The Manipulative Complexity of Lower Paleolithic Stone
*, Dietrich Stout
*, Jan Apel
, Bruce Bradley
1Department of Bioengineering and Department of Computing, Imperial College London, London, United Kingdom, 2Department of Anthropology, Emory University,
Atlanta, Georgia, United States of America, 3Department of Archaeology and Osteology, Gotland University College, Visby, Sweden, 4Department of Archaeology, Exeter
University, Exeter, United Kingdom
Early stone tools provide direct evidence of human cognitive and behavioral evolution that is otherwise
unavailable. Proper interpretation of these data requires a robust interpretive framework linking archaeological evidence to
specific behavioral and cognitive actions.
Here we employ a data glove to record manual joint angles in a modern experimental
toolmaker (the 4
author) replicating ancient tool forms in order to characterize and compare the manipulative complexity
of two major Lower Paleolithic technologies (Oldowan and Acheulean). To this end we used a principled and general
measure of behavioral complexity based on the statistics of joint movements.
This allowed us to confirm that previously observed differences in brain activation associated
with Oldowan versus Acheulean technologies reflect higher-level behavior organization rather than lower-level differences
in manipulative complexity. This conclusion is consistent with a scenario in which the earliest stages of human technological
evolution depended on novel perceptual-motor capacities (such as the control of joint stiffness) whereas later
developments increasingly relied on enhanced mechanisms for cognitive control. This further suggests possible links
between toolmaking and language evolution.
Citation: Faisal A, Stout D, Apel J, Bradley B (2010) The Manipulative Complexity of Lower Paleolithic Stone Toolmaking. PLoS ONE 5(11): e13718. doi:10.1371/
Editor: Michael D. Petraglia, University of Oxford, United Kingdom
Received February 3, 2010; Accepted September 27, 2010; Published November 3, 2010
Copyright: ß2010 Faisal et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by the Commission of the European Communities Research Directorate-General Specific Targeted Project number 029065,
Hand to Mouth: A framework for understanding the archaeological and fossil records of human cognitive evolution. The funders had no role in study design, data
collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: email@example.com (AF); firstname.lastname@example.org (DS)
Lower Paleolithic stone tools provide under-utilized evidence of
the timing and context of human cognitive evolution that is not
available from comparative studies. Of particular interest is the
relative contribution of perceptual-motor vs. cognitive adaptations
at various points in human technological evolution [e.g. 1,2,3,4].
Increasingly complex prehistoric toolmaking methods through
time offer insight into evolving capacities but require a robust
interpretive framework linking archaeological evidence to specific
behavioral and cognitive actions. We recorded the joint and
abduction angles of the hand digits of a modern experimental
toolmaker (the 4th author) replicating ancient stone tools in order
to better characterize the manipulative complexity of two major
Lower Paleolithic technologies, and to compare this with recent
functional brain imaging studies of the neural bases of these same
The technologies studied were Oldowan flake production and
Late Acheulean handaxe making Figure 1A–C), representing the
beginning and end of the Lower Paleolithic. The Lower Paleolithic
itself encompasses some 90% of human prehistory, beginning with
the first stone tools 2.6 million years ago (mya) and lasting more
than 2 million years. The earliest known tools are assigned to the
Oldowan Industry and consist simply of sharp stone flakes struck
from river cobbles through direct percussion with another stone
[7,8]. Nevertheless, their production involves considerable per-
ceptual-motor skill [5,9]. After about 1.7 mya , intentionally
shaped Acheulean tools began to appear, including large,
teardrop-shaped cutting tools known as ‘handaxes’ (Figure 1). By
the Late Acheulean some 0.5 mya, these forms had achieved a
high level of refinement and standardization, reflecting increas-
ingly elaborate and skill-intensive shaping techniques .
The development of intentional shaping in the Acheulean has
been considered a key event in human cognitive evolution,
reflecting new capacities for the ‘‘imposition of arbitrary form’’
 and the presence of more complex mental  and/or
procedural ‘‘templates’’ . Such observations have led various
researchers to consider possible links between stone toolmaking
and language evolution [e.g. 11,14,15]. More recently, functional
brain imaging studies of experimental Oldowan  and
Acheulean  toolmaking have provided direct evidence of neural
overlap between language and toolmaking in inferior frontal
cortex (Figure 1D). Oldowan toolmaking was associated with
activation of left ventral premotor cortex, a region known to be
involved in both manual grip coordination [16,17] and phono-
logical processing [18,19,20]. Acheulean toolmaking differed from
PLoS ONE | www.plosone.org 1 November 2010 | Volume 5 | Issue 11 | e13718
Oldowan in producing additional activity in the right hemisphere
(Figure 1D), including the supramarginal gyrus of the inferior
parietal lobule, the right ventral premotor cortex, and the right
hemisphere homolog of anterior Broca’s area (Brodmann area
[BA] 45) , a region that is bilaterally involved in higher-order
hierarchical cognition  and which is specifically implicated in
the processing of linguistic context and prosody (intonational
What these experiments did not make clear was whether
increased right hemisphere activation reflected increased demands
for grasp control in the contralateral hand, distinctive right
hemisphere contributions to the cognitive control of complex
action sequences [22,23], or both. During Paleolithic toolmaking
(Figure 1A–C), the non-dominant hand plays a critical role
supporting and orienting the stone ‘core’ from which flakes are
detached by relatively invariant ballistic strikes from a ‘hammer-
stone’ held in the dominant hand. We wanted to further
investigate the role of the non-dominant hand in Oldowan and
Acheulean technologies in order to clarify its relationship with the
observed contralateral brain activation.
Previous experimental studies of stone toolmaking have
provided insight into the manual grips [24,25,26], elementary
movements [27,28,29] and kinematic synergies  involved in
this complex perceptual-motor skill. To date, however, such
studies have either been based on qualitative grip typologies (as
characterized ‘by-eye’) or been confined to the movements of the
striking arm proximally from the wrist. We employed a data glove
to record digit joint angles in the left, core-holding hand of an
expert right-handed toolmaker during Oldowan flake production
and Late Acheulean handaxe shaping. The high precision
recording of manual joint angles enabled us to quantify grip
diversity and complexity in an objective, mathematically princi-
pled manner (Figures 2, 3, 4, 5). This allowed us to directly
compare the manipulative complexity of Oldowan and Acheulean
To confirm the capacity of this novel method to differentiate
manual activities, we collected additional data from two control
tasks (Figure 6), both involving manipulative actions of the left
hand. In the first task, small nut-sized objects (‘‘widgets’’) were
grasped and transferred between containers. This involved the use
of a 4-digit pad-to-pad grip, and loosely approximates the kind of
small object manipulation involved in primate manual foraging.
Functionally, it is analogous to the unimanual grasping and
positioning of nuts on an anvil during chimpanzee nut-cracking,
although the manual anatomy and preferred grips clearly differ
between species. In the second task, Styrofoam boxes were
repeatedly stacked and un-stacked in a carefully aligned column.
These open boxes afford a wide array of different grasps during
manipulation, and provide an example of structured interaction
with complex artifacts in the modern built environment.
To systematically quantify the grips we used a CyberGlove I
data glove that recorded the angles of the joints of each digit and
the abduction angles between digits. In total, the glove data has 18
sensors, spanning an 18-dimensional space of joint angles for the
hand. We found that throughout the toolmaking process of both
Oldowan and Acheulean techniques the proportion of time the
individual joints spend in specific configurations was similar and
displayed a Gaussian (‘‘Normal’’) distribution (Figure 3).
A major challenge in the quantitative analysis of behaviour is its
variability . This is clearly visible in the histograms of joint
angles during Oldowan and Acheulean toolmaking (Figure 3). For
this study, we had to consider that each toolmaking sequence has
unique, uncontrollable elements, such as the shape and internal
properties of the core, which are gradually revealed during the
toolmaking process. The grips would thus differ in detail between
cores, as well as gradually changing for a single core as flake
removal progressively alters core form. Furthermore, we might
expect systematic differences in variability across different
Figure 1. Lower Palaeolithic toolmaking. A stone ‘core’ (A) is struck with a hammerstone (B) in order to detach sharp stone ‘flakes’. In Oldowan
toolmaking (C, top) the detached flakes (left in photo) are used as simple cutting tools and the core (right in photo) is waste. In Acheulean toolmaking
(C, bottom), strategic flake detachments are used to shape the core into a desired form, such as a handaxe. Both forms of toolmaking are associated
with activation of left ventral premotor cortex (PMv), Acheulean toolmaking activates additional regions in the right hemisphere, including the
supramarginal gyrus (SMG) of the inferior parietal lobule, right PMv, and the right hemisphere homolog of anterior Broca’s area: Brodmann area 45
(BA 45) (Imaging data adapted from ).
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 2 November 2010 | Volume 5 | Issue 11 | e13718
toolmaking methods. A simple count of qualitatively defined grip
types would thus provide insufficient information to evaluate the
true manipulative complexity of this real-world activity.
We instead took a computational-statistical view of motor
behaviour to produce a quantitative estimate of the complexity of
hand configurations. Previous studies of human motor control
show that normal hand behavior uses only a small subset of the
possible hand configurations  and that we can extract these
using dimensionality reduction techniques. Dimensionality reduc-
tion techniques can be illustrated by considering the index finger,
which has 3 joints controlled by 5 muscles. Describing the flexing
behavior of this finger requires a priori 3 values (‘‘dimensions’’).
However, in specific movements like making a fist, as we flex one
joint of the index finger we flex the other two joints in a highly
coordinated manner. Thus, we would require, in principle, a single
value to describe the configuration of the full finger. If this were
systematically the case for all our movements, we would just
require 1 dimension to describe the configuration of the finger. In
reality this dimensionality varies somewhere between 1 and 3,
depending on the amount of coordination and correlation of the
hand’s joints in our movements and actions. This is what we
systematically measured for the actions of tool making using the
We used Principal Component Analysis (PCA) as dimension-
ality reduction technique, which is well suited given the Gaussian
distribution of the joint statistics (Figure 3). PCA reduces a set of
correlated variables, the joint angles of the hand, into a set of
uncorrelated variables called principal components (Figure 4,
inset). The first principal component accounts for as much of the
variability (as quantified by the variance) in the data as possible, as
does each succeeding component for the remaining variability. We
therefore use PCA as a measure for the complexity of hand
configurations, by measuring how many principal components can
explain how much variability in the data. For example, a simple
behavior, e.g. curling and uncurling a hand into a fist, would
reveal a single dominant principal component as all 5 fingers (and
each finger’s joint) move in a highly correlated manner. In
contrast, a complex behavior, such as expert typing on a keyboard
Figure 2. Stability of the core-holding (left hand) and velocity profile of the hammerstone wielding hand (right hand), as well as
manual video annotation of a ‘percussion’ event (Top). Time series of the 18 joint sensors on the data glove (photo left). Period of rapid hand
configuration changes (blue shaded region) frame a period of stability (clear region). This stability period corresponds to the period just before and
during a percussion strike, as obtained from the manually annotated video. (Bottom) Velocity profile of the hammerstone wielding hand aligned with
the plot above. The ‘percussion’ event is clearly visible in the velocity spike (top blue arrow).
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 3 November 2010 | Volume 5 | Issue 11 | e13718
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 4 November 2010 | Volume 5 | Issue 11 | e13718
would reflect five dominant principal components as each finger
moves independently from the others. In fact, a previous study that
measured human finger movements (i.e. joint angle velocities) over
several hours of normal modern-day activity showed that about
75% of the variance in the data could be explained by the first four
principal components .
From toolmaking experience and basic biomechanics, we
assume that grip stiffness of the hand holding the core (left hand)
should peak around the time of hammerstone impact (Figure 2).
Using a simple automatic annotation method, we performed
principal component analysis of the hand configuration during
high velocity impact events of the (right) hammerstone-wielding
hand. Plotting the variance explained versus principal components
shows (Figure 4), that across the five Acheulean and three
Oldowan sequences, the first principal component explained 25–
40% of the variance in the data. The first two principal
components explain about 50–62.5% of the variance. About
85% percent of the data can be explained by the first five principal
components. Moreover, for each principal component, there is no
significant difference between the Acheulean and Oldowan
variance explained. We tested this by conducting a t-test for each
principal component, assuming that the population of values for
that principal component from both Oldowan and Acheulean data
were drawn from a distribution with the same mean (DF = 7,
p,0.05 for all 18 principal components).
To further corroborate our findings, we manually annotated
video recordings of one Acheulean and one Oldowan tool
production episode. Eight different technical actions, including
striking actions, were recorded with sub-second precision. We
performed principal component analysis for the set of hand
configurations on which instantaneous ‘percussion’ events (fast,
hammer-strike like motions) or ‘light percussion’ events (fast,
chiseling-like motions) were annotated in the video. ‘Light
percussion’ is a form of edge modification used to facilitate the
removal of special ‘thinning’ flakes in Acheulean toolmaking and is
absent in Oldowan toolmaking. Plotting the variance explained
versus the number of principal components for Acheulean and
Oldowan toolmaking produced very similar results (Figure 5A).
The first principal component explains 30–40% of the variance;
the first and second principal components explain 55–60%. As in
the automatically annotated case, about 85% of the variance is
explained by the first five principal components.
Comparing the manual and the automated data directly
corroborates these findings (Figure 5B). The principal components
between the two methods of tagging the data are highly correlated,
as can be seen by plotting the variance explained from Acheulean
‘percussion’ (blue data), Acheulean ‘light percussion’ (black data),
and Oldowan ‘percussion’ (red data) from manual annotation data
against variance explained from automatically annotated data
from striking events in Acheulean (triangles) or Oldowan (circles)
(all pairwise correlation coefficients had r= 0.95). Thus, the results
indicate that the structure of non-dominant (i.e. core manipula-
tion) hand configurations for Acheulean vs. Oldowan percussions
is equally complex.
In addition to these findings and two confirm the validity of our
behavioral complexity measure, we performed two control
experiments involving everyday object manipulation tasks: small
object (‘‘widget’’) sorting and box stacking (Figure 6). We wanted
to compare the manipulative complexity of toolmaking with its
characteristic percussion events to these rather differently
structured control tasks. Therefore, to remove any bias we
analysed the full hand configuration time series of the toolmaking
sequence (and not just the subset of stable hand configurations)
Figure 4. Variability explained by principal components of
stable hand configurations (data from automatic annotation).
Curves show the cumulative sum of variance explained by increasing
numbers of principal components for 5 Acheulean and 3 Oldowan
reduction sequences. Principal component analysis is of Acheulean
(blue triangles, one for each of the corresponding 5 toolmaking
sequences and for each principal component) and Oldowan (red circles,
3 toolmaking sequences) hand configuration data of the core holding
(left) hand, when the hammerstone hand was moving faster than
0.5 m/s. Note, that some data points overlap for Principal Component 1
and higher order Principal Components. Inset: Conceptual drawing of
Principle Component Analysis (PCA). PCA is a linear transformation that
seeks to explain multiple, correlated dimensions (X1, X2) of variation in
the data (grey cloud) in terms of uncorrelated dimensions termed
principal components (PC1, PC2). This linearly uncorrelated represen-
tation can then be used to reduce the dimensionality of the 2-
dimensional data, e.g. describing the data set by using only PC1 as 1-
dimensional data set (effectively capturing the longitudinal character-
istic structure of the data). We do not reduce the dimensionality of the
data per se, but use the relative amount of variance explained by each
principle component as a characteristic value for the complexity of the
Figure 3. Probability distribution (or relative frequency or proportion of total time) of joint angles for 18 joint angles of the left
hand during Oldowan (blue dots) and Acheulean (red dots) toolmaking. Solid lines are Gaussian distributions with mean and standard
deviation matched to the empirical joint angle histograms. Plots have a logarithmic vertical axis, such that the data and the matching Gaussian
distributions appear as parabolas. The configuration of the hand was determined by the following joints (from left to right, top to bottom): the carpo-
metacarpal (ThumbTMJ), metacarpal-phalangeal (ThumbMPJ) and interphalangeal (ThumbIJ) joint angles for the thumb and the abduction angle
(ThumbAbduct) between the thumb and the palm of the hand, the metacarpal-phalangeal (MPJ) and proximal interphalangeal (PIJ) joint angles for
the four fingers, the three relative abduction angles between the four fingers (MiddleIndexAbduct, RingMiddleAbduct, PinkieRingAbduct), as well as
the arching and bending (during flexion/extension, ulnar/radial deviation) of the palm surface with respect to the wrist (PalmArch,
WristPitch,WristYaw). Note, the joint angle zero is relative to the joint angle on our calibration splints.
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 5 November 2010 | Volume 5 | Issue 11 | e13718
and compared these to the full time series of hand configurations
of the two control tasks. Results show that both control tasks have
lower complexity than stone toolmaking (Figure 6A). Widget
sorting was considerably less complex than the other 3
manipulation tasks, with the first 2 principal components
capturing 70% variance as compared to 50–53% in toolmaking
and box stacking. The box stacking task was somewhat more
complex, and closely approximates stone toolmaking across the
first 4 principle components. However, the lesser manipulative
complexity of box stacking is evident in the 5
which explain 5–10% more variance than in either Oldowan or
Acheulean toolmaking. Finally, we found that stone toolmaking
complexity as measured from the full time series and from the
percussion-triggered events was very similar (Figure 6B).
This series of results confirms that the manipulative complexity
of Oldowan and Acheulean toolmaking are indistinguishable.
Furthermore, we showed that complexity measures for stable hand
configurations during toolmaking and for the full time series
including manipulation between stable grips are very similar.
Finally, we demonstrated that toolmaking complexity is clearly
higher than the much simpler sorting task of nut-size ‘‘widgets’’
and box stacking tasks that we used as controls.
We used a data glove and electromagnetic position markers to
quantify the modern day reproduction of Lower Paleolithic stone
tools. To this end we used a novel, principled, and general
measure of behavioral complexity based on the statistics of
movements occurring during the task. The application of this
general technique to the specific investigation of the complexity
and diversity of left hand grips required to produce Acheulean and
Oldowan artifacts allowed us to make comparisons with previously
observed patterns of lateralized brain activation during stone
toolmaking (Figure 1). Increasing anatomical and functional
asymmetry is a key trend in human brain evolution, and apparent
left hemisphere dominance for both language and manual praxis
has inspired influential hypotheses linking human handedness, tool
use and language [e.g. 33]. Activation of left ventral premotor
cortex, a region also involved in phonological processing ,
during Oldowan toolmaking is consistent with these ideas and
suggests that early stone toolmaking (ca. 2.6 mya) could have
contributed to the evolution of neural substrates also important for
articulate speech . However, the increased right hemisphere
activity seen during Acheulean toolmaking is unexpected in this
An alternative framework [35,36] emphasizes the complemen-
tary roles of left and right hands in everyday human manipulative
behaviors (e.g. cutting bread, hammering a nail, writing on paper,
striking a match, washing dishes), in which the left hand typically
provides a stable postural support for the higher frequency actions
of the right hand. This may correspond to a similar hemispheric
‘‘division of labor’’ in the brain , with left hemisphere
preferentially involved with rapid, small-scale processing and right
hemisphere with larger-scale, longer-duration processing .
Indeed, it is becoming more widely appreciated that right
hemisphere plays a critical role in larger-scale prosodic and
contextual aspects of language processing [19,38] and in the
coordination of multi-step manual action sequences [22,39]. In
this framework, right hemisphere involvement in Acheulean
toolmaking suggests the presence of additional cognitive demands
for behavioral integration over time and the possibility of
Figure 5. (A) Cumulative variability (%) explained by principal components of stable hand configurations (from manual annotation). Data points
overlap for Principal Component(PC) 1 and higher order PCs. (B) Comparison of manually versus automatically annotated data. Plot of variance
explained by principal components analysis of stable hand configurations: Variance explained of all principal component calculated for manually
classified Acheulean percussion events (blue), Acheulean light percussion events (black) and Oldowan percussion events (red) plotted versus
corresponding principal components of automatically annotated Acheulean (triangle) and Oldowan (circle) percussion-like events. E.g. the curve with
black triangles, shows the principal components from 1 to 18 of the Acheulean light percussion events extracted by manual annotation (as horizontal
position of the black triangle) versus the corresponding principal components from 1 to 18 for the automatically extracted percussion events (as
vertical position of the black triangle). Note: automatic annotation did not distinguish between light percussion and percussion events. Correlation
coefficients were greater 0.95 for all curves confirms that manual and automatic annotation yield very similar results.
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 6 November 2010 | Volume 5 | Issue 11 | e13718
evolutionary interactions between (Late) Acheulean (ca. 0.5 mya)
toolmaking and discourse-level language processing .
A major challenge to such interpretations, however, is the
possibility that increased right hemisphere activity during
Acheulean toolmaking simply reflects an increased diversity/
complexity of left hand grips, without necessarily implicating
distinctive right hemisphere processing characteristics. In this case,
activation of right BA 45 would still imply increased demands for
the hierarchical organization  of (left hand) actions, however
the observed lateralization could simply be attributed to the
localization of cognitive control to the same hemisphere as task
execution . In order to test this hypothesis it was necessary to
quantify the diversity/complexity of left hand grips deployed
during Oldowan and Acheulean toolmaking in a principled,
quantitative manner allowing for direct comparison.
This is not a trivial undertaking because the details of the
required grips vary inherently as core shapes differ. Moreover, for
each core, the specific grip will have to vary during the
transformation of the core into a tool. The challenge was to
discover an underlying and quantitative simplicity which accounts
for and does not neglect variability in behavioral data, thus taking
a view we refer to as Bioinformatics of Behavior . We used a
Figure 6. (A) Cumulative variability explained by principle components (from the full set of the hand configuration time series data) for toolmaking
(Oldowan - circle, Acheulean - square) and control tasks (widget sorting - downward pointing triangle, box stacking - upward pointing triangle). Small
object manipulation is clearly less complex than toolmaking (greater variance explained by PC 1 and above). Box stacking is much more comparable,
but is still distinguished from PCs 6, 7,8 and onwards. (B) Comparison of using the full time series versus the event-triggered (percussion events) to
calculate the PCA variance explained demonstrate that they are very similar. Blue circles are the corresponding variance explained of the PCA
extracted from Oldowan trials (red squares Acheulean trials). (C,D) Schematic of the object manipulations performed in the control task. (C) individual
nut-sized widgets were picked from a box and placed in alternating adjacent boxes. (D) Boxes (styrofoam, top open) were stacked upon each other
and then unstacked again.
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 7 November 2010 | Volume 5 | Issue 11 | e13718
statistical measure to quantify the complexity of grips, Principal
Component Analysis, which ignores individual differences in the
grips but looks for a common structure to all grips in a data set.
Our approach to quantify behavioural complexity is not restricted
to hand joint angles, but can be applied to other behavioural time
series, such as full body joint movements or animal data.
Our behavioral complexity measure allowed us to compare
structure across all stable grips in Acheulean vs. Oldowan
toolmaking. We quantified this, and found little difference in the
number of principal components required to explain a given
amount of variability in the stable hand configurations, indicating
an equally complex structure. Moreover, the hand configuration
variance explained per principal component provides a measure of
the breadth of the statistical distribution of hand configurations.
Thus the variance explained reflects the diversity of grips, which
again overlaps between Acheulean and Oldowan toolmaking.
Interestingly, grip complexity in Lower Paleolithic toolmaking,
as quantified by variance explained per principal component, is
broadly comparable to modern-day daily manipulative complexity
measured with the same data glove technology (cf. data plotted
Figure 3.A in ). Although this previous study considered
dynamic hand movements rather than stable hand configurations,
the variance explained by a given number of principle components
in each case is typically within 5–10% of the other. The most
notable divergence is the lower percentage of variance (,10% less)
explained by the first two principle components in stone
toolmaking, which might suggest the presence of distinctive
manipulative patterns as compared to modern daily activities.
However, the tasks performed during the daily activity study were
unknown and the capacity of this method to distinguish more and
less complex tasks has yet to be directly demonstrated. To this end
we measured, for the first time, the manipulative complexity of
specific real-world tasks: small object sorting and box stacking. We
expected small object sorting to show a much lower complexity
than core manipulations, as it only involves picking up and
transferring identical objects – quite unlike manipulating a
constantly changing core during toolmaking. Similarly, we
expected the box stacking task would require an intermediate
complexity between small object sorting and core manipulation,
because more grasp variety and fine control is needed than in the
sorting task, yet the shape of the objects being manipulated does
not change as in stone toolmaking. These expectations were
corroborated by our results, showing both control manipulations
to be less complex than the two toolmaking tasks.
Having validated the method, the absence of differences in left
hand grip complexity between Oldowan and Late Acheulean
toolmaking remains a surprising result, especially considering the
substantial differences between the two technologies. Both require
that the core be properly positioned and firmly supported, yet
Acheulean toolmaking involves both a greater diversity of
technical actions and more substantial changes in core morphol-
ogy as the core is shaped into a pointed, thin and symmetrical
‘handaxe’. Furthermore, as the piece is thinned it becomes
increasingly important to properly brace it to prevent breakage, a
problem that is not present in Oldowan toolmaking. Nevertheless,
results indicate that the degree of grip complexity and diversity
already present in Oldowan toolmaking is sufficient to accommo-
date Later Acheulean toolmaking.
Stable hand configurations are essential for skilled tool making.
Recent work in human motor control has uncovered how finely
the nervous system controls the stability of mechanically unstable
objects during tool manipulation (e.g. when applying a screw
driver, or in our case when the hammerstone hits the edge of the
core). This is achieved through the orchestrated co-contraction of
antagonistic muscles so as to stiffen a joint against undesired
perturbation . Humans can tune the stiffness of joints (also
known as mechanical impedance) to optimally compensate for
internal or external perturbations, trading off the degree of muscle
co-contraction (which is metabolically expensive and promotes
muscle fatigue) with the required mechanical stability of a task
[43,44]. Thus, the stable hand configurations we observed in
toolmaking may arise due to a need to minimize the effects of
internal perturbations, e.g. maintaining the precise orientation and
location of the core’s edge for the percussion motions of the other
arm, and external perturbations, e.g. direction and force of the
Experimental data shows that the stiffness of the fingertips
depends on the specific configuration of the hand . This may
suggest the existence of a specific subset of hand configurations in
which stability of the core prior to and during striking is
maximized. A recent study on how the nervous system controls
force sharing in the fingers for object grasping and pickup, studied
an object that had a varying center of mass on each trial (akin to
the changing center of mass of the core as it is reduced) and
showed a stereotyped pattern of force production across the fingers
of the non-dominant hand. EMG recordings of Oldowan
stone toolmaking further suggest that the muscles stabilizing the
thumb (Flexor pollicis brevis and Adductor pollicis transverse) and little
finger (Flexor digiti minimi and Abductor digiti minimi) may be
particularly important in the case of Paleolithic toolmaking .
The thumb is of special interest in this context, first because the
biomechanics and neuronal control of the human thumb are
relatively independent from the other fingers, and second
because the anatomical relationship of the thumb to the other
fingers appears to be derived in humans compared to other apes
. Our results show that the variability of the thumb joint
configurations during toolmaking (Figure 3, top row) were among
the largest of all fingers and thus accounted for much of the
manipulative complexity of both Oldowan and Acheulean core
grasps. Current fossil and comparative evidence suggests that a
reduction in the lengths of the fingers compared to the thumb and
an increase in the robusticity of the thumb, both of which would
facilitate stable power grasps, occurred relatively early in hominin
evolution, prior to or during Oldowan times . Modern apes
without these adaptations appear to have difficulty stabilizing
stone cores for hand-held percussion, and may adopt alternative
strategies such as throwing, bracing against the torso, or even
grasping with a foot rather than a hand .
While the evolution of motor control for stable hand
configurations (stiffness control) is virtually unknown, we suggest
that the observed equal manipulative complexity of both Oldowan
and Acheulean toolmaking may be due to a common neuronal
motor control strategy by which the required mechanical stability
of the core is controlled during striking. Thus, skilled toolmaking
may have required the evolution and refinement of stiffness
control for the hand-arm system by the Early Stone Age, to
support tool production.
Previous work has shown that components of variation in
human grasp coordination patterns are organized along a gradient
from lower to higher finger individuation, with lower order
components reflecting coordinated opening and closing of the
whole hand and higher order components reflecting fine
adjustments of hand shape . As expected, these higher order
components distinguish stone toolmaking from even a relatively
complex manipulative control task (box stacking: Figure 6D). The
fact that they do not distinguish Oldowan from Acheulean
toolmaking indicates that differences between these technologies
instead lie at a superordinate level of behavioral organization. This
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 8 November 2010 | Volume 5 | Issue 11 | e13718
higher-level organization allowed the expert subject to achieve
different results and to adapt to highly variable object conditions
using a comparable array of subordinate grasp synergies.
The adaptive complexity of grasp regulation in Paleolithic
toolmaking invites comparison with the complex coordination of
articulators (vocal cords, tongue, palate, lips) in speech , which
must also form flexible synergies in the face of environmental
perturbations . This addresses a potential dis-analogy between
speech control and manual manipulation, in that the latter might
be seen as inherently less ‘arbitrary’ and more environmentally
determined [cf. 53]. Current results show that skilled toolmakers
are able to impose dynamically stable structure in their
manipulative behaviors, despite substantial environmental and
task-related variability. In combination with evidence of neural
overlap between tool use and phonetic processing in left ventral
premotor cortex [5,54], this supports the argument  that
Oldowan toolmaking may have provided a ‘preadaptive’ founda-
tion for the enhanced cortical control of vocalization emphasized
in many hypotheses of language evolution [37,51,55,56,57].
Increased right hemisphere activation during Late Acheulean
toolmaking (Figure 1D) indicates additional demands for the
cognitive control of action in this more complex technology,
including a functional/anatomic overlap with discourse-level
language processing. Because the structural diversity and com-
plexity of left-hand grips used during Oldowan and Acheulean
toolmaking are indistinguishable, these increases in right hemi-
sphere activation cannot be attributed to increases in the basic
complexity of contralateral grasp regulation and must instead be
attributed to recruitment of distinctive right hemisphere functions.
Hypothetically, these include task-set switching and inhibition of
contextually inappropriate actions  in right inferior frontal
cortex (BA45), and the regulation of complex action sequences in
right parietal cortex [39,58]. These functions are consistent with
the distinctive behavioral organization of Acheulean toolmaking,
which involves switching between different subordinate task-sets in
pursuit of superordinate goals to an extent that Oldowan
toolmaking does not [6,34]. For example, properly thinning a
Late Acheulean handaxe often requires the toolmaker to stop and
prepare edges and surfaces prior to an intended flake removal.
This ‘platform preparation’ is accomplished through the small-
scale chipping and/or abrasion of edges to alter their sharpness,
bevel, and placement relative to the midline  and involves its
own set of subordinate task goals, operations and tools. Insofar as
processing of linguistic context and prosody involves similar
demands for the integration of hierarchically structured informa-
tion over time in the right hemisphere , the anatomical
overlap of Late Acheulean toolmaking and right hemisphere
linguistic processing may reflect the flexible ‘‘mapping’’ of diverse
overt behaviors onto shared functional substrates in the brain .
This implies that: 1) selection acting on either language or
toolmaking abilities could have indirectly favored elaboration of
neural substrates important for the other, and 2) archaeological
evidence of Paleolithic toolmaking can provide evidence for the
presence of cognitive capacities also important to the modern
human faculty for language.
Materials and Methods
Data Acquisition and Manipulative Complexity Analysis
Hand and arm movements were recorded at a rate of 240 Hz
using a Polhemus Liberty electromagnetic tracking system
(POLHEMUS, Colchester (VT)). The measurement markers of
this system recorded horizontal, vertical and depth position of the
marker in a reference coordinate system, as well as the rotational
degrees of freedom (yaw, pitch and roll). The system was pre-
calibrated to within 3 centimeters precision over the large
workspace volume (a cube of about 1.5 by 1.5 by 1.5 meters).
Two measurement markers were mounted on a pair of leather
gardening gloves. Each marker was fixed on the back of each
gardening glove about 2 cm proximal from the knuckle of the
middle finger. Gardening gloves were required here to protect the
data glove from shear forces and dust/debris during the tool
production process. It must be considered that the gardening
gloves could possibly have constrained hand movements during
the experiment. However, it is not uncommon for such gloves to
be worn by experimental stone toolmakers for protection. In this
experiment, the gloves did not interfere with the successful
production of the intended tools on all eight trials (3 Oldowan, 5
Acheulean), showing that at least the technologically required
mobility was present.
The left thigh was used as a support platform for the knapping
process and to measure any potentially relevant movements, a
third measurement marker was positioned on the kneecap of the
left leg. A fourth marker was placed on the back of the tool-maker,
at the intersection between the spinal cord and the line connecting
the two shoulder joints, so as to account for changes in upper body
posture. The core was held and stabilized by the toolmaker’s left
hand. To record the left hand’s grip on the core an 18-sensor
CyberGlove I data glove (Cyberglove Systems, San Jose (CA)) was
worn underneath the gardening glove (see also Figure 2). The data
glove, made of thin cloth with embedded resistive sensors,
measures the degrees of freedom of the joints in the hand. Glove
sensors were polled at a rate of 80 Hz with a resolution of 8 bits
(256 different values) per sensor and calibrated to a splint (see also
Figure 3). Glove and marker data stream were resampled to
150 Hz, combined into a single consistent data stream, time-
stamped accordingly, and stored for offline analysis. All analysis
was conducted with MatLab (MathWorks, Natick (MA)). The full
glove time series data set contained between 101,424 and 133,110
data points for the 5 Acheulean toolmaking trials and between
20,840 and 33,486 data points for the 3 Oldowan trials. Automatic
annotation tagged about 1.8 to 3.5% of the Acheulean data points,
and 4.3% to 6.6% of Oldowan data points, as stable hand
configurations. Because all toolmaking data were collected from
one of the researchers (the 4
author, an expert stone tool-maker
with .40 years of experience) ethics approval and informed
consent were not sought.
The dimensionality of hand movements was examined by
means of Principal Component Analysis of the joint angles after
processing the data, similar to [32,60,61]. In our study the data for
both the Acheulean and Oldowan sequences was near Gaussian
distributed for all joints recorded (Figure 3). This made our dataset
ideally suited for Principal Component Analysis. Analysis was
performed on the relevant portions of the glove data time series,
which was normalized prior to computing the covariance matrix,
as our basis for PCA. The event-driven PCA analysis was based on
manually tagged or automatically tagged regions of the time series,
while the full analysis was based on the complete time series. In
contrast to a previous study  which considered joint
movements (joint velocities), we analyzed angular joint positions
(joint angles), as we were interested in the complexity and
variability of adaptive hand configurations rather than patterns of
movement. We note, that while we applied here PCA as algorithm
to calculate our complexity measure, we do not actually reduce the
dimensionality of the data, but use all dimensions of the data.
Although the hand joint statistics for our actions (Figure 3) are
indeed Gaussian distributed, our method would work irrespective
of the actual empirical data distribution, as we simply use the
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 9 November 2010 | Volume 5 | Issue 11 | e13718
amount of variance explained by each principal component (and
compute these for all dimensions) as characteristic to measure and
distinguish manipulative complexities.
The following two-step calibration was used. First, before each
trial the toolmaker placed his hand palm down against a flat
surface, with the four fingers parallel and the thumb aligned
against the side of the palm (without pressing into the gardening
glove’s resistance). A reading was then taken from the glove and
this served as the zero point for the joint angles in the subsequent
recording session. Then, a mechanical splint was placed between
the finger joints. Note, however that due to the protective nature of
the gardening glove a calibration of the joint sensors to the
theoretical maximum precision of ,1 degree was difficult, and we
operated with about 3 degree resolution – which is within the
linear response regime as specified by the manufacturer. After this
calibration the toolmaker curled his hand up into a fist. This
procedure was performed at the beginning of each trial, after
which toolmaking started within 10–30 seconds. In addition, each
toolmaking trial was video recorded at 30 Hz using an iSight
digital camera (Apple, Cupertino (CA)). A total of 5 Acheulean
handaxes and 3 sets of Oldowan stone flakes were produced.
Manual annotation and automated annotation of data
The focus of the present study was to investigate the complexity
of stable core grips. From our own toolmaking and biomechanical
experience, the stability and stiffness of the grip was expected to be
highest around the time when the hammerstone was going to
strike the core. Thus, we manually annotated the video of
Acheulean and Oldowan tool making sequences at subsecond
precision with the help of ethogram-production software (Etholog,
). The annotation recorded seven event types, including two
forms of percussion named ‘‘percussion’’(a hammer strike-like
motion) and ‘‘light percussion’’(a more chiseling-like motion). This
very laborious annotation process was only carried out for one
Acheulean and one Oldowan sequence. To complement this data
set, we developed a simple automated method for detecting these
two percussion events. Manual inspection of the position marker
data and the glove data synchronized to the video data (Figure 4),
produced a straightforward criterion: common to all tool
production sequences were characteristic spikes in the velocity
plots of the tool hand, of 0.7–2 meters/second for ‘‘percussion’’
and 0.5–1.2 meters/second for ‘‘light percussion’’ events. Thus for
all recorded tool making sequences we extracted the glove data
when the tool hand moved faster than 0.5 meters/second.
Control experiments. Control experiments in two
naturalistic tasks (small object sorting and box stacking) were
carried out with the same system as was used for the toolmaking
experiments. For small object sorting, small (nut-sized) complexly
shaped plastic objects (cable carriers: Bosch-Rexroth, Part
No. 3842526564) that we refer to here as ‘‘widgets’’ were filled
into a container. For box stacking, we used Styrofoam packaging
containers (17 cm W 610 cm H 619 cm D), with one open side.
Widget sorting task (Figure 5C): Individual widgets were picked
out with the left hand from a central container and placed
alternately in two containers to the left and right. The size and
shape of the widgets typically resulted in 4-finger grip for pick-up
and placement. After emptying the central container, objects were
picked up individually and alternately from the two adjacent
containers and placed back into the central container. This
procedure was repeated 3 times.
Box stacking task (Figure 5D): Three Styrofoam boxes were
repeatedly precisely stacked upon each other (using the left hand
only) and then unstacked. The boxes were normally grasped by the
top edge and grip remained effectively constant throughout
manipulation, resulting in mainly translatory arm movements
and rotation of the wrist. Due to the nature of our two control
tasks we could not use limb tracking as a straightforward measure
of behavioral actions of the hand during these tasks, because grip
phases overlapped broadly with considerable voluntary motion of
the hand. Therefore, we analyzed the complete data glove time
series of these behaviors and compared these to the complete time
series in the toolmaking tasks.
We would like to thank James Ingram for discussion and technical advice,
Manu Davies for logistical assistance, and Stuart Laidlaw for a photo used
in Figure 1. We thank two anonymous reviewers for helpful comments.
Conceived and designed the experiments: AAF DS. Performed the
experiments: AAF DS BB. Analyzed the data: AAF JA. Wrote the paper:
1. Bril B, Roux V (2005) Synthesis and Speculations. In: Roux V, Bril B, eds. Stone
knapping: the necessary conditions for a uniquely hominin behaviour.
Cambridge: McDonald Institute for Archaeological Research. pp 353–355.
2. Ambrose S (2001) Paleolithic technology and human evolution. Science 291:
3. Byrne R (2004) The manual skills and cognition that lie behind hominid tool use.
In: Russon AEB, David R, eds. The Evolution of Thought: Evolutionary Origins
Of Great Ape Intelligence. Cambridge: Cambridge University Press. pp 31–44.
4. Wynn T, McGrew W (1989) An ape’s view of the Oldowan. Man 24: 383–398.
5. Stout D, Chaminade T (2007) The evolutionary neuroscience of tool making.
Neuropsychologia 45: 1091–1100.
6. Stout D, Toth N, Schick K, Chaminade T (2008) Neural correlates of Early
Stone Age toolmaking: technology, language and cognition in human evolution.
Philosophical Transactions of the Royal Society B: Biological Sciences 363:
7. Semaw S (2000) The world’s oldest stone artefacts from Gona, Ethiopia: their
implications for understanding stone technology and patter ns of human
evolution 2.6-1.5 million years ago. Journal of Archaeological Science 27:
8. Stout D, Semaw S, Rogers MJ, Cauche D (2010) Technological variation in the
earliest Oldowan from Gona, Afar, Ethiopia. Journal of Human Evolution 58:
9. Schick KD, Toth N (1993) Making silent stones speak: human evolution and the
dawn of technology. New York: Simon & Schuster.
10. Clark JD (1994) The Acheulian Industrial Complex in Africa and Elsewhere. In:
Corruccini RS, Ciochon RL, eds. Integrative Paths to the Past: Paleoanthro-
pological Advances in Honor of F Clark Howell. New Jersey: Prentice Hall. pp
11. Holloway R (1969) Culture: a human domain.Current Anthropology 10: 395–412.
12. Gowlett JAJ (1984) Mental abilities of early man: A look at some hard evidence.
In: Foley R, ed. Hominid Evolution and Community Ecology. New York:
Academic Press. pp 167–192.
13. Gowlett JAJ (1996) Mental abilities of early Homo: elements of constrain and
choice in rule systems Modelling the early human mind. In: Mellars P,
Gibson KR, eds. Modelling the early human mind. Oxford: Oxbow Books.
14. Isaac GL (1976) Stages of cultural elaboration in the Pleistocene: possible
archaeological indicators of the development of language capabilities. In:
Harnad SR, Steklis HD, Lancaster J, eds. Origins and evolution of language and
speech. New York: New York Academy of Science. pp 275–288.
15. Greenfield PM (1991) Language, tools, and brain: The development and
evolution of hierarchically organized sequential behavior. Behavioral and Brain
Sciences 14: 531–595.
16. Rizzolatti G, Arbib M (1998) Language within our grasp. Trends Neurosci 21:
17. Chao LL, Martin A (2000) Representation of manipulable man-made objects in
the dorsal stream. NeuroImage 12: 478–484.
18. Hagoort P (2005) On Broca, brain, and binding : a new framework. Trends in
Cognitive Science 9: 416–423.
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 10 November 2010 | Volume 5 | Issue 11 | e13718
19. Bookheim er S (2002) Functional MRI of language: new appr oaches to
understanding the cortical organization of semantic processing. Annual Review
of Neuroscience 25: 151–188.
20. Sato M, Tremblay P, Gracco VL (2009) A mediating role of the premotor cortex
in phoneme segmentation. Brain and Language 111: 1–7.
21. Koechlin E, Jubault T (2006) Broca’s Area and the hierarchical organization of
human behavior. Neuron 50: 963–974.
22. Hartmann K, Goldenberg G, Daumuller M, Hermsdorfer J (2005) It takes the
whole brain to make a cup of coffee: the neuropsychology of naturalistic actions
involving technical devices. Neuropsychologia 43: 625–637.
23. Aron AA, Robbins TW, Poldrack RA (2004) Inhibition and the right inferior
frontal cortex. Trends in Cognitive Science 8: 170–177.
24. Marzke MW, Shackley MS (1986) Hominid hand use in the pliocene and
pleistocene: Evidence from experimental archaeology and comparative mor-
phology. Journal of Human Evolution 15: 439–460.
25. Marzke MW, Toth N, Schick K, Reece S, Steinberg B, et al. (1998) EMG study
of hand muscle recruitment during hard hammer percussion manufacture of
Oldowan tools. American Journal of Physical Anthropology 105: 315–332.
26. Marzke M (2005) Who made ston e tools? In: Roux V, Bril B, eds. Stone
knapping: the necessary conditions for a uniquely hominin behaviour.
Cambridge: McDonald Institute for Archaeological Research. pp 243–256.
27. Bril B, Roux V, Dietrich G (2005) Stone knapping: Khambhat (India), a unique
opportunity? In: Roux V, Bril B, eds. Stone knapping: the necessary conditions
for a uniquely hom inin behaviour. Cambridge: McDonald Institute for
Archaeological Research. pp 53–72.
28. Roux V, Bril B, Dietrich G (1995) Skills and learning difficulties involved in
stone knapping. World Archaeology 27: 63–87.
29. Bril B, Roux V, Dietrich G (2000) Habilites impliquees dans la taille des perles
en roches dure: characteristiques motrices et cognitives d’une action situee
complexe. In: Roux V, ed. Les perles de cambay: des practiques techniques aux
technosystemes de l’orient ancien. Paris: Editions de la MSH. pp 211–329.
30. Biryukova EV, Bril B, Dietrich G, Roby-Brami A, Kulikov MA, et al. (2005) The
organization of arm kinematic synergies: the case of stone-bead knapping in
Khambhat. In: Roux V, Bril B, eds. Stone knapping: the necessary conditions
for a uniquely hom inin behaviour. Cambridge: McDonald Institute for
Archaeological Research. pp 73–89.
31. Faisal AA, Selen LPJ, Wolpert DM (2008) Noise in the nervous system. Nat Rev
Neurosci 9: 292–303.
32. Ingram J, Kording K, Howard I, Wolpert D (2008) The statistics of natural hand
movements. Experimental Brain Research 188: 223–236.
33. Corballis MC (2003) From mouth to hand: gesture, speech, and the evolution of
right handedness. Behavioral and Brain Sciences 26: 199–260.
34. Stout D, Chaminade T (2009) Making tools and making sense: complex,
intentional behaviour in human evolution. Cambridge Archaeological Journal
35. Uomini NT (2009) The prehistory of handedness: Archaeological data and
comparative ethology. Journal of Human Evolution 57: 411–419.
36. Guiard Y (1987) Asymmetric division of labor in human skilled bimanual action:
the kinematic chain as a model. Journal of motor behavior 19: 486–517.
37. Deacon TW (1997) The symbolic species: The co-evolution of language and the
brain. New York: W.W. Norton.
38. Ross ED, Monnot M (2008) Neurology of affective prosody and its functional-
anatomic organization in right hemisphere. Brain and Language 104: 51–74.
39. Frey SH, Gerry V (2006) Modulation of neural activity during observational
learning of action and their sequential orders. The Journal of Neuroscience 26:
40. Stephan KE, Marshall JC, Friston KJ, Rowe JB, Ritzl A, et al. (2003) Lateralized
cognitive processes and lateralized task control in the human brain. Science 301:
41. Faisal A, Niven J, Rogers S (2009) Analysis of decision making in an insect’s gap-
crossing behavior using Markov models. COSYNE 2009. Salt Lake City (UT).
42. Shadmehr R, Wise SP (2005) The computational neurobiology of reaching and
pointing; Sejnowski TJ, Poggio TA, eds. Cambridge (MA): MIT Pre ss. 575 p.
43. Burdet E, Osu R, Franklin D, Milner T, Kawato M (2001) The central nervous
system stabilizes unstable dynamics by learning optimal impedance. Nature 414:
44. Selen LP, Franklin DW, Wolpert DM (2009) Impedance control reduces
instability that arises from motor noise. J Neurosci 29: 12606–12616.
45. Milner TE, Franklin DW (1998) Chara cterization of multijoint finger stiffness:
dependence on finger posture and force direction. IEEE Trans Biomed Eng 45:
46. Rearick M, Santello M (2002) Force synergies for multifingered grasping: effect
of predictability in object center of mass and handedness. Exp Brain Res 144:
47. Lang CE, Schieber MH (2004) Human finger independence: limitations due to
passive mechanical coupling versus active neuromuscular control. J Neurophysiol
48. Tocheri MW, Orr CM, Jacofsky MC, Marzke MW (2008) The evolutionary
history of the hominin hand since the last common ancestor of Pan and Homo.
J Anat 212: 544–562.
49. Savage-Rumbaugh S, Fields WM (2006) Rules and too ls: beyond anthropo-
morphism. In: Toth N, Shick KD, eds. The Oldowan: Case studies into the
earliest stone age. GosportIN: Stone Age Institute Press. pp 223–241.
50. Schieber MH, Santello M (2004) Hand function: peripheral and central
constraints on performance. J Appl Physiol 96: 2293–2300.
51. Studdert-Kennedy M, Goldstein L (2003) Launching language: the gestural
origins of discrete infinity. In: Christiansen MH, Kirby S, eds. Language
evolution. Oxford: Oxford University Press. pp 235–254.
52. Munhall KG, Lofqvist A, Kelso JAS (1994) Lip–larynx coordination in speech:
Effects of mechanical perturbations to the lower lip. The Journal of the
Acoustical Society of America 95: 3605–3616.
53. Chase PG (1991) Symbols and Paleolithic artifacts: style, standardization, and
the imposition of arbitrary form. Journal of Anthropological Archaeology 10:
54. Higuchi S, Chaminade T, Imamizu H, Kawato M (2009) Shared neural
correlates for language and tool use in Broca’s area. Neuroreport 20:
55. Liberman AM, Whalen DH (2000) On the relation of speech to language.
Trends in Cognitive Sciences 4: 187–196.
56. Lieberman P (2002) On the nature and evolution of the neural bases of human
language. American Journal of Physical Anthropology 45: 36–62.
57. MacNeilage PF (1995) Linguistic and manual evolution. Behavioral and Brain
Sciences 14: 568–569.
58. Jubault T, Ody C, Koechlin E (2007) Serial Organization of Human Behavior in
the Inferior Parietal Cortex. J Neurosci 27: 11028–11036.
59. Callahan E (1979) The basics of biface knapping in the Eastern Fluted Point
Tradition: A manual for flintknappers and lithic analysts. Archaeology of
Eastern North America 7: 1–172.
60. Santello M, Flanders M, Soechting J (1998) Postural hand synergies for tool use.
J Neurosci 18: 10105–10115.
61. Todorov E, Ghahramani Z (2004) Analysis of synergies underlying complex
hand manipulation. IEEE 26th Ann Intl Conf Engineering in Biology &
Medicine. San Francisco (CA): IEEE. pp 4637–4640.
62. Ottoni EB (2000) EthoLog 2.2: a tool for the transcription and timing of
behavior observation sessions. Behav Res Methods Instrum Comp ut 32:
Manipulation and Tool-Making
PLoS ONE | www.plosone.org 11 November 2010 | Volume 5 | Issue 11 | e13718