3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya

Abstract

Human evolutionary scholars have long supposed that the earliest stone tools were made by the genus Homo and that this technological development was directly linked to climate change and the spread of savannah grasslands. New fieldwork in West Turkana, Kenya, has identified evidence of much earlier hominin technological behaviour. We report the discovery of Lomekwi 3, a 3.3-million-year-old archaeological site where in situ stone artefacts occur in spatiotemporal association with Pliocene hominin fossils in a wooded palaeoenvironment. The Lomekwi 3 knappers, with a developing understanding of stone’s fracture properties, combined core reduction with battering activities. Given the implications of the Lomekwi 3 assemblage for models aiming to converge environmental change, hominin evolution and technological origins, we propose for it the name ‘Lomekwian’, which predates the Oldowan by 700,000 years and marks a new beginning to the known archaeological record.

Full text

ARTICLE doi:10.1038/nature14464
3.3-million-year-old stone tools from
Lomekwi 3, West Turkana, Kenya
Sonia Harmand
1,2,3
, Jason E. Lewis
1,3,4
, Craig S. Feibel
3,4,5
, Christopher J. Lepre
3,5,6
, Sandrine Prat
3,7
, Arnaud Lenoble
3,8
,
Xavier Boe
¨s
3,7
, Rhonda L. Quinn
3,5,9
, Michel Brenet
8,10
, Adrian Arroyo
2
, Nicholas Taylor
2,3
, Sophie Cle
´ment
3,11
, Guillaume Daver
12
,
Jean-Philip Brugal
3,13
, Louise Leakey
1
, Richard A. Mortlock
5
, James D. Wright
5
, Sammy Lokorodi
3
, Christopher Kirwa
3,14
,
Dennis V. Kent
5,6
&He
´le
`ne Roche
2,3
Human evolutionary scholars have long supposed that the earliest stone tools were made by the genus Homo and that this
technological development was directly linked to climate change and the spread of savannah grasslands. New fieldwork
in West Turkana, Kenya, has identified evidence of much earlier hominin technological behaviour. We report the
discovery of Lomekwi 3, a 3.3-million-year-old archaeological site where in situ stone artefacts occur in spatio-
temporal association with Pliocene hominin fossils in a wooded palaeoenvironment. The Lomekwi 3 knappers, with a
developing understanding of stone’s fracture properties, combined core reduction with battering activities. Given the
implications of the Lomekwi 3 assemblage for models aiming to converge environmental change, hominin evolution and
technological origins, we propose for it the name ‘Lomekwian’, which predates the Oldowan by 700,000 years and
marks a new beginning to the known archaeological record.
Conventional wisdom in human evolutionary studies has assumed
that the origins of hominin sharp-edged stone tool production were
linked to the emergence of the genus Homo
1,2
in response to climate
change and the spread of savannah grasslands
3,4
. In 1964, fossils looking
more like later Homo than australopithecines were discovered at
Olduvai Gorge (Tanzania) in association with the earliest known stone
tool culture, the Oldowan, and so were assigned to the new species:
Homo habilis or ‘handy man’
1
. The premise was that our lineage alone
took the cognitive leap of hitting stones together to strike off sharp flakes
and that this was the foundation of our evolutionary success.
Subsequent discoveries pushed back the date for the first Oldowan stone
tools to 2.6 million years ago
5,6
(Ma) and the earliest fossils attributable
to early Homo to only 2.4–2.3 Ma
7,8
, opening up the possibility of tool
manufacture by hominins other than Homo
9
before 2.6 Ma
10–12
.
The earliest known artefacts from the sites of Gona (,2.6 Ma)
6,12
,
Hadar (2.36 60.07 Ma
13
), and Omo (2.34 60.04 Ma
14
) in Ethiopia,
and especially Lokalalei 2C (2.34 60.05 Ma
15
) in Kenya, demonstrate
that these hominin knappers already had considerable abilities in terms
of planning depth, manual dexterity and raw material selectivity
14–19
.
Cut-marked bones from Dikika, Ethiopia
20
, dated at 3.39 Ma, has added
to speculation on pre-2.6-Ma hominin stone tool use. It has been
argued that percussive activities other than knapping, such as the
pounding and/or battering of plant foods or bones, could have been
critical components of an even earlier, as-yet-unrecognized, stage of
hominin stone tool use
21–25
. Any such artefacts may have gone unre-
cognized if they do not directly resemble known Oldowan lithics, occur
at very low densities or were made of perishable materials
10
.
In 2011, the West Turkana Archaeological Project (WTAP) began
an archaeological survey and excavation in the Lomekwi Member
26
(3.44–2.53 Ma) of the Nachukui Formation (west of Lake Turkana,
northern Kenya; Fig. 1) to search for evidence of early hominin lithic
behaviour. Several promising surface artefact concentrations and dis-
persed single finds were discovered. At the Lomekwi 3 archaeological
site, 28 lithic artefacts were initially found lying on the surface or
within a slope deposit, and one core was uncovered in situ. By the
close of the subsequent 2012 field season, excavation at LOM3 had
reached 13 m
2
, revealing an additional 18 stone tools and 11 fossils in
situ (Extended Data Table 1) within a horizon (approximately 80 cm)
of indurated sandy-granular sediments stratified in a thick bed of fine
silts (Fig. 2). A further 100 lithic artefacts and 22 fossil remains were
collected from the surface immediately around the site along with
two artefacts from the slope deposit (Extended Data Fig. 1). These
finds occur in the same geographic and chronological range as the
paratype of Kenyanthropus platyops (KNM-WT 38350)
27
, other
hominin fossils generally referred to cf. K. platyops
28
, and one unpub-
lished hominin tooth (KNM-WT 64060) found by WTAP in 2012
(Supplementary Information, part A and Supplementary Table 1).
Geochronological and palaeoenvironmental contexts
The chronological context of LOM3 derives from correlation with the
Lomekwi Member of the Nachukui Formation
26
and radiometrically
dated tuffs within it
29,30
, as well as from magnetostratigraphy of the
site and estimated sedimentation rates. The composite type section of
the Lomekwi Member, 2–5 km east of LOM3, is bracketed by the
a-Tulu Bor Tuff (3.44 60.02 Ma) at the base and the Lokalalei Tuff
(2.53 60.02 Ma) at the top
29,30
. Closer to LOM3, two new sections
provide additional context. Section 1 (CSF 2011-1; ,46 m thick,
located 1.44 to 1 km north of LOM3, Extended Data Fig. 2) includes
310 | NATURE | VOL 521 | 21 MAY 2015
1
Turkana Basin Institute, Stony Brook University, Stony Brook, New York 11794-4364, USA.
2
CNRS, UMR 7055, Pre
´histoire et Technologie, Universite
´Paris Ouest Nanterre La De
´fense, 21 alle
´ede
l’Universite
´, 92023 Nanterre Cedex, France.
3
West Turkana Archaeological Project, P.O. Box 40658-00100, Ngara Rd, Nairobi, Kenya.
4
Department of Anthropology and Center for Human Evolutionary
Studies, Rutgers University, New Brunswick, New Jersey 08901, USA.
5
Department of Earth and Planetary Sciences, Rutgers University, Piscataway, New Jersey 08854, USA.
6
Lamont-Doherty Earth
Observatory of Columbia University, Palisades, New York 10964, USA.
7
CNRS, UPR 2147, Dynamique de l’Evolution Humaine, 44 rue de l’Amiral Mouchez, 75014 Paris, France.
8
CNRS, UMR 5199 PACEA,
Universite
´de Bordeaux, 33615 Pessac, France.
9
Department of Sociology, Anthropology and Social Work, Seton Hall University, South Orange, New Jersey 07079, USA.
10
Inrap, Centre Mixte de Recherche
Arche
´ologique, Domaine de Campagne, 24620 Campagne, France.
11
Inrap, 34-36 avenue Paul-Vaillant Couturier, 93120 La Courneuve, France.
12
IPHEP, Institut de Pale
´oprimatologie, Pale
´ontologie
Humaine: E
´volution et Pale
´oenvironnements, CNRS, UMR 7262, Universite
´de Poitiers, Ba
ˆt. B35 – TSA 51106, 6 rue Michel Brunet, 86073 Poitiers Cedex 9, France.
13
Aix-Marseille Universite
´, CNRS, MCC,
UMR 7269, LAMPEA, 13094 Aix-en-Provence Cedex 2, France.
14
National Museums of Kenya, Department of Earth Sciences, Archaeology Section, P.O. Box 40658-00100 Ngara Rd, Nairobi, Kenya.
G2015 Macmillan Publishers Limited. All rights reserved

the a- and b-Tulu Bor Tuffs in the lower third (Supplementary
Information, part B). Composite Section 2 (upper CSF-2012-9,
,44 m thick, located 0.4 km south of LOM3 and lower CSF-2011-2
located 0.28 km north of LOM3, Fig. 3a, b and Extended Data Fig. 2)
includes at the base a lenticular tuff correlated geochemically with the
Toroto Tuff in the Koobi Fora Formation where it outcrops 10–12 m
above the a-Tulu Bor Tuff, and has been dated radiometrically to
3.31 60.02 (refs 29, 30). Both the two Tulu Bor Tuffs in Section 1
and the Toroto Tuff in Section 2 occur in normal polarity magneto-
zones, corresponding to the early part of the Gauss Chron C2An
(Fig. 3a and Supplementary Information, part C), while the overlying
sediments at both sites are in reversed polarity zones as are the sedi-
ments encompassing the in situ artefacts at LOM3, 10 m above the
Toroto Tuff (Fig. 3b). Thus, the artefacts were deposited after
3.31 60.02 Ma during the Mammoth reverse subchron C2An.2r
(3.33–3.21 Ma
31
). Based on extrapolation of sediment accumulation
rates between the levels of the a-Tulu Bor and Toroto Tuffs and the
onset of subchron C2An.2r, an age of 3.3 Ma is determined for LOM3
(Extended Data Fig. 3 and Supplementary Information, part C),
which accords with previous interpretations of the antiquity of fossils
from this locality
27–30
.
Stable carbon isotopic analyses of pedogenic carbonate nodules
located adjacent to and at LOM3 yielded a mean d
13
C
VPDB
value of
27.3 61.1%(Extended Data Fig. 4 and Supplementary Information,
part D), which indicates a mean fraction of woody cover (e
wc
)of
47 69% and positions the site within a woodland/bushland/thicket/
shrubland environment
32
. Our results are comparable to paleosol
d
13
C
VPDB
values of other East African hominin environments between
3.2 and 3.4 Ma but significantly woodier than the 2.6 Ma artefact site at
Gona, Ethiopia (Extended Data Figs 4b, c)
32,33
. The associated fauna
supports this interpretation (Supplementary Information, part E).
The Lomekwi 3 site
The LOM3 site is a low hill eroded into by a small ravine. The upper-
most sediments encountered during excavation form a plaque of slope
deposit which is a few centimetres thick (Fig. 2a). Under it, a series of
interdigitated lenses of sands, granules and silts are found. They corre-
spond to different facies of the same sedimentary environment related
to the distal fan deposit in which the artefacts are preserved (Fig. 2c
and Supplementary Information, part B). Sealed in situ in these
Pliocene sediments (Extended Data Fig. 5), the LOM3 archaeological
material is considered to be in a slightly re-distributed primary
archaeological context based on the following observations: (1) arte-
facts of different sizes, ranging from ,1 cm wide flake fragments to
very large worked cobbles and cores are present; (2) artefacts are
larger and heavier than could be carried by the energy of the alluvial
system that deposited the sediments (the maximal competence of the
transport flow can be inferred by the coarsest fraction of the bed load
deposited, that is, ,4 cm diameter granules); (3) many excavated
lithic pieces exhibit only slight abrasion, as reflected in the observation
of are
ˆte and edge widths measuring #100 mm. Moreover, although it
is not possible at present to link all surface finds to the excavated
context, the identification of a refit between a core recovered from
the dense stratified deposit and one surface flake clearly shows that at
least a portion of the surface material derives directly from the in situ
level (Fig. 4a). More precise interpretation of site preservation is based
on observations drawn from the excavation, with the most plausible
possibilities limited to either good preservation of the site and most of
the assemblage, or a slight redistribution in close proximity of the
original activity location (Supplementary Information, part B).
Technology of the Lomekwi 3 stone tools
Based on the lithic material recovered in 2011 and 2012, the current
total assemblage (n5149 surface and in situ artefacts) incorporates
1920 18 17
449.6
449.5
449.7
449.8
449.9
450.0
m above sea level
Lower sandy lenses
Silty layer
Upper sandy lens
Excavation oor
1 m
SW
Slope
deposit
Fan deposit
c
Slope deposit
Sands and granules
a
Artefact
included in
slope deposit
Slop
e de
posi
t
Artefact
included
slope depo
Artefacts in situ
in Pliocene bed
Pliocene beds
Desertic pavement
armouring old slope
Modern
alluvial
deposit
Present-day
channel
sands
Pliocene deposit outcropping
on currently eroding slope
Ex
GP
10 m
Ch
SW
447
446
448
449
450
451
452
453
Elevation (m above sea level)
vertical exaggeration ×3
Clays
Sandy silts
Sands and granules
Bedded gravels
and sands
b
Figure 2
|
LOM3 lithological context. a, View of the excavation, facing east,
showing relationship between surface, slope deposit, and in situ contexts
containing the artefacts and fossils. Scale in midground is 20 cm. Lower-
leftmost artefact is the anvil LOM3-2012-K18-2, shown in Fig. 5a.
b, Topographic profile and stratigraphic units at site level showing the
excavation zone (Ex), the geological trench made at the base of the section (GP);
the artefacts and fossils derive from a series of lenses of sand and granules
making up a ,1 m thick bed (Ch). c, Sectionat the excavation along bands I and
J (indicated by the black line in Extended Data Fig. 1a) showing the sediments
which form the fan deposits containing the artefacts.
Kerio River
Lake Turkana
3
4
2
8
12
9
13
6
61
6
Lomekwi 3
Lokalalei
1
2c
Kokiselei
1–6
Naiyena
Engol Kalochoro
Nachukui
Kalokodo
Kaitio
Nadung'a
1
2
110
11 1
34
5
3
2
Kangatukuseo
Lomekwi
Kaitio
Kalochoro
Nariokotome
Nachukui
Naiyena Engol
Kalokodo
Kokiselei
Lake Turkana
Kangaki
Lokalalei
N
04° 00′ N
02km
Koobi
Fora
Formation
Nachukui
Formation
Nachukui
Formation
Shungura
Formation
Shungura
Formation
Omo River
Turkwel River
0204060
km
Figure 1
|
Geographic location of the LOM3 site. Map showing relation of
LOM3 to other West Turkana archaeological site complexes.
21 MAY 2015 | VOL 521 | NATURE | 311
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

83 cores, 35 flakes (whole and broken), seven passive elements or
potential anvils, seven percussors (whole, broken or potential), three
worked cobbles, two split cobbles, and 12 artefacts grouped as inde-
terminate fragments or pieces lacking diagnostic attributes (Extended
Data Table 1a).
Cores are made predominantly from heavy and large-sized cobbles
or blocks of lava (mean of the cores: 167 3147.8 3108.8 mm, 3.1 kg;
Extended Data Table 2). Basalts (34.90%) and phonolites (34.23%) are
the dominant raw materials represented, followed by trachy-phonolite
(23.49%; Extended Data Table 1b), all of which were available in local
paleo-channels. Initial survey of a conglomerate source less than 100 m
from the site shows that cobbles and blocks of all sizes were available
locally, from which the largest were consistently selected. Most cores
were flaked from one striking platform onto one single surface, resulting
in several superposed and contiguous unidirectional removals (unifacial
partial exploitation), sometimes along a longer part of the perimeter. A
few specimens show unifacial partial exploitation by multidirectional
removals, while others show bifacial flaking. Significant knapping acci-
dents occurred during flaking, with numerous hinge and step flake
terminations visibleon cores (Fig. 4a), thoughmore invasive and feather
terminating flakes were also often successfully removed. In some cases,
cores display a series of shorter (,1 cm) contiguous small scars along a
more limited portion of the platform edge, although it is not yet clear
whether this results from the knapping techniques employed, or reflects
the utilization of some artefacts in heavy-duty tasks.
To reconstruct more accurately the techniques and reduction strat-
egies used to produce the LOM3 artefacts, an experimental program
was undertaken to replicate the lithics found at the site from the
same raw materials available locally at LOM3. Together with the
technological analysis of the archaeological material, these replication
experiments suggest that the LOM3 knappers were using techniques
including passive hammer
34,35
and/or bipolar
34
(Extended Data Fig. 6)
that have to-date rarely been identified in the Oldowan
21,22,36,37
. The
average size and weight of the LOM3 cores (Extended Data Table 2)
renders direct freehand percussion an arduous undertaking; however,
it cannot be ruled out for some of the smaller cores.
The technological features of flakes and flake fragments are clear,
unequivocal and seen repeatedly, demonstrating that they were inten-
tionally knapped from the cores. They range from 19 to 205 mm long
(Fig. 5d and Extended Data Table 2) and frequently present cortex on
their dorsal surfaces, sometimes on their striking platforms, or both.
Three pieces in particular bear localized battered areas on their dorsal
surfaces—including the specimen that refits onto the in situ core
(Fig. 4a)—showing that blanks were sometimes used for percussive
activities before flake removal and that at least some individual blocks
were involved in several distinctively different modes of use.
The largest and heaviest (up to 15 kg) pieces in the assemblage were
made on large blocks of basalt or coarse trachy-phonolite. They have
flat natural surfaces that could enable their stabilization for use
(Fig. 5a, b and Extended Data Fig. 7a). Comparisons with other
described anvils from the Early Stone Age and experiments suggest
these can be interpreted as anvils or passive elements
38,39
. Three of
these show a similar wear and fracture pattern. The largest piece
exhibits along one lateral plane a series of divergent step fractures
associated with crushing marks and an additional concentration of
impact damage on one horizontal surface (Fig. 5a). The other two
pieces have non-invasive step fractures along a greater or lesser por-
tion of their high-angled intersecting surfaces (edges) that are assoc-
iated with crushing and impact marks (Fig. 5b and Extended Data Fig.
7a). A further two cobbles show heavy battering marks concentrated
on a convex area and are interpreted as passive elements. Seven med-
ium-sized cobbles display battering marks and/or impact damage
associated with fractured surfaces and are interpreted as hand-held
percussors or active elements (Extended Data Figs 7b, c).
Discussion
LOM3 core and flake techno-morphology does not conform to any
observed pattern resulting from accidental natural rock fracture. On
the contrary, LOM3 cores and flakes bear all the techno-morpho-
logical characteristics of debitage products. Data reported on acci-
dental flakes from chimpanzee nut-cracking sites
40
falls closer to the
flake size spectrum observed at early Oldowan sites than to the size
range of LOM3 flakes (Extended Data Table 2). LOM3 knappers were
a
b
Lithology VGP lat. (°)
–90 –45 0 45 90
Section 1
α-TB: 3.44 ± 0.02 Ma
β-TB
α-TB (redeposited)
VVVVVV
VVVVVV
VVVVVV
–90 –45 0 45 90
Lithology
Section 2
VGP lat.(°)
Toroto
Tuff: 3.31 ± 0.02 Ma
VVVVVV
Kataboi Lomekwi
Members
0
10
20
30
40
50
60
70
(m)
–10
North South
~1km
Mudstones
Mudstones
and gravels
Tuff
Gravels
VVVVVV
3.00
3.05
3.10
3.15
3.20
3.25
3.30
3.35
3.40
3.45
3.50
3.55
3.60
C2An.2r
(Mammoth)
C2An.1r
(Kaena)
C2An.2n
C2An (Gauss)
Subchron Chron (Ma)
C2An.1n
C2Ar
(Gilbert)
C2Ar.1r C2An.3n
GPTS
?
Composite
polarity
2.3 2.5 2.7 2.9 3.1 3.3 3.5
(m)
0
20
40
60
80
100
120
140
160 C2An.3n C2An.2n C2An.1n
LOM3
archaeological site
α-TB to Lokalalei Tuff
sed. rate = 17.4 cm per kyr
α-TB
Ma
Lokalalei Tuff
Clays (vertic) and gravels series in the lower Lomekwi Member
Bentonite clay beds
Toroto Tuff 3.31 Ma
LOM3 Site
Palaeobeach
Lowermost vertic clays
Figure 3
|
Chronostratigraphic framework for LOM3.
a, Chronostratigraphic framework for LOM3 (star) with generalized
stratigraphic columns and magnetostratigraphic alignment to the geomagnetic
polarity time scale (GPTS) in context of dates of tuffaceous markers (61 s.d.)
and stratigraphicnomenclature for Members of the NachukuiFormation
26,30
.A
linearly interpolated date of 3.3 Ma for the in situ stone tools is consistent with
the site’s magnetostratigraphic position within the reverse polarity interval that
is correlated to reverse subchron C2An.2r (Mammoth Subchron) dated at
3.33–3.21 Ma
31
.b, Photograph facing north showing geographic and
stratigraphic relationship between Toroto Tuff, paleobeach, and LOM3.
312 | NATURE | VOL 521 | 21 MAY 2015
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

able to deliver sufficient intentional force to repeatedly detach series of
adjacent and superposed unidirectional flakes, sometimes invasive,
and then to continue knapping either by laterally rotating the cores
or by flipping them over for bifacial exploitation. However, though
multiple flakes were successfully detached, the majority of flake scars
terminate as hinge and step fractures. The precision of the percussive
motion was also occasionally poorly controlled, as shown by repeated
impact marks on core platforms caused by failed blows applied too far
from the striking platform edge to induce fracture. LOM3 lithics
(cores and flakes) are significantly larger in length, width, and thick-
ness than those from OGS7, EG10 and EG12 at Gona, A.L. 894 at
Hadar, and Omo 57 and Omo 123 in Ethiopia; Lokalalei 2C from
West Turkana, Kenya; and DK and FLK Zinj from Olduvai Gorge in
Tanzania (Extended Data Table 2). Furthermore, the LOM3
anvils and percussors are larger and heavier than those chosen for
nut-cracking by wild chimpanzees in Bossou
41
(southeastern Guinea;
Extended Data Table 3). The dimensions and the percussive-related
features visible on the artefacts suggest the LOM3 hominins were
combining core reduction and battering activities and may have used
artefacts variously: as anvils, cores to produce flakes, and/or as pound-
ing tools. The use of individual objects for several distinctive tasks
reflects a degree of technological diversity both much older than
previously acknowledged and different from the generally uni-
purpose stone tools used by primates
24,25
. The arm and hand motions
entailed in the two main modes of knapping suggested for the LOM3
assemblage, passive hammer and bipolar, are arguably more similar to
those involved in the hammer-on-anvil technique chimpanzees and
other primates use when engaged in nut cracking
42–44
than to the
direct freehand percussion evident in Oldowan assemblages. The
likely prevalence of these two knapping techniques demonstrates
a
b
c
d
2
1
12
10cm
10cm
10cm
5cm 5cm
Figure 4
|
Photographs of selected LOM3
artefacts. a,In situ core (LOM3-2011-I16-3,
1.85 kg) and refitting surface flake (LOM3-2011
surf NW7, 650 g). Unifacial core, passive hammer
and bipolar technique. Both the core and the flake
display a series of dispersed percussion marks on
cortex showing that percussive activities occurred
before the removal of the flake, potentially
indicating the block was used for different
purposes. b,In situ unifacial core (LOM3-2012-
H18-1, 3.45 kg), bipolar technique. See Extended
Data Fig. 6b for more details. c, Unifacial core
(LOM3-2012 surf 71, 1.84 kg), passive hammer
technique. d, Flakes (LOM3-2012-J17-3 and
LOM3-2012-H17-3) showing scars of previous
removals on the dorsal face. See Supplementary
Information part F for 3D scans of lithic artefacts.
21 MAY 2015 | VOL 521 | NATURE | 313
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

the central role that they might have played at the dawn of technology,
as previously suggested
21,22,36,37
.
LOM3 predates the oldest fossil specimens attributed to Homo in
West Turkana at 2.34 60.04 Ma
7
by almost a million years; the only
hominin species known to have been living in the West Turkana
region at the time is K. platyops
27
, while Australopithecus afarensis
is found in the Lower Awash Valley at 3.39 Ma in association with cut-
marked bones from Dikika
20
. The LOM3 artefacts indicate that their
makers’ hand motor control must have been substantial and thus that
reorganization and/or expansion of several regions of the cerebral
cortex (for example, somatosensory, visual, premotor and motor
cortex), cerebellum, and of the spinal tract could have occurred before
3.3 Ma. The functional morphology of the upper limb of Pliocene
hominins (especially A. afarensis, the only species for which contem-
poraneous fossil hand and wrist elements are known), particularly in
terms of adaptations for stone tool making, must be investigated
further if this important milestone in human evolution is to be under-
stood more fully (Supplementary Information, part A).
Critical questions relating to how the LOM3 assemblage compares
with the previously known earliest hominin stone tool techno-complex,
the Oldowan, remain. They are difficult to address because the term
Oldowan has been defined differently since it was first employed in
1934 (refs 16, 45–47). The simplest defining characteristics of the
Oldowan are that its knappers show the earliest evidence of a basic
understanding of the conchoidal fracture mechanics of stone and were
able to effectively strike flakes from cores, more often than not knapping
using ‘grammars of action’
48
and predominantly using the free-hand
knapping technique
11,17
. The LOM3 knappers’ understanding of stone
fracture mechanics and grammars of action is clearly less developed than
that reflected in early Oldowan assemblages and neither were they pre-
dominantly using free-hand technique. The LOM3 assemblage could
represent a technological stage between a hypothetical pounding-
oriented stone tool use by an earlier hominin and the flaking-oriented
knapping behaviour of later, Oldowan toolmakers. The term ‘Pre-
Oldowan’ has been suggested for modified stones if ever found in depos-
its older than 2.6 Ma, especially if they are different in terms of knapping
skill from the Oldowan sensu stricto
49
(thisisnottobeconfusedwith
previous uses of the same term by some authors to describe the early
Oldowan period between 2.6–2 Ma
50
). The LOM3 assemblage may
therefore concord with such a premise. We assert, however, that the
technological and morphological differences between the LOM3 and
early Oldowan assemblages are significant enough that amalgamating
them would mask important behavioural and cognitive changes occur-
ring among hominins over a nearly 2-million-year timespan. A separate
name for the LOM3 assemblage is therefore warranted. Given the para-
digmatic shift that LOM3 portends for models that aim to converge
environmental change, hominin evolution and technological origins,
the name Lomekwian is proposed. In any scenario, the LOM3 stone
tools mark a new beginning to the known archaeological record, now
shown to be more than 700,000 years older than previously thought.
Note added in proof: The recently described LD 350-1 partial mand-
ible from Ethiopia now provides the earliest evidence of the genus
Homo at 2.8 Ma (Villmoare, B. et al. Early Homo at 2.8 Ma from Ledi-
Geraru, Afar, Ethiopia. Science, 347, 1352–1355). The LOM3 artefacts
a
b
c
10cm
10cm
20cm
Figure 5
|
Photographs of selected LOM3
artefacts. a,In situ passive element/anvil (LOM3-
2012-K18-2, 12 kg). b, Passive element/anvil
(LOM3-2012 surf 60, 4.9 kg). Both anvils aand
bexhibit similar patterns of macroscopic wear
consisting of superposed step fracturing in
association with crushing and impacts marks. On
a, damage is localized on a single lateral face, with
battering marks present on one horizontal plane.
On b, damage is distributed along a greater portion
of the perimeter, but in this case no percussive
marks are identifiable on the horizontal plane. In
both cases, the intensity of the observed wear
signature indicates a use in heavy-duty activities.
c, Unifacial core (LOM3-2012 surf 90, 4.74 kg),
bipolar technique and semi-peripheral
exploitation. Inset shows crushing marks on the
proximal surface of the cobble related to battering
activities before or after the knapping of the core.
See Supplementary Information part F for three-
dimensional scans of lithic artefacts.
314 | NATURE | VOL 521 | 21 MAY 2015
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

still predate the known origins of Homo by half a million years and the
question of what hominin species made them remains.
Online Content Methods, along with any additional Extended Data display items
and SourceData, are available in theonline version of the paper;references unique
to these sections appear only in the online paper.
Received 1 November 2012; accepted 13 April 2015.
1. Leakey, L. S. B., Tobias, P. V. & Napier, J. R. A new species of the genus Homo from
Olduvai Gorge. Nature 202, 7–9 (1964).
2. Harris, J. W. K. Cultural beginnings: Plio-Pleistocene archaeological occurrences
from the Afar Rift, Ethiopia. Afr. Archaeol. Rev. 1, 3–31 (1983).
3. Quinn, R. L. et al. Pedogenic carbonate stable isotopic evidence for woodedhabitat
preference ofearly Pleistocene tool makersin the Turkana Basin. J. Hum. Evol. 65,
65–78 (2013).
4. Bobe, R. & Behrensmeyer, A. K. The expansion of grassland ecosystemsin Africa in
relation to mammalian evolution and the origin of the genus Homo. Palaeogeogr.
Palaeoclimatol. Palaeoecol. 207, 399–420 (2004).
5. Roche,H. & Tiercelin, J.-J. De
´couverted’une industrie lithiqueancienne in situ dans
la formationd’Hadar, Afar central,Ethiopie. C. R. Acad. Sci. Paris D 284, 1871–1874
(1977).
6. Semaw, S. et al. 2.5-million-year-old stone tools from Gona, Ethiopia. Nature 385,
333–336 (1997).
7. Prat, S. et al. First occurrence of early Homo in the Nachukui Formation (West
Turkana, Kenya) at 2.3–2.4 Myr. J. Hum. Evol. 49, 230–240 (2005).
8. Kimbel, W. H. et al. Late Pliocene Homo and Oldowan tools from the Hadar
formation (Kada Hadar member), Ethiopia. J. Hum. Evol. 31, 549–561 (1996).
9. Anto
´n, S. C., Potts,R. & Aiello, L. C. Evolution of earlyHomo: An i ntegrated biological
perspective. Science 345, 1236828 (2014).
10. Panger, M., Brooks, A. S., Richmond, B. G. & Wood, B. Older than the Oldowan?
Rethinking the emergence of hominintool use.Evol. Anth ropol. 11, 235–245 (2002).
11. Roche, H., Blumenschine, R. J. & Shea, J. J. in The First Humans — Origin and Early
Evolutionof the Genus Homo (eds Grine, F. E., Fleagle,J. G. & Leakey, R. E.) 135–147
(Springer, 2009).
12. Semaw, S. et al. 2.6-Million-year-old stone tools andassociated bones fromOGS-6
and OGS-7, Gona, Afar, Ethiopia. J. Hum. Evol. 45, 169–177 (2003).
13. Campisano,C. J. Geologicalsummary of theBusidima Formation (Plio-Pleistocene)
at the Hadar paleoanthropological site, Afar Depression, Ethiopia. J. Hum. Evol. 62,
338–352 (2012).
14. de la Torre, I. Omo revisited: evaluating the technological skills of Pliocene
hominids. Curr. Anthropol. 45, 439–465 (2004).
15. Roche, H. et al. Early hominid stone tool production and technical skill 2.34 Myr
ago in West Turkana, Kenya. Nature 399, 57–60 (1999).
16. Delagnes, A. & Roche, H. Late Pliocene hominid knapping skills: the case of
Lokalalei 2C, West Turkana, Kenya. J. Hum. Evol. 48, 435–472 (2005).
17. Stout, D., Semaw, S., Rogers, M. J. & Cauche, D. Technological variation in the
earliest Oldowan from Gona, Afar, Ethiopia. J. Hum. Evol. 58, 474–491 (2010).
18. Harmand, S. in Interdisciplinary Approaches to the Oldowan (edsHovers, E. & Braun,
D. R) 85–97 (Springer, 2009).
19. Goldman-Neuman, T. & Hovers,E. Raw materialselectivityin Late PlioceneOldowan
sites in the Makaamitalu Basin, Hadar, Ethiopia. J. Hum. Evol. 62, 353–366 (2012).
20. McPherron, S. P. et al. Evidence for stone-tool-assisted consumption of animal tissues
before 3.39 million years ago at Dikika, Ethiopia. Nature 466, 857–860 (2010).
21. Mora, R. & de la Torre, I. Percussion tools in Olduvai Beds I and II (Tanzania):
implicationsfor early human activities.J. Anthropol. Archaeol.24, 179–192 (2005).
22. Diez-Martı
´n, F., Sa
´nchez Yustos, P., Domı
´nguez-Rodrigo, M., Mabulla, A. Z. P. &
Barba, R. WereOlduvai hominins makingbutchering or batteringtools? Analysis of
a recentlyexcavated lithic assemblagefrom BK (Bed II, Olduvai Gorge,Tanzania). J.
Anthropol. Archaeol. 28, 274–289 (2009).
23. Blumenschine, R. J. & Selvaggio, M. M. Percussion marks on bone surfaces as a
new diagnostic of hominid behaviour. Nature 333, 763–765 (1988).
24. Marchant, L. & McGrew, W. in Stone Knapping: the Necessary Conditions for a
Uniquely Hominin Behavior (eds Roux, V. & Brill, B.) 341–350 (Cambridge
McDonald Institute, 2005).
25. Carvalho, S., Cunha, E., Sousa, C. & Matsuzawa, T. Chaı
ˆnes ope
´ratoires and
resource-exploitation strategies in chimpanzee (Pan troglodytes) nut cracking. J.
Hum. Evol. 55, 148–163 (2008).
26. Harris, J. M., Brown, F. H. & Leakey, M. G. Stratigraphy andpaleontology of Pliocene
and Pleistocene localities west of Lake Turkana, Kenya. Contr. Sci. Nat. Mus. Los
Angeles 399, 1–128 (1988).
27. Leakey, M. G. et al. New hominin genus from eastern Africa shows diverse middle
Pliocene lineages. Nature 410, 433–440 (2001).
28. Wood, B. & Leakey, M. G. The Omo-Turkana Basin fossil hominins and their
contribution to our understanding of human evolution in Africa. Evol. Anthropol.
20, 264–292 (2011).
29. McDougall, I. & Brown,F. H. Geochronology of thepre-KBS Tuff sequence,Turkana
Basin. J. Geol. Soc. Lond. 165, 549–562 (2008).
30. McDougall, I. et al. New single crystal
40
Ar/
39
Ar ages improve time scale for
deposition of the Omo Group, Omo-Turkana Basin, East Africa. J. Geol. Soc. Lond.
169, 213–226 (2012).
31. Lourens, L. J., Hilgen, F. J., Laskar, J., Shackleton, N. J. & Wilson, D. The Neogene
Period. A Geological Time Scale 2004 (edsGradstein, F. M., Ogg, J. G. & Smith, A. G.)
409–440 (Cambridge University Press, 2004).
32. Cerling, T. E. et al. Woody cover and hominin environments in the past 6 million
years. Nature 476, 51–56 (2011).
33. Quade, J. et al. Paleoenvironments of the earliest stone toolmakers, Gona, Ethiopia.
Geol. Soc. Am. Bull. 116, 1529–1544 (2004).
34. Crabtree, D. E. An introduction to flintworking. Part 1. An introduction to the
technology of stone tools. Occasional Papers of the Idaho State University Museum,
28, (Idaho State Univ. Museum, 1972).
35. Mourre, V., Jarry, M., Colonge, D. & Lelouvier, L.-A. Le de
´bitage sur enclume aux
Bosses (Lamagdalaine, Lot, France). Paleo (special issue) 49–62 (2010).
36. Diez-Martin, F. et al. New insights into hominin lithic activities at FLK North Bed I,
Olduvai Gorge, Tanzania. Quat. Res. 74, 376–387 (2010).
37. de la Torre, I. & Mora, R. A technological analysis of non-flaked stone tools in
Olduvai Beds I & II. Stressing the relevance of percussion activities in the African
Lower Pleistocene. Paleo (special issue) 13–34 (2010).
38. Alimen, M. H. Enclumes (percuteurs dormants) associe
´es a
`l’Acheule
´en supe
´rieur
de l’Ougartien (Oued Fare
`s, Sahara occidental). Bull. Soc. Pre
´hist. Fr. 60, 43–47
(1963).
39. Leakey, M. D. Olduvai Gorge, Vol. 3. Excavations in Beds I and II 1960–1963
(Cambridge Univ. Press, 1971).
40. Mercader, J. et al. 4,300-year-old chimpanzee sites and the origins of percussive
stone technology. Proc. Natl Acad. Sci. USA 104, 3043–3048 (2007).
41. Sakura, O. & Matsuzawa, T. Flexibility of wild chimpanzee nut-cracking behavior
using stone hammers and anvils: an experimentalanalysis. Ethology 87, 237–248
(1991).
42. Haslam, M. et al. Primate archaeology. Nature 460, 339–344 (2009).
43. Visalberghi, E., Haslam, M., Spagnoletti, N. & Fragaszy, D. Use of stone hammer
tools and anvilsby bearded capuchin monkeys over timeand space: construction
of an archeological record of tool use. J. Archaeol. Sci. 40, 3222–3232 (2013).
44. Matsuzawa, T. in Great Ape Societies (eds McGrew, W. et al.) 196–209 (Cambridge
Univ. Press, 1996).
45. Leakey, L. S. B.in Essays Presentedto C. G. Seligman (eds Evans-Pritchard,E. E., Firth,
R. Malinowski, B. & Schapera, I.) 143–146 (K. Paul, Trench, Trubner & Co, 1934).
46. de la Torre, I. The origins of stone tool technologyin Africa: a historical perspective.
Phil. Trans. R. Soc. B. 366, 1028–1037 (2011).
47. Shea, J. Lithic modes A-I: a new framework for describing global-scale variationin
stone tool technology illustrated with evidence from the East Mediterranean
Levant. J. Archaeol. Method Theory 20, 151–186 (2013).
48. Hovers, E. in Origins of Human Innovation and Creativity (ed. Elias, S.) 51–68
(Elsevier, 2012).
49. Toth, N., Schick, K. & Semaw, S. in The Oldowan: Case Studies into the Earliest Stone
Age (eds Toth, N. & Schick, K. D.) 155–222 (Stone Age Institute Press, 2006).
50. Piperno, M. in Hominidae: Proc. 2nd Intl Congr. Human Paleontol. 1987 189–195
(Jaca Books, 1989).
Supplementary Information is available in the online version of the paper.
Acknowledgements We thank the office of the President of Kenya, the Ministry of
Education, Science and Technology, the National Council for Science and Technology
(NCST/RCD/12B/012/25) and the National Museums of Kenya for permission to
conduct research. Funding was provided by the French Ministry of Foreign Affairs
(Nu681/DGM/ATT/RECH, Nu986/DGM/DPR/PRG), the French National Research
Agency (ANR-12-CULT-0006), the Fondation Fyssen, the National Geographic Society
(Expeditions Council #EC0569-12), the Rutgers University Research Council and
Center for Human Evolutionary Studies, and INTM Indigo Group France. We thank the
Turkana Basin Institute and Total Kenya Limited for logistical support and the GeoEye
Foundation for satellite imagery; the Turkana communities from Nariokotome,
Kokiselei and Katiko for field assistance, and the 2011-12 WTAP team, S. Kahinju,
P. Egolan, L. P. Martin, D. Massika, B. K. MulwaS. M. Musyoka, A. Mutisiya,J. Mwambua,
F. M. Wambua, M. Terrade, A. Weiss, R. Benitez, S. Feibel. M. Leakey and F. Spoor
supplied information on hominin fossils,and I. de la Torre and E. Hovers provided lithic
assemblage data. We are very grateful to A. Brooks, I. de la Torre, J. Shea, R. Klein and
M. Leakey for comments on earlier drafts. We also thank the Zoller & Fro
¨hlich GmbH
company, Ch. Fro
¨hlich and M. Reinko
¨ster, Autodesk and Faro (T. O’Mahoney,
K. Almeida Warren and T. Gichunge) for technical support with scanning and
J. P. Chirey for photographic assistance.
Author Contributions S.H. and J.E.L. directed field research and co-wrote the overall
paper. C.S.F., C.J.L., A.L. and X.B. recorded sedimentological and stratigraphic data,
conducted geological mapping, and wrote sections of the paper. C.S.F. interpreted
tephra data. C.J.L. interpreted paleomagnetic data. S.P., J.-Ph.B., S.L., C.K. and L.L.
conductedpaleontological survey.S.P., J.-Ph.B. and L.L.analysed and interpretedfossil
material.L.L. directed scanningof artefacts. S.P. laser scannedartefacts and excavation
surfaces, and wrote sections of the paper. R.L.Q. interpreted isotopic data and wrote
sections of the paper. C.S.F., C.J.L., R.L.Q., R.A.M., J.D.W. and D.V.K. analysed geological
samples. G.D. developed protocols for tool replication experiments and wrote sections
of the paper. S.H., H.R., N.T., M.B., S.C., S.L. and C.K. conducted archaeological survey
and excavation. S.H., H.R., A.A., N.T. and M.B. analysed and interpreted lithic material
and wrote sections of the paper. M.B. performed lithic replication experiments. S.C.
provided spatial data. S.L. discovered the LOM3 site.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the onlineversion of the paper. Correspondence
and requests for materials should be addressed to S.H.
(sonia.harmand@stonybrook.edu) or J.L. (jason.lewis@stonybrook.edu).
21 MAY 2015 | VOL 521 | NATURE | 315
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

METHODS
Paleomagnetic analyses. All samples from the Lomekwi outcrops were collected
from fresh surfaces uncovered by digging into the exposures for at least 20 cm.
Before each hand-cut block was extracted, in situ azimuths and dips were
recorded on a sample using a compass-inclinometer. Samples were taken typ-
ically at nominal 1 m vertical stratigraphic intervals, or as the distribution of fine-
grained strata allowed. Two sections were sampled, separated from each other by
about 1 km north to south across the landscape (Extended Data Fig. 2).
Overlapping Sections 1 and 2 (Fig. 3a) are each composed of a coarsening upward
succession of mudstones abruptly overlain by gravels and followed by a thick unit
of gravels and mudstones, which likely records a lacustrine regression and the
emplacement of a prograding alluvial fan. Inset in Fig. 3a shows stratigraphic
thickness of composite section plotted against key chronostratigraphic levels
(a-Tulu Bor (a-TB), 3.44 60.02; Toroto Tuff, 3.31 60.02 Ma; C2An.3n/.2r
boundary, 3.33 Ma
31
; Lokalalei Tuff, 2.53 60.02).
At Section 1 (Fig. 3a), sampling began at about 10 m below the lowermost
stratigraphic level of the a-Tulu Bor Tuff. Sampling continued upwardly from
the a-Tulu Bor Tuff for another 35 m, for a total of ,45 m sampled. At Section 2
(Fig. 3a), sampling commenced at the Tororo Tuff. Sampling started upwardly
from the Toroto Tuff for about 10 m to the level of the archaeological horizon at
LOM3, and then proceeded upwardly for another 35 m for a total sampled stra-
tigraphic thickness of about 45 m at Section 2.
For laboratory analyses, samples were cut into standard cube-shape specimens
(,10 cc) using a lapidary saw and sandpaper. All magnetic remanence measure-
ments were made with a 2G DCSQUID rock magnetometer in the shielded room
at the Paleomagnetics Laboratory of Lamont-Doherty Earth Observatory
(Columbia University). The natural remanent magnetization (NRM) of a spe-
cimen was subjected to progressive Thermal Demagnetization (TD) using 14 to
17 steps at 100, 50 and 25 uC increments in the temperature range of 100–700 uC.
Data from consecutive high-temperature steps were used for principal compon-
ent analysis (PCA
51
) to fit least-square lines tied to the origin for the final demag-
netization trajectories defining the characteristic remanent magnetization
(ChRM) as revealed on orthogonal projection plots (Extended Data Fig. 3a).
Magnetic susceptibility values were determined with a Bartington MS2B instru-
ment for each specimen initially and after each TD heating step to monitor any
laboratory-induced magnetochemical alteration. The virtual geomagnetic pole
(VGP) latitude corresponding to the ChRM direction was used to determine
the magnetostratigraphic polarity sequence. In Fig. 3a, filled black circles joined
by lines (isolated red squares) denote accepted (rejected) data with maximum
angular deviation (MAD) values ,15u(.15u) from principal component ana-
lyses. Characteristic remanent magnetizations were isolated after the removal of a
pervasive normal polarity overprint unblocked by a TD range of 600–670 uC for
the coarse alluvial fan strata (essentially all of Section 2 above the Toroto Tuff)
and a TD range of 400–550 uC for the finer strata (for example, mudstones from
the lower part of Section 1).
Pedogenic carbonate stable carbon isotopic analysis. Sedimentological field
analysis identified eleven paleosols with discernible preserved B
K
horizons. Ten
paleosols were sampled from Section 2011-1, and one from 2011-2 (Extended
Data Fig. 2). Carbonate nodules were extracted from paleosols .30 cm below the
contact with overlying stratum with vertic features within peds showing slick-
ensided surfaces. Twenty-four cross-sectioned nodules (five from one paleosol at
LOM3, 2011-2) were sampled witha 0.5 mm carbide drill bit (Foredom Series) and
loaded into v-vials for single acid baths (multi-prep device). Forty-seven isotopic
analyses were conducted on a Micromass Optima mass spectrometer in the
Department of Earth and Planetary Sciences at Rutgers University. Samples were
reacted at 90 uC in 100% phosphoric acid for 13 min. d
13
C
VPDB
values are reported
in the standard per mil (%)notation:5(R
sample
/R
standard
–1)31000, relative to
Vienna-Pee Dee Belemnite through analysis of laboratory standard NBS-19
(Extended Data Fig. 4). Analytical error is 60.05%. Using methods of ref. 32,
we subtracted14%from the d
13
C
VPDB
values of pedogenic carbonate to convert to
the isotopic equivalent of organic carbon (d
13
C
om
) and used the equation:
e
wc
5{sin[21.06688 20.08538(d
13
C
om
)]}
2
to generateestimates of fractionwoody
canopy cover for classification into UNESCO categories of African vegetation.
Categories were taken from White
52
and have the following d
13
C
VPDB
value ranges
of pedogenic carbonates
32
: (1) forest: continuous stand of trees at least 10-m tall
with interlocking crowns with greater than 80% woody cover (d
13
C
VPDB
:.211.5
%), (2) woodland/bushland/thicket/shrubland: woodland is anopen stand of trees
at least 8 m tall with woody cover .40% and a field layer dominated by grasses;
bushland is an open stand of bushes between 3m and 8m tall with woody cover
.40%; thicket is a closed stand of bushes and climbers between 3 m and 8m tall;
shrubland is an open or closed stand of shrubs up to 2 m tall (d
13
C
VPDB
:211.5 to
26.5%), (3) wooded grassland: land covered with grassland and has 10–40% tree
or shrub cover (d
13
C
VPDB
:26.5 to 22.3 %) and (4) grassland: land covered with
herbaceous plants with less than 10% tree and shrub cover (d
13
C
VPDB
:,22.3 %).
We also calculated percent C
4
biomass using a simple linear mixing model assum-
ing 212%and 226%as the C
4
and C
3
end members, respectively
53
.
Site scanning. To document the uncovering of the in situ artefacts and fossils
during the excavation, we took frequent 3D scans of the surface of individual
squares with the OptiNum RE handheld device (manufactured by Noomeo
Products, France) with a maximum spatial resolution of 300 mm. Additionally,
thanks to a collaboration between Zoller & Fro
¨hlich GmbH and Autodesk, we had
access to a recently developed high-resolution industrial 3D scanner operated by
M. Reinko
¨ster. This scanner was able to scan the entire site, registering 500,000
3D points each second, with a spatial resolution of ,3,000 mm, and recording for
several minutes continuously. After the laser scan a 3D photo was taken which
can be draped around the scan. The site was scanned in this manner after each day
of excavation. In this way, a high-resolution 3D digital model can be created for
the entire site, and individual squares, showing the evolution of the excavation
and the original context and gradual uncovering of the in situ artefacts and fossils.
Stone tool scanning. A representative sample of the LOM3 artefacts were
scanned at the National Museums of Kenya and the Turkana Basin Institute
facility in Turkwel, using a LMI Technologies R3 Advance portable structured
light scanner (LMI Technologies, Vancouver, Canada), calibrated to the size of
the objects in question, with the calibration grids being accurate to 50 mm. For
colour texture overlay, a Canon 600D/Rebel T3i SLR digital camera was also
calibrated with the scanner and images from this formed the base of the colour
texture. The textured files are saved in .obj format and non textured files (for 3D
printing or similar purposes) are saved in .stl format. These scans and 3-D digital
models are available at (http://africanfossils.org/search).
Sample size. No statistical methods were used to predetermine sample size
51. Kirschvink, J. L. The least-squares line and plane and the analysis of
palaeomagnetic data. Geophys. J. Int. 62, 699–718 (1980).
52. White, F. The vegetation of Africa, a descriptive memoir to accompany the
UNESCO/AETFAT/UNSO vegetation map of Africa. UNESCO. Nat. Resour. Res. 20,
1–356 (1983).
53. Fox, D. L. & Koch, P. L. Carbon and oxygen isotopic variability in Neogene paleosol
carbonates: constraints on the evolution of the C4-grasslands of the Great Plains,
USA. Palaeogeogr. Palaeoclimatol. Palaeoecol. 207, 305–329 (2004).
54. Levin, N. E., Quade, J., Simpson, S. W., Semaw, S. & Rogers, M. J. Isotopic evidence
for Plio-Pleistocene environmentalchange at Gona, Ethiopia. Earth Planet.Sci. Lett.
219, 93–110 (2004).
55. Levin, N. E. Compilation ofEast Africa soil carbonate stableisotope data. Integrated
Earth Data Applications http://dx.doi.org/10.1594/IEDA/100231 (2013).
56. Cerling, T. E., Bowman, J. R. & O’Neil, J. R. An isotopic study of a fluvial-lacustrine
sequence: the Plio-Pleistocene Koobi Fora sequence, East Africa. Palaeogeogr.
Palaeoclimatol. Palaeoecol. 63, 335–356 (1988).
57. Levin, N. E., Brown, F. H., Behrensmeyer, A. K., Bobe, R. & Cerling, T. E. Paleosol
carbonates from the Omo Group: isotopic records of local and regional
environmental change in East Africa. Palaeogeogr. Palaeoclimatol. Palaeoecol. 307,
75–89 (2011) CrossRef.
58. Wynn, J. G. Influence of Plio-Pleistocene aridification on human evolution:
evidence from paleosols from the Turkana Basin, Kenya. Am. J. Phys. Anthropol.
123, 106–118 (2004).
59. Kingston, J. D. Stable isotopic evidence for hominid paleoenvironments in East Africa.
Ph.D. Thesis, Harvard Univ. (1992).
60. Aronson, J. L., Hailemichael,M. & Savin, S. M. Hominid environments at Hadar from
paleosol studies in a framework of Ethiopian climate change. J. Hum. Evol. 55,
532–550 (2008).
61. Wynn, J. G. et al. Geological and palaeontological context of a Pliocene juvenile
hominin at Dikika, Ethiopia. Nature 443, 332–336 (2006).
62. Semaw, S., Rogers, M. J. & Stout, D. In The Cutting Edge: New Approaches to the
Archaeology of Human Origins (eds Schick, K. D. & Toth, N.) 211–246 (Stone Age
Institute Press, 2009).
63. Hovers, E. In The Cutting Edge: New Approaches to the Archaeology of Human Origins
(eds Schick, K. D. & Toth, N.) 137–150 (Stone Age Institute Press, 2009).
64. de la Torre, I. & Mora, R. Technological Strategies in the Lower Pleistocene at Olduvai
Beds I & II. ERAUL 112 (2005).
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

M
L
K
J
I
H
20
19
18
17
16
15
N
G
N
01 5 m
Fauna in situArtefacts in situ
Artefacts on surface
Artefacts from 2011 excavation
Retted core I16-3
Fauna on surface
Geological test pit
Geological section in Fig. 2c
450.00
449.50
449.00
449.00
450.50
448.50
448.00
MSL
452
450
448
449.65 MSL
MSL
452
450
448
2 m
NE SW
a
b
Extended Data Figure 1
|
Map and schematic section at LOM3. a, Map
showing xy coordinates of artefacts and fossils recovered in situ and from the
surface at the site in 2011 and 2012. b, Schematic section showing vertical
distribution of in situ artefacts and those located in the slope deposit at the
excavation. Key is the same for both figures.
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 2
|
Geology of the LOM3 site. a, Stratigraphic sections around LOM3 (locations in b), showing relationship of site to marker tuffs and
lithofacies. Sections aligned relative to top of flat-pebble conglomerate unit. b, GPS coordinates of stratigraphic sections (WGS84 datum).
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 3
|
Paleomagnetic data. a, Representative vector end-
point plots of natural remanent magnetism thermal demagnetization data from
specimen Toroto Tuff, tt2, wt59, wt50, wt45, wt36. Open and closed symbols
represent the vertical and horizontal projections, respectively, in bedding
coordinates. TD treatmentsteps: NRM, 100u, 150u, 200u, 250u, 300u, 350u, 400u,
450u, 475u, 500u, 525u, 550u, 575u, 600u, 625u, 650u, 660u, 670u, 675u, 680u, 690u,
and 700u. V/M 510 denotes a ,10 cc cubic specimen. b, Equal-area
projections for Section 1 (left) and Section 2 (right) of the lower Lomekwi
Member (see Fig. 3a). Open and closed symbols are projected onto the upper
and lower hemisphere, respectively, in bedding coordinates. Plotted are ChRM
sample-mean directions for accepted samples only (that is, those with MAD
values ,15u). Overall mean directions were calculated after inverting the
northerly (normal) directions to common southerly (reverse) polarity.
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 4
|
Paleoenvironmental reconstruction through
pedogenic carbonate stable carbon isotopic analysis. a, LOM3 paleosol
d
13
C
VPDB
values (%)61s, number of analyses, fraction woody canopy cover
(e
wc
) and percent C
4
biomass contribution to soil CO
2
. Asterisk denotes
nodules sampled at the LOM3 site, 2011-2b (see Extended Data Fig. 2a).
b, Schematic box and whisker plots of e
wc
from the LOM3 (3.3 Ma, this study)
and Gona
33,54,55
(Busidima Fm, 2.5–2.7 Ma) lithic sites and other East African
hominin localities from 3.2–3.4 Ma
54–61
relative to UNESCO structural
categories of Africanvegetation
32,52
. Grey box denotes 25th and 75th percentiles
(interquartile range); whiskers represent observations within upper and lower
fences (1.5 3interquartile range); black line shows mean value; grey line
equals median value; black circles indicate mild outliers. c, Summary statistics
of paleosol d
13
C
VPDB
values and e
wc
from LOM3 (3.3 Ma) and Gona
33,54,55
(2.5–2.7Ma) lithic sites and other East African hominin localities from
3.2–3.4 Ma
54–61
.LOM3d
13
C
VPDB
values are significantly lower than those
from the Busidima Formation at Gona (t-test, P,0.001) and have a mean
value that indicate 18% more woodycanopy cover. When compared to paleosol
d
13
C
VPDB
values of the Koobi Fora, Nachukui, Chemeron, and Hadar
formations from 3.2 to 3.4 Ma, LOM3 d
13
C
VPDB
values are not significantly
different (one-way ANOVA, P.0.05).
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 5
|
Gradual uncovering of core I16-3 from
in situ
pliocene sediment. a, Photograph showing square I16 at the beginning of
excavation. Yellow line indicates north wall of square (July 16, 2011, 12.14
p.m.). b, Close-up of square I16 indicating complete burial of as-yet-uncovered
artefact I16-3 (12.14p.m.). c, Square I16 after excavation had begun and artefact
I16-3 was initially exposed (2:11 p.m.). d, Close-up of artefact I16-3 after being
initially exposed (2.12 p.m.). e, Close-up of artefact I16-3 after further
excavation (3.02 p.m.). f, Square I16 after further excavation (5.32 p.m.).
g, Close-up of artefact I16-3 after further excavation (5.34 p.m.). h, Close-up of
artefact I16-3 after being completely freed from the surrounding matrix and
flipped over for inspection (5.36 p.m.). i, Close-up of impression from under
artefact I16-3 (5.47 p.m.).
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 6
|
Photos of selected LOM3 artefacts compared with
similar experimental cores. Together with the technological analysis of the
archaeological material, our replication experiments suggest that the LOM3
knappers were using passive hammer technique, in which the core,usually held
in both hands, is struck against a stationary object that serves as the percussor
34
(also referred to as on-anvil,block on block or sur percuteur dormant
35
) and/or
bipolar technique, in which the core is placed on an anvil and struck with a
hammerstone
34
.a, Unifacial passive hammer cores. Left is archaeological piece
LOM3-2012 surf 106 (2.04 kg); right is experimental piece Expe 55 (3.40 kg)
produced using thepassive hammer technique. Selectionof relatively flat blocks
with natural obtuse angles. The flake removal process starts from a slighly
prominent part of the block (white arrowsshow the direction of removals). The
removals tend to be invasive. The flakedsurface forms a semi-abrupt angle with
the platform surface. A slight rotation of the block ensures its semi-peripheral
exploitation. b, Unifacial bipolar cores. Left are archaeological pieces LOM3-
2012-H18-1 (left, 3.45kg) and LOM3-2012 surf 64 (right, 2.58 kg); right are
experimental pieces Expe 39 (left, 4.20 kg) and Expe 24 (right, 2.23 kg)
produced using the bipolar technique. The block selected are thicker and more
quadrangular in shape with natural angles <90u. Flakes are removed from a
single secant platform (white arrows show the direction of removals). The
flaked surface forms an abrupt angle with the other faces of the block. Impacts
due to the contrecoups (white dots) are visible on the opposite edge from the
platform.
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 7
|
Photographs of selected LOM3 artefacts.
a, Passive element/anvil (LOM3-2012 surf 50,15kg). Heavy sub-rectangular
block displaying flat faces and therefore a natural morphology and weight
which would enable stability. b, Hammerstone showing isolated impact points
(LOM3-2012 surf 33, 3.09 kg) and c, Hammerstone showing isolated impact
points (LOM3-2012 surf 54, 1.63 kg), associated with a flake-like fracture on
one end.
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Table 1
|
Numerical data on the LOM3 lithic assemblage (2011, 2012).
a
b
a, Initial categorisation of the lithic components. b, Breakdown of lithic raw materials in the LOM3 assemblage.
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Table 2
|
Comparison of whole flake and core dimensions between LOM3, early Oldowan sites and chimpanzee stone tool
sites
Length Width Thickness
Site Age (Ma) Ref. NMean Std Min Max Mean Std Min Max Mean Std Min Max Geo. Mean Mean Mass
FLAKES
LOM3 3.3 26 120 48.8 19 205 110.1 40.7 19 185 43.9 23.4 690 59.9 842.4 (N=26)
OGS7 2.6 62 73 39.1* 14.3 13 80 37.1* 14.1 13 74 12.7* 5.07 326 14.10 18.9 (N=76)‡
EG10 2.6 62 114 37.38* 15.34 14 78 34.63* 13.74 14 78 13.18* 6.26 333 13.74 24.9 (N=72)‡
EG12 2.6 62 62 34.5* 12.84 15 66 35.55* 13.23 19 66 12.13* 5.76 430 13.23 21.5 (N=61)‡
A
L894 2.36 63†1048 35.9* 23.63 6134 25.07* 17.57 2106 7.98* 6.4 145 17.1
LA2C 2.34 16 500 38* 15 12 96 35* 14 7128 11* 5 3 28 14.00
Omo57 2.34 14†44 24.75* 10.546 10 58 20.36* 6.851 10 44 7.73* 4.008 118 6.85
Omo123 2.34 14†110 20.8* 7.495 750 17.79* 6.485 638 5.9* 2.792 116 6.49
DK > 1.84 64 115 40.18* 14.803 18 111 37.41* 11.215 17 71 11.89* 5.404 429 11.22
FLKZinj 1.76-1.84 64 125 36.78* 12.13 16 82 32.88* 11.59 476 11.51* 5.45 436 11.59
Noulo§.0043 40 535* 20.62 15 70 48* 27.06 15 90 11.6* 3.21 815 20.15
CORES
LOM3 3.3 83 167 23.4 132 260 147.8 23.1 90 210 108.8 21.8 61 170 139 3096.4 (N=81)
OGS7 2.6 62 744.14* 13.68 28 67 59* 8.54 45 70 37* 8.2 22 49 45.85 78‡
EG10 2.6 62 16 83.33* 10.34 69 105 60.9* 9.18 44 80 45.27* 12.36 30 69 61.25 232‡
EG12 2.6 62 774.45* 8.72 58 93 59.73* 8.06 49 77 43.73* 77.4 25 53 57.94 194 (N=9)‡
A
L894 2.36 63†38 75.01* 30.32 19.31 136.3 55.33* 22.54 12.21 94.9 35.87* 18.1 7.92 78.2 53.00
LA2C 2.34 16 70 66* 18 39 123 52* 14 32 95 32* 12 12 78 47.9
Omo 57 2.34 14†737.4* 8.81 25 52 28.8* 7.313 22 40 16.5* 4.721 11 24 26.10
Omo 123 2.34 14†11 30.5* 12.193 17 56 22.27* 8.186 13 42 13.5* 4.569 924 20.93
DK > 1.84 64 69 67.93* 19.146 30 117 62.78* 17.992 25 100 48.25* 14.435 18 81 59.04
FLK Zinj (lava only) 1.76-1.84 64 49 76.35* 12.57 53 95 78.85* 16.26 49 112 59* 12.3 37 87 70.82
Dimensions are in mm, Mass in g. *Denotes significant difference with LOM3 (t-test, one-tailed, P,0.0001). Given the small sample sizes and potential non-normal nature of stone tool measurements, a non-
parametric test such as Mann–Whitney would be preferable, but this would require access to the raw measurement data from the other Oldowan sites, access to which is currently beyond the scope of this work. The
Student’s t-test is very robust, however, as deviations from normality do not affect it very much, and it is currently the only option when working with published data summaries.
{The summary data from this publication was not in the correct format for direct comparison with LOM3, so information for this table was provided directly by the authorin the form of personal communication.
{Data from ref. 17, hence differing sample sized from ref. 62.
1Dimensions of accidentally produced flakes from chimpanzee nut-cracking activity are included here for comparative purposes, although a direct technological comparison would be inappropriate as those
pieces are not the result of intentional flake manufacture and do not bear the classic technological flake characteristics like those from LOM3 and early Oldowan sites.
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Table 3
|
Comparison of anvils and percussors dimensions found at LOM3 site with anvils and percussors used by non-human
primates in Bossou (wild chimpanzees, Pan troglodytes verus from ref. 41)
Dimensions are in cm, Mass in g. *Denotes significant difference with LOM3 (t-test, two-tailed, P,0.0199). Given the small sample sizes and potential non-normal nature of stone tool measurements, a non-
parametric test such as Mann–Whitney would be preferable but this would require access to the raw measurement data from ref. 41, access to which is currently beyond the scope of this work. The Student’s t-test is
very robust, however, as deviations from normality do not affect it very much and it is currently the only option when working with published data summaries.
{N531.
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved