See discussions, stats, and author profiles for this publication at: [Link]

net/publication/234149058

Finite Element Analysis and Understanding the Biomechanics and

Evolution of Living and Fossil Organisms

Article in Annual Review of Earth and Planetary Sciences · April 2007

DOI: 10.1146/[Link].35.031306.140104

CITATIONS READS
446 5,243

1 author:

Emily J Rayfield
University of Bristol
257 PUBLICATIONS 6,991 CITATIONS

SEE PROFILE

All content following this page was uploaded by Emily J Rayfield on 06 June 2014.

The user has requested enhancement of the downloaded file.

ANRV309-EA35-18 ARI 20 March 2007 17:2

Finite Element Analysis

and Understanding the
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Biomechanics and Evolution

of Living and Fossil
by Monash University on 05/27/08. For personal use only.

Organisms
Emily J. Rayfield
Department of Earth Sciences, University of Bristol, Wills Memorial Building,
Queens Road, Bristol, BS8 1RJ, United Kingdom; email: [Link]field@[Link]

Annu. Rev. Earth Planet. Sci. 2007. 35:541–76 Key Words

First published online as a Review in Advance on biomechanics, stress, strain, functional morphology, skull, feeding
January 29, 2007

The Annual Review of Earth and Planetary Sciences is Abstract

online at [Link]
Finite element analysis (FEA) is a technique that reconstructs stress,
This article’s doi: strain, and deformation in a digital structure. Although common-
10.1146/[Link].35.031306.140104
place in engineering and orthopedic science for more than 30 years,
Copyright  c 2007 by Annual Reviews. only recently has it begun to be adopted in the zoological and pa-
All rights reserved
leontological sciences to address questions of organismal morphol-
0084-6597/07/0530-0541$20.00 ogy, function, and evolution. Current research tends to focus on ei-
ther deductive studies that assume a close relationship between form
and function or inductive studies that aim to test this relationship,
although explicit hypothesis-testing bridges these two standpoints.
Validation studies have shown congruence between in vivo or in vitro
strain and FE-inferred strain. Future validation work on a broad
range of taxa will assist in phylogenetically bracketing our extinct
animal FE-models to increase confidence in our input parameters,
although currently, FEA has much potential in addressing questions
of form-function relationships, providing appropriate questions are
asked of the existing data.

541
ANRV309-EA35-18 ARI 20 March 2007 17:2

INTRODUCTION
The finite element analysis (FEA) is a technique that reconstructs stress, strain, and
FEA: Finite element deformation in structures. Its methodology centers on the representation of a struc-
analysis; refers to the ture as a series of discrete problems that are readily solvable by mathematical analysis.
process of analysis Its roots lie in mathematical finite difference approximations (Richardson 1910) and
FEM: finite element engineering elastic continuum problems (Turner et al. 1956), but not until the 1960s
method; refers to the were these strands drawn together into the FEM (Clough 1960, Zienkiewicz et al.
methodology
2005). In the 1970s and 1980s, the FEM rose to prominence as an analytical technique
in engineering (Zienkiewicz 1971) and orthopaedic medicine (Huiskes & Chao 1983,
Huskies & Hollister 1993), its success dependent in part on the exponential advances
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

in computing power and availability in the latter part of the twentieth century. FEA
is now widely used to assess the biomechanics of the human musculoskeletal sys-
tem, including soft tissue mechanics, heat transfer, and computational fluid dynamic
problems such as blood flow. Until very recently, however, its potential to engage
in questions of vertebrate biomechanics and evolution remained largely unexplored.
by Monash University on 05/27/08. For personal use only.

This review provides a benchmark for the current state of the application of FEA in
organismal morphology. Although nonvertebrate examples are considered, the bulk
of research performed so far centers around problems in vertebrate morphology,
and nearly all focus on the skull. The aim of this review is to chart perhaps not the
explosion, but the steady stream of papers that have appeared in the past 10 years
or so, utilizing FEA in this area. Here I provide a critical assessment of a field that
still remains in its infancy. I assess the philosophical and scientific basis of applying
stress-strain analysis in vertebrate morphology, review and assess current research
directions, and discuss future programs of research in a field that, if considered care-
fully, has much promise and is increasingly shaping our understanding of vertebrate
morphology, function, and evolution.

STRESSES AND STRAINS

Applied load, or force [measured in Newtons (N)], generates stress within a structure
(Figure 1), where stress (σ) equals force per unit area (N m−2 ). The deformation
resulting from applied load is known as strain (ε) and is defined as the change in
length divided by original length (l/l). Strain is therefore a ratio, but is frequently
assigned the unit microstrain (με), where one με equals a 0.000001 or 1e-106 change in
dimension. By convention, tensile stresses and strains are defined as positive, whereas
compressive stresses and strains bear a negative prefix.

How Bone Responds to Stress and Strain

The orientation, distribution, and magnitude of stress and strain within a structure are
dependent on the applied load, but are also influenced by material properties (Young’s
modulus and Poisson’s ratio) and structural organization (Figure 1). The skeleton is
a dynamic structure and is capable of modifying its shape and biology in response to
applied loads; the central tenet of bone modeling and remodeling. The relationship

542 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

Deformation

Externally
applied Structural
Stress Strain organization
forces (N) Load per Stretch per
Muscle, bite forces, unit area unit length Gross morphology
joint loadings, ground microstructure
reaction force mineralization

Material properties
Stiffness, strength, density, tendency
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

to change width when stretched

Figure 1
Factors affecting how stress and strain are generated and withstood by the skeleton. Modified
from Thomason (1995).
by Monash University on 05/27/08. For personal use only.

between load, resulting stress trajectories, and bony morphology was famously noted
by Julius Wolff (Wolff 1892, translated into English 1986); a formulation that has
evolved into Wolff’s Law, a general paradigm that states that bones are formed as
mechanically optimal structures of maximal strength and minimal weight, according
to a set of particular mathematical rules (see Roesler 1987). Subsequent authors have
pointed out methodological (Huiskes 2000) and experimental (Bertram & Swartz
1991) problems with Wolff ’s hypothesis, yet they do not denounce the concept of
adaptive bone remodeling per se, rather they suggest that it proceeds in an alternative
manner; that of a quantitative self-regulating mechanism (Huiskes 2000). Ultimately,
this is more in tune with our current understanding of how bone remodels: osteocyte
cells trapped within the bony matrix detect a strain signal and elicit a response such as
deposition, removal, or realignment of tissue (see Currey 2002 for a comprehensive
review). Structural modifications can occur at different hierarchical levels within bony
tissue: gross morphology, microscopic morphology (trabecular, osteonal, and collagen
fiber orientation), and at the level of crystalline alignment and percent mineralization
(Currey 2002) (Figure 1). One tends to think of remodeling as an epigenetic process,
yet there is no doubt that bone’s ability to adapt to load occurs at the genetic level
also, and that skeletal shape, material properties, mineral content, or parameters that
control remodeling (e.g., the magnitude of strain that initiates resorption) are selected
for in an evolutionary sense.
Crucially, a signature of loading history and hence function is recorded within
bony tissue. Therefore, any technique, such as FEA, that enables us to reconstruct
stress and strain within the skeleton allows us to explore questions of how that skeleton
functioned and why evolution shaped it in a particular manner. The process by which
loading history influences skeletal geometry is known as mechanical adaptation. Many
classic experiments have demonstrated bone remodeling in response to imposed load
(Goodship et al. 1979, Lanyon et al. 1982; see Currey 2002 for a review), but whether
bones are optimized to loading is a controversial point. As discussed below, how far
one believes that bone can adapt crucially dictates how FEA is applied in a research

[Link] • Finite Element Analysis in Vertebrate Morphology 543

ANRV309-EA35-18 ARI 20 March 2007 17:2

program; a dichotomy exists between those who follow the extreme adaptationist
idea that bone is optimally adapted for function versus those who focus on testing
the actual optimality of skeletal tissues. Some in the latter camp argue that actual
EMG: electromyography
skeletal morphology may represent a compromise between optimal (adult) function
and phylogenetic, ontogenetic, and/or constructional constraints.

Measuring Stress and Strain within the Skeleton

There are various methods available to measure stress and strain within the skeleton.
In studies on extant animals including, to a certain degree, humans, we can perform
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

experimentation, direct observation, and measurement. These are briefly described

below, before I move to discussion of FEA.

Strain gauges. Strain gauges are small, variable, resistive elements bonded to the sur-
face of bone that change electrical resistance in proportion to localized deformation
by Monash University on 05/27/08. For personal use only.

or strain (see Biewener 1992b for further information). In vivo, gauges are occa-
sionally used in concert with electromyography (EMG), which records muscle-firing
patterns. In addition, force transducers (bite bars, foot plates) are used to quantify
loading conditions (see work by authors such as Herring & Teng 2000, Hylander
et al. 1987, and Rafferty et al. 2003 for examples of this approach in skulls). Gauges
have the advantage of being the only method to directly measure strain in vivo during
function and have greatly enhanced our understanding of functional strains in ani-
mals and, on rare occasions, humans (see Currey 2002). However, gauges can only be
applied to the surface of bony tissue and are constrained to accessible regions of bone.
The technique is also invasive and, of course, their use is restricted to living animals
and not fossils, in either in vivo or ex vivo experiments performed on defleshed bone
loaded in hydraulic testing machines.

Geometrical abstraction. An alternate methodology suitable for studies on living

and extinct taxa is to employ a set of partial differential equations derived from con-
ditions of static equilibrium and the theory of elasticity to calculate either stresses or
displacements within a structure. This technique works well when applied to geomet-
ric structures or to those experiencing quantifiable loads (i.e., under the conditions of
beam theory), but the technique becomes complicated to the point of intractability
for complex geometries, loading conditions, or material properties. Hence, complex
structures must be reduced to simple geometric entities, which in some cases may
defeat the purpose of the analysis.

FEA. FEA is an attractive tool for the reconstruction of the mechanical behavior of
the musculoskeletal system in zoological and paleontological studies because it (a) is
noninvasive, (b) reconstructs stress at multiple sites and depths through the skeleton,
(c) is applicable to extinct taxa, and (d ) is capable of dealing with complex geometries
and loading conditions. FEA combines the ability to assess complex scenarios and to
tightly control the loads applied to the structure and to reasonably conduct sensitivity
analyses. Further, because models are digital representations of structures, they may

544 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

be digitally manipulated to test particular hypotheses of function or hypothetical

morphologies.
CAD: computer-aided
design
HOW DOES FEA WORK?
The FEM began life as a mathematical technique but now nearly all FE-analyses
are run on commercially available software packages (e.g., Ansys, Abaqus, Cosmos)
or occasionally users write their own specific code. The first step in the process
is to create a digital representation of the structure of interest. This is termed the
preprocessing stage (Figure 2) and may take place within a computer-aided design
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

(CAD) component of the finite element solution software, or in a separate CAD

package with data transfer capabilities. The structure of interest exists in life as a
continuum, consisting of an infinite number of degrees of freedom. In FEA, by a
process termed discretization, the structure is divided into a finite number of discrete
subregions, called elements, that are interconnected at nodal points, or nodes. This
by Monash University on 05/27/08. For personal use only.

interconnected network of elements and nodes constitutes the finite element mesh
(Figure 2). Elements are then assigned specific material properties that represent the
elasticity of the real structure. Finally, virtual loads are added to the model, typically to
nodal points. Constraining anchors are also determined at this step and mobility may
be restricted to particular degrees of freedom. These applied loads and constraints are
collectively termed the boundary conditions. On running the analysis step (Figure 2),
nodal displacements are calculated in response to the applied boundary conditions,
taking into account structural geometry and the predefined elasticity of the structure
(see The Mathematics of FEA, follow the Supplemental Material link from the Annual
Reviews home page at [Link] Nodal displacements are
used to calculate structural strain, and subsequently stress, which is used to build
a composite picture of the mechanical behavior of the object. After analysis, the
final postprocessing step consists of the representation and interpretation of FEA
results. Most FE software packages contain graphics-based postprocessing capabilities
for such a purpose. Stress-strain magnitudes and displacements are represented as
scaled color plots; orientation as vector plots and frequently models can be animated
to represent structural deformations (Figure 2). Results may be used to assess the
accuracy and appropriateness of mesh generation and boundary conditions, at which
point modifications may be made and the analysis rerun. At this point, a convergency
test may be performed to assess how well the discrete model approximates the true
continuum structure; the analysis is rerun with successively smaller elements until
the stress-strain results converge on the same value.

Early Application of FEA to Musculoskeletal Mechanics

The first application of the FEM in orthopaedic medicine came soon after the tech-
nique was widely adopted in engineering science. Brekelmans et al. (1972) and Rybicki
et al. (1972) used FEA to examine the mechanical behavior of skeletal tissues for the
first time (Figure 3). The commercial application of FEA as a method for examining
the behavior of artificial joints (particularly hip prostheses) ensured a regular source

[Link] • Finite Element Analysis in Vertebrate Morphology 545

 Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]
by Monash University on 05/27/08. For personal use only.
ANRV309-EA35-18

Figure 2
ARI

finite element analysis.

Stepwise description of

546
20 March 2007

Rayfield
17:2
ANRV309-EA35-18 ARI 20 March 2007 17:2

Femoral shaft Greater trochanter

Tension

Compression 88.5 psi

Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Joint
Finite element analysis 0 2000 4000 loading
Beam analysis Stress scale, psi
Femoral head

Figure 3
by Monash University on 05/27/08. For personal use only.

One of the earliest FE-models in orthopedic medicine: a comparison between the FEM and
beam analysis in deducing stress patterns in the proximal third of the femur during joint
loading at the femoral head. Distance of dashed or continuous line from edge of bone
indicates magnitude and for FEA, type of stress at the bones’ outer surface (arrows pointing
toward bone indicate compression, pointing away indicate tension). Beam analysis predicted
higher stress magnitudes than FEA. The authors conclude whilst beam models produced an
accurate representation of stress in the femoral shaft, FEA was more effective at
reconstructing stress in the femoral head. Modified from figure 7, Rybicki et al. (1972).

of funding for this field of research (Rab 1982). This resulted in a huge body of
work examining the practical applications of FEA within orthopaedic biomechanics
(Huiskes & Chao 1983, Huiskes & Hollister 1993) and influential theoretical studies
investigating a wider range of general skeletal processes, such as the role of mechan-
ical stress in skeletal and tissue growth and development and mechanical adaptation
(see studies by Beaupré & Carter 1992, van Rietbergen et al. 1995). The use of FEA to
characterize mechanosensation and transduction at the level of trabeculae (Mullender
et al. 1994, Huiskes et al. 2000) and whole bone (Beaupré et al. 1990) has greatly en-
hanced our understanding of the fundamental factors controlling bone morphology
(see Carter & Beaupré 2001 for further information).

The First Examples of FEA in Zoology and Paleontology

It was some years until the FEM was adopted by zoologists and paleontolgists, and
the first few analyses were a rather eclectic mix of single-taxon studies, with little
holistic focus on methodology, broad-scale evolutionary questions, or comparative
analysis, with the exception of the study of Carter et al. (1998). Carter et al. (1998)
employed FE-methodology devised from studies of human long-bone growth and
development to demonstrate that in vertebrates where perichondral ossification pre-
cedes endochondral ossification during long-bone development (in extant bony fish,
basal tetrapods, amphibians, dinosaurs, and birds) it is the distribution of mechanical
stresses in the ossifying cartilage that tends to inhibit the appearance of secondary

[Link] • Finite Element Analysis in Vertebrate Morphology 547

ANRV309-EA35-18 ARI 20 March 2007 17:2

ossification centers. This innovative marriage of orthopaedic and zoological knowl-

edge holds much potential, but has yet to be followed up by subsequent studies.
To my knowledge, the first example of the FEM in vertebrate morphology is that
of Guillet et al. (1985), who built a simple (approximately 200 elements) 3-D model of
the bill of a shoebill (Balaeniceps rex), an African wading bird (Figure 4a). Stress plots
and displacements were displayed for two bill-loading regimes; however, actual loads
and material properties were not specified. The authors emphasized the potential
of the technique in the mechanical analysis of complex biological entities, yet no
follow-up work is recorded in the literature.
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]
by Monash University on 05/27/08. For personal use only.

Figure 4
Early zoological and paleontological FE-models. (a) Shoebill (Balaeniceps rex) beak (Guillet
et al. 1985); (b) idealized models of the extant foraminifera Marginopora (Song et al. 1994);
(c) echinoid test (Echinus esculentus) (Philippi & Nachtigall 1996); (d ) the relationship between
ammonite septal morphology and safety factor as deduced by FE-modeling (Daniel et al.
1997); (e) FE-model of the wing of the Southern Aeshna dragonfly (Aeshna cyanea) (Kesel et al.
1998). Note the diverse range of taxa and morphological complexity represented by these
models. This is in stark contrast to recent models that have focused almost exclusively on
higher vertebrate skulls. Figure 4a from Guillet et al. (1985, figure 6a); Figure 4c from
Philippi & Nachtigall (1996, figure 3b), reproduced with kind permission of Springer Science
and Business Media. Figures 4b and 4d reproduced with permission from Paleobiology;
Figure 4e reprinted from Kesel et al. 1998, with permission from Elsevier.

548 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

Other relevant studies were carried out on nonvertebrate organisms. Collectively,

these studies tend to utilize sophisticated FE-models with (a) complex 3-D geometry
[e.g., Philippi & Nachtigall (1996) demonstrating that tests of regular echinoids are
well adapted to the mechanical activity of the tube feet; Figure 4c], (b) complex
boundary conditions [e.g., using water depth pressures, Daniel et al. (1997) challenged
the assertion that increased complexity observed through time in ammonite septa
permitted the occupation of increased habitat depth; Figure 4d ], and (c) nonlinear
material properties [examining vein morphology and mechanical strength in the wings
of dragonflies (Kesel et al. 1998); Figure 4e]. In other studies of note, Song et al. (1994)
used simple, stylized 3-D FE-models to investigate morphological optimization in
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

gigantic living benthic foraminifera, concluding that discoid growth patterns were
an adaptation for increased stress resistance while preserving surface area to volume
ratios consistent with depth and photosynthetic potential (Figure 4b) and Hassan
et al. (2002) developed larger, more complex ammonite FE-models that contradicted
Daniel and colleagues’ (1997) claims, and argued once more for increased sutural
by Monash University on 05/27/08. For personal use only.

complexity associated with depth. Such models and counter-models highlight how
FEA results are inherently dependent on the initial model parameters.

SO WHY NOT MORE STUDIES? PROBLEMS

AND ASSUMPTIONS
If FEA has the potential to uniquely advance our understanding of mechanical be-
havior in living and extinct taxa, then why has it not been adopted in zoology and
paleontology with the same fervor that it has been in orthopaedic medicine over the
past 30 years? This reticence is almost certainly a reflection of: (a) the perceived
difficulty in applying the technique, especially by those not mathematically trained;
(b) the lack of availability and cost of FE-software and associated hardware (although
use of FEA is commonplace in engineering sciences); (c) as alluded to in the am-
monite example above, the real concern that there are simply too many problems
and assumptions associated with modeling biological, and especially paleontological,
material effectively and reproducibly with FEA; and (d ) a difference in philosophical
and methodological approaches between researchers.
Point a is obviously a matter of personal choice. Point b is becoming less of a prob-
lem with the advent of cheaper computing power, although advances in FE-software
that can facilitate much larger models, in the realm of millions of elements requiring
parallel processing power may be restrictive or unappealing to the casual or early
stage user. As for the concerns associated with the effectiveness and reproducibility
of FEA, as well as how best to employ FEA, the major issues are detailed in the
following sections.

Model Creation
The first issue faced when creating a FEM is what level of morphological complexity
should or can the model reflect, and how can appropriate morphology be captured
for the digital environment.

[Link] • Finite Element Analysis in Vertebrate Morphology 549

ANRV309-EA35-18 ARI 20 March 2007 17:2

What dimension? From the outset, the user must decide on whether a 1-, 2-, or
3-D model is more appropriate. This decision must be based on both the question to
be addressed and the resources available, as 3-D models contain appreciably greater
numbers of elements and degrees of freedom, therefore solution times are longer
and, depending on the size of the model, sometimes beyond the capabilities of high-
end desktop machines. In situations where loading occurs mostly in a single plane, a
2-D model may suffice, and 2-D models have been employed with some success in
orthopedic and paleontological problems (Carter et al. 1998; Rayfield 2004, 2005a).
When out-of-plane and torsional loads become important, the added costs of complex
model generation and computing time must be weighed against the benefits of a more
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

accurate representation of the mechanical behavior of the structure.

Capturing 2-D geometry. Two-dimensional models can be effectively created by

digitizing photographs or other electronic images (e.g., Rayfield 2004). Freeware
such as ImageJ ([Link] is perfectly adequate for the capture of
by Monash University on 05/27/08. For personal use only.

x, y coordinates that can be fed into FE-software to form a 2-D planar structure
(Figure 5a). Stand-alone or integrated CAD graphics packages can be utilized to
create 2-D or 3-D FE-model geometry, although by nature the resulting models
tend to be rather geometrical in their construction. Whether this, and the use of 2-D
models in general, is appropriate depends on the questions asked.

Capturing 3-D geometry. Laser surface-scanning technologies capture the exter-

nal geometry of a structure in digital form, yet they lack information on internal

Figure 5
Model generation choices. (a) Two-dimensional FE-model of Allosaurus, created by digitizing
photographs; (b) three-dimensional FE-model of Allosaurus, created by assembling CT-scan
data. Note the similarity in stress patterns, except in the posterior skull where the 2-D model
is fixed, demonstrating the utility of the 2-D model as a first approximation in this case.
Examples of element types: (c) 2-D linear triangular and quadrilateral; (d ) 2-D higher-order
quadratic elements; (e) 3-D linear and quadratic element types.

550 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

features (medullary cavity, pneumatic chambers, etc.) and are not wholly appropriate
for 3-D modeling (unless one can be sure that the specimen is solid and homogenous
or accurately represented by a surface shell). Instead, computed tomography (CT)
CT: computed tomography
scanning is currently the best method available to capture both external and inter-
nal geometry of zoological and paleontological specimens (see Rayfield et al. 2001)
(model in Figure 5b is created from CT-scan data). CT-scanning works by detect-
ing the attenuation of an X-ray source passed through a specimen. Grayscale units
(named Hounsfield Units) demonstrate the density of the specimen within a series of
cross-sectional slices taken through the specimen, with black as lowest density (typ-
ically air) and white as highest density. Slices are typically output as native DICOM
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

format, or TIFF files, and can be assembled into 3-D volumetric data. For further
information on CT and MRI scanning, the reader is directed to Spoor et al. (2000a,b).
The resolution of the image depends on the slice thickness and the pixel size (the
latter in turn depends on the size of the field of view: dimensions of the scan window
in meters or millimeters and size of the 2-D slice, typically 512 × 512 or 1024 × 1024
by Monash University on 05/27/08. For personal use only.

pixels). Therefore, image resolution is dependent on the size of the specimen and the
capabilities of the scanner and the contrast between bone and matrix in paleontolog-
ical specimens. Normal medical or veterinary scanners should be capable of scanning
specimens up to perhaps a meter in length. However, they are limited by specimen
width, are typically not capable of reaching micrometer slice thickness, and may have
limited grayscale resolution to distinguish density contrast. High-resolution scanning
facilities such as the University of Texas High Resolution X-ray Computed Tomogra-
phy Facility ([Link] and micro-CT scanners can achieve finer
resolution to hundreds or tens of micrometers in thickness, but specimen size may
be limited. Finer resolution means more slices and longer processing time, therefore
a compromise may have to be reached between resolution and practicality. Other
available scanning techniques include neutron scanning (Schwarz et al. 2005), and
for specimens on the submillimeter scale, synchotron X-ray micro-tomography has
proven extremely effective for reconstructing specimen geometry (Donoghue et al.
2006); both techniques may offer a means to construct FE-models on the macro- or
microscale, although this remains to be explored.

Thresholding. CT data frequently requires thresholding, which involves the re-

moval of unwanted material from the scan slices, such as rock matrix or contain-
ers used to hold specimens in the scanner. Thresholding is frequently performed
by purpose-built software packages such as Amira ([Link] and
Mimics ([Link] [Link]). When the CT
contrast is good, thresholding can be done by eye or, for extant specimens, by us-
ing preprogrammed recognition criteria to distinguish bone or soft tissue. Some
workers have used outline recognition software to trace external and internal speci-
men outlines from good contrast zoological CT data, which are then imported and
linked in CAD software (e.g., Solidworks), and then exported to FEA software (Strait
et al. 2005). On the other hand, some paleontological specimens in which there is
little contrast between matrix and bone may require extensive and difficult thresh-
olding. In paleontology, these techniques currently are used primarily for anatomical

[Link] • Finite Element Analysis in Vertebrate Morphology 551

ANRV309-EA35-18 ARI 20 March 2007 17:2

reconstruction (see Robinson et al. 2005 for an example), yet they have much potential
for the creation of FE-models.

Conversion of CT data to FE-models. This is a major rate-limiting step in the

analysis process. FE-models are created either by surface-based or voxel-based re-
construction techniques (van Rietbergen et al. 1995). The distinction lies in how CT
information is transferred to FE-software. Early paleontological studies employed
surface-based approaches, where individual 2-D CT slices were digitized and the x,y
coordinates imported into FE-software where 3-D models were manually created by
joining the dots (Rayfield et al. 2001) in a crude surface-based approach. This process
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

permits greater manual control over specimen geometry but is extremely time con-
suming, requiring months to create a single model. Other surface-based approaches
involve exporting the thresholded CT data as a 3-D wireframe dxf or stl file (e.g.,
Dumont et al. 2005). This approach may be effective if the FE-software can manage
the many thousands of triangular surfaces that form the wireframe; if not, decimation
by Monash University on 05/27/08. For personal use only.

(reduction of surface triangles) must be performed taking care not to lose important
model resolution. In practice, these steps may need to be performed in a series of
different software applications with different capabilities, including mesh-repairing
tools to fix frequent errors in the 3-D triangular surface.
In contrast, the voxel-based reconstruction method offers much promise in quickly
generating geometrically accurate FE-models. The method works on the principle
that (after thresholding, if appropriate) each individual voxel is converted into a 3-D
finite element. This results in an extremely accurate, but also extremely large (in terms
of the number of elements), FE-model that may reach or exceed the processing ca-
pabilities of individual desktop machines. Trials with the voxel-based reconstruction
software Simpleware ([Link] have shown the
technique to be extremely effective (Rayfield & Milner 2005), and this is a methodol-
ogy that will increase in popularity as computing costs drop and powerful machines
or networked clusters are cheaper to purchase. It should be noted, however, that there
is still a real problem with lack of compatibility between CT to CAD, CAD to FEA,
CT to FEA, and even CAD to CAD data transfer. Different file formats, or even dif-
ferent “flavor” of the same file format (dxf and iges being particular culprits) present
huge obstacles to successful and accurate FE-model formation. Additionally, as most
software packages are marketed at the design industry or medical orthopaedic labs,
evolutionary biology or paleontological research groups with typically more mod-
est budgets may find software costs and yearly maintenance contracts difficult or
even prohibitive to meet. Collaboration with engineers and pooling resources and
knowledge offer a step toward resolving this problem.

Meshing. When discretizing the model to create the FE-mesh, the user must choose
whether to use triangular or quadrilateral (2-D), or tetrahedral or cuboidal (3-D), el-
ements, and also if linear (two nodes per edge) or quadratic (three nodes per edge)
elements are required (Figure 5c–e). Quadrilateral and cuboidal elements are gen-
erally preferred over triangular and tetrahedral elements, as these are more accurate
and less likely to develop dangerously high aspect ratios that generate erroneously

552 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

high element stresses (Beaupré & Carter 1992). In practice, however, tetrahedral el-
ements are sometimes able to mesh the complex geometrical shapes of some skeletal
tissues more effectively than cuboidal elements. Quadratic elements are also more
accurate, as they enable stress and strain to vary over each element rather than linear
elements that have a constant value of stress and strain per element. As they bear more
nodes, quadratic elements demand much greater computational power and solution
time. Dumont et al. (2005) found mean stress values were within 10% on comparing
4-noded linear and 10-noded quadratic element versions of the same FE-model, so
it might be the case that simpler elements might suffice for many problems.
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Material Properties
An extremely important issue for FE-models of any biological tissue—zoological
(including human), or paleontological—is the representation of material properties:
elasticity (Young’s modulus), Poisson’s ratio (the change in width after a given change
by Monash University on 05/27/08. For personal use only.

in length), shear properties (shear modulus), density, and bone mineral fraction. Bone
is a composite and inhomogenous material, and its elastic properties may differ in
different directions rendering it either anisotropic or orthotropic in nature. The FE-
modeler faces the challenge of capturing this variability, and/or determining whether
it is important for the problem at hand. For zoological (including human) tissues, it is
of course possible to directly machine test the elastic properties of bone using either
the classic Instron or a variant hydraulic loading rig, using ultrasonic techniques (e.g.,
Wang & Dechow 2006), or nanoindentation (e.g., Kupczik et al. 2007). We now have
a fairly comprehensive understanding of the material properties of human and bovine
mammal long bones and it has been established that these bones display transverse
isotropy, where radial and transverse orientations share very similar elastic properties
(Reilly & Burstein 1975). Material properties of nonhuman or nonbovine species are
less well understood, although data do exist for a wide range of femora (Erickson
et al. 2002) and other long bones for other species (Currey 1987, Blob & Biewener
1999). Unfortunately, this information has been of limited use in FEA; until recently,
most FE-models have focused on the cranium. Furthermore, until very recently, we
had barely any data on the material properties of cranial bone, with the exception
of the human cranium (e.g., Peterson & Dechow 2003). However, recent studies
on the Rhesus monkey Macaca mulatto and the baboon Papio hamadryas anubis show
that regional variation in material properties within the cranium of these species is
consistent with the human data (Wang & Dechow 2006, Wang et al. 2006), although
data presented by Kupczik et al. (2007) suggest some material property values may
alter with age. Much work still needs to done to gain a comprehensive understanding
of cranial and postcranial material properties in extant vertebrate taxa.

Material properties of fossil skeletons? FE-models of fossil taxa are even more
difficult to develop and “ground-truth” than those of extant species because the ma-
terial properties of the bone of extinct taxa cannot be directly determined for fossil
specimens; soft tissues have degraded away and mineral infill renders elastic proper-
ties more akin to rock than bone. One approach has been to assume that similar

[Link] • Finite Element Analysis in Vertebrate Morphology 553

ANRV309-EA35-18 ARI 20 March 2007 17:2

histological morphology means similar material properties (e.g., thin sections of

Allosaurus cranial bone depict extensive secondary remodeling (Reid 1996) akin to
that observed in the bones of fast-growing bovine mammals (Rayfield et al. 2001,
EPB: extant phylogenetic
bracketing Rayfield 2001). However, with fossil bone it is not possible to effectively characterize
anisotropy or the bone mineral fraction, both of which have important influences on
stress and strain distribution within bone, so working with fossil bone will always be
more problematic than working with living species.
One approach to assessing the most likely material properties of fossil bone is
to use material property data from extant taxa; but should one chose closely related
species or functional analogues? Wang et al. (2006) demonstrated that interspecific
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

differences in material properties in closely related species do exist; however species-

specific patterns were more similar between macaques and baboons than between
either of these and the more distantly related humans (Wang et al. 2006). This is an
extremely promising observation for our studies on extinct taxa; Wang et al. note,
“reasonable estimates of cortical bone elastic properties should be possible for extinct
by Monash University on 05/27/08. For personal use only.

species through the study of phylogenetically related and functionally similar modern
forms,” thus hopefully increasing the accuracy of our extinct FE-models. To address
the question of whether phylogenetic or functional analogues are more appropriate,
more data are required, and it is likely that an extant phylogenetic bracketing (EPB)
approach (see Witmer 1995) would help clarify material property variables in extinct
taxa.

Boundary Conditions
Boundary conditions must represent the loads, displacements, and constraining an-
chors that the structure experiences during function. For extant taxon studies, through
synthesizing experimental data such as bite and ground reaction forces, EMG muscle
recruitment patterns, and muscle fiber dissections, we are able to generate reason-
able estimates of boundary conditions to apply to our models. Extinct animals offer a
greater challenge. Muscle reconstruction offers a means to predict the load imposed
on the skeleton during function, and methodologies such as the extant phyloge-
netic bracket approach (Witmer 1995), wherein inferences are made about fossil taxa
based on their closest living relatives that braacket them phylogenetically, describe a
framework on which levels of confidence or inference can be applied to our recon-
structions. In some cases, muscle scars are prominent and where muscle is contained
by the skeleton itself [such as in the diapsid (i.e., dinosaurs, crocodilians, lizards, and
snakes among other groups) adductor chamber], volumes and cross-sectional areas
can be estimated. As there is a relationship between the cross-sectional surface area
of a muscle and the force it can produce, we can take surface-area measurements and
use known values muscle stress (typically around 300 KPa in vertebrates) to calcu-
late muscular forces. These in turn can be fed into lever arm equations to calculate
absolute bite forces and joint moments. In some cases, either muscle scars are absent
or cross-sectional areas are difficult to predict, which leads to less-reliable absolute
estimates. However, whether we need absolute estimates of muscle, bite, and joint
force for our FE-models of extinct taxa depends on the questions asked. If we aim

554 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

to reconstruct absolute, realistic function, then absolute input values are key; how-
ever, if the aim is to conduct a comparative analysis, test a particular key innovation
or function, or if we are using a reverse engineering approach to deduce load from
strain patterns, then absolute values are of lesser importance. The key point here is
that how appropriate FEA is for extinct taxon analysis depends on the questions being
addressed.

STRESS-STRAIN ANALYSIS IN VERTEBRATE

MORPHOLOGY: APPROACHES AND METHODOLOGY
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Stress-strain analysis in engineering and orthopaedic implant research has clearly de-
fined objectives: fatigue assessment, failure criteria, and overall design testing. Fur-
thermore, the theoretical branches of orthopaedic medicine undertake FE-modeling
with the overarching aim of characterizing bone modeling/remodeling behavior:
How do tissues respond to load and what stress or strain signal initiates a response?
by Monash University on 05/27/08. For personal use only.

Within studies of vertebrate morphology, a broader range of questions may be ad-

dressed under the umbrella of the function and evolution of the musculoskeletal
system. For example, we can investigate what particular functions a structure is or is
not capable of undertaking: a reverse engineering approach. We can test the function
of specific features or key innovations and digitally manipulate our models to explore
the mechanical and functional effect of morphological change. We can investigate
questions of optimality and adaptation and constraint in the skeleton, and stress-strain
analysis can be used in a comparative sense to test the effect of modifying, adding, or
removing structures that evolve within a lineage to decipher the possible mechanical
impetus for evolutionary change.
Which of these questions are addressed and the approach taken depends on the
philosophical and methodological standpoint of the individual researcher. Do we
believe that form entirely represents function? If so, then by employing a reverse
engineering approach we can deduce the loads and therefore function and behavior
of skeletons. Or, alternatively, do we question the rigid marriage of form and function
and seek to employ FE-methods to test how optimized the skeleton is for a particular
function? We might employ FEA to explore the constraints inherent in skeletal form,
imposed by classic developmental, architectural, physical, and phylogenetic factors.
Or, we might want to examine function in individual taxa, or use a comparative
approach to test broad-scale evolutionary trends. Do we wish to use existing skeletal
morphologies, or do we want to employ hypothetical morphologies that enable us to
test evolutionary hypotheses of form? Finally, we might wish to engage in validation
attempts to establish the accuracy of FEA as a technique.
A brief survey of the current FE-vertebrate morphology literature soon reveals
that fundamental differences in approach influence the purpose, methodology, and
even the results of our FEAs. It is most appropriate, therefore, to characterize the
current status of FEA research in vertebrate morphology in the context of current
methodological and philosophical approaches. Current research can be divided into
two broad camps: deductive and inductive programs.

[Link] • Finite Element Analysis in Vertebrate Morphology 555

ANRV309-EA35-18 ARI 20 March 2007 17:2

Deductive Models
Deductive FEA models essentially follow two approaches to deductive logic. The
work of Preuschoft & Witzel can be said to use deductive logic in the sense of applying
general rules (in this case of bone remodeling) to specific cases. Other approaches
conform to the hypothetical-deductive or experimental approach of hypothesis testing
and observational science.

The structuralist approach. After the isolated work of Guillet et al. (1985), some
of the first published work using FE in vertebrate morphology was that of Witzel &
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Preuschoft (1999). These authors follow a highly structuralist, deductive program of

research. Their work has its roots in the theories of Wolff (1892), whereby a set of
mathematical rules govern form, and, following an adaptionist program, it is assumed
that form accurately reflects function, in particular, mechanical function. Thus, this
approach asserts that reverse engineering, and indeed forward engineering, based on
by Monash University on 05/27/08. For personal use only.

set rules can be used to reconstruct accurate form. This approach begins with simpli-
fied models of the structure of interest, to which representative boundary conditions
(bite and muscle forces, constraining points) are applied, and a series of analyses run
in which unstressed elements are sequentially removed, resulting in a fully optimized
model in which all elements are stressed in response to load. In their first paper
describing this approach, Witzel & Preuschoft (1999) studied the effect of the general
proportions of the facial skeleton on the distribution of facial stresses in primates.
A series of extremely simplified planar, or 3-D, models were created in ANSYS FE-
software to represent either high face pongids and some cercopithicoids with orbits
dorsal to the nasal opening, or low face genus Homo (orbits level with the dorsal nasal
cavity). Models were loaded with bite forces of unspecified magnitude and bone was
treated as a homogenous, isotropic material. Results suggest that the prominence of
the hominid nasal skeleton is an adaptation to contain bite-induced cranial stresses.
This work formed the basis for future studies using deductive methodology: start-
ing with abstract geometric models, applying loads, and removing unstressed regions.
In all cases, the net result is a transmutation from a geometric block of elements to
a structure representing the skeletal part under study. These papers led to the devel-
opment of an overarching argument that the shape of the skull is purely determined
by a set of specific biomechanical and biological factors. The length and width of the
jaw and the relative position of the jaw in relation to the braincase and nasal capsules,
the location of the nasal opening and orbits, and the size of the braincase and nasal
capsules are acted on by selection and are responsible for the basic parameters of skull
form. All other aspects of cranial form are fully controlled by mechanical behavior, and
stress flows resulting from functional loads dictate cranial shape. FE-models incor-
porating these predetermined parameters are all that are needed to reproduce cranial
form (Preuschoft & Witzel 2002a). Accessory fenestra (cranial openings) for weight
reduction, protection of the brain and sense organs, and even armor are dismissed as
secondary functions of cranial bone.
These parameters, they argue, are at work in a diverse array of vertebrate taxa,
ranging from basal synapsids (from which mammals are derived) to dinosaurs. For

556 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

example, sinuses and pneumatized spaces in hominid skulls are a result of low and
negligible stress concentrations (Preuschoft & Witzel 2002b), and the skull of the her-
bivorous sauropod dinosaur Diplodocus is optimally constructed for maximal strength
with minimal material (Figure 6; Witzel & Preuschoft 2005). Moreover, element re-
moval shows a correlation between resulting stress flows and bony material in humans
and other primates (Preuschoft & Witzel 2002c), and Neanderthals and australop-
ithecines (Preuschoft & Witzel 2004a). For example, they argue that if the temporalis
muscle takes its origin from the posterior wall of the orbit, then the supraorbital
torus will form similar to the condition seen in the skull of Gorilla or Paranthropus
(Preuschoft & Witzel 2004b). Furthermore, they conclude that the postorbital bar is a
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

constructional element of the skull (Preuschoft & Witzel 2005) and is not protective;
cranial fenestra appear in unstressed regions, as stress flows are concentrated primar-
ily to the dorsal and ventral surfaces of nonmammalian taxa. They also argue that the
by Monash University on 05/27/08. For personal use only.

Figure 6
An example of the deductive method. (a) The skull of Diplodocus. (b) The original FE-model
loaded with muscle force and bite restraints. (c) Compressive surface stress after the first model
iteration. Note how the occipital region (supraoccipital: soc; basioccipital: boc) is highly
stressed, whereas the anterior rostrum (snout) is stress free. Elements experiencing <−0.8 N
mm−2 will be subsequently removed. (d ) Final model of successive iterations and element
removal. Reproduced with permission from Witzel & Preuschoft (2005), John Wiley & Sons
Inc.

[Link] • Finite Element Analysis in Vertebrate Morphology 557

ANRV309-EA35-18 ARI 20 March 2007 17:2

size of the braincase dictates whether the skull is composed of a shell-type structure
(as seen in mammals) or a series of rods and plates (as seen in lizards) (Preuschoft &
Witzel 2002).
Implicit to all these studies is that: (a) constraints (physical, developmental, phylo-
genetic and those imposed by functions other than feeding), aside from those clearly
laid out in the biological parameters, are not present, and thus, models are optimally
constructed; (b) the loading conditions acting on the skull are known a priori; (c) bone
remodels in response to compressive stress; and (d ) cranial form is controlled by
function and is independent of phylogeny. Other studies employing similar ideology
but using different parameters as the bone remodeling trigger (strain energy density,
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

axial strain, or strain rate) have reproduced detailed, realistic models of femoral ossi-
fication and trabecular architecture under load (see Carter & Beaupré 2001). Witzel
& Preuschoft argue the subsequent modification of FE-models is similar to an adap-
tive, evolutionary process, but the process does not take into account phylogenetic
history: the ancestral morphologies and modifications that ultimately constrain form
by Monash University on 05/27/08. For personal use only.

and thus evolutionary history is not considered.

The hypothetical-deductive approach. A recent application of FE-modeling has

been to employ traditional hypothesis-testing methodology. A particular hypothesis
of function or loading conditions is derived, and then models are used to test the
hypothesis. The benefits of such an approach are that FE-models can be digitally
manipulated to alter loading conditions or add or remove pieces of bone, free joints,
or fill in fenestra. A requirement of these studies is to determine how differences
between the models will be quantified and how the hypothesis will be falsified. As
bone fails first in tension rather than compression (Reilly & Burstein 1975), some
previous studies have chosen peak tensile stress as an indicator of comparative skull
strength (Rayfield 2004).
Jenkins et al. (2002) created two approximately 350-element models of the snout
of a gorgonopsid and a therocephalian synapsid, to which they applied either bilateral
or unilateral canine vertical and shear bites. Bite-force magnitudes were based on data
of extant canine tooth biting in dogs (Thomason 1991). The models were geometrical
approximations and the chosen material properties were those of mild steel, but even
under these conditions, circumorbital scarf joints and palatal mobility appear well
adapted to contain the feeding-induced stresses predicted by the FE-model, and
hence sustain impact loads produced by an active carnivorous lifestyle (Figure 7a).
Rayfield (2004) examined the mechanical behavior of the Tyrannosaurus rex skull.
Evidence from tooth-marked prey bones suggested that T. rex fed using a puncture-
pull feeding strategy, in which a powerful, bone-crushing bite was followed by drag-
ging teeth across prey flesh and bone. This was despite the paradoxical nature of a
series of open (patent) sutures between the facial bones. A series of FE-models were
generated with and without a break in the mesh, forming a crude representation of
a sutural contact. Loads were applied in a vertical fashion to represent puncturing
and in a horizontal manner for pulling. Results showed that the open sutures acted to
lower potentially damaging tensile stress, thus with the addition of soft tissues, they
may have functioned as shock-absorbers. Furthermore, peak compressive and shear

558 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]
by Monash University on 05/27/08. For personal use only.

Figure 7
Hypothesis testing with FEA. (a) Therapsid rostra (snouts) experiencing
compression-inducing unilateral canine bite load, most clearly seen in the yellow element
blocks depicting a compressive trajectory from tooth to rear snout (top) and mild compression
in red in the palate (bottom) (Jenkins et al. 2002); (b) bite-induced compressive stress in
Tyrannosaurus rex without (top) and with (bottom) an open (patent) maxilla-jugal suture; black
arrows indicate stress orientation (Rayfield 2004); (c) Von Mises (VM) stress (a function of
principal stresses σ1 , σ2 and σ3 that measures how stress distorts a material; failure of ductile
material is estimated when VM stress equals the yield strength of the material in uniaxial
tension) in Alligator mississippiensis rostra (snout) before (top) and after (bottom) the introduction
of an antorbital fenestra (E.J. Rayfield & A. C. Milner, unpublished data). Figure 7a
reproduced with permission from Jenkins et al. (2002), Schweizerbart Publishers
([Link]

stresses localized to the nasal bones (Figure 7b), which are characteristically fused
from an early ontogenetic stage in all tyrannosaurids. Further comparative studies
between the taxonomically disparate theropod dinosaurs Coelophysis, Allosaurus, and
Tyrannosaurus demonstrated that although T. rex localized compression and shear to
the nasal regions, the other taxa localized compression and shear to the caudally (pos-
teriorly) placed fronto-parietal region (Rayfield 2005a). These results were in good
agreement with the morphology of the skull roof: Allosaurus has fused frontals and
parietals but gracile nasals, whereas T. rex has robust nasals and gracile frontals. Thus,
a link between morphology and functional load is strongly suggested. Both studies
used 2-D FE-models, which generally work well when approximating a planar struc-
ture with constant cross-sectional geometry that is relatively long in the in-plane
dimension, but are less likely to accurately represent the mediolaterally expanded
rear of the skull, especially as these models were constrained in this region, about
the lateral temporal fenestra. Isotropic material properties were employed, based on
the assumptions of histology presented in Rayfield et al. (2001). These models are a
useful first approximation and can reconstruct most stress distribution patterns seen

[Link] • Finite Element Analysis in Vertebrate Morphology 559

ANRV309-EA35-18 ARI 20 March 2007 17:2

in the 3-D models (E.J. Rayfield, personal observation for Allosaurus). Further 3-D
models are needed to confirm these interesting results.
A recent study (Rayfield & Milner 2006) exploited new developments in FE-model
generation with Simpleware—a voxel-based FE-generation piece of software. The
skull of the unusual theropod dinosaur Baryonyx walkeri is convergently similar to the
large extant crocodilians and fossilized gut contents reveal the animal had at least a
partially piscivorous (fish) diet. Rayfield & Milner (2006) used plots of peak cranial
stress at successive bite positions along the jaw to test which animal, extant gavials,
alligators, or extinct generalized theropods, formed the most appropriate mechanical
analogue for Baryonyx. They also performed a series of hypothesis-testing models
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

to add or remove the bony secondary palate separating the nasal and buccal (mouth
cavity) passages, and the antorbital fenestral opening, both key crocodylian innova-
tions (Figure 7c). Baryonyx was mechanically most similar to the modern piscivorous
gavial, and in both creatures the secondary palate and reduced antorbital fenestra
strengthened the skull against bending. Interestingly, the palate performed a torsion-
by Monash University on 05/27/08. For personal use only.

resisting role in the alligator, which is more akin to alligator grab-and-roll feeding
observed in the wild. In this study, we can make inferences about the feeding behavior
of extinct animals and also assess the mechanical impetus for morphological change
at the base of the modern crocodylian lineage. With these phylogenetically broad
analyses, hypothesis-testing models serve their greatest purpose.

Inductive Models
Witzel & Preuschoft (2005) claim that inductive FE-models “do not allow the expla-
nation of skull shape as an absolute necessity and a mechanical answer to mechanical
loading”, meaning they do not allow one to include within a FE-analysis their de-
ductive methodology of applying general rules, i.e., that skeletal morphology is com-
pletely dictated by functional loading. This is indeed true, but should skull shape and
mechanical loading be viewed so intimately linked in the first place? The inductive
approach has the benefit of being able to test the assumption that form and function
are tightly linked and to investigate form without a set of predetermined assumptions.
On a more modest level, inductive methodology forms the backbone of studies into
FEA validation and the representation of reality.
The inductive method has been employed in studies of both extinct and extant
organisms. Three rough categories of analysis can be defined: analyses of individual
taxa, comparative behavior, validation/sensitivity analysis.

Analyses of individual taxa. The first FE-study in vertebrate morphology that at-
tempted to represent accurate geometry and estimate loading conditions and realistic
material properties was that of Rayfield et al. (2001), based on Rayfield (2001, unpub-
lished thesis). A model of the extinct theropod dinosaur Allosaurus fragilis was created,
using CosmosM software (Figure 8a). External and internal geometry was recon-
structed using CT scan data, via a laborious process of digitizing individual 2-D slices
in TIFF format using ScionImage (a forerunner of ImageJ freeware), finally creating
a 3-D model in excess of 200,000 elements. The assumption was made that similar

560 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

Figure 8
Single taxon studies.
(a) Allosaurus fragilis:
compressive stress after bite
at maxillary teeth 3–5 (black
arrows) (modified from
Rayfield et al. 2001);
(b) deformation in a
pterosaur skull when
feeding at the tips of the
jaws (reproduced with
permission from Fastnacht
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

et al. 2002, Schweizerbart

Publishers, [Link]
[Link]).
by Monash University on 05/27/08. For personal use only.

histology reflected similar material properties, and the model was assigned the mate-
rial properties of bovine bone based on similarities between the extent of Haversian
remodeling (the degree to which the canals that surround the blood vessels and nerves
PCSA: physiological
within the bone are remodeled) in dinosaur and bovine bone. Owing to the difficulties cross-sectional area
of designating orthotropic properties a priori to the cranium, isotropy was assumed.
SF: safety factor
The skull was constrained at the occipital condyle (where the skull and backbone meet)
and at the insertions of the cervical (neck) musculature on the occiput. Muscular and
bite loads were estimated using a variation of the physiological cross-sectional area
(PCSA) approach of Weijs & Hillen (1985) and the dry-skull method of Thomason
(1991), in which adductor muscles were reconstructed in clay within a life-size cast
of the skull, the maximum cross-sectional area measured, and muscle and bite forces
calculated and applied to the model. Results depicted stress trajectories through the
skull during a bite. Stresses looped around the fenestral openings and central struts
of bone (contra Preuschoft & Witzel 2002). By increasing bite loading at the teeth
until the yield strength of bone was reached, it was possible to calculate a safety factor
(SF) for the skull. This ranged from 9 to 69 depending on loading conditions, but in
nearly all cases was much greater than existing safety factor measurements for extant
crania and long bones. An alternative is that the skull is simply over-engineered for
normal feeding loads. Rayfield et al. (2001) reasoned to the skull could accommodate
large dynamic loads generated by the animals feeding strategy if it were to slam the
head into prey (in the process obviating the need for the weaker lower jaw in feeding).
Frazzetta & Kardong (2002) questioned whether the basicranial joint could withstand
such an impact, yet Rayfield et al. (2002) drew their attention to a system of sliding
joints preserving basicranial integrity during biting. Although under-estimation of
muscle and bite force loads applied to the model would make it erroneously stronger
and wrongly elevate the SF, loads would have to be quadrupled to reduce SF to levels

[Link] • Finite Element Analysis in Vertebrate Morphology 561

ANRV309-EA35-18 ARI 20 March 2007 17:2

observed in living taxa. Most likely, underestimation of load (static and dynamic), in
combination with representing the skull as a single entity rather than collection of
sutured elements, coupled with an over-engineered construction are responsible for
elevated safety factors.
Using the same 3-D Allosaurus model, it was discovered that during many load-
ing conditions, cranial stresses generated by feeding loads were aligned along sutural
contacts, such that sutural morphology appeared adapted to accommodate cranial
stress and strain: tensile stresses appeared aligned with sliding scarf joints, whereas
compressive and shear stresses appeared in areas of highly interdigitated sutures
(Rayfield 2005b). Wang et al. (2007) demonstrate how sutures dampen and change
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

the magnitude of strain acting across them in vitro in rhesus monkeys [following
similar observations made in vivo in miniature pigs (e.g., Rafferty et al. 2003) and al-
ligators (Metzger et al. 2005)] and emphasize the importance of introducing sutures
into FE-models to build an accurate picture of cranial strain.
CT data was used to create a first approximation, 3-D, low-resolution, 2400-shell
by Monash University on 05/27/08. For personal use only.

element model of the Alligator mississippiensis cranium (Daniel & McHenry 2001) and
subsequently six FE-models of the snout of five typical longirostrine or brevirostrine
crocodylian species (McHenry et al. 2006). In the 2001 model, proportional adductor
muscle loads were applied to the posterior skull, but instead of applying bite loads,
the teeth and jaw joint were constrained from movement to approximate a static
rigid-body load. Major pathways of compressive stress correlate to the orientation of
tracts of thickened ornament observed in older individuals, with interesting implica-
tions for many reptiles and amphibians that possess such characteristic thickenings.
Isotropic properties were modified to orthotropic, with the stiffest axis parallel with
stress trajectories; overall cranial stresses increased in response. The comparative
models of the 2006 paper show a good correlation between beam analysis and FE-
model results. Both approaches demonstrate how different shaped snouts performed
better under different loading regimes, and that the broad flat platyrostral skull of
modern crocodiles was not optimized for feeding and thus may have been subject to
hydrodynamic constraints.
Fastnacht et al. 2002 used pterosaur skulls (Figure 8b) and teeth to examine the
effect of optimizing element size on model deformation and accuracy and to conduct a
sensitivity analysis of the effect of introducing enamel and dentine material properties,
which is a key issue that deserves further attention in FE-studies.
The only FEA so far performed on the postcranial skeleton, Snively & Russell
(2002) created an FE-model of the left metatarsus of the theropod dinosaur
Gorgosaurus libratus that suggests axial compressive stress is transmitted through the
metatarsus and ligaments assist in damping rotation of the distal elements.

Comparative behavioral studies. A series of papers have used FEA to investigate

the food processing capabilities of primates and other mammals, through studies of
the mechanics of teeth at the gross morphological level and at the level of enamel
crystalline structure, and the interaction of food particles with tooth geometry. Rens-
berger pioneered the study of enamel microstructure and FE-models, discovering,
among other observations, that the highly modified wave-like decussation planes in

562 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

the canines of the spotted hyena Crocuta crocuta were optimized in areas of peak
stress (Rensberger 1995). Spears & Crompton (1996) investigated the relationship
between angulation of intercuspal surfaces in great ape molars and the mechanism of
food breakdown during mastication. They focused on the loads employed to the food
particle by tooth geometry, creating a circular 2-D FE-model of the food particle,
comprising 108 elements with hyperelastic (large deformation) properties. Particles
were loaded by a stiff beam element, parallel to the occlusal surface. Models predict
that the sloping occlusal surfaces of Pongo molars are more likely to crush food parti-
cles, whereas Homo, Pan, and Gorilla fracture small food particles by shear. Subsequent
analyses examining stresses within teeth, and modeling the food particle with the ma-
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

terial properties of Mezzettia seeds, a known foodstuff of primates, found differences

in the biomechanical behavior of human, Pan, and Pongo molars, suggesting molars
are adapted to differently directed loads depending on their position in the mouth;
in particular, maxillary molars function as load-receiving mortars, whereas mandibu-
lar molars are able to dissipate the loads associated with a pestle-type action (Spears
by Monash University on 05/27/08. For personal use only.

& Macho 1998, Macho & Spears 1999). Introducing accurate FE-representation of
teeth is very difficult owing to the varying material properties of teeth and their
surrounding alveolar bone and periodontal ligament; however, these authors were
able to model teeth as sophisticated 2-D elements, with anisotropic properties of
enamel, isotropic dentine, and frictionless contact between tooth and food parti-
cle (Figure 9a). Although these studies do not take into account the behavior of
the alveolus, Chen & Chen (1998) in an FE-model of the third molar of a gorilla
(Gorilla gorilla) concluded that although properties of alveolar bone are important in
determining the strain field, the exact stiffness of periodontal ligament is less impor-
tant. Furthermore, Macho et al. (2005) created four highly detailed (3-D, >200,000
element), micro-FE models of the enamel prism microstructure of human, chimp,

Figure 9
Comparative behavioral studies. (a) Finite element mesh of a cross-section of a human
maxillary second molar M2 with meshed Mezzettia seed and enamel-dentine boundary (in red )
clearly seen (Spears & Macho 1998); (b) Von Mises stress during unilateral molar biting (left)
and bilateral canine biting (right) in the skull of the fruit bat Artibeus jamaicensis (Dumont et al.
2005) Von Mises stress is a function of principal stresses σ1 , σ2, and σ3 , that measures how
stress distorts a material. Failure of ductile material is estimated when VM stress equals the
yield strength of the material in uniaxial tension. Figure 9a reproduced with permission from
Spears & Macho (1998), John Wiley & Sons, Inc. Figure 9b reproduced with permission from
Dumont et al. (2005), John Wiley & Sons, Inc.

[Link] • Finite Element Analysis in Vertebrate Morphology 563

ANRV309-EA35-18 ARI 20 March 2007 17:2

gorilla, and Australopithicus anamensis teeth. Properties were assigned as before, and a
load of 1 MPa applied to the structure. Models were validated by statistically compar-
ing the Young’s modulus as derived from an FE-model of enamel against published
experimental data. The structural arrangement of prismatic enamel in A. anamensis
is more effective at load dissipation under compressive loads and implies this species
was adapted for puncture-crushing a hard-tough diet.
Dumont et al. (2005) used FEA models to test the hypothesis that two species of
bats that have converged to a similar ecological niche, Artibeus and Cynopterus, have
converged upon the same craniofacial morphology and mechanical performance.
Previous in vivo observations revealed that each bat exhibits a highly preferential
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

mode of feeding, and Dumont et al. (2005) discovered that each skull was most
resistant to their preferred loading regime, although Cynopterus less so than Artibeus.
Interestingly, in both cases, the skulls were not wholly optimized for feeding loads
because large strain gradients were observed within cranial bone. Dumont et al.
(2005) used a sophisticated set of software programs to convert and repair digital stl
by Monash University on 05/27/08. For personal use only.

(stereolithograhy) surface mesh files obtained from CT data to a refined, volumetric

10-noded tetrahedral, quadratic element mesh (Figure 9b). The many steps in the
chain, utilizing often expensive software packages, emphasize the obstacles one must
overcome to produce accurate, reliable FE-models from CT data.

Validation studies and sensitivity analysis. A key question applicable to all the
studies described so far is how accurately do our FE-results reflect reality? This is
a central question in FEA in vertebrate morphology, and one that has so far been
explored in only a handful of taxa, primarily mammals, and the alligator. The two
central issues of validation are precision and accuracy. Precision is the problem of
representing a continuum structure as a series of discrete elements. Precision can be
assessed by performing a convergency test, as explained in How Does FEA Work?,
above. All users of the FEM are encouraged to perform convergence tests (Huiskes
& Chao 1983), but in practice many do not. A common problem is lack of computing
power as element numbers exponentially increase. A potential solution is adaptive
meshing, in which areas of predicted high strain or complex geometry are assigned
greater element density than less-complex regions. The ease of this approach gener-
ally depends on the software employed and user experience. In contrast, the issue of
accuracy focuses on the reproducibility of biological reality: the estimation of geom-
etry, boundary conditions, and material properties. On the whole, the results of vali-
dation studies have been encouraging: FE strain orientations are generally confluent
with in vivo recorded strain, and as the levels of accuracy are increased (incorporating
orthotropic elastic properties, etc., into our models), so strain magnitudes become
increasingly realistic.
In vivo bone strain in the facial skeleton of the crab-eating macaque (Macaca fas-
cicularis) has been well characterized (e.g., Hylander et al. 1987 and other papers),
therefore it is of no surprise that the most extensive validation studies so far have
focused on the macaque cranium. Strait et al. (2005) and Kupczik et al. (2007) have
created a series of CT-derived FE-models to investigate the effect of material prop-
erties on FE-generated strain patterns. After applying physiological muscle loads

564 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

adjusted with EMG data for activation patterns, Strait et al. (2005) found that in
nearly all cases, FE-models overestimated average in vivo shear strains (Figure 10a)
and the maximum:minimum principal strain ratio (Figure 10b), yet the pattern of
strain magnitude and orientation is mostly reproduced reasonably accurately by FE-
models (Figure 10a–c). Both studies found that FE-models with either isotropic or
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]
by Monash University on 05/27/08. For personal use only.

Figure 10
FEA validation in the crab-eating macaque, Macaca fascicularis: effect of material properties.
Comparison of (a) maximum shear stress between in vivo experiments (mean and ± 2 S.D.)
and (b) max:min principal strain ratio, and (c) maximum principal strain orientation for
FE-models with varying material properties. Analysis 1: isotropic human properties; Analysis
2: isotropic macaque properties; Analysis 3: isotropic macaque properties with regional
variation; Analysis 4: orthotropic macaque properties with regional variation. Reproduced
with permission from Strait et al. (2005), John Wiley & Sons, Inc.

[Link] • Finite Element Analysis in Vertebrate Morphology 565

ANRV309-EA35-18 ARI 20 March 2007 17:2

orthotropic macaque craniofacial material properties depicting regional variation re-

produce in vivo strain most effectively (Analysis 3 and 4, respectively, Figure 10). It is
encouraging that even models using isotropic, homogenous human elastic properties
in the Strait et al. (2005) analysis produce strain data that is, on all but one occasion,
within the spread of data recorded in vivo in a macaque (Analysis 1, Figure 10).
Thus, more appropriate properties increase accuracy, yet largely realistic models can
be achieved with isotropy in this case.
A further study by the same authors (Ross et al. 2005) performed a similar analysis
on the effects of varying muscle forces on FE-model results. They constructed 36
different models exhibiting variation in (a) sizes of muscles (different PCSA measure-
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

ments), (b) relative firing pattern of muscles, and (c) timing of muscle recruitment
prior to peak bite force. In vivo recorded strain data show some variability in ori-
entation and magnitude, whereas FE yields a more precise estimate that in most
cases lies within the spread of in vivo measurements ± 2 S.D. Results showed that
most of the variation in model behavior was accounted for by differences in the
by Monash University on 05/27/08. For personal use only.

overall magnitude of muscle force, rather than relative amounts of force generated
by each muscle. Differences in working versus balancing-side muscle recruitment
patterns influence model strain also, as would be expected. In general, the results
demonstrated that to accurately recreate in vivo strain in an FE-model, knowledge
of muscle recruitment patterns is important; yet “a lot can be learned about patterns
of skull deformation in fossil species. . .by applying external forces proportional to
the estimated relative PCSAs of the jaw adductors” (Ross et al. 2005). Interestingly,
Kupczik et al. (2007) found that strains were more realistic when muscle forces were
applied directly to the skeletal nodes, rather than via a digitally modeled muscle. This
is helpful, as it reduces the need to model soft tissues primarily for load application,
although the role of soft tissues as tensile members during function must not be
underestimated.
It is worth noting that to gather workable data on strain at nodal points, Ross et al.
(2005) and Strait et al. (2005) recorded strain values for a network of nodes covering
the cranial surface, with which they used geometric morphometrics to explore changes
in shape. This approach helps distil the strain signal from the many tens or hundreds
of thousands of nodal strain values produced by the FEM output.
With their validated FE-model, Strait et al. (2007) are now able to deduce macaque
cranial function and to use this data to make dependable predictions as to the influ-
ence of cranial form in human evolution, such that derived cranial features in robust
australopithecines may be adaptations to withstand premolar loads generated by mas-
tication of large, resistance food items.
Further preliminary validation of FE-models has been found with the replication
of Buckland-Wright’s (1978) classic work on patterns of strain in the cat skull; the same
five zones of peak strain and comparable strain magnitudes found by Buckland-Wright
were replicated in silico by FE-models (Curtis et al. 2005). Thomason et al. (2005)
expressed some surprise that the complex morphology and elastic properties of horses
hooves were replicated with a high level of accuracy by FE-models incorporating
tissue-specific, but isotropic, material properties. Silva et al. (2005) found that strains
in micro-CT-generated FE-models of mice lower hindlimbs in three-point bending

566 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

were within 15% of in vitro measured values. In contrast, Marinescu et al. (2005)
conducted a specimen-specific validation of a Macaca fascicularis mandible, and found
generally quite poor correlation between tensile strain and the ratio between principle
strains on comparing in vitro and FE-results. The authors point out the danger of
overconstraining FE-models, erroneously exaggerating model stiffness and reducing
strain values. This study once more highlights the importance of orthotropic and
heterogeneous material properties if striving for accurate FE-reproduction of cranial
strain. Wilcox (2007) also highlights the importance of element-specific material
properties and also morphologically accurate geometry, in her in vitro strain validation
of subject-specific FE-models of porcine vertebra.
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Nearly all validation studies have been performed on mammalian skulls. The
exception is the study of Metzger et al. (2005) on the cranium of an Alligator mis-
sissippiensis. Using Daniel & McHenry’s (2001) FE-model of the alligator, Metzger
et al. found that there were clear differences in absolute strain magnitudes and the
patterns of strain gradients between in vivo recorded cranial strain, strain calculated
by Monash University on 05/27/08. For personal use only.

by treating the skull as a simple beam, and FE-derived strain; however, the FE-model
was able to reproduce strain orientations. This is in contrast to the bulk of mam-
malian validation studies that have shown good correlation between inferred strain
from FE-models and strains measured in vivo. The low-resolution, shell element,
isotropic alligator model, with unquantified muscle forces, may be responsible for
this lack of correlation, perhaps exacerbated by the fact that cranial sutures were not
modeled. However, macaque models reached reasonable correlation without sutures,
even though Kupczik et al. (2007) demonstrate that introducing sutures increases
FE-model accuracy. Future, more accurate models are needed to explore crocodylian
validation further.

Summary
So to summarize, FEA has, in the past ten or so years, slowly risen to become a
potential tool of choice for biomechanists and evolutionary anatomists to decode
the functional and evolutionary significance of skeletal form. The technique has ad-
vanced through the first studies of Rensberger (1992) and Spears & Macho (1998)
employing 2-D models of teeth, to the more complex, volumetric, 3-D models of
Rayfield et al. (2001), utilizing increasingly available fossil CT data, but facing prob-
lems of model creation, to the more recent studies of Dumont et al. (2005) and Ross
et al. (2005), which used a battery of software programs to convert CT data into
accurate, volumetric models. Conceptually, applications of the technique range from
the mechanistic deductive studies of Preuschoft & Witzel 1999 (onwards) to those
beginning to explore optimization, constraint and mechanical adaptation (Rayfield
2005b), and direct correlation between functional demands of feeding and cranio-
facial morphology (Dumont et al. 2005). Sophisticated validation studies in living
animals are enabling us to place error bars and confidence levels on our models of
both living and extinct taxa, and to deduce what parameters matter the most in accu-
rate model reconstruction. Strikingly, models have almost exclusively focused on the
cranium.

[Link] • Finite Element Analysis in Vertebrate Morphology 567

ANRV309-EA35-18 ARI 20 March 2007 17:2

PROGNOSIS AND FUTURE DIRECTIONS

Significant advances have been made in the application of FEA to questions in verte-
brate morphology and evolution. Many more specimens are subject to CT scanning,
and scanners are becoming more freely available and less expensive to purchase.
Technological advances are helping to facilitate the transfer of CT information to
FE-models. Functional morphology is witnessing somewhat of a revival, based in
part on the realization that to understand the link from genotype to phenotype,
one must understand how selection operates on the phenotypic characters an or-
ganism presents and how it interacts with the environment. This in turn is driving
basic research into comparative musculoskeletal anatomy and histology, and material
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

properties and experimental studies of extant animal function, the basic data required
for FE-validation and modeling.

Methodological and Technological

by Monash University on 05/27/08. For personal use only.

There are many methodological and technological avenues that are open to explo-
ration in order to improve the status of finite element modeling in fossil and living
animal morphology.

Validation. The most important challenge facing FE-modelers is validation. Al-

though work on macaques is exemplary and provides a benchmark for other studies,
much more needs to be done to validate other living taxa, and from a paleontological
perspective to understand those taxa that phylogenetically bracket extinct animals,
such as crocodilians and birds. In doing so, we will be able to place error bars on our
understanding of extant and extinct animal function. In vivo studies can characterize
real-time strain during function, but in vitro studies can play an important role, as
loading parameters can be carefully controlled and thus reproduced accurately in the
FE-model, strain gauges can be applied in some difficult to access sites, and a global
synchronous strain pattern recovered (Wang et al. 2007). As validation studies must
incorporate material property testing, PCSA estimation, EMG, and in vivo and in
vitro strain gauge data collection, they require immense collective effort.

Scanning technologies. Micro-CT and increases in computing power are enabling

the micromillimeter reconstruction of trabecular architecture and hence incredibly
detailed models. Synchotron X-ray tomographic microscopy offers the possibility
to reconstruct geometries on the submillimeter, even micrometer scale. Thus, FE-
model resolution will most likely increase in the coming years as even the smallest
specimens become analyzable.

From scans to FE. A further technological advance required is the seamless inte-
gration of CT data to FE-models. This is happening with the advent of voxel-based
FE-model generation software such as Simpleware ([Link]
which also permits heterogeneous material property application based on grayscale
algorithms that link Hounsfield Units (pixel grayscale values) to human material

568 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

properties. Further work is needed to see if this technique is appropriate for living
taxa, and how it can be adapted for fossil taxa—especially when contrast and grayscale
may vary stepwise through a single scan of a fossil as a result of differential mineral-
ization and preservation, rather than due to true changes in the fossil’s bone material
properties.

Soft tissues. All published FE-models, with the exception of Kupczik et al. (2007),
have focused exclusively on reconstructing only skeletal geometry. Although muscu-
lar loads have been imposed on the model, soft tissues have not been reconstructed.
It is well known that in some cases soft tissues act as key tension members of facil-
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

itate kinesis within the skeleton, and therefore play a key role in function. In these
taxa it will become increasingly important to model the action of soft tissues, either
geometrically or through finite elements that may act, for example, as springs. Under
normal loading conditions, bone displays linear elastic behavior (i.e., stress and strain
increase in proportion). Soft tissues behave with nonlinear elasticity and therefore
by Monash University on 05/27/08. For personal use only.

warrant complex material properties and model conditions.

Conceptual Advances
Once a level has been reached where a degree of confidence can be ascribed to
our FE-model results, and model creation time has been substantially reduced, this
opens the door to many possibilities for biomechanical and evolutionary studies. In
an evolutionary sense, this would permit the study of broad-scale evolutionary trends
in musculoskeletal form across groups and the origin of higher groups: comparative
analysis can test the effect of modifying, adding, or removing structures that evolve
within a lineage to decipher the possible mechanical impetus for evolutionary change,
and there is the potential to explore hypothetical and ancestral morphologies. The
integration of FEA with techniques such as geometric morphometrics offers the
possibility of exploring shape change and placing morphological variation within a
biomechanical context, and the study of optimization and constraint on form can
be opened up, investigating the influence of mechanical load of skeletal form and
adaptation. Furthermore, the evolution of key innovations and the function of unusual
animals or autapomorphic features can be decoded. Intraspecific and interspecific
variation can be accounted for, rather than relying on a single representative model.
Some general points applicable here that workers must bear in mind are exactly
what variables are to be used to test mechanical hypotheses. Previous studies have, for
example, taken peak stress as an indicator of comparative skull strength (Dumont et al.
2005, Rayfield & Milner 2006); but should one choose tensile stress or strain, as bone
fails first in tension, or Von Mises stress, which is a good predictor of failure under
ductile fracture, as commonly occurs in bone. Von Mises (VM) stress is a function of
principal stresses σ1 , σ2 , and σ3 that measures how stress distorts a material. Failure
of ductile material is estimated when VM stress equals the yield strength of the
material in uniaxial tension. Some commonality of approach is required to normalize
comparative results. With increased computational power, just how much model
sophistication is required? Furthermore, even with appropriate validation, FEA will

[Link] • Finite Element Analysis in Vertebrate Morphology 569

ANRV309-EA35-18 ARI 20 March 2007 17:2

probably never enable us to exactly predict the precise level of strain in fossils. Yet it
will allow us to constrain our predictions within certain limits.
The crucial overarching point, which is applicable to all FE-studies, is that the
effectiveness of FEA depends on the quality and appropriateness of the questions
being asked. Refinement of questions and increasingly targeted hypotheses will help
immensely, however there will always be a role for inductive studies using detailed
parameters for simple stress-strain recreation and performance testing. At all times we
must be aware of the possibilities, but also the limitations of this extremely powerful
tool, a tool which will no doubt increasingly influence our understanding of living
and extinct vertebrate morphology.
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

SUMMARY POINTS
1. Finite element analysis is a very promising technique for understanding
of vertebrate morphology, function, and evolution in extinct and extant
by Monash University on 05/27/08. For personal use only.

animals.
2. Comparison of FE-models to in vivo and in vitro strain shows that with well-
characterized material properties and functional load, FEA can replicate
bony stress-strain patterns, orientation, and, with some accuracy, magnitude.
3. Current FE-research is driven by either deductive or inductive reasoning.
Deductive processes tend to assume the skeleton is mechanically adapted to
functional loads, whereas inductive approaches seek to test skeletal optimal-
ity and constraint.
4. The effectiveness of how we use fossil FE-models depends on the questions
we ask. With our current knowledge of validation, we are unlikely to uncover
absolute stress-strain values, but hypothesis testing and comparative analysis
offer much potential in understanding extinct animal function.
5. For fossil studies, phylogenetically similar taxa possess more similar material
property values and intracranial variation than more distantly related taxa.
With a broader taxonomic understanding of FE-validation in living taxa, we
will be able to phylogenetically bracket and place reliable error bars on our
extinct FE-models greatly increasing their utility.

FUTURE ISSUES
1. A broader understanding of extant animal in vivo and in vitro strain, material
property variation, and loading conditions experienced during function is
needed. This will lead to an increased understanding of how FE-models
reflect reality in a variety of extant taxa, particularly those that bracket extinct
taxa.
2. Improved transfer of scan data to FE-models is needed.

570 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

3. We must develop clearer statements of aims and methodology, and derive

appropriate questions for the species or taxon under study.
4. FEM should be applied to the postcranial skeleton and incorporation of
soft tissues; current studies have almost exclusively focused on the skeletal
component of the cranium.

LITERATURE CITED
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Beaupré GS, Carter DR. 1992. Finite element analysis in biomechanics. See Biewener
1992a, pp. 149–73
Beaupré GS, Orr TE, Carter DR. 1990. An approach for time-dependent bone mod-
eling and remodeling - application: a preliminary remodelling simulation. J.
Orthop. Res. 8:662–70
by Monash University on 05/27/08. For personal use only.

Bertram JEA, Swartz SM. 1991. The ‘law of bone transformation’: a case of crying
Wolff ? Biol. Rev. 66:245–73
Biewener AA, ed. 1992a. Biomechanics: Structures and Systems. Oxford: Oxford Univ.
Press
Biewener AA. 1992b. In vivo measurement of bone strain and tendon force. See
Biewener 1992a, pp. 123–48
Blob RW, Biewener AA. 1999. In vivo locomotor strain in the hindlimb bones of
Alligator missippiensis and Iguana iguana: implications for the evolution of limb
bone safety factors and nonsprawling limb posture. J. Exp. Biol. 202:1023–46
Brekelmans WA, Poort HW, Slooff TJ. 1972. A new method to analyze the mechan-
ical behavior of skeletal parts. Acta Orthop. Scand. 43:301–7
Buckland-Wright JC. 1978. Bone structure and the patterns of force transmission in
the catskull (Felis catus). J. Morph. 155:35–62
Carter DR, Beaupré GS. 2001. Skeletal Form and Function. Cambridge: Cam-
A key book describing
bridge Univ. Press. 318 pp. mechanical adaptation in
Carter DR, Mikic B, Padian K. 1998. Epigenetic mechanical factors in the evolution bones.
of long bone epiphyses. Zoo. J. Linn. Soc. 123:163–78
Chen X, Chen H. 1998. The influence of alveolar structures on the torsional strain
field in a gorilla corporeal cross-section. J. Hum. Evol. 35:611–33
Clough RW. 1960. The finite element in plane stress analysis. Proc. ASCE Conf. Elec-
tronic Comput., 2nd, Pittsburgh, PA
Currey JD. 1987. The evolution of the mechanical properties of amniote bone. J.
Biomech. 20:1035–44
Currey JD. 2002. Bones: Structure and Mechanics. Princeton: Princeton Univ.
A very accessible study of
Press. 436 pp. bone biology and
Curtis K, Kupczik K, Fagan MJ, Evans SE. 2005. Finite element modelling of the cat biomechanics.
skull. Symp. Vertebr. Palaeontol. Comparative Anat., 53rd, Nat. History Mus.,
London
Daniel TL, Helmuth BS, Saunders WB, Ward PD. 1997. Septal complexity in am-
monoid cephalopods increased mechanical risk and limited depth. Paleobiology
24:470–81

[Link] • Finite Element Analysis in Vertebrate Morphology 571

ANRV309-EA35-18 ARI 20 March 2007 17:2

Daniel W, McHenry C. 2001. Bite force to skull stress correlation - modelling the
skull of Alligator mississippiensis. In Crocodilian Biology and Evolution, ed. GC Grigg,
F Seebacher, CE Franklin. Surrey, UK: Beatty & Sons
Donoghue PCJ, Bengston S, Dong X, Gostling NJ, Huldtgren T, et al. 2006. Syn-
chrotron X-ray tomographic microscopy of fossil embryos. Nature 442:680–83
Dumont ER, Piccirillo J, Grosse IR. 2005. Finite-element analysis of biting behavior
and bone stress in the facial skeletons of bats. Anat. Rec. 283A:319–30
Erickson GM, Catanese J, Keaveny TM. 2002. Evolution of the biomechanical ma-
terial properties of the femur. Anat. Rec. 268:115–24
Fastnacht M, Hess N, Frey E, Weiser HP. 2002. Finite analysis in vertebrate palaeon-
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

tology. Senckenbergiana lethaea 82:195–206

Frazzetta TH, Kardong KV. 2002. Prey attack by a large theropod dinosaur. Nature
416:387–88
Goodship AE, Lanyon LE, McFie H. 1979. Functional adaptation of bone to in-
creased stress. J. Bone Joint Surg. 61A:539–46
by Monash University on 05/27/08. For personal use only.

Guillet A, Doyle WS, Ruther H. 1985. The combination of photogrammetry and

finite elements for a fine grained functional analysis of anatomical structures.
Zoomorphology 105:51–59
Hassan MA, Westermann GEG, Hewitt RA, Dokainish MA. 2002. Finite-element
analysis of simulated ammonoid septa (extinct Cephalopoda): septal and sutural
complexities do not reduce strength. Paleobiology 28:113–26
Herring SW, Teng S. 2000. Strain in the braincase and its sutures during function.
Am. J. Phys. Anthropol. 112:575–93
Huiskes R. 2000. If bone is the answer, then what is the question? J. Anat. 197:145–56
Huiskes R, Chao EYS. 1983. A survey of finite-element analysis in orthopedic
A key review of FEA in
orthopaedics; represents biomechanics—the first decade. J. Biomech. 16:385–409
the level morphological Huiskes R, Hollister SJ. 1993. From structure to process, from organ to cell: re-
studies have currently cent developments of FE-analysis in orthopaedic biomechanics. J. Biomech. Eng.
reached. 115:520–27
Huiskes R, Rulmerman R, van Lenthe GH, Janssen JD. 2000. Effects of mechan-
ical forces on maintenance and adaptation of form in trabecular bone. Nature
405:704–6
Hylander WL, Johnson KR, Crompton AW. 1987. Loading patterns and jaw move-
ments during mastication in Macaca fascicularis: a bone-strain, electromyo-
graphic, and cineradiographic analysis. Am. J. Phys. Anthropol. 72:287–314
Jenkins I, Thomason JJ, Norman DB. 2002. Primates and engineering principles:
applications to craniodental mechanisms in ancient terrestrial predators. Senck-
enbergiana Lethaea 82:223–40
Kesel AB, Philippi U, Nachtigall W. 1998. Biomechanical aspects of the insect wing:
an analysis using the finite element method. Comp. Biol. Med. 28:423–37
Kupczik K, Dobson CA, Fagan MJ, Crompton RH, Oxnard CE, O’Higgins P. 2007.
Assessing mechanical function of the zygomatic region in macaques: validation
and sensitivity testing of finite element models. J. Anat. 210:41–53
Lanyon LE, Goodship AE, Pye CJ, MacFie JH. 1982. Mechanically adaptive bone
remodelling. J. Biomech. 15:141–54

572 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

Macho GA, Spears IR. 1999. Effects of loading on the biochemical behavior of molars
of Homo, Pan and Pongo. Am. J. Phys. Anthropol. 109:211–27
Macho GA, Shimizu D, Jiang Y, Spears IR. 2005. Australopithecus anamensis: a finite
element approach to studying the functional adaptation of extinct hominins.
Anat. Rec. 283A:310–18
Marinescu R, Daegling DJ, Rapoff AJ. 2005. Finite element modeling of the anthro-
poid mandible: the effects of altered boundary conditions. Anat. Rec. 283A:300–9
McHenry CR, Clausen PD, Daniel WJT, Meers MB, Pendharkar A. 2006. Biome-
chanics of the rostrum in crocodilians: a comparative analysis using finite element
analysis. Anat. Record Part A 288A:827–49
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Metzger KA, Daniel WJT, Ross CF. 2005. Comparison of beam theory and finite-
element analysis with in vivo bone strain data from the alligator cranium. Anat.
Rec. 283A:331–48
Mullender MG, Huiskes R, Weinans H. 1994. A physiological approach to the sim-
ulation of bone remodelling as a self-organizational control process. J. Biomech.
27:1389–94
by Monash University on 05/27/08. For personal use only.

Peterson J, Dechow PC. 2003. Material properties of the human cranial vault and
zygoma. Anat. Rec. 274A:785–97
Philippi U, Nachtigall W. 1996. Functional morphology of regular echinoid tests
(Echinodermata, Echinoida): a finite element study. Zoomorph. 116:35–50
Preuschoft H, Witzel U. 2002a. Biomechanical investigations on the skulls of reptiles
and mammals. Senckenbergiana lethaea 82:207–22
Preuschoft H, Witzel U. 2002b. Pneumatized spaces, sinuses and spongy bone in the
skulls of primates. Anthropol. Anz. 60:67–79
Preuschoft H, Witzel U. 2002c. The functional shape of the human skull, as docu-
mented by three-dimensional FEM studies. Anthropol. Anz. 60:113–35
Preuschoft H, Witzel U. 2004a. Functional structure of the skull in hominoidea. Folia
Primat. 75:219–52
Preuschoft H, Witzel U. 2004b. A biomechanical approach to craniofacial shape in
primates, using FESA. Ann. Anat. 186:397–404
Preuschoft H, Witzel U. 2005. Functional shape of the skull in vertebrates: which
forces determine skull morphology in lower primates and ancestral synapsids.
Anat. Rec. 283A:402–13
Rab GT. 1982. An orthopaedist looks at finite element analysis. In Finite Elements
in Biomechanics, ed. RH Gallagher, Simon BR, Johnson, PC, Gross JF, pp. 1–8.
Chichester: Wiley & Sons
Rafferty KL, Herring SW, Marshall CD. 2003. Biomechanics of the rostrum and the
role of facial sutures. J. Morph. 257:33–44
Rayfield EJ. 2001. Cranial form and function in a large theropod dinosaur: a study using
finite element analysis. PhD thesis. Cambridge Univ. 300 pp.
Rayfield EJ. 2004. Cranial mechanics and feeding in Tyrannosaurus rex. Proc. R. Soc.
London B 271:1451–59
Rayfield EJ. 2005a. Aspects of comparative cranial mechanics in the theropod di-
nosaurs Coelophysis, Allosaurus and Tyrannosaurus. Zoo. J. Linn. Soc. 144:309–16
Rayfield EJ, Milner AC. 2006. The evolution of piscivory in theropod dinosaurs. J.
Vert. Paleontol. 26:114A

[Link] • Finite Element Analysis in Vertebrate Morphology 573

ANRV309-EA35-18 ARI 20 March 2007 17:2

Rayfield EJ, Norman DB, Upchurch P. 2002. Prey attack by a large theropod di-
nosaur, reply. Nature 416:388
Rayfield EJ, Norman DB, Horner CC, Horner JR, May Smith P, et al. 2001.
The first comprehensive
study of vertebrate Cranial design and function in a large theropod dinosaur. Nature 409:1033–
morphology using FEA 37
incorporating material Reid REH. 1996. Bone histology of the Cleveland-Lloyd dinosaurs and of dinosaurs
properties and boundary in general. Brigham Young Univ. Geol. Stud. 41:25–71
condition estimates and
Reilly DT, Burstein AH. 1975. The elastic and ultimate properties of compact bone
CT-reconstructed
geometry. tissue. J. Biomech. 8:393–405
Rensberger JM. 1992. Relationship of chewing stress and enamel microstructure
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

in rhinocerotoid cheek teeth. In Structure, Function and Evolution of Teeth, ed.

P Smith, E Tchernov, pp. 163–83. London: Freund Publ. House
Rensberger JM. 1995. Determination of stresses in mammalian dental enamel and
their relevance to the interpretation of feeding behaviours in extinct taxa. In
Functional Morphology in Vertebrate Paleontology, ed. JJ Thomason, pp. 151–72.
by Monash University on 05/27/08. For personal use only.

Cambridge: Cambridge Univ. Press

Richardson LF. 1910. The approximate arithmetical solution by finite differences of
physical problems. Trans. R. Soc. London A 210:307–57
Richmond BG, Wright BW, Grosse I, Dechow PC, Ross CF, et al. 2005. Finite
Useful review of FEA
from a technical element analysis in functional morphology. Anat. Rec. 283A:259–74
perspective. Robinson J, Ahlberg PE, Koentges G. 2005. The braincase and middle ear region of
Dendrerpeton acadianum (Tetrapoda: Temnospondyli). Zoo. J. Linn. Soc. 143:577–
97
Roesler H. 1987. The history of some fundamental concepts in bone biomechanics.
J. Biomech. 20:1025–34
Ross CF, Patel BA, Slice DE, Strait DS, Dechow PC, et al. 2005. Modeling mas-
Key FEA validation paper
exploring the effect of ticatory muscle force in finite element analysis: sensitivity analysis using
boundary conditions. principal coordinates analysis. Anat. Rec. 283A:288–99
Rybicki EF, Simonen FA, Weis EB. 1972. On the mathematical analysis of stress in
the human femur. J. Biomech. 5:203–15
Schwarz D, Vontobel P, Lehmann EH, Meyer CA, Bongartz G. 2005. Neutron
tomography of internal structures of vertebrate remains: a comparison with X-
ray computed tomography. Palaeontol. Elect. 8:30A, 11 pp.
Silva MJ, Brodt MD, Hucker WJ. 2005. Finite element analysis of the mouse tibia: es-
timating endocortical strain during three-point bending in SAMP6 osteoporotic
mice. Anat. Rec. 283A:380–90
Snively E, Russell A. 2002. The Tyrannosaurid metatarsus: bone strain and inferred
ligament function. Senckenbergiana Lethaea 82:35–42
Song Y, Black RG, Lipps JH. 1994. Morphological optimization in the larest living
foraminifera: implications from finite element analysis. Paleobiology 20:14–26
Spears IR, Crompton RH. 1996. The mechanical significance of the occlusal geom-
etry of great ape molars in food breakdown. J. Hum. Evol. 31:517–35
Spears IR, Macho GA. 1998. Biomechanical behavior of modern human molars:
implications for interpreting the fossil record. Am. J. Phys. Anthropol. 106:467–
82

574 Rayfield
ANRV309-EA35-18 ARI 20 March 2007 17:2

Spoor F, Jeffery N, Zonneveld F. 2000a. Imaging skeletal growth and evolution.

In Development, Growth, and Evolution: Implications for the Study of the Hominid
Skeleton, ed. P O’Higgins, MJ Cohn., pp. 123–61. San Diego: Acad. Press Linn.
Soc. London
Spoor F, Jeffery N, Zonneveld F. 2000b. Using diagnostic radiology in human evo-
lutionary studies. J. Anat. 197:61–76
Strait DS, Wang Q, Dechow PC, Ross CF, Richmond BG, et al. 2005. Modeling
Key FEA validation paper
elastic properties in finite element analysis: how much precision is needed exploring the effect of
to produce an accurate model? Anat. Rec. 283A:275–87 material properties.
Strait DS, Wright B, Richmond BG, Ross CF, Dechow PC, et al. 2007. Craniofacial
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

strain patterns during premolar loading: implications for human evolution. In

Primate Craniofacial Function and Biology, ed. CJ Vinyard, MJ Ravosa, CE Wall.
New York: Springer. In press
Thomason JJ. 1991. Cranial strength in relation to estimated biting forces in some
mammals. Can. J. Zool. 69:2326–33
Thomason JJ, McClinchey HL, Faramarzi B, Jofriet JC. 2005. Mechanical behav-
by Monash University on 05/27/08. For personal use only.

ior and quantitative morphology of the equine laminar junction. Anat. Rec.
283A:366–79
Turner MJ, Clough RW, Martin HC, Topp LJ. 1956. Stiffness and deflection analysis
of complex structures. J. Aero. Sci. 23:805–23,54
van Rietbergen B, Weinans H, Huiskes R, Odgaard A. 1995. A new method to deter-
mine trabecularbone elastic properties and loading using micromechanical finite
element models. J. Biomech. 28:69–81
Wang Q, Dechow PC. 2006. Elastic properties of external cortical bone in the cran-
iofacial skeleton of the rhesus monkey. Am. J. Phys. Anthropol. 131:402–15
Wang Q, Dechow PC, Wright BW, Ross CF, Strait DS, et al. 2007. Surface strain on
bone and sutures in a monkey facial skeleton: an in vitro method and its relevance
to Finite Element Analysis. In Primate Craniofacial Function and Biology, ed. CJ
Vinyard, MJ Ravosa, CE Wall. New York: Springer
Wang Q, Strait DS, Dechow PC. 2006. A comparison of cortical elastic properties
in the craniofacial skeletons of three primate species and its relevance to human
evolution. J. Hum. Evol. 51:375–82
Weijs WA, Hillen B. 1985. Cross-sectional areas and estimated intrinsic strength of
the human jaw muscles. Acta Morphol. Neerlando-Scand. 23:267–74
Wilcox RK. 2007. The influence of material property and morphological parame-
ters on specimen-specific finite element models of porcine vertebral bodies. J.
Biomech. 40:669–73
Witmer LM. 1995. The extant phylogenetic bracket and the importance of recon-
structing soft tissue in fossils. In Functional Morphology in Vertebrate Paleontology,
ed. JJ Thomason, pp. 19–33. Cambridge: Cambridge Univ. Press
Witzel U, Preuschoft H. 1999. The bony roof of the nose in humans and other
primates. Zool. Anz. 238:103–15
Witzel U, Preuschoft H. 2005. Finite-element model construction for the virtual syn-
thesis of the skulls in vertebrates: case study of Diplodocus. Anat. Rec. 283A:391–40
Wolff J. 1892. Das Gesetz der Transformation der Knochen. (Transl. The Law of Bone
Remodelling). Berlin: Springer-Verlag

[Link] • Finite Element Analysis in Vertebrate Morphology 575

ANRV309-EA35-18 ARI 20 March 2007 17:2

Zienkiewicz OC. 1971. The Finite Element Method in Engineering Science. New York:
McGraw-Hill. 521 pp.
Zienkiewicz OC, Taylor RL, Zhu JZ. 2005. The Finite Element Method: Its
The introductory-level
textbook on Finite Basis and Fundamentals. Amsterdam: Elsevier Butterworth-Heinemann.
Element Analysis. 733 pp.
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]
by Monash University on 05/27/08. For personal use only.

576 Rayfield
 ANRV309-FM ARI 4 April 2007 13:29

Annual Review of
Contents Earth and
Planetary Sciences

Volume 35, 2007

Frontispiece
Robert N. Clayton p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p xiv
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

Isotopes: From Earth to the Solar System

Robert N. Clayton p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Reaction Dynamics, Molecular Clusters, and Aqueous Geochemistry
William H. Casey and James R. Rustad p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 21
by Monash University on 05/27/08. For personal use only.

The Aral Sea Disaster

Philip Micklin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 47
Permo-Triassic Collision, Subduction-Zone Metamorphism, and
Tectonic Exhumation Along the East Asian Continental Margin
W.G. Ernst, Tatsuki Tsujimori, Ruth Zhang, and J.G. Liou p p p p p p p p p p p p p p p p p p p p p p p p p p p 73
Climate Over the Past Two Millennia
Michael E. Mann p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p111
Microprobe Monazite Geochronology: Understanding Geologic
Processes by Integrating Composition and Chronology
Michael L. Williams, Michael J. Jercinovic, and Callum J. Hetherington p p p p p p p p p p p137
The Earth, Source of Health and Hazards: An Introduction to Medical Geology
H. Catherine W. Skinner p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p177
Using the Paleorecord to Evaluate Climate and Fire Interactions in Australia
Amanda H. Lynch, Jason Beringer, Peter Kershaw, Andrew Marshall,
Scott Mooney, Nigel Tapper, Chris Turney, and Sander Van Der Kaars p p p p p p p p p p p p p p p215
Wally Was Right: Predictive Ability of the North Atlantic “Conveyor
Belt” Hypothesis for Abrupt Climate Change
Richard B. Alley p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p241
Microsampling and Isotopic Analysis of Igneous Rocks: Implications
for the Study of Magmatic Systems
J.P. Davidson, D.J. Morgan, B.L.A. Charlier, R. Harlou, and J.M. Hora p p p p p p p p p p273
Balancing the Global Carbon Budget
R.A. Houghton p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p313
Long-Term Perspectives on Giant Earthquakes and Tsunamis at
Subduction Zones
Kenji Satake and Brian F. Atwater p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p349
vii
 ANRV309-FM ARI 4 April 2007 13:29

Biogeochemistry of Glacial Landscape Systems

Suzanne Prestrud Anderson p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p375
The Evolution of Trilobite Body Patterning
Nigel C. Hughes p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p401
The Early Origins of Terrestrial C4 Photosynthesis
Brett J. Tipple and Mark Pagani p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p435
Stable Isotope-Based Paleoaltimetry
David B. Rowley and Carmala N. Garzione p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p463
Annu. Rev. Earth Planet. Sci. 2007.35:541-576. Downloaded from [Link]

The Arctic Forest of the Middle Eocene

A. Hope Jahren p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p509
Finite Element Analysis and Understanding the Biomechanics
and Evolution of Living and Fossil Organisms
by Monash University on 05/27/08. For personal use only.

Emily J. Rayfield p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p541

Chondrites and the Protoplanetary Disk
Edward R.D. Scott p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p577
Hemispheres Apart: The Crustal Dichotomy on Mars
Thomas R. Watters, Patrick J. McGovern, and Rossman P. Irwin III p p p p p p p p p p p p p p p p621
Advanced Noninvasive Geophysical Monitoring Techniques
Roel Snieder, Susan Hubbard, Matthew Haney, Gerald Bawden,
Paul Hatchell, André Revil, and DOE Geophysical Monitoring Working Group p p p p p653
Models of Deltaic and Inner Continental Shelf Landform Evolution
Sergio Fagherazzi and Irina Overeem p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p685
Metal Stable Isotopes in Paleoceanography
Ariel D. Anbar and Olivier Rouxel p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p717
Tectonics and Climate of the Southern Central Andes
M.R. Strecker, R.N. Alonso, B. Bookhagen, B. Carrapa, G.E. Hilley,
E.R. Sobel, and M.H. Trauth p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p747

Indexes

Cumulative Index of Contributing Authors, Volumes 25–35 p p p p p p p p p p p p p p p p p p p p p p p p789

Cumulative Index of Chapter Titles, Volumes 25–35 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p793

Errata

An online log of corrections to Annual Review of Earth and Planetary Sciences

chapters (if any, 1997 to the present) may be found at [Link]

viii Contents

View publication stats