Quantitative Assessment of Interproximal Wear Facet
Outlines for the Association of Isolated Molars
Stefano Benazzi,1
* Luca Fiorenza,2
Stanislav Katina,1,3
Emiliano Bruner,4
and Ottmar Kullmer2
1
Department of Anthropology, University of Vienna, 1090 Vienna, Austria
2
Department of Palaeoanthropology and Messel Research, Senckenberg Research Institute,
D-60325 Frankfurt am Main, Germany
3
Department of Applied Mathematics and Statistics, Faculty of Mathematics, Physics and Informatics,
Comenius University, 842 48 Bratislava, Slovakia
4
Centro Nacional de Investigacio´n sobre la Evolucio´n Humana (CENIEH), Paseo Sierra de Atapuerca s/n,
09002 Burgos, Espan˜a
KEY WORDS dental remains; 3D digital models; elliptic Fourier analysis; geometric
morphometrics
ABSTRACT The determination of the minimum number
of individuals can be very challenging, especially in an
assemblage of fragmentary bones and isolated teeth. Similarities
in tooth morphology, degree of wear, and interproximal
wear facets (IPWF) are generally used to associate
isolated teeth qualitatively. However, no quantitative
method has yet been established for an objective identification
and matching of isolated tooth crowns. In this study,
we analyze the IPWF morphology of adjacent mandibular
molars (17 M1/M2 pairs), applying both qualitative and
quantitative methods to test a reproducible approach for
crown association. The surfaces of distal (for M1) and
mesial (for M2) IPWF were surface-scanned and digitally
selected. Three-dimensional (3D) and two-dimensional (2D)
outlines of IPWF were analyzed using elliptic Fourier analysis
(EFA) and geometric morphometrics methods (GMM).
Additionally, teeth were qualitatively associated by visual
evaluation of the IPWF outline and by physical matching.
Unsatisfactory results with less than 50% of tooth pairs
correctly associated were obtained by using both methods,
shape analysis (digital approach) and the visual evaluation
(qualitative assessment) of the IPWF outline. The physical
matching of the crowns showed highly variable accuracy
ranging between 53% and 77%. The quantitative formspace
analysis of 2D IPWF outlines provided the best
results (82% of correctly associated teeth), but no statistically
significant differences were recorded when compared
with the manual matching. Since three tooth pairs out of
17 could not be quantitatively associated, we suggest that
the quantitative analysis of IPWF should be used only in
addition with other approaches. Am J Phys Anthropol
144:309–316, 2011. VVC 2010 Wiley-Liss, Inc.
Teeth are an important source of biological information,
and this is of particular interest in paleontological
and bioarchaeological studies. Tooth crowns are the most
abundant remains in the fossil and archaeological record
because of the hardness and resistance of enamel. Determining
if two or more isolated teeth belong to one or
more individuals is of particular importance for the
reconstruction of individual dentitions and the evaluation
of the minimum number of individuals recovered.
One approach to face the problem of isolated tooth specimens
is a simple qualitative assessment by trying to fit
them into a jaw bone discovered in the same stratigraphic
layer (Weidenreich, 1937). However, this method
is limited by the fact that jaw bones are often too fragmentary
or absent at fossil sites. Even internal micromorphology
has been used for matching adjacent teeth
of the same individual. For example, the contour lines of
Owen in the dentine and the striae of Retzius in the
enamel were used for tooth association (Gustafson, 1947;
Lunt, 1974). The latter method, however, is usually invasive,
requiring tooth sectioning, and consequently these
techniques are not applicable in a rare fossil assemblage.
Consequently most tooth associations in fossil hominin
collections are done using qualitative identifications. The
Neanderthal dental sample from Krapina, Croatia was
analyzed by using the orientation and morphology of the
interproximal wear areas, the continuity of occlusal
wear, and some unique features present in some crowns
(Wolpoff, 1978, 1979; see also Radovcˇic´ et al., 1988). The
same method was applied to identify seven human teeth
of uncertain origin from the Tabun Cave in Israel (Coppa
et al., 2005). Photographs of interproximal wear facets
(IPWF) were used to superimpose outline drawings visually
on acetate transparencies to reconstruct tooth rows
of the hominoids from Pasalar in Turkey (Genc¸turk et
al., 2008). In all these cases, the authors used the IPWF
to fit adjacent teeth together in the same tooth row.
Interproximal tooth wear is commonly an age-related
process (Kieser et al., 1985) produced by tooth-to-tooth
contact and crown movements occurring during chewing
(Fig. 1). Two main forces are involved during masticatory
The first two authors contributed equally to this work.
Grant sponsor: EU Marie Curie Training Network; Grant number:
MRTN-CT-2005-019564 EVAN; Grant sponsor: NSF Hominid Grant
2007; Grant number: NSF 01-120; Grant sponsor: Italian Institute
of Anthropology.
*Correspondence to: Dr. Stefano Benazzi, Department of Anthropology,
University of Vienna, Althanstraße 14, 1090 Vienna, Austria.
E-mail: stefano.benazzi@univie.ac.at
Received 28 January 2010; accepted 27 August 2010
DOI 10.1002/ajpa.21413
Published online 10 November 2010 in Wiley Online Library
(wileyonlinelibrary.com).
VVC 2010 WILEY-LISS, INC.
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 144:309–316 (2011)
movements: a lateral force that directs bucco-lingually, and
a mesial force that pushes the teeth making them migrate
anteriorly (Picton, 1962). One can expect that these movements
generate a strong correlation in size, morphology,
orientation, and angulations between contiguous IPWF,
providing a suitable way to determine tooth associations.
Two different approaches based on the IPWF are routinely
used to determine tooth associations: the IPWF outline
can be qualitatively evaluated as an independent feature
for tooth association (e.g., Genc¸turk et al., 2008) or
IPWF can be directly employed for the physical matching
of adjacent teeth (Wolpoff, 1978, 1979; Coppa et al., 2005).
In the latter approach, besides the IPWF surface, other
dental features, such as tooth morphology and dental
wear stage, are considered. These two approaches are
based on a qualitative assessment of the IPWF.
Quantitative data based on measures of shape and size
of IPWF provide an objective and reproducible method for
tooth association, whereas the results of a qualitative analysis
are more subjective, mainly depending on the experience
of the worker. In the present study, we tested the utility
of IPWF outlines for individual association of isolated
molars by means of quantitative analysis. Distal (for M1)
and mesial IPWF (for M2) were selected and analyzed
using elliptic Fourier analysis (EFA) and geometric morphometric
methods (GMM). These results were compared
with those produced by traditional visual approaches.
MATERIALS AND METHODS
The sample consists of 34 first and second mandibular
molars (17 M1/M2 pairs) of modern humans belonging to
15 individuals, 11 of which were collected at the Department
of History and Methods for the Conservation of
Cultural Heritage, University of Bologna in Ravenna,
Italy. Four individuals were collected at the Department
of Paleoanthropology and Messel Research in the Senckenberg
Research Institute in Frankfurt am Main, Germany.
For each specimen, distal IPWF of M1 and mesial
IPWF of M2 were considered. Only molars with a slight
or moderate degree of wear were included in the analysis
(wear stage 5 and less; Smith, 1984). We considered
the left and right sides of the same individual, when
available, as two separate tooth pairs.
The essential steps followed for the digital approach of
the present research (casting the teeth, three-dimensional
(3D) model generation, IPWF segmentation, outline orientation,
and statistical analysis) are summarized in Figure
2 and described in detail in the following paragraphs.
Scanning and digitalization of the models
The transparency and reflection of tooth enamel usually
make the direct scanning of original specimens
Fig. 1. Lower right second molar (RM2), mesial view. The
arrow points out the mesial interproximal wear facet (IPWF).
The scale in the figure is equivalent to 1 cm.
Fig. 2. Methodological steps included in the analysis: B,
buccal; L, lingual; O, occlusal; b, buccal point; GMM, geometric
morphometric methods; EFA, elliptic Fourier analysis.
Abbreviations
CS centroid size
EFA elliptic Fourier analysis
GMM geometric morphometrics methods
GPA generalized procrustes analysis
IPWF interproximal wear facets
OFA occlusal fingerprint analysis
310 S. BENAZZI ET AL.
American Journal of Physical Anthropology
impossible. The use of talc or an ammonium chloride
coating to cover the tooth surface was avoided because
small particles of the powder might remain on the surface
even after cleaning and could prevent future microwear
analyses. For this reason, we produced high resolution
molds of isolated teeth using an impression material
based on A-silicones (Provil Novo Light C.D.2, Heraeus
Kulzer GmbH) (Fiorenza et al., 2009). Afterwards, casts
with nonreflective, superhard gypsum (Everest Rock,
KaVo), a replica material specially developed for surface
scanning, were produced. We surface-scanned the casts
of the original specimens with a white-light scanner
(smartSCAN3D, Breuckmann GmbH) with a x/y-resolution
of 55 lm. The acquisition and alignment of the
scan-data were processed with optoCAT software
(Breuckmann GmbH), whereas PolyWorks1
10.1 (InnovMetric
Software Inc.) was used for the final 3D image
analysis. We used left molars as a reference while right
molars were mirrored digitally.
The shape and size of the IPWF is usually age-related,
resulting in concave, convex, and sinuous three-dimensional
surfaces in advanced age (Kieser et al., 1985). Thus,
3D and 2D analysis of the IPWF outlines were carried out
to test whether the latter might involve a relevant loss of
information when compared with the 3D outlines.
Orientation of the IPWF
The digital models were approximately oriented manually
on the screen by rotating the occlusal surface parallel
to the xy-plane and the mesio-distal and bucco-lingual
crown diameters parallel to the x-axis and y-axis of
the Cartesian coordinate system (Fig. 2). We considered
the distal IPWF for M1 and the mesial IPWF for M2
(Fig. 3a). In the IMEdit module of PolyWorks1
the outline
of the facets was marked and segmented with the
polyline tool (Fig. 3a). Subsequently, independent digital
models of the corresponding facets were created (Fig.
3b). To investigate the outline shape and size correspondence
of adjacent teeth, the M2 IPWF was rotated 1808
along the z-axis (Fig. 3b). The facets were imported into
the IMInspect module of PolyWorks1
while maintaining
the previous orientation defined in IMEdit, with the buccal
side on the left, lingual side on the right, and the occlusal
side on the top of screen. Depending on the masticatory
forces and on the degree of wear, the molar’s
IPWF show differences in shape. A circular shape, in
which a major axis is not distinguishable, is often dominant
in unworn or moderately worn teeth, whereas in
teeth characterized by advanced wear stages the IPWF
tend to be bucco-lingually elongated and occluso-cervically
narrowed (subrectangular shape). For this reason
the first and second eigenvector of the facet were computed
by fitting all the data points of the digital model
using the singular value decomposition of the data
points distribution (Fig. 2). Each facet was subsequently
rotated until the plane identified by the first two eigenvectors
was parallel to the xy-plane of the Cartesian
coordinate system. The 2D outlines were obtained by
projecting the 3D outlines orthogonally onto the xyplane,
and the buccal endpoint of the first eigenvector
allows identification of the buccal point, b (Fig. 2). For
the identification of the similar points in the 3D outline,
the buccal endpoint of the first eigenvector was orthogonally
projected onto the 3D facet outline.
Afterwards, the outlines (2D and 3D) were imported
into Rhinoceros 3.0 (Robert McNeel & Associates, Seattle,
WA) and standardized for rotation and translation.
Each outline was oriented with its first two eigenvectors
onto the Cartesian coordinate system, rotating the first
eigenvector parallel to the x-axis and the second eigenvector
along the y-axis. Accordingly, the outlines were
translated to overlap the buccal point from which 30
equidistant points were created (Fig. 2).
Statistical method
Statistical analyses were performed in R software (R
Development Core Team, 2008). Shape and form analyses
of the outlines were performed using EFA and GMM.
EFA provides an orthogonal decomposition of each curve
into a sum of harmonically related ellipses that progressively
describe the outlines in detail (Kuhl and Giardina,
1982; Ferson et al., 1985; Rohlf, 1990). As mentioned
above, the outlines were already invariant for translation,
rotation, and starting point. Invariance of size was
obtained by dividing each point’s coordinate (the 30 equidistant
points) with the outline’s centroid size (CS). The
first seven harmonics of the elliptical Fourier series were
sufficient to provide a comprehensive description of
IPWF outlines, explaining almost 98% of the IPWF outlines.
A set of 30 EFA coefficients in 2D (two coefficients
for the zeroth harmonic and four coefficients for each
other harmonic) and 45 EFA coefficients in 3D (three
Fig. 3. (a) Distal facet of M1 and mesial facet of M2 are visualized; (b) facets are virtually detached from the digital dental models,
distal facet of M1 provides the reference orientation. To obtain the same orientation for the mesial facet of M2, the facet is
rotated 1808 along the z-axis, and the internal aspect is considered.
311INTERPROXIMAL FACET ANALYSIS FOR TOOTH ASSOCIATION
American Journal of Physical Anthropology
coefficients for the zeroth harmonic and six coefficients
for each other harmonic) was obtained.
Moreover, the equidistant points (semilandmarks) on
the target outlines were iteratively slid along the outline
(both in 2D and 3D), based on bending energy, to create
geometrically homologous points (Bookstein, 1991, 1997;
Gunz et al., 2005). The sliding was done considering
points on a reference outline. Using this algorithm, the
bending energy between the reference and target outline
was minimized and artificial deformation removed. In
the first step, the first outline was chosen as a reference
outline and the rest of the outlines (target outlines) were
slid according to this one. In the second step, GPA (Generalized
Procrustes Analysis) was performed to obtain a
Procrustes mean outline as a second-step-reference outline,
and the sliding procedure was repeated considering
this reference. Furthermore, GPA and sliding were
repeated iteratively until convergence. Accordingly, final
alignment among outlines was performed by GPA, which
standardizes the size of the objects and optimizes their
rotation and translation so that the distances between
corresponding (semi)landmarks are minimized (Bookstein,
1991; Dryden and Mardia, 1999; Zelditch et al.,
2004). The centered Procrustes shape coordinates (Procrustes
fit coordinates) were augmented by natural logarithm
of centroid size, log CS.
EFA coefficients, squared Procrustes shape, and form
distances were submitted to hierarchical cluster analysis
(complete, single, average, and median linkage) (Everitt,
1974) to quantify the percentage of correctly associated
teeth. Since this analysis is not a scale free method, the
result depends on the units of the variables. Therefore,
EFA coefficients cannot be augmented by log CS. The
combination between EFA coefficients (artificial distances)
and log CS do not create any reasonable space that
could be used for hierarchical cluster analysis.
While intraobserver errors can arise during facet
marking, the quality of the dental 3D models is assumed
from the feature accuracy information of the smartSCAN3D
(612 lm) provided by the Breuckmann GmbH
developer (www.breuckmann.com). Once the IPWF outline
has been marked and segmented, the computation
of the first two eigenvectors and the standardization process
(for translation, rotation, and scaling) of the outlines
is a highly reproducible method. Accordingly, for
the intraobserver error evaluation during facet outline
digitization, the IPWF of eight teeth were marked and
segmented twice. The obtained outlines were compared
with the sample variance (where the variance is calculated
for all teeth together, n 5 34). We found that the
mean intraobserver error rate (mean intraobserver
squared error divided by the sample variance, in percent)
is equal to 2.44%, an error that could be considered
acceptable. For this reason, repeating the entire procedure
to test the results was not required.
Visual approach
The 34 isolated lower molars were analyzed by one observer
using traditional visual methods. The first method
is based only on the visual assessment of IPWF outlines.
The isolated teeth were arranged randomly on the table
with the correct IPWF oriented upwards. In addition, the
molar roots were covered with adhesive paper tape, used
to hide the specimens’ label. The observer was only
allowed to look at the tooth wall of the IPWF under study.
Instead, the second visual method is based on the physical
contact of the isolated molars, where, beyond the
IPWF profiles, continuity in dental morphology, tooth
size, and occlusal wear is also important. In both methods,
the analyses were carried out at different time intervals
of one week and repeated three times to avoid a
familiarization with the teeth during consecutive days.
Comparison of the methods
Finally, to quantify the difference between all the
methods, we used the McNemar chi-square test of symmetry.
It is a nonparametric method for dichotomous (binary)
data (good classification/misclassification) applied
to 232 contingency tables for matched-pair problems. In
our setting, rows and columns of 232 contingency tables
correspond to compared methods (for each visual
approach, the best result was used in the analysis). The
tested hypothesis regards ‘‘marginal homogeneity’’ (compared
methods are given the same results), which is
equivalent to the test of equality of row and column marginal
frequencies (McNemar, 1947).
RESULTS
Based on the visual assessment of the IPWF outline,
not more than eight individuals (47%) were correctly
associated in all three independent tests (Fig. 4). The percentage
of correctly associated teeth increases through
the physical-contact matching of the molars, even if the
results are highly variable: in the first test, 13 individuals
(77%) were correctly recognized, but in the second and
the third experiments, the number of correctly matched
molars decreases to nine individuals (53%).
For the digital approach, the cluster analysis method
with the largest number of true tooth pairs at the single
cluster (the so called ‘most similar’) was chosen. Using
the four hierarchical cluster analyses, we found that
complete linkage presented the highest number of correctly
matched individuals.
Shape-space sliding (semi)landmarks and EFA (for 2D
and 3D outlines) provide almost the same results, with
not more than eight individuals correctly coupled (47%).
The number of correctly associated teeth increases to 14
individuals (82%) in 2D form-space analysis when log CS
is included (Fig. 4). In general, the association of crowns
showed no improvement when using 3D outlines compared
with 2D projected outlines.
Regarding these considerations, we only present the
cluster dendrogram obtained when Procrustes form distances
were submitted to hierarchical cluster analysis.
Among the three misclassified pairs, it is worthwhile
pointing out that only one pair is totally mismatched
(pair 16), the other two pairs being partially clustered
(pairs 9 and 13).
The McNemar chi-square test emphasizes that the
result of hierarchical cluster analysis for 2D form-space
data is significantly different from all the other methods
except the physical manual matching of the molars (Table
1). The matching success rates obtained for the latter
are significantly different from those obtained by EFA
(both 2D and 3D), 3D shape-space, and visual matching
based on IPWF outlines, and are near the significant
level for 2D shape-space.
Finally, it is important to emphasize that the results
are less consistent when only the size (area of the IPWF,
Table 2) is considered in the hierarchical clustering. In
312 S. BENAZZI ET AL.
American Journal of Physical Anthropology
fact, in this case, only four individuals were correctly
associated (11.8%).
DISCUSSION AND CONCLUSIONS
Interproximal wear facets are considered a valuable
feature for the association of isolated teeth (Wolpoff,
1978, 1979; Radovcˇic´ et al., 1988; Coppa et al., 2005;
Genc¸turk et al., 2008). Currently, two different methods
Fig. 4. The bar plots display the percentage of correctly matched teeth in the traditional (top) and digital (bottom) approach,
respectively. Number of associated tooth pairs are inside the bars.
TABLE 1. McNemar chi-square test of symmetry among the methods used for the association of isolated molars
List of
methods
2D
EFA
3D
EFA
2D
shape-space
2D
form-space
3D
shape-space
3D
form-space
Visual
matchinga
Physical
contactb
2D EFA 0 0.655 0.655 0.020 1.000 0.317 0.739 0.034
3D EFA – 0 1 0.034 0.705 0.527 1.000 0.025
2D shape-space – – 0 0.014 0.564 0.480 1.000 0.059
2D form-space – – – 0 0.008 0.046 0.014 0.564
3D shape-space – – – – 0 0.18 0.763 0.034
3D form-space – – – – – 0 0.414 0.180
Visual matching1
– – – – – – 0 0.025
Physical contact2
– – – – – – – 0
a
Visual matching of IPWF outlines.
b
Physical contact of the molars.
Significant P-values \0.05 are given in bold.
TABLE 2. Descriptive statistics of the IPWF area (mm2
)
N 34
Mean 4.98
SD 1.67
Min 1.87
Max 8.64
1st
Qu. 3.82
Median 4.9
3rd
Qu. 6.26
313INTERPROXIMAL FACET ANALYSIS FOR TOOTH ASSOCIATION
American Journal of Physical Anthropology
for tooth association are in use: 1) the visual evaluation
of the IPWF outlines, 2) the manual matching of the
teeth by physical contact of two dental walls, plus the
evaluation of dental features, such as general dental
morphology, tooth size, and the occlusal wear. Nevertheless,
almost nothing is known about the correctness of
IPWF outline matching by the above mentioned methods.
In all the previous works, the IPWF outline was
only qualitatively evaluated (e.g., Wolpoff, 1978, 1979;
Coppa et al., 2005; Genc¸turk et al., 2008). In physical
anthropology, the necessity to support qualitative
descriptions with quantitative data is well recognized
(e.g., Pietrusewsky, 2008). In this respect, a new digital
approach does not only provide the opportunity to quantify
IPWF outlines and to compare them objectively for
tooth associations, it also allows a statistical description
and interpretation for the use of IPWF in the future
determination of the minimum number of individuals in
an assemblage of isolated teeth.
In our pilot study, we observed that the IPWF of adjacent
teeth belonging to the same individual can show
clear differences in outline shape, and an almost similar
shape can be shared between teeth belonging to different
individuals. The results of the shape analysis using a
digital approach (EFA and GMM) are relatively consistent
with the results obtained by visual assessment of the
IPWF outline. In both cases less than 50% of the individuals
were correctly associated. Tooth pairs not correctly
associated (the residual 50%) have different IPWF outlines.
A higher rate of correct matches was obtained by
manually testing the molar fitting, although this
approach showed a high variation in success (between
53% and 77% in the correctness of the results). This variation
could depend on subjective evaluations, where the
success of tooth association is influenced by individual
experience and knowledge of interproximal wear. In
spite of this, the physical manual matching is a simply
applied method which, in contrast to the computer-based
Fig. 5. Complete linkage cluster analysis of the form differences between 2D IPWF outlines. Shadowed specimens are mismatched,
and their outlines are shown to describe the degree of form discrepancy (outlines are at scale). The 17 specimens were labeled
using progressive numbers: specimen 1 5 1.1 (M1) and 1.2 (M2); specimen 2 5 2.1 (M1) and 2.2 (M2); specimen 3 5 3.1 (M1)
and 3.2 (M2); etc.
314 S. BENAZZI ET AL.
American Journal of Physical Anthropology
approach, is less time consuming and does not require
expensive equipment. Consequently, one has to consider
why a more careful and objective evaluation of the IPWF
outline, using both shape and size information, will be
advantageous for tooth associations. In our study, the
percentage of correctly associated teeth increased to 82%
when applying 2D form-space analysis, including log CS.
Accordingly, about 35% of the individuals (the percentage
of individuals obtained by subtracting the result of
2D shape-space (47%) from 2D form-space (82%)) were
matched mainly with regard to their IPWF outline size,
though they present a different shape. In addition, it is
important to mention that 18% of the tooth pairs possess
a completely different IPWF outline (Fig. 5). Accordingly,
based on our preliminary results, we suggest that in a
general molar sample, at least 20% of IPWF pairs could
be completely different, limiting the performance of any
digital approach for tooth association. This limit does not
regard the digital method itself but depends on the
intrinsic biological difference of some IPWF belonging to
adjacent teeth. Why complementary IPWF display differences
in their shape and size were not the focus of this
analysis, but our digital results perhaps will spark some
interest for future investigation to answer this question.
It is worthwhile emphasizing that a quantitative morphometric
analysis can be useful for investigations of
IPWF development and variation (Wolpoff, 1971). Shape
and size variation in IPWF outlines of adjacent teeth
could provide interesting information regarding the relationship
between IPWF development and masticatory
forces that are dependent on dietary and nondietary
habits (Hinton, 1982).
Additionally, we have observed that a 2D digital
approach provides fairly good results because the facets
possess only a shallow concavity (at least in our modern
human sample), with IPWF outlines usually located on a
plane. The 3D outline variability can be decomposed into
two parts: first, the variability in the plane defined by
the first two eigenvectors, which basically accounts for
variability of the 2D outline (as described above, for each
facet the 2D outline is obtained projecting the 3D outline
orthogonally in the plane identified by the first two eigenvectors);
second, the information perpendicular to this
plane. We suggest that adding information from the third
dimension to the 2D outlines creates noise that could also
affect the sliding algorithm, which in the end might interfere
with the number of correctly associated teeth.
As discussed above, interproximal wear is an agerelated
process (Kieser et al., 1985), depending also on
the magnitude of the masticatory forces because of diet
and food preparation (Hinton, 1982). Unworn or slightly
worn teeth are characterized by small and subcircular
IPWF. In this circumstance, the positive coupling mainly
depends on the facet’s size, whereas the IPWF shape
provides small contributions. The digital approach cannot
overcome this problem. When facets are small with a
subcircular shape, the first eigenvector can have unpredictable
directions, affecting the orientation of the IPWF
outline. Consequently, for both manual matching and a
digital approach, a low success rate in correctly associated
teeth is to be expected in young individuals and in
populations with soft diets.
On the contrary, both methods work better with moderately
or heavily worn teeth, where IPWF usually have
subrectangular shapes, which are bucco-lingually
extended. This is the case with adult individuals, and in
populations, such as modern and ancient hunter-gatherers,
where diets required powerful mastication. Nevertheless,
both the approaches show advantages and disadvantages.
It is an advantage of form-space analysis that
results are reproducible and variation in shape and size
can be quantitatively evaluated. On the other hand, the
manual approach accounts for important details in
crown morphology, dental wear, etc., which are not considered
in the digital approach. Accordingly, we suggest
combining qualitative and quantitative approaches for
tooth association derived from manual matching and
form-space analysis of 2D outlines.
As argued by Genc¸turk et al. (2008) and supported by
our results, the correspondence and similarity of two
IPWF outlines is possibly not sufficient to justify an association.
This becomes even more apparent if we look at
the distribution of correctly identified adjacent tooth pairs
in our sample. Only one tooth pair was consistently recognized
correctly in all the methods applied. The results of
our study suggest caution when using IPWF analysis as a
stand-alone feature for determining tooth association.
Instead, IPWF analysis should be combined with other
approaches such as occlusal fingerprint analysis (OFA)
(Kullmer et al., 2009; Fiorenza et al., 2010), morphological
features, and enamel hypoplasia occurrence.
ACKNOWLEDGMENTS
The authors would like to thank Prof. Giorgio Gruppioni
(Department of History and Method for the Conservation
of Cultural Heritage, University of Bologna, Italy)
for providing the dental sample used in this work. Many
thanks also to Dick Byer, Jeremy Tausch, and Amanda
Smith for proof-reading this manuscript. The authors
are grateful to the two anonymous reviewers and the associate
Editor for their important comments, which
improved the quality of this paper. Finally, we would
like to thank Jasaman Faridfar for her invaluable help
regarding tooth association based on a visual approach.
LITERATURE CITED
Bookstein FL. 1991. Morphometric tools for landmark data. Geometry
and biology. New York: Cambridge University Press.
Bookstein FL. 1997. Landmark methods for forms without landmarks:
morphometrics of group differences in outline shape.
Med Image Analysis 1:225–243.
Coppa A, Gru¨n R, Stringer C, Eggins S, Vargiu R. 2005. Newly
recognized Pleistocene human teeth from Tabun Cave. Israel
J Hum Evol 49:301–315.
Dryden IL, Mardia KV. 1999. Statistical shape analysis. New
York: Wiley.
Everitt B. 1974. Cluster analysis. London: Heinemann Educational
Books.
Ferson SF, Rohlf FJ, Koehn RK. 1985. Measuring shape variation
of two-dimensional outlines. Syst Zool 34:59–68.
Fiorenza L, Benazzi S, Kullmer O. 2009. Morphology, wear and
3D digital surface models: materials and techniques to create
high-resolution replicas of teeth. J Anthropol Sci 87:211–218.
Fiorenza L, Benazzi S, Tausch J, Kullmer O, Schrenk F. 2010.
Identification reassessment of the isolated tooth Krapina D58
through occlusal fingerprint analysis. Am J Phys Anthropol
143:306–312.
Genc¸turk I, Alpagut B, Andrews P. 2008. Interproximal wear
facets and tooth associations in the Pasalar hominoid sample.
J Hum Evol 54:480–493.
Gunz P, Mitteroecker P, Bookstein FL. 2005. Semilandmarks in
three dimensions. In: Slice DE, editor. Modern morphometrics
in physical anthropology. New York: Kluwer. p 73–98.
315INTERPROXIMAL FACET ANALYSIS FOR TOOTH ASSOCIATION
American Journal of Physical Anthropology
Gustafson G. 1947. Microscopic examination of teeth as a
means of identification in forensic medicine. J Am Dent
Assn 35:720.
Hinton R. 1982. Differences in interproximal and occlusal tooth
wear among prehistoric Tennessee Indians: implications for
masticatory function. Am J Phys Anthropol 57:103–115.
Kieser JA, Groeneveld HT, Preston CB. 1985. Patterns of dental
wear in the Lengua Indians of Paraguay. Am J Phys Anthropol
66:21–29.
Kuhl FP, Giardina CR. 1982. Elliptic Fourier features of a
closed contour. Comput Graph Image Process 18:236–258.
Kullmer O, Benazzi S, Fiorenza L, Schulz D, Bacso S, Winzen
O. 2009. Technical note: occlusal fingerprint analysis: quantification
of tooth wear pattern. Am J Phys Anthropol 139:600–
605.
Lunt DA. 1974. Identification and tooth morphology. J Forensic
Sci Soc 14:203–207.
McNemar Q. 1947. Note on sampling error of the difference
between correlated proportions or percentages. Psychometrika
12:153–157.
Picton DCA. 1962. Tilting movements of teeth during biting.
Arch Oral Biol 7:151–159.
Pietrusewsky M. 2008. Metric analysis of skeletal remains:
methods and applications. In: Katzenberg MA, Saunders SR,
editors. Biological anthropology of the human skeleton. Hoboken,
NJ: Wiley. p 487–532.
R Development Core Team. 2008. R: a language and environment
for statistical computing. R Foundation for Statistical
Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://
www.r-project.org.
Radovcˇic´ J, Smith FH, Trinkaus E, Wolpoff MH. 1988. The Krapina
hominids. An illustrated catalog of skeletal collection.
Zagreb: Croatian Natural History Museum.
Rohlf FJ. 1990. Fitting curves to outline. In: Rohlf FJ, Bookstein
FL, editors. Proceedings of the Michigan Morphometric Workshop.
The University of Michigan and Museum of Zoology,
MI: Ann Arbor. p 167–178.
Smith BH. 1984. Patterns of molar wear in hunter-gatherers
and agriculturalists. Am J Phys Anthropol 63:39–56.
Weidenreich F. 1937. The dentition of Sinanthropus pekinensis: a
comparative odontography of the hominids. Paleont Sinica D
1:1–180.
Wolpoff MH. 1971. Interstitial wear. Am J Phys Anthropol
34:205–228.
Wolpoff MH. 1978. The dental remains from Krapina. In: Malez
M, editor. Krapinski Pracˇovjek I Evolucija Hominida. Zagreb:
Yugoslav Academy of Sciences and Arts. p 119–144.
Wolpoff MH. 1979. The Krapina dental remains. Am J Phys
Anthropol 50:67–114.
Zelditch ML, Swiderski DL, Sheets DH, Fink WL. 2004.
Geometric morphometrics for biologists: a primer. London:
Elsevier.
316 S. BENAZZI ET AL.
American Journal of Physical Anthropology