Direct Dating of Human Fossils
Rainer Gru¨n*
Research School of Earth Sciences, Research School of Pacific and Asian Studies,
The Australian National University, Canberra ACT 0200, Australia
KEY WORDS 14C; U-series; ESR; Border Cave; Tabun; Skhul; Qafzeh; Vindija; Banyoles; Mungo
ABSTRACT The methods that can be used for the
direct dating of human remains comprise of radiocarbon,
U-series, electron spin resonance (ESR), and amino acid
racemization (AAR). This review gives an introduction to
these methods in the context of dating human bones and
teeth. Recent advances in ultrafiltration techniques have
expanded the dating range of radiocarbon. It now seems
feasible to reliably date bones up to 55,000 years. New
developments in laser ablation mass spectrometry permit
the in situ analysis of U-series isotopes, thus providing a
rapid and virtually non-destructive dating method back to
about 300,000 years. This is of particular importance
when used in conjunction with non-destructive ESR anal-
ysis. New approaches in AAR analysis may lead to a ren-
aissance of this method. The potential and present limita-
tions of these direct dating techniques are discussed for
sites relevant to the reconstruction of modern human evo-
lution, including Florisbad, Border Cave, Tabun, Skhul,
Qafzeh, Vindija, Banyoles, and Lake Mungo. Yrbk Phys
Anthropol 49:2–48, 2006. V
C
2006 Wiley-Liss, Inc.
When reconstructing human evolution, it is necessary
to know how old the human fossils are. This information
is usually extracted from a variety of sources, including
the general chronological frameworks of the local geol-
ogy, the flora, fauna, and artifacts found in association
with the human fossils as well as numerical dating stud-
ies on these associated materials. This indirect dating
approach, with respect to the human fossils, is in many
cases not satisfactory, because:
i. the human remains are often buried into the sedi-
ments and the association with other materials is
uncertain (e.g. Skhul, Qafzeh, etc.);
ii. faunal remains or minerals from the sediment are
reworked from older deposits (see e.g. present discussion
of the age of the Homo erectus remains in Indonesia);
iii. the hominid specimens were discovered at a time
when no careful excavations were carried out and it
has become impossible to correlate the human re-
mains with other datable material (nearly 90% of all
paleoanthropological specimens).
Direct dating of human remains would, of course, allevi-
ate many of these problems (see also Trinkaus, 2005).
Until recently, human fossils could only be directly dated
by radiocarbon. This method reaches back to about 50,000
years. As a consequence, all older fossils did not yield
meaningful chronological results and many important
questions in our understanding of human evolution could
not be addressed. Furthermore, most dating techniques
are destructive. Human remains are scarce and extremely
valuable, therefore any sort of destruction has to be kept
to an absolute minimum. This is of particular importance
in Australia, where any human fossils are sacred to the
Aboriginal communities. This, of course, also applies to
other areas, such as parts of North America. It is therefore
necessary to develop and apply more or less non-destruc-
tive techniques for the analysis of human material. New
technical developments, particularly in U-series and elec-
tron spin resonance (ESR), now allow the virtually non-
destructive analysis of human remains.
The dating methods that can be used for dating fossil
bones and teeth consist of radiocarbon, U-series, ESR, and
amino acid racemization (AAR). These methods can gener-
ally be applied on a wide range of materials, but in this pa-
per only their application for dating human remains is
critically appraised (for general reviews on dating techni-
ques, see e.g. Noller et al., 2000 and references therein).
Because of analytical and technical limitations, each
dating technique has a certain age range, to which it can
be applied (Fig. 1). It is obvious that radiocarbon, includ-
ing the most advanced pretreatment techniques (Bird
et al., 1999; Bronk Ramsey et al., 2004b) can only address
chronological issues relating to relatively recent fossils,
mainly Homo sapiens, and perhaps the youngest Neander-
thal and Homo erectus specimens as well as Homo flore-
siensis (Brown et al., 2004; Morwood et al., 2004, 2005).
On the other hand, U-series and AAR can cover, in princi-
ple, all chronological aspects of modern human evolution;
ESR could be used to explore chronological relationships
of earlier human groups. However, for the age estimation
of older fossils, including most Australopithecus and Par-
anthropus species, there is presently no direct dating
technique available.
METHODS
The underlying principles of scientific methods do not
rapidly change. Thus, the following introductions to and
descriptions of the dating methods are updated versions of
Gru¨n (in press; submitted-a,b) and Gru¨n et al. (in prepa-
ration), tailored to the topic and audience of this review.
*Correspondence to: Rainer Gru¨n, Research School of Earth Sci-
ences, Research School of Pacific and Asian Studies, The Australian
National University, Canberra ACT 0200, Australia.
E-mail: Rainer.Grun@anu.edu.au
DOI 10.1002/ajpa.20516
Published online in Wiley InterScience (www.interscience.wiley.com).
VC
2006 WILEY-LISS, INC.
YEARBOOK OF PHYSICAL ANTHROPOLOGY 49:2–48 (2006)

Before describing the dating techniques, it is necessary
to delve ever so slightly into the tedious subject of error
calculations, and their meaning in the assessment and
interpretation of dating results. It is the aim of every dat-
ing method to produce highly precise, highly correct, and
therefore accurate results (see Fig. 2 (after Wagner, 1998),
which is based on German phraseology and has the
advantage of distinguishing more clearly between preci-
sion and accuracy). Note that in some statistics books, ac-
curacy is used for overall systematic error, (correctness in
Fig. 2, e.g., Saunders and Fleming, 1957), in others for the
combination of random and systematic errors (as used in
Fig. 2, e.g., Lyon, 1970).
Each measurement that is carried out for a quantitative
analysis has a degree of uncertainty. There are two
sources of error: random errors (determining the preci-
sion) and systematic errors (responsible for the correct-
ness of the result, see Fig. 2). Only if both error sources
are small, can accurate results be obtained.
Random errors
These are introduced by the degree of inability to mea-
sure the same quantity exactly in repeated measure-
ments. When carrying out an experiment, e.g. weighing
with a high precision balance, one will notice that there
are slight differences in the measured weights. The best
estimate of the result is provided by the mean value,
which is obtained by adding up all individual measure-
ments, and dividing them by the number of measure-
ments. When a large number of these measurements are
plotted on a frequency diagram, they follow a normal, or
Gaussian distribution. The Gaussian curve is bell shaped
and is defined by two parameters, the mean value and the
standard deviation, r
r
.
The standard deviation (SD) means that there is a
68.3% probability that a single measurement falls within
the range of mean value 6 r
r
. However, about one-third of
all measurements will fall outside this range. The 2r
r
range contains 95.4% of all measurements (one in twenty
results will still fall outside this range!) and the 3r
r
range
contains 99.7%. To be explicit, if the dating result of a sin-
gle sample of human material is 10,000 6 1,000 years, its
age has about a two-third chance to be anywhere within
9,000–11,000 years, and a one-third chance to either be
younger than 9,000 years or older than 11,000 years. No
one betting on horses would ignore such odds. Also, when
analyzing larger sample sets, it lies in the nature of statis-
tics that some samples lie outside the 2-r range (here
8,000–12,000 years). These are not necessarily outliers
and should not be ignored when assessing and interpret-
ing analytical results. Data sets can be analyzed with
appropriate tests to check whether certain results are out-
liers or not (using t or v
2
tests, for more details, see e.g.,
Lyon, 1970).
The standard error, S
r
, defines the confidence interval
for the mean and is the standard deviation divided by the
square root of the number of measurements. This means
that the confidence interval of a mean value is critically
dependent on the number of measurements that have
been carried out, i.e., the uncertainty in the mean can be
reduced by increasing the number of measurements. This
relationship can be used to improve the age determination
of an object or a stratigraphic unit (provided the samples
have the same age). Compared to a single measurement,
the age uncertainly is halved, when measuring four
samples.
Systematic errors
The difference between random and systematic errors
are shown in Figure 2. The systematic error may have
Fig. 1. Approximate dating ranges of the methods that can
be used for the direct dating of human remains. [Color figure
can be viewed in the online issue, which is available at www.
interscience.wiley.com.]
Fig. 2. Relationship between random and systematic errors.
Random errors govern the precision, systematic errors the cor-
rectness. Accurate results can only be obtained if both error sour-
ces are small (after Wagner, 1998). [Color figure can be viewed
in the online issue, which is available at www.interscience.
wiley.com.]
3DIRECT DATING OF HUMAN FOSSILS
American Journal of Physical Anthropology—DOI 10.1002/ajpa