tests: Quality of Flaking, Bonobo v. Human,
2 categories
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 8.873b1 .003
Continuity Correctiona7.373 1 .007
Likelihood Ratio 9.258 1 .002
Fisher’s Exact Test .004 .003
N of Valid Cases 64
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 10.66.
Chi-Square Tests: Quality of Flaking, Gona v. Human,
2 categories
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square .015b1 .902
Continuity Correctiona.000 1 1.000
Likelihood Ratio .015 1 .902
Fisher’s Exact Test 1.000 .592
N of Valid Cases 54
a. Computed only for a 2x2 table.
b. 1 cells (25.0%) have expected count less than 5.
The minimum expected count is 3.83.
Discussion
Examination of the distribution of fl aking quality
among the samples shows, in fact, a gradient among
the three samples. The bonobo sample, characterized
by cores that showing primarily fair (~49%) and good
(~46%) fl aking quality, showed the lowest overall level
of skill of the three samples. The Gona sample showed a
preponderance of good fl aking (~78%), and the human
sample was characterized by cores showing excellent
(~45) and good (~39%) fl aking quality. Thus, in terms of
overall quality of fl aking, the Gona sample was interme-
diate between the bonobo and human samples.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 177
10. Cores: Edge Battering
Rationale
Core edges exhibiting heavy battering (hammer-
stone percussion marks, crushing, and small-scale step
aking) may be an indication of less skilled fl aking and
numerous unsuccessful attempts to remove fl akes with a
percussor. This criterion was developed after examina-
tion of the bonobo cores appeared to show much more
edge battering than had been observed on archaeologi-
cal cores or on experimental cores produced by humans.
Edge battering was divided into four categories: none,
low, moderate, and high.
Results
The bonobo sample is characterized by an apprecia-
ble percentage of cores (~33%) showing high levels of
edge battering from hammerstone percussion. The mod-
al value for Gona cores was a low degree of edge batter-
ing (~44%), but appreciable quantities of also showed
moderate battering (~26%). The great majority of cores
in the human sample (~84%) exhibited no edge batter-
ing. Interestingly, none of the cores from the Gona and
human samples showed a high degree of edge battering.
Each of the three tool-making populations was signifi -
cantly different from the other two (at the .05 confi dence
level) in terms of core edge battering.
Core Edge Battering by Tool-maker
Edge battering, microstepping,
crushing
none low mod-
erate high Total
Tool-maker bonobo Count 10 6 6 11 33
% within
Tool-maker 30.3% 18.2% 18.2% 33.3% 100.0%
gona Count 7 10 6 0 23
% within
Tool-maker 30.4% 43.5% 26.1% .0% 100.0%
human Count 2641031
% within
Tool-maker 83.9% 12.9% 3.2% .0% 100.0%
Total Count 43 20 13 11 87
% within
Tool-maker 49.4% 23.0% 14.9% 12.6% 100.0%
humanGonabonobo
Tool-maker
100.0
80.0
60.0
40.0
20.0
0.0
Percent
High
Moderate
Low
None
Edge Battering by Tool-maker
Chi-Square Tests: Core Edge Battering, Bonobo v. Gona
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 11.098a3 .011
Likelihood Ratio 14.997 3 .002
N of Valid Cases 56
a. 2 cells (25.0%) have expected count less than 5.
The minimum expected count is 4.52.
Chi-Square Tests: Core Edge Battering, Bonobo v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 22.042a3 .000
Likelihood Ratio 26.918 3 .000
N of Valid Cases 64
a. 3 cells (37.5%) have expected count less than 5.
The minimum expected count is 3.39.
Chi-Square Tests: Core Edge Battering, Gona v. Human
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 16.254a2 .000
Likelihood Ratio 17.071 2 .000
N of Valid Cases 54
a. 2 cells (33.3%) have expected count less than 5.
The minimum expected count is 2.98.
Discussion
Over half of the bonobo cores (~52%) showed either
a moderate or high degree of edge battering. This indi-
cates high numbers of unsuccessful hammerstone blows
by the bonobos in attempts to remove usable fl akes and
a signifi cantly lower level of skill than the Gona and hu-
man samples. Also, ~49% of bonobo cores showed little
or no edge battering compared to ~74 % for Gona and
178 The Oldowan: Case Studies Into the Earliest Stone Age
almost all of the human cores (~97%). In terms of the
criterion of edge battering, then, the Gona hominins are
intermediate in skill level relative to the bonobos and hu-
mans.
11. Cores: Modi ed Leakey Typology
Rationale
Mary Leakey’s (1971) typology of Oldowan cores
is one of the most widely used systems, especially if one
views these forms as core morphologies rather than func-
tional classes. Her major categories of cores made on
cobbles (her “heavy-duty tools”) are choppers, discoids,
heavy-duty scrapers, proto-bifaces, polyhedrons, dis-
coids, heavy-duty scrapers, and spheroids. In this study,
there were no polyhedrons, protobifaces, or spheroids in
any of the samples. We added one new category, “casual
core,” for very minimally-fl aked cobbles (probably clos-
est to Mary Leakey’s “modifi ed and battered nodules and
blocks” category). In addition, we decided to separate
choppers into their two major categories, side choppers
and end choppers, based upon whether the fl aked edge
was along the end or side of the cobble.
Results
The bonobo cores were characterized by an abun-
dance of end choppers (~64%), in comparison with the
Gona and human cores, which had predominantly side
choppers (~65% and 71% respectively). The bonobo
sample was signifi cantly different from the other two
samples in the numbers of end v. side choppers, while
the Gona and human samples did not differ signifi cantly
at the .05 level. In view of the high proportion of chop-
pers in each of the three tool-making samples (bonobos
82%, Gona ~74%, and humans 100%), they would all
be assigned to the “Oldowan Industry” in Leakey’s clas-
sifi cation system (1971). The bonobo sample also has a
number of cores assigned to the “casual core” category,
and the Gona sample has small numbers of other core
types (discoids, end choppers, heavy-duty scrapers, and
one casual core).
Leakey Core Type by Tool-maker
Mary Leakey core type
Side
chopper End
chopper Discoid
Heavy-
duty
scraper Casual
core Total
Tool-
maker bonobo Count 6 2200533
% within
Tool-maker 18.2% 66.7% .0% .0% 15.2% 100.0%
gona Count 15232123
% within
Tool-maker 65.2% 8.7% 13.0% 8.7% 4.3% 100.0%
human Count 22900031
% within
Tool-maker 71.0% 29.0% .0% .0% .0% 100.0%
Total Count 43 3332687
% within
Tool-maker 49.4% 37.9% 3.4% 2.3% 6.9% 100.0%
Chi-Square Tests: Leakey Core Types, All Tool-makers
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 43.131a8 .000
Likelihood Ratio 46.596 8 .000
N of Valid Cases 87
a. 9 cells (60.0%) have expected count less than 5.
The minimum expected count is .53.
Chi-Square Tests: End and Side Choppers, Bonobo v. Gona
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 18.968b1 .000
Continuity Correctiona16.379 1 .000
Likelihood Ratio 20.771 1 .000
Fisher’s Exact Test .000 .000
N of Valid Cases 45
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 7.93.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 179
Chi-Square Tests: End and Side Choppers, Bonobo v. Human
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 14.479b1 .000
Continuity Correctiona12.561 1 .000
Likelihood Ratio 15.191 1 .000
Fisher’s Exact Test .000 .000
N of Valid Cases 59
a. Computed only for a 2x2 table.
b. 0 cells (.0%) have expected count less than 5.
The minimum expected count is 13.29.
Chi-Square Tests: End and Side Choppers, Gona v. Human
Value df
Asymp.
Sig.
(2-sided)
Exact
Sig.
(2-sided)
Exact
Sig.
(1-sided)
Pearson Chi-Square 1.853b1 .173
Continuity Correctiona1.005 1 .316
Likelihood Ratio 2.007 1 .157
Fisher’s Exact Test .284 .158
N of Valid Cases 48
a. Computed only for a 2x2 table.
b. 1 cells (25.0%) have expected count less than 5.
The minimum expected count is 3.90.
Discussion
Although all three samples are dominated by core
forms that would be designated as choppers in Mary
Leakey’s classifi cation, the bonobo assemblage stands
out in its high proportion of end rather than side chop-
pers. This is a result of bonobos reducing a cobble sig-
nifi cantly less, resulting in a preponderance of end chop-
pers rather than side choppers. This is in contrast to the
cores produced by the humans and Gona tool-makers,
which tend to be more heavily reduced; many of these
cores almost certainly started as end choppers, but with
further, more effi cient reduction, they were transformed
into side choppers (in the human sample, in which re-
duction was held to approximately 50% of the original
cobble, as well as in the Gona archaeological sample).
12. CORES: WEIGHT
Rationale
Core weights in a given raw material can help show
differences between assemblages that may be due to
differences in the weights of the original cobble forms
and/or differences in the amount of core reduction. For
both of the experimental samples, an equivalent range
of sizes and shapes of pre-selected cobbles (discussed
above) was made available to the subjects.
Results
The bonobos produced the heaviest cores, averag-
ing about 687 g, with a much larger standard deviation
than the other two samples as well. The Gona cores were
the smallest (~226 g average), with the modern human
sample intermediate (~385 g). Each tool-maker sample
differed signifi cantly from the other two in terms of fi nal
core weight at the .000 level.
Core Weight by Tool-maker
Weight in grams
Tool-maker Mean N Std. Deviation
bonobo 686.82 33 328.682
gona 226.22 23 111.925
human 384.55 31 117.893
Total 457.34 87 290.953
humangonabonobo
Tool-maker
700
600
500
400
300
200
100
0
Mean Weight in grams
Cores: Mean Weight by Tool-maker
Test Statistics: Core Weight,
Bonobo v. Gonaa
Weight in grams
Mann-Whitney U 35.000
Wilcoxon W 311.000
Z -5.738
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Weight,
Bonobo v. Humana
Weight in grams
Mann-Whitney U 179.500
Wilcoxon W 675.500
Z -4.460
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
180 The Oldowan: Case Studies Into the Earliest Stone Age
Test Statistics: Core Weight,
Gona v. Humana
Weight in grams
Mann-Whitney U 105.000
Wilcoxon W 381.000
Z -4.400
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Discussion
The bonobo cores are markedly heavier than the
Gona or human samples. This is due to the fact that bono-
bos almost invariably chose larger cobbles for fl aking,
since their hands are large with long fi ngers and a very
short thumb and are less dexterous than those of mod-
ern humans (and also, presumably, than those of Gona
hominins as well), and also because the bonobo cobble
cores were much less reduced than the Gona or human
samples. The human cores are intermediate in weight
between the bonobo and Gona cores, but, on average,
are closer to the mean weight of Gona cores. A number
of other attributes examined (above and below) suggest
that the human cores are not as extensively reduced as
the Gona cores; if they were, the mean weights would
likely be very similar.
13. Cores: Length (Maximum Dimension)
Rationale
This measurement gives an indication of the size of
cores based on the largest linear measurement. It also
shows the minimum size of cobble blanks selected by
early hominins. This measurement is also used to calcu-
late core breadth/length ratios, as well as the ratio of the
largest fl ake scar/core maximum dimension (discussed
below).
Results
As with weight, the bonobo core maximum dimen-
sions were the largest (mean ~116 mm), with humans
intermediate (~93 mm), and the Gona cores the small-
est (~81 mm); the bonobo cores also showed a much
higher standard deviation. Each tool-making population
produced cores whose mean maximum dimension dif-
fered signifi cantly (at the .000 level) from the other two
samples.
Cores: Maximum Dimension by Tool-maker
Length
Tool-maker Mean N Std. Deviation
bonobo 115.85 33 18.171
gona 81.09 23 10.816
human 93.45 31 11.331
Total 98.68 87 20.083
humangonabonobo
Tool-maker
120
100
80
60
40
20
0
Mean Length
Cores: Mean Length by Tool-maker
Test Statistics: Core Maximum
Dimension, Bonobo v. Gonaa
Length
Mann-Whitney U 39.000
Wilcoxon W 315.000
Z -5.673
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Maximum
Dimension, Gona v. Humana
Length
Mann-Whitney U 138.000
Wilcoxon W 634.000
Z -5.019
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Maximum
Dimension, Gona v. Humana
Length
Mann-Whitney U 147.000
Wilcoxon W 423.000
Z -3.668
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Discussion
As with weight, the differences seen in core maxi-
mum dimension in the three tool-making samples can
be explained. As the bonobos tended to choose larger
cobbles (because of their hand morphology) and also re-
duced the cobble cores less extensively, their resultant
cores tended to be larger. The human core sample was
closer in size to the Gona cores. As previously discussed,
the human cores are somewhat less reduced overall than
the archaeological cores and thus somewhat larger.
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 181
14. Cores: Breadth
Rationale
Breadth was defi ned as the dimension of the core
measured at a right angle to the length or maximum di-
mension. This measurement is also useful in determining
breadth/length ratios and thickness/breadth ratios.
Results
The mean breadth of cores in the three samples dif-
fered greatly. As with core length, the bonobo core sam-
ple showed the greatest mean breadth (~90 mm), with
the human sample intermediate (~75 mm) and the Gona
sample the smallest (~61 mm). Each population differed
from the other two at the .000 level of signifi cance.
Cores: Breadth by Tool-maker
Breadth
Tool-maker Mean N Std. Deviation
bonobo 90.06 33 16.948
gona 60.52 23 9.375
human 74.94 31 8.675
Total 76.86 87 17.182
Test Statistics: Core Breadth,
Bonobo v. Gonaa
Length
Mann-Whitney U 37.500
Wilcoxon W 313.500
Z -5.698
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Breadth,
Bonobo v. Humana
Length
Mann-Whitney U 191.000
Wilcoxon W 687.000
Z -4.309
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Breadth,
Gona v. Humana
Length
Mann-Whitney U 89.500
Wilcoxon W 365.500
Z -4.675
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Discussion
As with core length (maximum dimension), the
larger core breadth seen in the bonobo sample seems to
be primarily a function of their selecting larger cobbles
for fl aking. The greater breadth seen in the human cores
relative to the Gona sample is probably due to the fact
that the human cores are not as extensively reduced as
the Gona cores; with further reduction of especially side
choppers, breadth measurements would probably de-
crease accordingly.
15. Cores: Thickness
Rationale
Thickness was measured at a right angle to breadth,
and represents the smallest of the three length/breadth/
thickness measurements. Interestingly, this measure-
ment on Oldowan cobble cores probably often represents
a reasonable estimate of the thickness of the original
cobble blank that was fl aked.
Results
The Gona cores were the thinnest (~44 mm), while
the bonobo and human samples were almost identical in
terms of absolute thickness (~57 mm and ~56 mm re-
spectively). Statistical tests indicated that the Gona sam-
ple differed from the bonobo and human samples (at the
.001 and .000 confi dence levels, respectively), while the
bonobo and human samples were not signifi cantly differ-
ent (.824 signifi cance level).
Cores: Thickness by Tool-maker
Thickness
Tool-maker Mean N Std. Deviation
bonobo 56.76 33 14.908
gona 43.87 23 11.022
human 56.39 31 9.222
Total 53.22 87 13.238
humangonabonobo
Tool-maker
60
50
40
30
20
10
0
mm
Cores: Mean Thickness
182 The Oldowan: Case Studies Into the Earliest Stone Age
Test Statistics: Core Thickness,
Bonobo v. Gonaa
Length
Mann-Whitney U 184.000
Wilcoxon W 460.000
Z -3.257
Asymp. Sig. (2-tailed) .001
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Thickness,
Bonobo v. Humana
Length
Mann-Whitney U 495.000
Wilcoxon W 991.000
Z -.222
Asymp. Sig. (2-tailed) .824
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Thickness,
Gona v. Humana
Length
Mann-Whitney U 113.000
Wilcoxon W 389.000
Z -4.264
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Discussion
Core thicknesses on the Gona sample suggest that
the early hominin tool-makers were selecting thinner
cobbles than the experimental samples. This may have
been done in part because thinner cobbles tend to be eas-
ier to fl ake, with thinner edges more amenable to fl ake
detachment with a hammerstone. It should be noted
that, although the bonobo cores had a mean thickness
equivalent to the human cores, their absolute length was
considerably greater. These point to an overall size and
shape difference in the bonobo cores, a pattern that will
be explored further below.
16. Cores: Ratio Breadth/Length
Rationale
This ratio is an indication of how relatively elongat-
ed a core is with regard to breadth. A low ratio indicates
a more elongated core, whereas a higher ratio indicates a
more equilateral core.
Results
Although all three populations have roughly similar
breadth/length mean ratios (from about 75 to 81%), the
difference between the means of the Gona sample (with
the smallest ratio) and human sample (with the largest
ratio) was signifi cant at the .05 confi dence level. There
were no signifi cant differences between the bonobo sam-
ple (with an intermediate breadth/length value) and the
other two samples.
Cores: Breadth to Length Ratio
b_by_l
Tool-maker Mean N Std. Deviation
bonobo .7826 33 .11858
gona .7493 23 .08815
human .8077 31 .09408
Total .7827 87 .10407
Test Statistics: Breadth to Length Ratio,
Bonobo v. Gonaa
b_by_l
Mann-Whitney U 298.000
Wilcoxon W 574.000
Z -1.357
Asymp. Sig. (2-tailed) .175
a. Grouping Variable: Toolmkrcod
Test Statistics: Breadth to Length Ratio,
Bonobo v. Humana
b_by_l
Mann-Whitney U 448.000
Wilcoxon W 1009.000
Z -.853
Asymp. Sig. (2-tailed) .394
a. Grouping Variable: Toolmkrcod
Test Statistics: Breadth to Length Ratio,
Gona v. Humana
b_by_l
Mann-Whitney U 229.000
Wilcoxon W 505.000
Z -2.230
Asymp. Sig. (2-tailed) .026
a. Grouping Variable: Toolmkrcod
Discussion
This attribute is actually somewhat problematic as a
point of comparison between different assemblages: this
ratio can change dramatically as the axes of length and
breadth change dimension in the process of core reduc-
tion, and can even reverse direction of change when con-
tinued fl aking transforms the “length” into the “breadth.”
It should be noted that although the bonobo mean was
not signifi cantly different than the Gona and human sam-
ples, the bonobo sample was composed mainly of end
choppers, with the worked edge roughly perpendicular
to the long axis of the core, while the Gona and human
samples were composed largely of side choppers, with
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 183
the worked edge roughly parallel to the long axis of the
core. As a hypothetical case, the reduction of an elongate
cobble could give the following progression of breadth/
length ratios and core typology as fl aking proceeded:
.74 (end chopper), .84 (end chopper), .95 (end chopper),
.96 (side chopper) [after the worked “end” becomes the
worked “side”], .79 (side chopper), .60 (side chopper),
.51 (an extremely reduced side chopper). Thus, an end
chopper can progress upward in breadth/length ratio un-
til it is equidimensional, and then further working of the
same edge as a side chopper will continue to reduce this
ratio.
As the Gona and human samples were dominated by
side choppers, continued fl aking of their primary edges
would tend to decrease the breadth/length ratio (as fl ake
removals remove additional material from the breadth
of the core). Thus, the fact that the Gona cores have a
somewhat smaller mean breadth/length ratio than the hu-
man cores may refl ect the more intensive reduction of
the archaeological cores (with further reduction of the
human side choppers tending to reduce their mean and
moving it further toward the Gona value). Reduction of
the bonobo end choppers, on the other hand, is acting to
increase their breadth/length ratio to artifi cially place it
within the human-Gona range; further fl aking along the
cobble end would tend to increase it further (until the
core is equidimensional).
17. Cores: Modi ed Breadth/Length Ratio
Rationale
Perhaps a better indicator of core morphology would
be to defi ne length as the dimension perpendicular to the
aked edge and breadth as the dimension parallel to the
aked edge. This attribute would then discriminate fully
between side choppers and end choppers. By doing this,
the modifi ed breadth:length ratio of end choppers would
always be less the 1.0, while the modifi ed breadth:length
ratio of side choppers would be greater than 1.0. This
new ratio was applied to the choppers from the three
samples.
Results
The bonobo cores, with a preponderance of end
choppers, had the lowest mean modifi ed breadth/length
ratio (~0.92), and the Gona cores, with a preponderance
of side choppers, the highest (~1.27). The human cores
were intermediate at ~1.16. Statistical tests showed a
signifi cant difference between the bonobo sample and
both the Gona and human samples (at the .000 signifi -
cance level), but not between the Gona and human core
samples at the .05 level of signifi cance.
Cores: Modi ed Breadth by Length
Mod breadth by length
Tool-maker Mean N Std. Deviation
bonobo .9211 27 .26246
gona 1.2745 17 .26293
human 1.1554 31 .24763
Total 1.0981 75 .28977
humangonabonobo
Tool-maker
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
Modified Breadth/Length
Cores: Modified Breadth/Length by
Tool-maker
Test Statistics: Modi ed Breadth by
Length, Bonobo v. Gonaa
Mod breadth
by length
Mann-Whitney U 80.000
Wilcoxon W 458.000
Z -3.603
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Modi ed Breadth by
Length, Bonobo v. Humana
Mod breadth
by length
Mann-Whitney U 194.500
Wilcoxon W 572.500
Z -3.492
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Modi ed Breadth by
Length, Gona v. Humana
Mod breadth
by length
Mann-Whitney U 182.000
Wilcoxon W 678.000
Z -1.757
Asymp. Sig. (2-tailed) .079
a. Grouping Variable: Toolmkrcod
184 The Oldowan: Case Studies Into the Earliest Stone Age
Discussion
These results show a likely continuum in the reduc-
tion of somewhat elongated cobbles from end choppers
to side choppers. The bonobo sample, whose cores are
characterized by less reduction, show a relatively low
ratio indicating a dominance of end choppers, while the
Gona sample and, to a lesser extent, the human sample,
have higher ratios, in line with being more heavily re-
duced and having a dominance of side choppers. Thus,
this modifi ed breadth/length ratio more accurately re-
ects technological differences in the core assemblages
in the samples than does the simple breadth/length ratio.
18. Cores: Ratio Thickness/Breadth
Rationale
This ratio was analyzed to help quantify the rela-
tive shape of core forms made on cobbles. As with the
breadth/length ratio, this ratio treats a core as a geological
clast irregardless of the orientation of the fl aked edge.
Results
The bonobo core sample has the lowest thick-
ness/breadth ratio (~.64), the human sample the highest
(~.76), and the Gona sample an intermediate ratio (~.73).
Statistical tests showed that only the bonobo and human
samples differed at the .05 confi dence level.
Cores: Thickness to Breadth Ratio by Tool-maker
th_by_b
Tool-maker Mean N Std. Deviation
bonobo .6439 33 .18172
gona .7283 23 .15469
human .7559 31 .10749
Total .7061 87 .15796
Test Statistics: Thickness to Breadth
Ratio, Bonobo v. Gonaa
th_by_b
Mann-Whitney U 280.500
Wilcoxon W 841.500
Z -1.649
Asymp. Sig. (2-tailed) .099
a. Grouping Variable: Toolmkrcod
Test Statistics: Thickness to Breadth
Ratio, Bonobo v. Humana
th_by_b
Mann-Whitney U 310.000
Wilcoxon W 871.000
Z -2.707
Asymp. Sig. (2-tailed) .007
a. Grouping Variable: Toolmkrcod
Test Statistics: Thickness to Breadth
Ratio, Gona v. Humana
th_by_b
Mann-Whitney U 316.500
Wilcoxon W 592.500
Z -.700
Asymp. Sig. (2-tailed) .484
a. Grouping Variable: Toolmkrcod
Discussion
The low ratio for bonobo cores relative to the human
cores is probably a function of the bonobos using larger
cobbles with larger absolute breadth and length than the
human sample, although with similar thicknesses. The
bonobos are also tending to make more end choppers
and thus are tending to reduce the length, rather than the
breadth, of these larger clasts as reduction proceeds.
19. Cores: Maximum Dimension
Largest Scar
Rationale
The maximum dimension of the largest scar was
measured; this was an absolute measurement, without
consideration as to whether the scar was truncated by
later fl aking. This measurement can give an indication of
the size of fl akes being detached, although as reduction
proceeds, it is likely that some large fl ake scars are trun-
cated by later scars and thus their maximum dimension
reduced.
Results
The bonobo fl ake scars had the highest mean (~62
mm), with human cores intermediate (~58 mm), and the
Gona sample showing the lowest mean (~46 mm). Sta-
tistical tests showed that the means of the Gona fl ake
scars differed signifi cantly from both the other samples,
while the human and bonobo samples showed very little
difference.
Cores: Mean Length Largest Scar by Tool-maker
Length of largest scar
Tool-maker Mean N Std. Deviation
bonobo 61.73 33 23.018
gona 45.91 23 10.352
human 57.84 31 13.902
Total 56.16 87 18.248
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 185
Test Statistics: Core Mean Length
Largest Scar, Bonobo v. Gonaa
Length of
largest scar
Mann-Whitney U 204.500
Wilcoxon W 480.500
Z -2.918
Asymp. Sig. (2-tailed) .004
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Mean Length
Largest Scar, Bonobo v. Humana
Length of
largest scar
Mann-Whitney U 505.500
Wilcoxon W 1066.500
Z -.081
Asymp. Sig. (2-tailed) .936
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Mean Length
Largest Scar, Gona v. Humana
Length of
largest scar
Mann-Whitney U 167.500
Wilcoxon W 443.500
Z -3.309
Asymp. Sig. (2-tailed) .001
a. Grouping Variable: Toolmkrcod
Discussion
The mean largest fl ake scar dimension for cores ap-
pears in part to be a function of the size of the cobble
that was chosen as the blank and the size of the largest
akes removed, in combination with the amount of trun-
cation of these scars by subsequent fl aking (i.e., intensity
of reduction). This latter factor, the truncation and even
removal of earlier fl ake scars by later core reduction, ap-
pears to be a major factor reducing the mean of this at-
tribute in the human core sample, as the largest fl akes
removed in the human sample for each core (77.5 mm)
averaged approximately 15 mm larger than the largest
bonobo fl akes per core (62.2 mm). Thus, the dimensions
of fl ake scars on the bonobo cores tend to be representa-
tive of the largest fl akes removed (both approximately
62 mm), largely because their cores are less intensively
reduced and thus fl ake scars tend to be more complete.
The human sample had largest mean scar value
(57.8 mm), an underestimate for the largest fl ake from
each core, which averaged approximately 77.5 mm. The
relatively smaller size of fl ake scars on the Gona cores
is likely due to the smaller core size and their more in-
tensive reduction, which would have tended to truncate
and reduce the size of fl ake scars on the cores. For more
heavily reduced cores, then, the absolute size of the larg-
est fl ake scar on a core has more limited utility in pre-
dicting dimensions of the fl ake populations removed.
20. Cores: Ratio of Largest Scar/Core
Maximum Dimension
Rationale
This ratio is used to arrive at a quantifi able assess-
ment of the invasiveness of the fl aking. A ratio that is ap-
proaches 1.0 indicates that the largest fl ake scar is almost
the same length as the maximum dimension of the core,
in other words, denoting a very invasively fl aked core.
Results
The bonobo cores were the least invasively fl aked by
this measure (largest scar-to-length ratio of approximate-
ly .53); the human sample was the most invasive (.62),
with the Gona sample intermediate (.57). Statistical test-
ing indicated that only the groups with the smallest and
largest means, i.e. the bonobo and human samples, dif-
fered at the .05 level of signifi cance.
Cores: Ratio of Largest Scar to Length by Tool-maker
lscar_l
Tool-maker Mean N Std. Deviation
bonobo .5278 33 .15330
gona .5708 23 .12649
human .6195 31 .13539
Total .5718 87 .14420
Test Statistics: Core Ratio Largest
Scar to Length, Bonobo v. Gonaa
lscar_l
Mann-Whitney U 286.500
Wilcoxon W 847.500
Z -1.549
Asymp. Sig. (2-tailed) .121
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Ratio Largest
Scar to Length, Bonobo v. Humana
lscar_l
Mann-Whitney U 287.000
Wilcoxon W 848.000
Z -3.016
Asymp. Sig. (2-tailed) .003
a. Grouping Variable: Toolmkrcod
186 The Oldowan: Case Studies Into the Earliest Stone Age
Test Statistics: Core Ratio Largest
Scar to Length, Gona v. Humana
lscar_l
Mann-Whitney U 262.000
Wilcoxon W 538.000
Z -1.653
Asymp. Sig. (2-tailed) .098
a. Grouping Variable: Toolmkrcod
Discussion
Although the size of the bonobo cores was the larg-
est of the three samples, they were the least invasively
aked as assessed by this attribute. Even though the
human cores were somewhat larger than Gona cores in
terms of maximum dimension, their largest scar/length
ratio was slightly higher. This may point to slightly more
optimal fl ake production in the human sample (optimiz-
ing fl ake size per core), along with more intensive reduc-
tion of the Gona sample (truncating, and thus minimiz-
ing, the size of fl ake scars on the core).
21. Cores: Percentage of Original Cobble
(Estimate)
Rationale
The technology exhibited in early Oldowan cores
often makes it possible to arrive at reasonable estimates
of the size of the original cobble clast and what percent-
age of that cobble mass remains in the fi nal core form.
This estimate is based on the presence and curvature of
cortex on cores and extrapolating the continuation of
cortical surface in areas that have been fl aked, thereby
reconstructing the original size and shape of the cobble.
Obviously, cores that retain much of their cortex are bet-
ter candidates for more accurate reconstruction of origi-
nal cobble weight.
This subjective method is based on a great deal of
experience in the analysis of Oldowan lithic technology.
A blind test was conducted on an experimental sample of
25 cores (with known original cobble weights) to check
our accuracy in inferring the original cobble weight of
cores similar to those found at the Gona sites. The mean
margin of error was plus or minus 5%, suggesting that
this method is a useful guide to the intensity of core re-
duction.
As previously mentioned, the mass of the cobbles
in the human experimental sample was intentionally re-
duced by approximately 50% to provide a baseline when
analyzing the archaeological sample. This is refl ected
in the mean percentage of original cobble in the human
sample.
Results
The bonobo cores were the least reduced (~70% of
original cobble) and the Gona cores were the most re-
duced (~37%), with the human sample (~54%) interme-
diate. Statistical testing indicated that each of the three
samples were signifi cantly different from the other two
at the .00 confi dence level.
Cores: Percentage Original Cobble Remaining (Estimate)
by Tool-maker
Percentage of original core left
Tool-maker Mean N Std. Deviation
bonobo 70.30 33 16.490
gona 36.50 23 11.120
human 53.90 31 10.860
Total 55.50 87 18.850
humangonabonobo
Tool-maker
80
60
40
20
0
Percentage
Cores: Estimate Percentage of Original Cobble
Remaining
Test Statistics: Percentage Original
Cobble, Bonobo v. Gonaa
Percentage of
original core left
Mann-Whitney U 51.500
Wilcoxon W 327.500
Z -5.520
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Percentage Original
Cobble, Bonobo v. Humana
Percentage of
original core left
Mann-Whitney U 199.500
Wilcoxon W 695.500
Z -4.259
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 187
Test Statistics: Percentage Original
Cobble, Gona v. Humana
Percentage of
original core left
Mann-Whitney U 99.000
Wilcoxon W 375.000
Z -4.612
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Discussion
The high mean percentage of original cobble re-
maining in the bonobo core sample is a clear indication
of the low intensity of reduction (removing ~ 30% of
the cobble mass), while the low mean percentage in the
Gona sample is a clear indication of much higher intensi-
ty of reduction (removing 63% of the cobble mass). The
ability to effi ciently reduce a cobble and produce larger
populations of usable fl akes and fragments for a given
range of raw materials is one measure of skill. Clearly
the Gona hominins were able to reduce their cores more
than twice as effi ciently as the bonobos, suggesting bet-
ter mastery of stone reduction.
22. Cores: Number of Flake Scars
Rationale
The number of fl ake scars (10 mm maximum di-
mension) on cores gives a minimum estimate of the num-
ber of fl akes that have been removed from a core. Cores
with low scar counts tend to be less reduced; cores with
high scar counts often are heavily reduced. Of course,
on more heavily reduced cores, later core reduction may
remove earlier fl ake scars, so the total number of fl akes
removed can be considerably greater than the fl ake scar
count estimate.
Results
The bonobo cores showed a much lower mean fl ake
scar count (~5.5) than the Gona (~9.3) or human (~9.1)
sample. Statistical testing showed that the bonobo core
sample differed signifi cantly from both the Gona sample
and the human sample at the .00 confi dence level. The
Gona and human samples, however, were almost identi-
cal regarding mean fl ake scar number.
Cores: Number of Flake Scars by Tool-maker
Number of ake scars
Tool-maker Mean N Std. Deviation
bonobo 5.48 33 3.537
gona 9.30 23 3.948
human 9.13 31 3.074
Total 7.79 87 3.903
humangonabonobo
Tool-maker
10
8
6
4
2
0
Number
Cores: Mean Number of Flake Scars by
Tool-maker
Test Statistics: Core Flake Scars,
Bonobo v. Gonaa
Number of
ake scars
Mann-Whitney U 145.000
Wilcoxon W 706.000
Z -3.926
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Flake Scars,
Bonobo v. Humana
Number of
ake scars
Mann-Whitney U 178.000
Wilcoxon W 739.000
Z -4.510
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Flake Scars,
Gona v. Humana
Number of
ake scars
Mann-Whitney U 354.000
Wilcoxon W 630.000
Z -.044
Asymp. Sig. (2-tailed) .965
a. Grouping Variable: Toolmkrcod
Discussion
The bonobo cores were much less heavily fl aked,
with between three and four fewer fl ake scars on aver-
age than the Gona and human samples. This is yet an-
other indication of the low degree of reduction of bonobo
cores relative to the Gona and human cores.
It is also useful to compare the actual number of
188 The Oldowan: Case Studies Into the Earliest Stone Age
akes removed from a core to the number of core fl ake
scars, which is possible to do on the experimental sam-
ples. As an indication of the total number of fl akes pro-
duced in a sample, the number of whole fl akes, number
of split fl akes (judged from the number of left or right
splits, whichever is greater), and the number of proxi-
mal snaps were added together to derive the minimum
number of whole fl akes produced. Via this procedure, the
bonobo sample produced 177 reconstructed fl akes, the
Gona sample 216, and the human sample 368. The ratio
of reconstructed fl akes to cores was 5.36 for bonobos,
9.39 for Gona, and 11.87 for humans. It is likely that the
Gona fl ake counts would be higher if all the debitage
was represented, since the Gona cores are more heav-
ily reduced than the human cores and because there is
evidence (to be discussed below) that certain fl akes in
the reduction history of cores are not represented at the
excavated sites in expected numbers.
23. Cores: Ratio of Steps &
Hinges/Total Scars
Rationale
Flake scars terminating in steps and hinges are of-
ten viewed as representing misguided fl aking, since the
akes did not feather off neatly from the core. (Steps are
ake scars that terminate at a right angle break, usually
producing a piece of debitage that would be classifi ed
as a proximal snap, while hinges are fl ake scars that ter-
minate in a curved termination). The ratio of steps and
hinges to total number of fl ake scars has been used by
some archaeologists as an assessment of skill level. A
low number and low ratio of steps and hinges to other
scars has sometimes been interpreted as an indication
of greater skill in fl aking, whereas a higher number and
higher ratio of steps and hinges has been interpreted as
an indication of a lower level of skill.
Results
Surprisingly, the human 50% reduction sample had
the highest ratio of steps and hinges to scars (~.31); the
bonobos were intermediate (~.26) and the Gona sample
the lowest (~.18). Statistical testing found that only the
Gona and human samples showed signifi cant differences
at the .05 confi dence level.
Cores: Step to scar ratio by tool-maker
step_scr
Tool-maker Mean N Std. Deviation
bonobo .2570 33 .21476
gona .1839 23 .14908
human .3083 31 .16444
Total .2560 87 .18616
humangonabonobo
Tool-maker
0.30
0.20
0.10
0.00
Step to Scar Ratio
Cores: Mean Step to Scar Ratio by
Tool-maker
Test Statistics: Core Step-to-Scar Ratio,
Bonobo v. Gonaa
step_scr
Mann-Whitney U 304.000
Wilcoxon W 580.000
Z -1.271
Asymp. Sig. (2-tailed) .204
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Step-to-Scar Ratio,
Bonobo v. Humana
step_scr
Mann-Whitney U 435.000
Wilcoxon W 996.000
Z -1.033
Asymp. Sig. (2-tailed) .302
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Step-to-Scar Ratio,
Gona v. Humana
step_scr
Mann-Whitney U 202.000
Wilcoxon W 478.000
Z -2.709
Asymp. Sig. (2-tailed) .007
a. Grouping Variable: Toolmkrcod
Discussion
Initially, analysis of the results of this attribute was
perhaps the most surprising of the entire study. If a high
proportion of steps and hinges to fl ake scars is indeed
an indication of lesser skill, then why would the human
sample exhibit the highest proportion of step and hinge
scars, when other attributes investigated here indicate
greater fl aking skill in the human sample than in the
Gona or bonobo samples? There appear to be a number
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 189
of factors that would help explain the observed pattern.
It is likely that the shape of the cobble chosen as a
core blank could have a strong effect on the incidence
of steps and hinges in an assemblage. Tool-makers se-
lecting thinner cobbles or wedge-shaped cobbles with a
thin edge that are much easier to fl ake might produce
signifi cantly lower proportions of steps and hinges. This
could be the case with the Gona assemblage, in which
relatively thinner, easier-to-fl ake cobbles may have been
selected relative to those cobbles in the experimental
samples. If there were enough type 1 fl akes (fl akes with a
cortical striking platform and total cortex dorsal surface,
normally the fi rst ake to be removed from a cobble),
this might be tested: the morphology of the cortex of
that fl ake could be indicative of the ease of fl aking of the
cobble edge; unfortunately, there is only one type 1 fl ake
from the Gona sites. Another test of clast shape would be
refi tting entire cobbles, but unfortunately, this is rarely
possible in signifi cant numbers at Oldowan sites.
In fact, investigation of the difference in core thick-
ness may yield information pertinent to the patterns ob-
served in experimental samples. The human cores tended
to be thicker than the Gona cores, not only absolutely but
also relative to core length (see below). This difference
in relative thickness would tend to provide steeper edges
for fl aking and thus more opportunity to produce step
or hinge fl akes, increasingly so as a core is reduced and
aking forces are directed more steeply into the mass
of the core. Although similar in absolute thickness to
the human cores, the bonobo cores are thinner relative
to their length and are not as heavily reduced, both of
which factors may reduce the tendency for stepping to
occur relative to the human sample.
As will be discussed below, in order to see how pref-
erentially later stages of fl aking might infl uence core and
debitage characteristics (as this appears to be a major
factor differentiating the human and Gona core samples),
a subset of the human core sample (n=10, out of the total
sample of 33 cores) was further reduced to a level more
comparable to the Gona cores. Interestingly, these more
heavily reduced cores had a step-to-scar ratio of .1565
(S.D. of .08), a ratio even lower than at Gona (.1839).
The Mann-Whitney test did not fi nd signifi cant differ-
ence in step-to-scar ratios between the Gona sample and
the human later stage sample.
Test Statistics: Core Step-to-Scar Ratio,
Gona v. Human Later Stagesb
step_scr
Mann-Whitney U 111.000
Wilcoxon W 166.000
Z -.157
Asymp. Sig. (2-tailed) .875
Exact Sig.
[2*(1-tailed Sig.)] .893a
a. Not corrected for ties.
b. Grouping Variable: Toolmkrcod
It would thus appear that later stages of Oldowan
cobble core reduction produce less step and hinge frac-
tures, possibly because fl akes are able to travel down
established core ridges produced by previous fl ake re-
movals. Earlier stages of cobble reduction, however,
appear to produce relatively higher proportions of steps
and hinges, possibly because fl akes travel down curved
cortical surfaces of cobble cores less successfully and
terminate more frequently in steps or hinges.
In sum, the ratio of steps and hinges to total fl ake
scars is probably not a reliable indicator of skill, but is
likely infl uenced by a number of variables. Of special
importance is the degree of reduction of the core, with
more heavily reduced cores tending to exhibit fewer step
and hinge scars. In addition, step and hinge scar frequen-
cies are likely related to the overall ease of fl aking the
cobble, infl uenced in turn by variables such as presence
of a thin edge for fl aking, the absolute thickness of the
core, and core thickness relative to length.
24. Cores: Edge Angle
Rationale
This measurement gives an indication of potential
functional qualities of a core edge as well as an indica-
tion as to whether a given core could be easily fl aked
further. A fl aked core edge will yield different angles de-
pending on where you measure, so for this analysis the
minimal edge angle was recorded. An edge angle of 70°,
for example, suggests a possible cutting/chopping tool
and a core that could still be reduced, assuming the core
was still large enough for further reduction, whereas an
angle of 95° suggests a non-functional edge for cutting
or chopping and an exhausted core form.
Results
The bonobo cores showed the steepest angles (~83
degrees), the human sample showed the most acute
angles (~69 degrees) with the Gona cores intermediate
(~78 degrees). Statistical testing indicated that the hu-
man sampled differed from both the bonobo and Gona
samples at the .000 confi dence level, but that the bonobo
and Gona sample were not different at the .05 level.
Cores: Core angle by tool-maker
Core angle, nearest 5 degrees
Tool-maker Mean N Std. Deviation
bonobo 82.88 33 13.231
gona 78.26 23 6.676
human 68.87 31 9.722
Total 76.67 87 12.120
190 The Oldowan: Case Studies Into the Earliest Stone Age
humangonabonobo
Tool-maker
100
80
60
40
20
0
Core Angle
Cores: Mean Edge Angle by Tool-
maker
Test Statistics: Core Edge Angles,
Bonobo v. Gonaa
Core angle,
nearest 5
degrees
Mann-Whitney U 305.500
Wilcoxon W 581.500
Z -1.246
Asymp. Sig. (2-tailed) .213
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Edge Angles,
Bonobo v. Humana
Core angle,
nearest 5
degrees
Mann-Whitney U 207.000
Wilcoxon W 703.000
Z -4.126
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Core Edge Angles,
Gona v. Humana
Core angle,
nearest 5
degrees
Mann-Whitney U 148.500
Wilcoxon W 644.500
Z -3.682
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Discussion
The ability of a stone knapper to maintain acute
angles as fl aking proceeds is one index of skill. By this
criterion, the human sample showed the most skill, with
the most acute angles, with the Gona cores intermediate
(though not signifi cantly different from the bonobo sam-
ple), and the bonobo cores showing the least skill. The
exterior platform angle on whole fl akes, to be discussed
below, is perhaps an even better indicator of core edge
angle, as every fl ake bears evidence of the core angle im-
mediately prior to this fl ake detachment and can repre-
sent all stages of core reduction rather than just the cores
nal form (at which time it was abandoned or lost).
25. Cores: Percentage of Surface Cortex
Rationale
The amount of surface cortex on Oldowan cobble
cores, relative to fl aked core surface, can be a gross es-
timate of the amount of reduction of these cobbles. (A
possible exception might be a roller-shaped cobble that
is unifacially reduced along its long axis; even if the
mass is reduced by over 50 percent, the surface cortex
may still be ca. 80 percent). This attribute is estimated to
the nearest 5% based on visual examination.
Results
The bonobo cores had the greatest mean cortex val-
ue at ~75 percent; the human cores were intermediate at
~62 percent, with the Gona core surfaces the most heav-
ily fl aked at ~53 percent. Statistical tests indicated that
each of the three samples signifi cantly differed from the
other two samples at the .05 confi dence level.
Cores: Percentage cortex remaining by tool-maker
Percentage of cortex
Tool-maker Mean N Std. Deviation
bonobo 74.50 33 9.710
gona 53.48 23 16.130
human 61.61 31 11.280
Total 64.37 87 14.840
humangonabonobo
Tool-maker
80
60
40
20
0
Percentage Cortex
Cores: Mean Percentage Cortex
A Comparative Study of the Stone Tool-Making Skills of Pan, Australopithecus, and Homo sapiens 191
Test Statistics: Percentage Cortex,
Bonobo v. Gonaa
Percentage
of cortex
Mann-Whitney U 86.000
Wilcoxon W 362.000
Z -5.009
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Percentage Cortex,
Bonobo v. Humana
Percentage
of cortex
Mann-Whitney U 202.000
Wilcoxon W 698.000
Z -4.300
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod
Test Statistics: Percentage Cortex,
Gona v. Humana
Percentage
of cortex
Mann-Whitney U 247.000
Wilcoxon W 523.000
Z -2.009
Asymp. Sig. (2-tailed) .045
a. Grouping Variable: Toolmkrcod
Discussion
Percentage of cortex on core surfaces is an index
of how reduced a given cobble core is: 90 percent cor-
tex would suggest minimal reduction, while 10 percent
cortex would suggest very heavy reduction. The bonobo
cores are the least reduced, and the Gona cores the most
reduced. The extent of Gona core reduction, based on the
low percentage of surface cortex, suggests a higher level
of skill in the ability to remove substantial debitage as
reduction proceeded.
26a. Flakes: Flake Types
Rationale
Classifying whole fl akes into six distinct types (with
a seventh indeterminate category) helps identify differ-
ent stages of cobble reduction and preferential modes of
aking (Figure 23). These include:
Type 1: cortex platform; total cortex dorsal surface
Type 2: cortex platform; partial cortex dorsal surface
Type 3: cortex platform; non-cortex dorsal surface
Type 4: non-cortex platform; total cortex dorsal surface
Type 5: non-cortex platform; partial cortex dorsal
surface
Type 6: non-cortex platform; non-cortex dorsal surface
Type 7: indeterminate; blown or punctiform platform or
too weathered
As discussed by Toth (1982, 1985), examination of
proportions of these fl ake types can also be employed
to make predictions about what fl ake type populations
would be expected in early stages (e.g. types 1, 2, 4, and
5 fl akes) vs. later stages (e.g. types 3 and 6 fl akes) of
Oldowan cobble reduction, as well as about expected
changes in fl ake type proportions in the event of hydro-
logical winnowing during sedimentation and burial (e.g.
preferentially winnowing away from a fl aking locale the
lighter types 3 and 6 fl akes and leaving the heavier types
1, 2, 4, and 5). It is also possible to predict what fl ake
types might be depleted in an assemblage if highly func-
tional fl akes were to be transported by hominins away
from a fl aking area (discussed below in section 27b). For
the following analysis, the type 7 indeterminate fl akes
are omitted, since they are lacking technological infor-
mation, usually missing identifi able striking platforms.
Results
The bonobo fl ake population is characterized by
high proportions of type 2 fl akes, with moderate propor-
tions of fl ake types 3, 1, and 5 and low proportions of
ake types 4 and 6. The Gona population has a predomi-
nance of fl ake types 3 and 2, with moderate proportions
of type 6 fl akes and low proportions of fl ake types 5, 4,
and 1. The human sample, like the bonobo samples, has
high proportions of fl ake type 2, but with more bifacial
reduction also shows moderate proportions of types 5
and 3 fl akes as well as type 1 fl akes, and low proportions
of fl ake types 6 and 4. Chi square tests indicated that all
samples differed from one another at the .05 level of sig-
nifi cance. The Kolmogorov-Smirnov test of cumulative
frequency distributions indicate a signifi cant difference
between Gona and the other two samples, but found the
human and bonobo sample were not signifi cantly dif-
ferent at the .05 confi dence level. The Chi-square test,
which detected a signifi cant difference between these
two samples, was likely more sensitive to the specifi c
features of the fl ake type distributions, particularly the
higher frequencies of type 5 fl akes and lower frequen-
cies of type 1 fl akes in the human sample relative to the
bonobo sample.
192 The Oldowan: Case Studies Into the Earliest Stone Age
Flake Type Frequencies by Tool-maker
Flake Type
ake
type 1 ake
type 2 ake
type 3 ake
type 4 ake
type 5 ake
type 6 Total
Tool-
maker bonobo Count 25 60 29 4 15 9 142
% within
Tool-maker 17.6% 42.3% 20.4% 2.8% 10.6% 6.3% 100.0%
gona Count 1 49 79 2 13 19 163
% within
Tool-maker .6% 30.1% 48.5% 1.2% 8.0% 11.7% 100.0%
human Count 24 90 45 8 54 13 234
% within
Tool-maker 10.3% 38.5% 19.2% 3.4% 23.1% 5.6% 100.0%
Total Count 50 199 153 14 82 41 539
% within
Tool-maker 9.3% 36.9% 28.4% 2.6% 15.2% 7.6% 100.0%
Chi-Square Tests: Flake Types 1-6, Bonobo v. Gona
Value df Asymp. Sig.
(2-sided)
Pearson Chi-Square 49.582a5 .000
Likelihood Ratio 55.761 5 .000
N of Valid Cases 305
a. 2 cells (16.7%) have expected count less than 5.
The minimum expected count is 2.79.
Kolmogorov-Smirnov Test: Flake Types 1-6,
Bonobo v. Gonaa
Flake type
number
Most Extreme Absolute .292
Differences Positive .000
Negative -.292
Kolmogorov-Smirnov Z 2.542
Asymp. Sig. (2-tailed) .000
a. Grouping Variable: Toolmkrcod