Page 1
A Geometric Morphometric Analysis
of Ruminant (Ungulata: Artiodactyla)
and Ornithopod (Dinosauria:
Ornithischia) Snouts: Comparative
and Functional Ecomorphology
A report submitted in partial fulfilment of the requirements for the
degree of MSc Advanced Methods in Taxonomy and Biodiversity,
Imperial College London and Natural History Museum, London
Jonathan Tennant 00661116
Natural History Museum Cromwell Road
London SW7 5BD
Page 2
1
Contents
[4] 1. 0 Abstract
[5] 2. 0 Introduction
[6] 2.1 Previous Attempts to Infer Feeding Strategy from Morphology
[12] 3.0 Materials and Methods
[20] 4.0 Results
[20] 4.1 Procrustes Superimposition
[21] 4.2 Ecological Classification
[28] 4.3 Taxonomic Classification
[34] 4.4 Shape Classification
[58] 5.0 Discussion
[63] 6.0 Conclusions
[66] 7.0 Acknowledgements
[67] 8.0 References
[83] 9.0 Appendices
Page 3
2
Figures and Tables
[10] Figure 1 - Mean shape and range envelopes of browsing, grazing, and intermediate
ruminants (modified from Solounias et al., 1988)
[13] Figure 2 - Exemplar ruminant profile shapes
[14] Figure 3 - Grid plot of Corythosaurus casuarius specimens in Procrustes-scaled space
[20] Figure 4 - Procrustes-scaled mean shape and overlay plot for all ruminant specimens
[20] Figure 5 - Procrustes-scaled mean shape and overlay plot for all ornithopod specimens
[22] Figure 6 - Two-dimensional PCA score plot for ruminants classified according to their
feeding type
[23] Figure 7 - Three-dimensional PCA score plot for ruminants classified according to their
feeding type
[24] Figure 8 - Two-dimensional CVA score plot for ruminants classified according to their
feeding type
[24] Figure 9 - Three-dimensional CVA score plot for ruminants classified according to their
feeding type
[25] Figure 10 - Modelled CVA axes in two-dimensional CVA space for ruminants classified
according to their feeding type
[26] Figure 11 - Back-projection of modelled CVA axes into two-dimensional PCA space
[27] Figure 12 - Χ2 probability density function plot for ruminants classified according to
feeding type in CVA space
[27] Figure 13 - Strobe plot of modelled CVA axes in PCA space for ruminants classified
according to feeding type
[29] Figure 14 - Two-dimensional PCA score plot for ruminants classified according to their
family
[29] Figure 15 - Three-dimensional PCA score plot for ruminants classified according to their
family
[30] Figure 16 - Two-dimensional PCA score plot for ruminants classified according to their
sub-family
[30] Figure 17 - Three-dimensional PCA score plot for ruminants classified according to their
sub-family
[31] Figure 18 - Two-dimensional CVA score plot for ruminants classified according to their
family
[31] Figure 19 - Three-dimensional CVA score plot for ruminants classified according to their
family
Page 4
3
[32] Figure 20 - Two-dimensional CVA score plot for ruminants classified according to their
sub-family
[33] Figure 21 - Three-dimensional CVA score plot for ruminants classified according to their
sub-family
[34] Figure 22 - Two-dimensional PCA score plot for ruminants classified according to the
primary shape descriptor (‘blunt’ or ‘pointed’)
[35] Figure 23 - Three-dimensional PCA score plot for ruminants classified according to the
primary shape descriptor
[36] Figure 24 - Two-dimensional CVA score plot for ruminants classified according to the
primary shape descriptor
[37] Figure 25 - Two-dimensional PCA score plot for ruminants classified according to the
secondary shape descriptor (‘pointed’ and ‘blunt’ appended with either ‘concave’, ‘convex’
or ‘linear’)
[37] Figure 26 - Three-dimensional PCA score plot for ruminants classified according to the
secondary shape descriptor
[38] Figure 27 - Two-dimensional CVA score plot for ruminants classified according to the
secondary shape descriptor
[39] Figure 28 - Strobe plot of modelled CVA axis in PCA space for ruminants classified
according to the primary shape descriptor
[39] Figure 29 - Strobe plot of modelled CVA axes in PCA space for ruminants classified
according to the secondary shape descriptor
[42] Figure 30 -Projection of the ornithopod PCA scores into the PCA space defined by PC-1
and PC-2 for the ruminants when classified according to feeding type
[43] Figure 31 - Projection of the ornithopod PCA scores when classified according to their
family into the PCA space defined by PC-1 and PC-2 for the ruminants when classified
according to feeding type
[44] Figure 32 - Two-dimensional PCA score plot for ornithopods when classified according
to their family
[45] Figure 33 - Two-dimensional CVA score plot for ornithopods classified according to
their family
[46] Figure 34 - Two-dimensional CVA score plot for ornithopods classified according to
their chronostratigraphic age
[47] Figure 35 - Back-projection of modelled CVA axes into two-dimensional PCA space for
ornithopods classified according to their chronostratigraphic age
Page 5
4
[48] Figure 36 - Strobe plot of modelled CVA axes in PCA space for ornithopods classified
according to their chronostratigraphic age
[50] Figure 37 - Two-dimensional PCA score plot for ornithopods classified according to the
primary shape descriptor
[50] Figure 38 - Three-dimensional PCA score plot for ornithopods classified according to
the primary shape descriptor
[51] Figure 39 - Two-dimensional CVA score plot for ruminants classified according to the
primary shape descriptor
[52] Figure 40 - Two-dimensional PCA score plot for ornithopods classified according to the
secondary shape descriptor
[53] Figure 41 - Two-dimensional CVA score plot for ornithopods when classified according
to the secondary shape descriptor
[54] Figure 42 - Back-projection of modelled CVA axis into two-dimensional PCA space for
ornithopods classified according to the primary shape descriptor
[54] Figure 43 - Strobe plot of modelled CVA axis in PCA space for ornithopods classified
according to the primary shape descriptor
[55] Figure 44 - Strobe plot of modelled CVA axes in PCA space for ornithopods classified
according to the secondary shape descriptor
[21] Table 1 – Calculated principal component eigenvalues for both the ruminant and
ornithopod Procrustes-transformed data sets
[40] Table 2 - Results of the chi-squared likelihood ratio test for all ruminant groups
[47] Table 3 - Results of the chi-squared likelihood ratio test for all ornithopod groups
Page 6
5
Abstract
Snout shape is a prominent aspect of herbivore ecology with respect to feeding strategy,
affecting both forage selectivity and intake rate. Within ruminants, feeding classes are
delimited based on snout shape, with grazing species attributed ‘blunt’ snouts and
browsing species ‘pointed’ snouts. These functional varieties have yet to be tested in a
robust and quantifiable structure. In this study, using a variety of geometric morphometric
techniques, this aspect of functional ecology is analysed in a statistically rigorous geometry-
based framework, principally testing to see if feeding strategy is consistent with snout
morphology and using a two-dimensional profile of the premaxilla in ventral aspect as a
proxy. The secondary objective is to assess this approach using ornithopod dinosaurs, the
putative extinct analogues of modern ruminants, to see whether similar patterns in shape
variation can reveal their feeding strategies. The results here reveal that when ruminants
are classified ecologically based on a priori assigned feeding strategies, they cannot be
discriminated on the basis of their premaxilla shape profile, instead forming a shape
variation continuum. Moreover, previously used terminology such as ‘pointed’ and ‘blunt’
are deemed inadequate for delimiting snout shape varieties, also lacking the descriptive
power to define the morphological disparity demonstrated. These results are statistically
significant, and found not to be an artefact of phylogenetic similarity. Conversely,
ornithopods are found to exhibit a strong shape variation pattern between ‘blunt’ and
‘pointed’ snout shapes. These patterns exist in both a phylogenetic and temporal context,
and may relate to both browsing height and forage selectivity as has been previously
proposed for diplodocoid sauropods. The pattern shows a trend from a plesiomorphic
selective ‘browsing’ condition in the Jurassic and Lower Cretaceous for many clades,
through a transitional experimental phase with iguanodontids in the Lower Cretaceous,
culminating in the large-bodied hadrosaurs in the Upper Cretaceous, which have
distinctively flat and laterally-widened snouts for non-selective, ground-level ‘grazing’. This
transformation is proposed to be related to increased nutritional requirements relating to a
systematic increase in body size, and possibly associated with increased angiosperm
diversity during the Upper Cretaceous.
Page 7
6
2.0 Introduction
The idea that ungulates and ornithopods may be comparable in terms of ecological and
macroevolutionary significance has long been of interest (Dodson, 1975; Molnar, 1977;
Weishampel, 1985; Norman and Weishampel, 1985; Carrano et al., 1999; Rybczynski et al.,
2008). Parallels have been drawn in the context of their sexual dimorphism, masticatory
apparatus and cranial kinetics, herding, and parental care (Carrano et al., 1999), as well as
with their digestive systems (Farlow, 1987). Within ungulates, it is often claimed that snout
morphology performs an integral role in both defining various ecological guilds, as well as
playing a significant interactive role during ingestion. However, the claim that there is a
close association between snout morphology and feeding strategy has not been subjected
to a rigorous, geometry-based quantitative analysis to date. The proximate target of the
present investigation is a determination of whether the shape of the premaxilla can be used
to delimit feeding strategies successfully in ruminants, and then whether the same patterns
can be used to make inferences about the feeding habits of their apparent extant
analogues, the ornithopod dinosaurs. This question is examined within a geometric
morphometric framework, due to its inherent analytical strength when studying the shape
of morphological structures.
Members of the Suborder Ruminantia (Order Artiodactyla) are a clade of even-toed
ungulate mammals defined uniquely by possession of a two-step digestion system involving
the fermentation chamber in the foregut of the stomach, and by the presence of a
reticulorumen, the structure from which the clade takes its name. This Suborder includes
many well-known ungulate taxa including antelope, deer, giraffe, and cattle, together
occupying and regulating many modern terrestrial ecosystems. Currently some 200 extant
species are recognised (Fernández and Vrba, 2005). Ornithopods (Order: Ornithischia)
exhibit a temporal phylogenetic hierarchy, with camptosaurs and dryosaurs occupying the
Late Jurassic, iguanodontid forms and hypsilophodont-grade taxa dominating the Lower
Cretaceous, and hadrosaurids radiating in the Upper Cretaceous. This correlates to a
systematic increase in body mass, as well as masticatory complexity through time. Along
with then contemporaneous sauropods, they formed a major constituent of Mesozoic
Page 8
7
terrestrial communities, dominating herbivorous niches until their extinction during the
end-Cretaceous mass extinction.
Ruminant feeding strategies are reflected in their craniodental and gastrointestinal
morphophysiological diversity, and have been conventionally categorised into ‘browsers’
and ‘grazers’, with an ‘intermediate’ sub-group, where browsers are considered to be
obligate non-grazers, but not vice-versa (Clauss et al., 2003). For example, Bodmer (1990),
Mendoza et al. (2002) and Gordon (2003) include variants of frugivores, high-level
browsers, and fresh grass grazers in this group in an attempt to encapsulate the full
theoretical range of feeding strategies. Variations in feeding strategy also explicitly occur in
a spatiotemporal context. These relate to the ‘environmental stresses’ of Bourtière and
Hadley (1970) and plausibly a hierarchical grazing succession related to species’ migration
patterns, geomorphology, resource partitioning or forage quality (Gwynne and Bell, 1968;
Fryxell, 1991; Voeten and Prins, 1999). No attempt to classify their putative ornithopodan
analogues in such a manner has been undertaken to date.
2.1 Previous Attempts to Infer Feeding Strategy from Morphology
Van Zyl (1965) was the first to define a classification for ungulates explicitly based on
feeding strategy. Following this, Hofmann and Stewart (1972) and Hofmann (1973, 1988,
1989, 1991, and 1999) extended Van Zyl’s definitions to contain a morphophysiological
underpinning, specifically in ruminants pertaining to their particular ecological roles. This
modified feeding classification scheme has maintained its value ever since based on its
heuristic value. Its popularity is nevertheless somewhat counter-intuitive insofar as, until
recently, few studies attempted to empirically or theoretically validate these initial putative
classifications (Clauss et al., 2008b).
The typical dichotomy of ‘browsers’ and ‘grazers’ relies on a botanical foundation. Browsers
typically consume berries and dicotyledonous leaves (e.g., Hofmann and Stewart, 1972;
Jarman, 1974; Janis, 1995). Grazers consume monocotyledonous grasses, and intermediate
feeders vary depending on season and geography (e.g., Shipley, 1999; Sanson, 2006). The
Page 9
8
significance of this variation is that the physio-mechanical and biochemical properties of
the different plant types are adequate to potentially drive this functional dichotomy in
ruminants. Clauss et al. (2003) argued convincingly that the properties of respective forages
exert strong control on the evolution of the masticatory apparatus and gastrointestinal
tract; specifically the reticulorumen physiology within ruminants (see also Clauss et al.,
2010).
The botanical definitions of browsers and grazers have a complicated history, with various
authors unable to settle on a consistent threshold of forage consumption for either class.
Janis and Ehrhardt (1988), Janis (1990), Janis et al. (2000) and Pérez-Barbería et al. (2001a)
regarded browsers as ruminants that consume <10% grass, and grazers as those consuming
>90% grass per annum, with all other species being ranked as intermediate. These authors
provide no justification for these seemingly arbitrary thresholds. Conversely, Gordon and
Illius (1996), Pérez-Barbería and Gordon (1999), Pérez-Barbería et al. (2001b) and Mendoza
et al. (2002) selected >75% grass per annum as the threshold criterion for their grazer class,
and >75% browse for browsers. There is further confusion, as Clauss et al. (2009a) defined
grazers as those consuming >80% monocot material, and strict browsers as those with a
“very low intake of monocot forage” (p. 399), whilst others use natural diet as a continuous
variable (e.g., Sponheimer et al., 2003). This variation is partially summarised in Clauss et al.
(2008b). In many other studies, feeding strategy delimitation is purely qualitative (e.g.,
Owen-Smith, 1997), where grazers are classified as those “consuming primarily grasses,
sedges and other graminoids”, (p. 178). Gordon and Illius (1994) find that different
thresholds of classification give different results in ecological analyses. Defining these
thresholds in congruence with functional or ecological significance remains a problematic
issue.
There are numerous morphophysiological parameters affecting digestive rate and
efficiency, as well as masticatory efficiency. However, the first part of an organism to
interact with forage (excluding perhaps the tongue and prehensile lips) is the snout. Snout
shape is a highly prominent aspect of herbivore ecology, defining initial intake rate,
Page 10
9
chewing efficiency and forage selection ability (Gordon and Illius, 1988; Murray and Brown,
1993; Shipley et al., 1994; Pérez-Barbería and Gordon, 1998; Clauss et al., 2007).
The anterior section of the snout is predominantly formed by the premaxillae, the shape of
which is an important ecomorphological feature, or trait. It is commonly noted that
browser mammal species have pointed premaxillae and grazers a more squared or blunt
shape representing the derived cropping condition (e.g., Janis and Ehrhardt, 1998; Pérez-
Barbería and Gordon, 2001). Intermediate feeding strategies have an intermediate form,
often mediolaterally compressed into a club-like shape (Solounias and Moelleken, 1993).
The functional significance of snout shape lies in selectivity; i.e. a more pointed rostrum
allows for increased selection sensitivity, whereas a wider or blunter form confers a more
random cropping process, with greater oral intake (Gordon and Illius, 1988; Janis and
Ehrhardt, 1998). Following this, Janis et al. (2010) inferred that, theoretically, a higher
intake rate should drive covariation within the mandible, forcing the evolution of stronger
anatomical structures (strengthening or fusion of sutures, increased muscle attachment
area, and decreasing pleurokinesis and resistance to strain). This should not be taken to
suggest that, as snout shape varies, it forces covariation of other morphophysiological
parameters. Rather, it simply controls the initial parameter with which all other domains
interact. Janis et al. (2010)’s suggestion was corroborated by Fletcher et al. (2010), who
determined that evolution of the masticatory apparatus has a functional or adaptational
origin, challenging other studies where it is deemed to be a phylogenetic artefact (e.g.
Pérez-Barbería and Gordon, 1999, 2001; Figuerido et al., 2010; Raia et al., 2010).
Irrespective of these process-level speculations and investigations, the functional varieties
of snout shape in ruminants have yet to be tested in a robust quantifiable framework
beyond simple exemplar outlines. Solounias et al. (1988) and Solounias and Moelleken
(1993), following the methodology of Walker (1984), used snout shape primarily to aid
reconstruction of palaeodiets in ruminants. Their assessment is based on a quantitative
interpretation of exemplar taxa, with the method requiring construction of the anterior
snout curve using a cubic spline-fit function framed to assess intraspecific variation. The
construction method uses a somewhat arbitrary system of vectors to encapsulate the
Page 11
10
majority of premaxillary shape variation. These authors use photographs in dorsal aspect
stating that there is “no homologous point” on the premaxillary outline (Solounias and
Moelleken, 1993; p. 1063). This is why the ventral aspect should be preferred, due to the
easily traceable premaxillary-maxillary suture, and the fact that this is the interactive
surface of the oral aperture. The main drawback with their method is that the a priori
classification of specimens into functional feeding guilds - with no statistical testing or
evidence-based support for assignment - will clearly exert a degree of subjectivity in the
mean shape and shape variation envelopes. Classification should ideally be determined a
posteriori, once distinct variations between sub-groups has been discovered, if it all. For
example, in Figure 3 of Solounias et al. (1988), the intermediate shape looks considerably
closer to the grazer class (not ‘half-way’, as stated; Fig. 1). This highlights the problem of
independent author bias whenever a rudimentary qualitative method of discrimination is
applied. The images depicted only serve to emphasize the imprecision of already arbitrarily
bound categories, and the mean shapes are not a useful guide due to the overlapping
shape range envelopes. The statistics provided in Table 1 of Solounias et al. (1988)
confounds matters further, as their ‘intermediate’ forms are clearly more similar to grazers.
Janis and Ehrhardt (1988) and Gordon and Illius (1988) posit snout width as a proxy for
distal snout shape, with measurements taken at the ventral maxilla-premaxilla intersection
on the lateral margin. When describing the geometry of complex shapes, it is usually
inadequate to attempt to describe shape with a single linear metric, as equal
measurements can often describe completely disparate geometries of varying complexity.
These authors used this measurement, and the palatal width to define a ‘relative muzzle
width ratio’, which they used to represent the ratio between body size and the oral
aperture, as well as possibly representing oral intake and processing rate (note, that
‘muzzle’ describes the flesh covering the snout, not the cranial bones, as misconstrued
here). A ratio is a poor measure of shape when used singularly, since all the simplest shape
a ratio can represent is an ellipse, if the two measurements represent orthogonal axes, as in
the method used. This may be adequate for partially representing extremes of the
‘browser’ end of the continuum, but can just as easily describe a blunt form, as grazers are
Page 12
11
postulated to have. In fact, the shape a ratio can represent can be infinitely complex. For
example, imagine trying to model a sine-wave with a two-dimensional ratio! Hence ratios
are completely inappropriate proxies for snout shape (contra Janis and Ehrhardt, 1988).
Furthermore, ratios will also almost always fail to account for the ubiquity of allometric
growth patterns in organisms. Their results however, did find a general relationship
between muzzle width and the defined dietary categories. Nevertheless, these authors
failed to account for phylogeny, despite acknowledging that phylogenetic affinity may
constrain morphological design in terms of cranial plasticity, based on the general
presumption that organisms with closer shared ancestry are more likely to have increased
morphological similarity.
It is apparent that many underlying hypotheses regarding the browser-grazer dichotomy
have been rejected based on insufficient data, a lack of statistical support, or simply
correction based on more recent analysis (see Gordon and Illius, 1994, and Robbins et al.,
1995). However, Codron et al. (2007) reject the results of both of these studies, due to the
over-simplification of the biomechanical properties of associated plants. Alternatively,
Codron et al. (2007) suggested that dietary variation occurs on a spatiotemporal scale for
all browsers and grazers, and retains an intraspecific signal, conforming to Owen-Smith
(1997) and du Toit (2003).
In phylogenetic analysis, estimated cladograms are typically rejected without consideration
if no methods of statistical support are applied, in the context of sensitivity to data
Figure 1 – The mean shape (dark lines) and range envelopes (stippled area) of the anterior snout in
browsing, intermediate and grazing ruminants, modified from Solounias et al. (1988). The numbers
indicate the number of species used for the reconstructions.
Page 13
12
perturbation or decay tests. Ecological hypotheses clearly require similar scrutiny, given the
availability and ease of application of numerous relevant statistical procedures. In spite of
this, several distinctions are becoming apparent between browsing and grazing ruminants,
and are supported within a statistical framework (e.g., Clauss et al., 2008b). Note that, in
this context, support is not synonymous with accuracy. In the respect of the underlying
signals being recovered, support describes results that can be maintained and replicated
with high explanatory power, as revealed by statistical analysis, given the precision of the
underlying data.
The history of ecological classification in ruminants is convoluted, with only marginal
progress in the way of clarity or consensus. Based initially on a simple botanical
underpinning, the problem became increasingly multifaceted as new functional traits were
exposed with new methods of analysing them, and new theoretical revisions. This problem
can be framed as, what, if any, is the best method of classifying ruminants in a functional
ecological context, and what will be the traits that define these discrete categories. The
principle aim of this study, then, is to determine whether snout shape varieties can be
supported in a formally statistical framework using geometric morphometrics, with defined
morphotypes covarying between feeding strategies. If it can be demonstrated that snout
shape does not affect feeding strategy in ruminants, a secondary objective will be to further
test this approach on ornithopod dinosaurs to determine whether similar patterns of shape
variation – that might reflect feeding strategy – are evident.
The null hypothesis for this investigation relates to the conclusions of Pérez-Barbería and
Gordon (1999), among others, that feeding strategy is incongruent with premaxilla
morphology. The alternative hypothesis is that the shape of specimen snouts will form a
continuum, with ‘browser-type’ and ‘grazer-type’ morphologies forming the end-members.
This hypothesis is based on the inference that classifying what are intrinsically
morphologically diverse organisms into discrete clusters is problematic and somewhat
counter-intuitive, if purely for the purposes of having an antecedent framework on to
which new hypotheses of functional morphology can be built.
Page 14
13
3.0 Materials and Methods
Geometric morphometrics is a procedure involving the multivariate statistical analysis of
two- or three-dimensional Cartesian co-ordinate data, typically involving discrete spatially-
defined landmarks (i.e. topographically homologous loci on a structure; see Mitteroecker
and Gunz (2009) for a review of techniques). Recognition that morphology is essentially the
physical manifestation of geometry is vital as a theoretical basis for biological
morphometrics. Zoological and palaeontological studies are using these techniques
increasingly due to their ability to analyse form in many different systematic contexts (e.g.,
functional morphology, sexual dimorphism, ontogenic development, allometry, and
phylogenetics) within a statistically coherent framework (see Adams et al., 2004). For more
specific examples of the recent application of geometric morphometrics in the field of
vertebrate palaeontology, see Hadley et al. (2009), Goswami et al. (2010), Jones and
Goswami (2010) and Barden and Maidment (2011).
The ruminant data set used in this investigation consisted of 121 extant species
representing 6 families and 13 sub-families. The majority of specimens are bovids and
cervids (115 out of 121), as these are the most taxonomically diverse groups. The
ornithopod data set consisted of 66 specimens comprising 34 species, ranging in age from
the Lower Jurassic to the latest Cretaceous. These were sub-divided into families; however,
the status of Heterodontosauridae and Hypsilophodontidae is equivocally unresolved
presently (see Butler et al., 2008). Taxa in these clades probably represent grades (i.e.,
basal ornithischians, and basal ornithopods respectively, both with a small, gracile body
form). The groupings into which these data are sub-divided are based on a range of
sources. For the taxonomic groupings, Butler et al. (2008), Boyd et al. (2009), and Prieto-
Márquez (2010) were used for the ornithopod data set, and Wilson and Cole (2000) for the
ruminants. Ruminant ecology was also based on a number of sources, the individual criteria
and references for which are provided in Appendix 1.
The shape groupings for both data sets were delineated arbitrarily. Initially, pointed or
blunt profiles were decided upon, simply based on whether a specimen was more blunt, or
more pointed. Exemplars are presented in Fig. 2, showing the typical pointed and blunt
Page 15
14
forms, as well as what constituted the arbitrary threshold between the two. Specimens
with profiles similar to the threshold shape were delimited based on how broad the
premaxillae were before culminating in the distal end (i.e., broader ones were more
generally flatter and wider at the end opposed to those with medially-sloping lateral
edges).
No intermediate feeding group character state was used due to difficulty in assigning
group-limit thresholds (see above). A second grouping was also created in an attempt to
represent the nature of lateral shape variation in the premaxillae; pointed and blunt seem
to refer purely to the distal rostrum, without incorporating variation in the curvature of the
adjoining lateral margins. A simple qualitative addition to the shape descriptors was made
where these edges are described as concave, convex, or the intermediate shape, linear, the
three of which are largely independent of the distal geometry. Breaking down the shape in
this manner is theoretically a more precise method of describing the snout, and also breaks
down the initial groupings, which may alleviate weighting issues when using large groups.
The shape of the ruminant ‘horny premaxillary pad’ is reflected in the shape of the
underlying premaxilla (Solounias et al., 1988). It is expected that intraspecific shape
differences will differentiate less than interspecific shape differences; this is only a problem
when numerous specimens of one species are used. Therefore, only a single specimen per
Figure 2 – Exemplar species representing what constitutes a typical pointed profile (A. Litocranius
walleri), blunt profile (B. Alcelaphus buselaphus), and the transitional threshold between the two
(C. Hippotragus equinus; note that this specimen was classified as blunt).
A B C
Page 16
15
species is utilised here for the ruminants. In this study, this is adequate, as the intention is
to account for the between-species variation.
For the ornithopod data set, choices of which specimens to include were more complex.
Where multiple specimens of the same species are used, there did not appear to be any
consistent geometric configurations in intraspecific premaxillary shape in several of the
samples (e.g., Lambeosaurus lambei). To address these issues, putative morphotypes
groups were identified, and then analysed using canonical variates analysis (see below).
Where discrete clusters of specimens were identified and successfully discriminated, the
mean shapes of each cluster were calculated and used to replace the associated specimens
in the Procrustes transformed data set. For example, for Corythosaurus casuarius,
specimens were reduced into four discrete sub-groups based on unique morphs (specimens
1, 6 and 2, 3 and 5, and 4, 7 and 8; Fig. 3). This halves the number of specimens for this
species, thereby reducing the effective weighting of the analysis towards larger groups,
while retaining the maximum possible intraspecific shape variation and unique morph
shapes. Specimens for species such as Heterodontosaurus tucki, Hypsilophodon foxii, and
Hypacrosaurus altispinus were deemed similar enough in shape to be regarded as single
morphotypes, with their mean shapes used for all subsequent analyses.
Figure 3 – Specimens 1 to 8 of Corythosaurus casuarius in Procrustes-scaled shape space. Note the
profile similarity between the specimens outlined in the text, hence the grouping as individual
morphotypes.
Page 17
16
According to Pérez-Barbería and Gordon (2000), muzzle width is strongly significantly
correlated between sexes in ruminants. However, when body mass is controlled for, this
trait scales isometrically between sexes (see also Gordon and Illius, 1994). Thus, sexual
dimorphism is not a particular problem for this analysis, at least for ruminants. Also note
that this simple metric is an over-simplification of the premaxillary morphology and so
(again) cannot be used as an accurate proxy for overall snout shape. Adult specimens were
all chosen for the ruminants (i.e., a fully erupted 3rd molar), to account for any possible
ontogenetically-coupled allometric variation (e.g., Herring, 1985).
Ontogenetic stages are unknown for the ornithopod data set (with the exception of the
sub-adult Tenontosaurus specimen; Tennant, 2010), and it is unknown if the premaxilla
exhibits allometric variation in shape through development. It is also currently unknown
whether ornithopods were sexually dimorphic with respect to premaxillary shape, although
this is an unlikely feature. Speculations on identifying sexual morphs typically revolve
around the sacral count (e.g., in Hypsilophodon foxii; Galton, 1974) cranial crest
morphology, as well as in the strength and configuration of the pelvic girdle (e.g., the length
of the post-pubic rod may be dimorphic; Chapman et al., 1997). No details regarding
taphonomic deformation are known, and it is common for fossilised skulls to be variably
compressed or distorted in the mediolateral plane. This could create the illusion of a
narrower skull, a problem suspected for at least several of the specimens here (e.g.,
Agilisaurus louderbacki; R. Butler, pers. comm.).
Regarding specimen completion, crania in which at least one half of the anterior premaxilla
was complete were selected. As crania are symmetrical about the sagittal plane, if only a
single half was preserved, the landmark co-ordinates were simply be reflected about the
sagittal plane to reconstruct the entire premaxilla (in dorsal and ventral aspects).
Body mass was not considered in this investigation as the evolution of this trait probably
reflects numerous extrinsic factors such as geography and climate (Cooper and Purvis,
2010). However, previously a relationship known as the Jarman-Bell Principle (Bell, 1974;
Geist, 1974; Jarman, 1974) found that intake rate scales to body-mass0.75, with the
Page 18
17
implication that larger animals feed on a lower-quality (i.e., high-fibre, folivorous) diet
(Clauss et al., 2009). Nevertheless, Shipley et al. (1994) found that this ratio varies between
‘grazing’ and ‘browsing’ classes in extant herbivores and in both cases was lower than that
predicted by the Jarman-Bell Principle, as well as in all analysed herbivores (rate scales to
body-mass0.72). Body mass is likely to influence the intake rate as a function of oral aperture
size, and general size of an organism. As the only aspect of form of interest in this study is
pure shape, size is removed via a method known as Procrustes superimposition (see below),
and is not considered further here. Furthermore, Solounias et al. (1988) cite several
examples of organisms with an order of magnitude separating their body masses, but
similar-shaped premaxillae (e.g., wildebeest (275kg) and oribi (21kg); p. 294).
Snout profile outlines were redrawn based on the initial photographs using Adobe
Illustrator CS4. The starting point for the ruminant outline was defined as where the suture
between the maxilla and premaxilla intersects the lateral margin, from a ventral aspect.
Images provided by Norman MacLeod for the majority of the ornithopod section of this
study however were not constructed in such a manner. Nonetheless, these included the
general geometry of the distal snout, so may still provide data for analysis.
Outlines were digitized using a chain of semi-landmarks collected from the images using
Media Cybernetics’ Image-Pro Plus 7.0 software, a graphical image analysis package. The
starting landmark in the chain was always on the sinistral maxilla-premaxilla junction,
ensuring that all subsequent semi-landmarks were interpolated to be topographically
homologous with respect to a significant morphological feature.
One hundred equally spaced semi-landmarks were collected along the outlines for both
data sets, as this provided a geometrically faithful representation of the profiles. Each semi-
landmark has a defined x-y position with respect to the co-ordinate system origin. As the
purpose of this investigation is to analyse pure shape variation in the peripheral margins of
the sample premaxillae, no inferences can be made about the internal geometry of the
structures since they are not covered by the semi-landmarks. Size was not intended as a
target for analysis, so scale information was not recorded. A three-dimensional approach
Page 19
18
where snout depth would be analysed was also not deemed necessary, as this is not
considered to interact in any way with gape and hence oral intake.
All programs for the geometric morphometric analysis were created in Wolfram’s
Mathematica (vers. 8.0). The initial landmark data for both ornithopods and ruminants
were subjected to a two-dimensional Procrustes (GLS) transformation (vers. 3.8; Rohlf and
Slice, 1990). Procrustes superimposition forms the core of analysis of pure shape, by
removing the extraneous variation in scale, orientation and position for all specimens’ semi-
landmark constructions (see Box 2 of Klingenberg, 2010). Optimising the fit of all specimens
to each other was achieved by the rigid rotation iterating until the distance between
successive mean landmark configurations fell below 0.0001. This provided the ability for
progression of analysis in shape space as opposed to form space. The output was a series of
Procrustes-scaled shape co-ordinates for each specimen, as well as the mean shape co-
ordinates and centroid size values. The specimens at this stage were sub-divided into their
various sub-groupings for each subsequent analysis (see the Results section).
These superposed data were subject to a covariance-based principal components analysis
(PCA vers. 3.7; MacLeod, 2005), a method of eigenanalysis that preserves the Procrustes
distances among specimens (Mitteroecker and Gunz, 2009). Principal component analysis is
a rigid rotation of the initial ordination to new orthogonal principal component (PC) axes,
with scores for each axis that totally represent 100% of the maximum initial variance. For
geometric simplicity, the arbitrary threshold of 95% variance is used (i.e. using the
minimum number of PC axes that describe 95% of the total variation), as it characteristically
provides adequate explanatory power with substantially lower dimensionality (the total
dimensionality for the data set is 200). Here, four principal component axes (eigenvalues)
explained greater than 95% of the variance for both data sets, so were selected on this
basis (Table 1).
The principal component scores for both data sets were then subjected to a canonical
variates analysis (CVA, vers. 1.8; MacLeod, 2007). This multivariate technique transforms
the data to re-ordinate it against the axes which can discriminate best between group
Page 20
19
centroids by finding new shape variables that maximise the inter-group variation with
respect to the intra-group variation (Campbell and Atchley, 1981; MacLeod, 2007;
Klingenberg, 2010). The purpose of this procedure is to firstly, using PCA, reduce the shape
variation into a smaller, manageable number of variables, and then use these as a basis for
discriminating between groups. A chi-squared likelihood ratio test was performed to test
group distinctiveness (i.e. group dispersion structure) of the data, with respect to the
sample that defines the discriminant space (MacLeod, 2007a). The resulting Χ2 probability is
a validation test of the between-groups covariance structure; i.e. a low probability (<0.05,
traditionally) reflects statistically significant difference in the covariance, or dispersion,
structure with respect to the defined groups. This implies that the groupings are the
products of some extrinsic factor, such as biogeography, phylogeny, functional constraints,
or ecology, as opposed having a stochastic distribution. A group distance table is also
provided, showing the distance of all objects (specimens) from the group centroids. This is a
representation of the consistency of the pre-defined groups and the accuracy in which they
are constructed (i.e. the reliability of being able to reconstruct and discriminate the groups;
MacLeod, 2007a). These distances are also summarised in the form of a ‘confusion matrix’,
a tabulation that summarises the accuracy that specimens will be correctly assigned to their
a priori allocated groups through a Mahalanobis distance analysis (MacLeod, 2007b).
The number of orthogonal canonical variates axes corresponding to the number of pre-
defined groups minus one (i.e. the minimum number of axes required to demarcate
groups), with five modelled points per axis, were retrograded through the canonical
transformations (i.e., scalings and rotations) and back-projected into the space of the raw
principal components (MacLeod, 2007b). These points represent that two extreme points,
the central point, and two medially-interpolated points between these on the CV axes. The
result is a set of non-orthogonal canonical variates (i.e. discriminant axes) oriented with
respect to the data within Procrustes-scaled PCA space. Each model axis was plotted using
the program Axis Models (Co-ordinate point and strobe plots, vers. 1.1; MacLeod, 2009) in
a representation of deformation along the CV axes in the PCA shape space. These ‘strobe
Page 21
20
plots’ show a transformation sequence through the data based on hypothetical successive
models of the snout profiles in a space defined by maximum shape variation.
Page 22
21
4.0 Results
4.1 Procrustes Superimposition
Procrustes transformation provides a user with a data set of superimposed co-ordinates for
analysis of pure shape. Three ordinations were produced for each data set – the mean
shape, an overlay plot for all specimens, and a grid of all Procrustes-scaled semi-landmark
constrained representations of the specimens snout profiles. The mean shape and
complete datasets are illustrated for the ruminants (figs 4a and b) and ornithopods (figs 5a
and b).
Figure 4 – The Procrustes-scaled mean premaxilla shape (A) and overlay plot (B) for all 121
ruminant specimens. Note the almost continuous variation between a ‘blunt’ and ‘pointed’ shape
in the distal tip in B.
A
A B
A B
Figure 5 – The Procrustes-scaled mean premaxilla shape (A) and overlay plot (B) for all 66
ornithopod specimens. Note the similar variations in shape to the ruminant specimens, and also
the mediolateral compression of the mean shape. The slight lateral distortion of the mean shape
may be a taphonomic artefact (i.e., post-burial deformation).
A
Page 23
22
4.2 Ecological Classification
Covariance trends within these data were summarised using PCA with the resulting scores
being plotted in both two dimensional and three dimensional spaces. All two dimensional
ordinations plot PC-1 against PC-2. The three-dimensional plots additionally show PC-3. For
the ruminants, five different primary groupings were defined, resulting in five analyses
being performed. This was to see the degree to which premaxilla shape covaries with
different taxonomic and ecological groupings, as assessed by a CVA.
When ruminants are grouped according
to feeding strategy, there is no clear
pattern or structure distinguishing the
shapes between either of the feeding
categories in the space formed on the
first two PCA axes, although grazers do appear relatively confined in this PC space,
compared to the completely indistinguishable browsers, frugivores, and intermediate
feeders (Fig. 6). Inclusion of PC-3 shows that frugivores clearly occupy a subspace apart
from grazers, but are indistinguishable from adjacent browsers and intermediate-grade
species (Fig. 7).
Grazers appear to form a well-defined wedge, with substantial overlap between the
remaining categories. Intermediate forms exceed the browser shape envelope in three
directions (all but the negative PC-1 direction). Several intermediate species not only
increase the range of shape variation, but also occupy more extreme shapes than browsers.
Species classed as having ‘intermediate’ feeding strategies clearly do not occupy
‘intermediate’ positions with respect to the premaxilla shape space as represented by PC-1
and PC-2 (and PC-3). The overlap is also not distributed evenly between the two end-
member groups; both appear to be more similar to browsers. Even if intermediates are
removed, ruminants classed as either ‘browsers’ or ‘grazers’ maintain about 40% overlap in
their shape domains. However, this result may reflect the wide range of methods or
characteristics that were used to create the a priori classifications. Only five browsers (Alces
pulmatus, Rhynchotragus domorensis, R. guentheri, R. kirkii, and Madoqua phillipsi), five
PC-1 PC-2 PC-3 PC-4 Total
96.662
68.735 20.184 3.831 2.691 95.441
73.430 16.031
Data Set
4.039 3.161
% Variance Explained
Ruminants
Ornithopods
Table 1 – Calculated principal component
eigenvalues for both the ruminant and ornithopod
Procrustes-transformed data sets.
Page 24
23
intermediates (Hydropodus inermis, Rangifer tarandus, Gazella gazella, Antidorcas
marsupialis, and Dorcotragus megalotis) and six grazers (Oryx leucoryx, Connochaetes
taurinus, C. gnou, Damaliscus albifrons, Kobus leche, and Beatragus hunter) can be
unequivocally resolved in Figure 6 as belonging to their assigned groups, a 14% success
rate. The fact that all 3 major groupings exhibit significant overlap in the centre of the
ordinations indicates that within each group, there are numerous individuals with their
premaxilla shape essentially identical to that of the mean shape for the combined data set.
The range and continuous nature of the overlapping region is indicative of continuous
within-group variation about this region of the grand mean. Both of these factors suggest
that group separation based on premaxillary profile is problematic. This is not to say that
group separation is not present, it simply is not being conveyed with respect to the first two
principal components (theoretically they can be separated in any combined plot of PC-a
against PC-b).
Figure 6 – Two-dimensional PCA score plot for ruminants classified according to feeding type.
The information for this plot is based on numerous sources (see Appendix 1). The amount of
variance each PC axis describes is given adjacent to the respective axis labels.
Page 25
24
Figures 8 and 9 show the results of a canonical variates analysis (CVA) of the PC scores in
two and three dimensions, respectively. In Figure 8, browsers and intermediates occupy
similar canonical variate (CV) space regions with respect to the first two axes. Grazers
occupy a distinct region, apart from the other two major groups, occupying higher values
along the CV-1 axis. The overlapping nature of the groups in figures 8 and 9 implies that the
within-groups shape variation is distributed in such a manner that there is little or no
evidence for between-group shape discontinuities existing for these data. The introduction
of CV-3 into the ordinations appears to make the groupings increasingly convoluted, with
no apparent between-group discrimination (Fig. 9). The ordinations resulting from these
analyses demonstrate that for premaxilla shape there is a lack of consistency between the
groupings in terms of group-specific shape differences, when grouped according to a
secondary criterion that defines their feeding strategy (Appendix 1). Figure 9 shows that
discontinuities cannot be found within the groups in CVA space, as the variation shown
between groups is distributed in too broad a manner. This leads to rejection of the
hypothesis that ruminants develop different snout shapes according to their feeding
strategies, and acceptance of the null hypothesis that there is no significant difference
Figure 7 – Three-dimensional PCA score plot for ruminants classified according to feeding type.
Note that PCA is not designed to pull apart groups, but to re-orientate data along axes that
describe the highest shape variation. Legend as Figure 6.
Page 26
25
between feeding strategy groups in terms of premaxilla shape, when the profile is used as a
proxy for gross snout morphology.
Figure 8 – Two-dimensional canonical variates score plot for ruminants classified corresponding to
feeding type. The quantity of variance each canonical variate describes is shown adjacent to the
respective axes.
Figure 9 – Three-dimensional CVA score plot. There is little in the way of group discrimination, based
on a priori classification according to feeding strategy. Considering CVA is specifically designed to re-
orientate the data along axes of maximum group discrimination, the intricate relationships here
indicate that the classification system used here is inadequate. Legend as Figure 8.
Page 27
26
The slight structure grazers exhibit in Figure 8 is lost as they are pulled into the
amalgamation of intermediates and browsers. Moving on, the next phase was plotting the
CVA results for this grouping set. In Figure 10, the CV axes are marked by black points
representing the locations of along axis shape models in the CVA ordination space. Three
axes were modelled for the ecology-based analysis (the third not shown in Figure 10). But
when all discriminant axes are back-projected into their corresponding PC-space
orientations, the relative orientations of the CVA axes becomes apparent (Fig. 11). Figure
12 shows the results of a likelihood ratio test of the separation of group means relative to
their within-group dispersions (see Satorra and Saris, 1985). The resulting ratio here is 0.0,
indicating that the likelihood of these groups occupying their positions in the overall CVA
space as a result of the effect of random sampling of a single, underlying population is 0.0.
Accordingly, the alternative hypothesis - that these data were drawn from different shape
populations with different characteristics - is accepted.
Figure 10 – Orientation of CV-model axis 1 (horizontal) and CV-model axis 2 (vertical) in two-
dimensional CV-shape space. Each individual modelled point (in black) represents a specific
point on its CV axis. Legend as Figure 8.
Page 28
27
One of the more appealing aspects of geometric morphometrics is its ability to support
visualisation of the changes that are occurring within the ordinations; in this context, a
‘strobe plot’ (Lohmann and Schweitzer, 1990; MacLeod, 2009) of the CV-models in PC space
shows the progressive deformation from one end of the shape spectrum within the
maximum shape envelope described by the ruminant specimens’ premaxillae (Fig. 13). The
shape variation described by these three axes can evidently not be described as simply as a
continuum from ‘blunt’ to ‘pointed’. The first CVA axis shows progressive deformation of
the premaxilla, with a distal compression into a slight bowing, concurrent with
anterolaterally-directed expansion into a laterally-deflecting curvature. The second CVA
axis describes similar variation, but on a smaller scale, and with a slight lateral deflection of
the proximal region adjacent to the maxilla-premaxilla suture. The third axis describes the
least variance, but describes an alteration from a fairly broad-tipped shape, with
posterolaterally directed margins to a distally expanded and increasingly laterally
compressed shape. The result is that snout shape, with a two-dimensional profile in ventral
aspect as a proxy, is not sufficient to differentiate between and classify ruminants into
grazers, browsers and intermediates. The implication, is that snout shape is unrelated to
other putative functional traits that distinguish between the feeding types (e.g., the
Figure 11 – Back-projection of the CV-model axes into the respective PC shape space. Note that
the length of the model axes is inconsistent here, due to their respective orientations and
restriction to two-dimensional PCA space. Legend and axes as Figure 6.
Page 29
28
hypsodonty index, percent of grass consumed; see Appendix 1; note that where conflict
arose, priority was given to the most recent classification), or that there is some subsidiary
function that it serves. It could equally be that ruminants are so uniquely diverse, and have
adapted to maximise resource exploitation in their respective ecosystems, that they exhibit
widespread morphological convergence on the various snout shapes, forming a continuum
of shape variation with each particular species occupying a defined point relating to a
specific suite of ecomorphological characteristics.
Figure 13 – Strobe plot of the CV-model axes in PCA space. The rows correspond to the respective
axes (e.g. row 1 represents model axis 1), and the columns are each point on that axis. The right-
hand column is an overlay plot, showing the progressive shape deformation between end-points
on the axis. The more external models are increasingly further from any specific specimens (Fig.
11), and represent a hypothetical point in shape space encapsulated by the shape envelope for all
specimens.
Figure 12 – Χ2 probability density function plot. The result describes a statistically significant
likelihood of non-random data (i.e., that some alternative signal is controlling the distribution
of the data).
Page 30
29
4.3 Taxonomic Classification
Figures 14 and 15 are ordinations of the ruminant specimens at the taxonomic level of
Family (see Appendix 1). One would typically expect, that based on common ancestry,
organisms within the same family would exhibit more similarity than those between
different families (see Figure 12.3 in Jones and Goswami, 2010). In both plots, it is apparent
that both major families (Bovidae and Cervidae) show significant overlap between their
premaxillary shapes. Shape variation in both groups is spread continuously along the major
axis of shape variation, PC-1. Unfortunately, three of the studied families are only
represented by a single specimen (Giraffidae, Antilocapridae, and Moschidae), so it is
impossible to assess intra-family shape variations for them. Tragulids occupy a small niche
of shape space, nested within the amalgamation of cervids and bovids, with PC-2 and PC-3
somewhat distinguishing them from the other specimens (Fig. 15).
When studying functional morphology, a portion of the covariance between organisms may
be the result of phylogeny, were the patterns being produced are governed by the fact that
closely related organisms tend to be more similar than further related ones. Ideally, shape
variation is structured in a manner that is adaptational and reflects functional groupings
(e.g., based on feeding strategy). If the former was prevalent, one would expect to see
patterns in shape variation reflected at various levels in higher taxonomic groupings. In this
case, this type of patterning would that would be reflected as a well-structured
differentiation between Cervidae, Bovidae, and the other families, a pattern that is not
observed in the principal component ordinations at either the family level (Figs. 14 and 15)
or sub-family level (figs. 16 and 17). Instead of there being evidence for a confounding
phylogenetic signal, what is apparent is that there is widespread convergence, and possibly
parallelism, between all ruminants into a continuum of shape variation, with neither
forming discrete clusters based on premaxillary shape. This implies that an adaptational or
functional signal is prevalent with regards to this trait.
Page 31
30
Figure 14 - Two-dimensional PCA score plot for ruminants classified corresponding to their
family. Note that the PC eigenvalues are identical to the other PCA plots. Even individual genera
cannot be constrained into morphospace groupings (e.g., Gazella).
Figure 15 - Three-dimensional PCA score plot for ruminants classified corresponding to their
family. Legend as Figure 14.
Page 32
31
Figure 16 – Two-dimensional PCA score plot for ruminants grouped according to their sub-
family (or family for Tragulidae, Moschidae, and Giraffidae, as they only contain one specimen).
Figure 17 – Three-dimensional PCA score plot for ruminants grouped according to sub-family.
Legend as in Figure 16. With little or no discernible clustering of similarly classed organisms in
shape space, it can be inferred that phylogenetic relationships are acting only in a minor way, if
at all, to occlude true adaptational patterns. Members of Antilopinae however are clustering
towards the higher end of PC-1, which may be due to the high specimen to genus ratio (i.e., few
genera, many species).
Page 33
32
Figure 19 – Three-dimensional CVA score plot for ruminants classified according to their family.
Bovids and cervids maintain significant overlap with respect to all three axes, indicating that the
shape variations exhibited are more homoplastic in nature than phylogenetic (i.e., related to
ecology or functional constraints). Legend as in Figure 18.
Figure 18 – Two-dimensional CVA score plot for ruminants grouped classified according to their family.
Page 34
33
The absence of any controlling phylogenetic signal is observed when the above data are
subject to a CVA. If premaxilla shape was being controlled by phylogeny, then the
specimens would be drawn into their taxonomic clusters reflecting similarity based on
common ancestry. Figure 17 shows the overlap between Bovidae and Cervidae more clearly
– they are indistinguishable at this level based on premaxilla shape, indicating that there is
an extrinsic factor excluding phylogeny that is controlling their distibution. This is also seen
in the three-dimensional ordination (Fig. 18). Tragulids occupy a small domain in the overall
canonical variate space, suggesting that the three species of the genus Tragulus have
adapted to occupy a similar ecological role. These three browsers also occupy similar space
when grouped according to feeding strategy. This pattern of subspace occupation may be
associated with the fact that these species are georaphically constrained to southern and
south-east Asia, are an ancient lineage forming the sister group to all other extant
ruminants, and are the smallest extant ruminants known (Meijaard and Groves, 2004).
Their ecological role is significant for this study, as similarly to key basal ornithischians, they
have premaxillary ‘tusks’, analogous to species such as Heterodontosaurus tucki. This does
however indicate that, with respect to tragulids, a minor aspect of phylogenetic patterning
exists within the ordinations.
Figure 20 – Two-dimensional CVA score plot for ruminants classified according to their sub-family.
Whilst there does appear to be a degree of clustering for several groups (e.g., Reduncinae), the
amount of overlap between groups suggests that the phylogenetic signal is being over-ruled by a
subsidiary factor.
Page 35
34
When classified according to sub-family (figs. 19 and 20), ruminants exhibit similar patterns
to when they are classified according to family, but with slightly tighter clustering for
several groups (e.g., Hippotraginae). This is to be expected, as when taxonomic resolution
increases at lower levels, the organisms by default will be more closely related than at
higher taxonomic levels. There is, however, the problem that comparatively, the
hierarchical taxonomic levels used here may not be comparable in terms of definition (i.e.,
currently there is no strict definition describing what constitutes the family or sub-family
boundaries). It follows that the various clades represented within Ruminantia may not be
directly comparable, due to their differences in phylogenetic history, which is not reflected
in their taxonomy.
Figure 21 – Three-dimensional CVA score plot for ruminants classified according to their sub-
family. Legend as Figure 20.
Page 36
35
4.4 Shape Classification
This aspect of the analysis required a priori classification of all specimens based on the
general shape of their premaxillae. The aim was to see initially if the simple classification
scheme of ‘pointed’ and ‘blunt’ was, firstly, adequate to encapsulate the entire geometry of
the ruminant snout (excluding size), and secondly to see if visual assignment of species to
these groups could be supported in a statistical framework.
When classified according to whether the snouts were blunt or pointed, little resolution
was acquired through PCA (Fig. 22). The third PC axis appears to increase the complexity of
relationships between the two groups (Fig. 23). The implication is that this terminology is
largely inadequate to describe the total range of shape variation that ruminant snouts
exhibit. Many ruminants classed as either pointed or blunt occupy the PCA morphospace of
the opposite group. This may be due to either incorrect a priori classification, or that the
shape the PC axes are describing goes beyond the curvature of the distal snout tip. When
the data are subject to a CVA, there is little in the way of additional resolution. This is
somewhat surprising, as CVA is designed to pull pre-defined groups apart by maximising the
distance between the group centroids relative to the group dispersions. The reorientation
of the data into the CV space maintains a similar level of inter-group distinction to the
Figure 22 – Two-dimensional PCA score plot for ruminants based on the a priori classification of
their general snout shape (the primary shape descriptor).
Page 37
36
ordination in PC space (Fig. 24). A possible reason for this is that, again, the level of
classification is insufficient to encapsulate the complex morphological diversity of ruminant
premaxillae. As mentioned before, ‘pointed’ and ‘blunt’ profiles only form the end-member
classifications, supposedly representing the extremes of ‘browsing-’ and ‘grazing-type’
ruminants respectively, with intermediates between forming a shape continuum between
the two. This intermediate phase has been neglected here, to reduce the impact of an
additional arbitrary threshold – without a secondary criterion directly linked to snout shape
that can also discriminate between feeding strategies, any arbitrary threshold created here
cannot be empirically validated. The lack of group-resolution in the CVA space defined by
the first two CV axes indicates that the level of intra-group variation is overpowering the
inter-group variation to an extent that their distribution clouds are unable to be recovered
as completely discrete.
Figure 23 - Three-dimensional PCA score plot for ruminants based on the a priori classification of
their general snout shape. Legend as Figure 22.
Page 38
37
The above ordinations demonstrate the deficiency in descriptive power that the terms
‘pointed’ and ‘blunt’ contain when the snout profiles are delimited in this fashion, albeit in
a qualitative and somewhat subjective manner. Increasing the quality of the descriptive
terms in the secondary shape descriptor increases their explanatory power as well as
increasing their precision. Due to the relative ease in identifying whether a curve is straight,
convex, or concave is straightforward, this should not increase the uncertainty for the
qualitative delimitations. These additional terms were combined with the primary profile
shape descriptors to increase the number of groupings to six; this should in theory reduce
the intra-group variation and lead to increased clarity in the discrimination analysis by
increasing the inter-group variation. A 2D ordination in PCA space is shown for this new
grouping in Figure 25. The first two PC axes here describe almost 90% of the variance. The
plot shows the relative positions of the six groups in the space defined by the first two
principal components, and the complexity of the overlaps is rather apparent. Inclusion of
the third principal component further increases the convoluted distributions, although it
does seem to restrict the pointed-class specimens to the positive end of PC-2 (Fig. 26). No
single group is entirely captured within the space envelope of another, providing slight
resolution.
Figure 24 - Two-dimensional CVA score plot for ruminants based on the a priori classification of their
general snout shape. The shape envelopes in this space maintain a similar degree of overlap to the
associated PCA ordination.
Page 39
38
Figure 25 – Two-dimensional PCA score plot for ruminants classified according to the secondary
shape description.
Figure 26 – Three-dimensional PCA score plot for ruminants classified according to the secondary
shape description. Note the broad distribution of all groups with respect to PC-1, except for the
‘blunt-convex’ group, which appears to be restricted to negative values.
Page 40
39
When subjected to a CVA, the new groupings are ordinated in space with high explanatory
power (97% covariance in the first two CV axes). As in the PCA ordination, ruminants with a
‘blunt-concave’ (in the format ‘distal tip-lateral margin) profile occupy a large space
domain, with most of the specimens well discriminated from the other groups (although
still with a high degree of overlap for many others). The ‘blunt-linear’ and ‘blunt-convex’
groups occupy a highly similar position in the space, and are clearly more similar to each
other than to the third ‘blunt’ grouping. All three are fairly well constrained with respect to
CV-1, mostly with scores greater than 0.05 on this axis. With the exception of several
specimens grouped as ‘pointed-convex’, the ‘pointed’ domain appears confined to scores
less than 0.05 on CV-1. The ‘pointed-linear’ and ‘pointed-concave’ groups occupy similar CV
space, and are both largely confined within the ‘pointed-convex’ space. All ‘blunt’ and
‘pointed’ groups maintain a degree of overlap with their opposites. This ordination provides
further evidence that, despite an increase in resolution and descriptive power, snout shape
cannot be easily discriminated into discrete groups, and instead forms a continuum, likely
with ‘pointed-convex’ and ‘blunt-concave’ profiles occupying end-members of the shape
spectrum. All other specimens can be placed within this continuum, but restricting them to
discrete groups is not only unsupported, but acts to reduce the distinct morphological
complexity exhibited in the ruminant premaxillae.
Figure 27 – Two-dimensional CVA score plot for ruminants classified according to the secondary
shape description.
Page 41
40
Following the method outlined above, strobe plots were constructed for both profile shape
categories. For the first category, only a single discriminant axis was possible to model due
to the presence of only two sub-groups (Fig. 28).
The resemblance of Figure 28 to the first modelled axis when the ruminants are classified
according to their feeding strategy is intriguing (Fig. 13). The axis is reversed, a
phenomenon that can be ascribed to the arbitrary direction of the principal component
axes (MacLeod, 2005). The differences between the two plots are subtle: in Figure 13, the
anterolateral expansion tapers more gradually towards the maxilla-premaxilla suture,
whereas in Figure 28, the expansion rapidly pinches posteriorly, developing into a
secondary proximal lateral expansion. The expansion of the distal tip is also slightly less
Figure 28 – Single strobe plot of the CV model axis in its respective PCA space.
Figure 29 – Strobe plot for the 5 modelled CV axes in PCA space when the shape descriptor
complexity is increased.
Page 42
41
pronounced in Figure 28. One implication of this similarity is that the transformation
sequence seen here may be enough to describe variations in feeding strategy. One end of
the spectrum would be occupied by specimens with a proximally compressed, distally
expanded, and relatively blunt snout, the other by a proximally expanded, distally rounded
and compressed snout. This is slightly more complex than simply ‘blunt’ and ‘pointed’, and
describes the shape variation along the length of the premaxillary profile. However, this
does not mean that either of these shapes adequately describes or defines the archetypal
snout profile for either browsing or grazing ruminants; this dichotomy remains unresolved
in terms of snout shape. The modelled strobe plot for the second shape category exhibits
much the same sequence of deformation for the first axis as the initial category, although
the variation is less pronounced (Fig. 29).
The second modelled axis exhibits strong deformation between an initially
anteroposteriorly compressed and laterally broad profile, into a progressively lengthened
and mediolaterally compressed profile (Fig. 29). The third axis retains uniformity around the
distal half, with deformation occurring proximally with increasing lateral expansion towards
the maxilla-premaxilla suture. The fourth axis is unique among all modelled axes, in that it
clearly lacks the deformational symmetry of the others, and is distinctly skewed both in the
mediolateral and anteroposterior dimensions. The causes for this are possibly due to error
in the profile reconstructions, but perhaps attributable to the fact that these represent
hypothetical models, not genuine specimens, and perhaps represent an unrealistic and
extreme deformity of the profile
shape. The fifth axis displayed no
visible deformation patterns.
Table 2 shows the results of a
likelihood ratio test performed on the
data subsequent to their respective
canonical variates analyses. The
statistical significance of the ecology-based analysis has been mentioned previously.
Statistically significant results were also recovered when data were grouped based on their
Table 2 – Results of the chi-squared likelihood test for
all ruminant groups. Note how all analyses are
statistically significant in CVA space, except for when
the specimens are grouped by Family.
GroupingObserved
Φ- value
Degrees of
FreedomProbability (%)
Family 30.77 15 0.942
Subfamily 134.8 45 6.764E-09
Shape I 43.29 3 2.132E-07
Shape II 91.15 20 4.054E-10
Ecology 55.84 12 0
Page 43
42
sub-family and both shape categories. Somewhat perturbingly, almost the exact opposite
was found for the data when grouped by family. This likelihood value may be the product of
the way the data was grouped, as the classifications were dominated by the Families
Bovidae and Cervidae. If so, then it is somewhat paradoxical why a similar result was not
recovered for the first shape category (i.e., ‘blunt’ or ‘pointed’ profiles).
When ruminants are classified ecologically as browsers, grazers, intermediates, or
frugivores based on a range of secondary criteria, they cannot be discriminated based on
the shape of their premaxillary profile. Instead, snout shape appears to be homoplastic in
nature, with a broad range of geometries present in each of the feeding style groups.
Furthermore, previously used descriptive terms such as ‘pointed’ and ‘blunt’ are not only
inadequate to partake in defining these groups, they also like the descriptive power to
define the shape variation that ruminant snouts so visibly demonstrate. Refinement of
these terms to describe a more complex geometry leads to increased resolution of more
precisely-defined groups, but still a lack of demarcation into spatially discrete clusters.
These results are confirmed as statistically significant using a chi-squared likelihood
assessment, and are also found not to be an artefact of phylogenetic affinity, although this
inference is somewhat dubious (i.e., statistically non-significant data when grouped by
Family, and qualitative assessment of phylogenetic relationships in shape space).
Given the above inconsistencies with previous assumptions, to what extent can the
patterns found here be extrapolated back to ornithopods, given that little can be explicitly
defined in terms of classifying feeding strategy in terms of snout morphology? It is evident
that individual ruminants occupy specific positions in a continuum of shape variation,
irrespective of phylogenetic affinity and putative pre-defined feeding strategy. Although
demarcating specific thresholds for groups such as ‘browsers’ and ‘grazers’ is problematic,
given their overlapping nature with respect to a premaxilla-based profile shape analysis, it
is apparent that specific shapes represent the end-members of the continuum.
Page 44
43
To see exactly how ornithopods are related to ruminants in terms of premaxillary shape,
the PCA scores for the reduced ornithopod data set were projected into the PC space
created by the first two principal components for when ruminants are classified according
to feeding strategy (Fig. 30). There is both significant overlap with for many specimens, and
definition of a particular ornithopodan morphospace.
Closer inspection shows that this pattern has a strong structure relating to the phylogenetic
positioning of the ornithopods (Fig. 31). Although every morphospace defined by each
particular class of ruminants exhibits strong overlap with the others, it is noteworthy that
the only ornithopods that similarly occupy this space are the more primitive species (e.g.,
heterodontosaurs, hypsilophodonts, iguanodonts). Hadrosaurs occupy their own defined
space away from ruminants, along with several iguanodonts and, surprisingly,
Heterodontosaurus tucki. This strongly challenges the viewpoint that ungulates and
hadrosaurs are ecologically analogous (e.g., Carrano et al., 1999), in spite of the fact that
snout shape is a poor reflection of feeding strategy. Regarding the species that do fall into
ruminant space, there is a slight structure evident in their distribution. Only two can be
unequivocally assigned putative feeding strategies; Zalmoxes robustus is placed within
Figure 30 – Projection of the ornithopod PCA scores into the space defined by the first two PC
axes for the ruminants, with ruminants shown classified according to their feeding strategy (see
Figure 6).
Page 45
44
‘grazer’-defined space, and Zephyrosaurus schaffi in ‘intermediate’-defined space.
Specimens placed within space where all three groups overlap are all of hypsilophodont-
grade (e.g., Oryctodromeus cubicularis, Changchunsaurus parvus), and those placed within
either ‘browser’ or ‘intermediate’ groups include most of the iguanodont-grade specimens
(including also Camptosaurus dispar and Dryosaurus lettowvorbecki), the remaining
hypsilophodonts, and the heterodontosaur-grade Agilisaurus louderbacki. Tenontosaurus
dossi, Heterodontosaurus tucki, and Telmatosaurus transsylvanicus (the most basally-placed
hadrosaur analysed here) are all placed close to the ‘grazer’-defined space, but within the
predominantly hadrosaur-defined ornithopod domain. This may open up pathways for
further investigations into what can be inferred about the digestive or ingestive physiology
of these organisms, but given the ambiguity surrounding the browser-grazer dichotomy,
any extrapolations made will only be tentative at best.
BrowserCamptosauridaeDryosauridaeFrugivoreGrazerHadrosauridaeHeterodontosauridaeHypsilophodontidaeIguanodontidaeIntermediateRhabdodontidae
Figure 31 - Projection of the ornithopod PCA scores into the space defined by the first two PC
axes for the ruminants, with ruminants shown classified according to their feeding strategy.
Ornithopods are shown categorised according to their Family. PC axes as Figure 6.
Page 46
45
When the ornithopod data set is exclusively ordinated, the hierarchical pattern between
taxonomic groupings becomes apparent (Fig. 32).
The structure in Figure 32 shows a transition along PC-2 from heterodontosaur-grade
species, sequentially through hypsilophodont-grade and iguanodont-grade species, and
finally hadrosaurs which occupy the broadest region of shape space. If iguanodontids are
removed, the spatial structure each group occupies is very distinct. The iguanodontians
located in the hypsilophodont domain are Fukuisaurus tetoriensis, Theiophytalia kerri, and
the second morph of Tenontosaurus tillettorum. The former is considered a relatively small
species, which may be cause for its position here. Theiophytalia, in a recent analysis by
McDonald (2011) was found to be a basal member of the clade Styracosterna, placing it as a
close relative to Camptosaurus; indeed, this species was long considered a camptosaur until
reassignment by Brill and Carpenter (2007). The current taxonomic status of Tenontosaurus
tillettorum is equivocal, with authors unable to decide if it should be placed within
Iguanodontia, outside the group with the other basal ornithopods in a hypsilophodont-
grade melange, or even assigned its own Family with the congeneric T. dossi. Given this
information, it is possible to infer that snout shape here forms a continuum, analogous to
ruminants, and strongly reflects a phylogenetic signal. Dryosaurs and camptosaurs appear
Figure 32 – Two-dimensional PC score plot for ornithopods when classified according to their
family level.
Page 47
46
to occupy the middle ground between hypsilophodonts and iguanodonts, as expected.
However, their spaces are poorly constrained due to lack of specimens. The substantial
overlap between iguanodontids and hadrosaurs is postulated here to be the result of a
moderate component of ontogenetic shape variation, with physically younger hadrosaurs
morphologically representing the phylogenetically more basal iguanodontians. This pattern
of phylogenetically-related shape change is also evident when these data are subject to a
canonical variates analysis. The groupings are discriminated marginally more, with the
exception of Zephyrosaurus schaffi, which again occupies a place between hadrosaurs and
iguanodonts (Fig. 33).
This apparent transitional structure between progressively more advanced groups may be
an artefact of the ages of the species. Although ornithopods exhibit a defined temporal
phylogenetic hierarchy, this may enhance or distort the signals reflected in snout shape
observed here. Accordingly, the specimens were categorised in broad groups according to
their chronostratigraphic geological age. If similar patterns are observed between
phylogenetic and temporal trends, then this would provide strong support for previous
Figure 33 – Two-dimensional canonical variates score plot for ornithopods when classified
according to their family. The transition between groups is constrained to CV-1.
Page 48
47
assertions that, not only did masticatory complexity in ornithopods increase through time,
but also that it occurred in a step-wise manner with a strong phylogenetic structuring.
Figure 34 shows a discriminant plot for the ornithopods when they are coarsely classified
into chronostratigraphic groups. What is immediately apparent is the under-representation
of species from the Lower and Middle Jurassic. Upper Jurassic species (i.e., Camptosaurus,
Dryosaurus) are moderately constrained to the space defined by Lower Cretaceous
specimens (dominantly iguanodonts and hypsilophodonts). Despite the poor temporal
sampling, there is an apparent pattern of increasing morphospace occupation from the
Upper Jurassic to the Upper Cretaceous, although this may purely be an artefact of sample
size. If this is a genuine pattern however, it implies that at some point between the Lower
and Upper Cretaceous, there was a remarkable increase in snout shape variation. The
distal-most species creating this broad space include the lambeosaurine hadrosaurs,
Oryctodromeus cubicularis, and Zalmoxes robustus, indicating that this morphological
radiation is not purely confined phylogenetically. The possibility that an external factor is
driving, or being driven by, the increased snout shape variation can, therefore, not be
excluded.
Figure 34 – Two-dimensional canonical variates score plot for ornithopods broadly classified
according to their chronostratigraphic ages. Unfortunately, Jurassic species are poorly represented
in this aspect.
Page 49
48
To observe the shape change occurring along the CV axes, the procedure applied to the
ruminant data set was followed, with five modelled points per CV axis (for four axes) being
back-projected into the PC space defined by the first two principal components, which
together explained 88% of the total variance (Fig. 35).
The pattern of morphological
variation is even more apparent in
Figure 35. Each subsequently older
group is entirely constrained within
the space occupied by the younger
group (from the Upper Jurassic to
Upper Cretaceous). This is almost
certainly representative of an
increase in morphological diversity as opposed to a sampling artefact, especially regarding
the Lower to Upper Cretaceous transition. Both PC-1 and PC-2 explain between-group
variation. One problem with this data is that it is not deemed statistically significant using
Figure 35 – Two-dimensional PCA score plot for ornithopods classified according to their
chronostratigraphic age. Note the progressive increase in morphospace occupation as seen in
Figure 34. Back-projected CV axis models are represented by the black points.
GroupingObserved
Φ- value
Degrees of
FreedomProbability (%)
Family 61.82 24 0.004
Age 35.05 16 0.391
Shape I 21.95 4 0.02
Shape II 49.6 20 0.025
Table 3 – Results of the Χ2 likelihood test for the four
groups in the context of the sub-groups in CVA space.
The results imply that all distributions are statistically
significant, except for when ornithopods are grouped
according to their family.
Page 50
49
the traditional threshold of significance when the chi-squared likelihood test is applied.
These results are summarised in Table 3, along with the results for when this test is applied
to the other three canonical variates analyses conducted on the ornithopod data set. The
reason for this low probability in the age-defined groups is possibly attributed to that fact
that two out of the five groups only contain one member. Nonetheless, the temporal
pattern observed in Figure 35 is still clear, even if not recovered as statistically significant.
The strobe plots along the modelled CV axes show some rather peculiar results (Fig. 36).
The first modelled axis shows an almost direct vertical reflection of a distorted variation of
the mean shape, with the two end points clearly skewed distally in the lateral aspect. This
lack of symmetry seems to imply that the axis of greatest discrimination describes
taphonomic deformation in PC-defined space.
Figure 36 – Strobe plot representing the four modelled CV axes in PC-defined space. Note the lack
of symmetry in the transformations along CV-1 (and marginally CV-4), but the strong symmetry
exhibited by CV-2 and CV-3.
Page 51
50
The second and third canonical variates axes describe a general lateral broadening of the
snout, proximally and distally respectively, with the respective distal and proximal regions
remaining relatively unchanged. This is also observed for CV-4, with the whole snout
broadening but with accompanying distal compression. What this illustrates is that through
time, there was a transition between a markedly ‘pointed’ snout shape to a broader, more
‘blunt’ shape, the style of transition previously posited for ruminants, but not supported in
this study. However, for CV-2, this transformation sequence actually is the reverse to both
CV-3 and CV-4, with the proximally broadened forms belonging to the Lower Jurassic
Heterodontosaurus tucki mean shape, and those with a distal mediolateral compression
being typical of hadrosaurs. The third CV axis models the transition between the Lower
Cretaceous iguanodontians and Upper Cretaceous hadrosaurs with distally-flattened
snouts, such as Gryposaurus latidens. The fourth CV axis models a transect along hadrosaur-
dominated shape space, with a variety of conspecific morphs being represented.
To observe this shape variation without the interaction of Family level groupings, the same
shape classifications used for ruminants were applied to the ornithopod data set. When
initially classified as either ‘blunt’ or ‘pointed’, a remarkably stronger pattern of variation is
found than that for the ruminant data set. A slight overlap is observed when subject to a
principal components analysis in the space defined by PC-1 and PC-2 (Fig. 37), but with the
inclusion of PC-3 in the ordination, the distinction between the two groups becomes
strikingly clear (Fig. 38). Ornithopods with a ‘blunt’ snout profile are confined to higher
values on the PC-2 axis, and occupy a broad range with respect to PC-1. Those with a
‘pointed’ profile are tightly confined in both axes to a smaller morphospace, with PC-2
differentiating moderately well between the two groups. The ornithopods occupying the
overlapping zone, classified as ‘pointed’ are not constrained to any particular taxonomic
group, whereas those classified as ‘blunt’ and occupy the overlapping zone are
iguanodontians, Dryosaurus lettowvorbecki, and the basal hadrosaur Telmatosaurus
transsylvanicus. The only hadrosaurs classified as ‘pointed’ include the mean shape for
Hypacrosaurus altispinus and the second morphotype of Prosaurolophus maximus,
Page 52
51
suggesting that that this shape is uncommon for hadrosaurs, supporting the assertions
made above regarding taxonomic distinctions between hadrosaurs and the other groups.
Figure 37 – Two-dimensional PCA score plot for ornithopods classified according to the shape of
their snout.
Figure 38 – Three-dimensional PCA score plot for ornithopods classified according to their primary
snout shape. Note the almost complete distinction between the groups, contrary to the results
obtained for ruminants.
Page 53
52
Somewhat counter-intuitively, this distinction is confounded when these data are subject
are subject to a CVA (Fig. 39). This is plausibly due to the fact there is a single discriminant
axis trying to differentiate between two morphologically complex groups, and the within-
groups variation is exceeding the between-groups variation to a such degree that the group
centroids are unable to be pulled apart further in CVA space. The majority of this overlap is
in fact created by the placement of the mean shape of Hypacrosaurus altispinus, which is
placed within a strong cluster of ‘blunt’-profiled hadrosaurs. Both of the specimens used to
generate the mean shape clearly exhibit a ‘pointed’ profile, which suggests that, as with the
ruminants, this classification system, despite achieving significantly higher resolution, is
partially inadequate to capture the full geometry of morphological variation in the
ornithopod snout profile.
For ruminants, the profile shape classification was moderately enhanced when the lateral
geometry was incorporated into the shape descriptors. Accordingly, the ornithopods were
subject to categorisation according to these more complex descriptions. In a plot of PC-1
against PC-2, what is instantly apparent is an increased complexity in the relationships
between the groups.
Figure 39 – Two-dimensional CVA score plot for ornithopods classified according to their primary
snout shape. Note the increased number of specimens classified as ‘blunt’ in the overlapping zone
between the two groups.
Page 54
53
Those classed as either ‘pointed-linear’ or ‘pointed-concave’ are well constrained, occurring
towards the lower end of PC-2. However, those classified as ‘pointed-convex’ extend
beyond this range into a space dominated by the ‘blunt’ sub-groupings. This is possibly
attributable to the observation that, if the lateral geometry is convex-outwards, it will
generally lead to a broadening of the snout, which will give it the illusion of seeming more
blunt. This could imply that, more so than the distal shape, it is the geometry of the lateral
margins of the snout that controls the distribution of the specimens here in shape space, by
secondarily inflicting control upon the distal snout geometry. It also suggests that PC-2 is
describing shape variation acting upon the lateral margins of ornithopod snouts. When
subject to a CVA, there is little in the way of increased between-group discrimination (Fig.
41). This is possibly accountable again to high within-group variation inhibiting the ability of
the canonical variates axes to successfully discriminate completely between the group
centroids.
Figure 40 – Two-dimensional PCA score plot for ornithopods when classified according to a
secondary shape descriptor incorporating the lateral snout geometry.
Figure 40 – Two-dimensional PCA score plot for ornithopods when classified according to a
secondary shape descriptor. Note how all the specimens that create the overlap in Figure 37 are
classed here as ‘pointed-convex’.
Page 55
54
The group dispersion pattern shown in Figure 41 is interesting, as it shows a tendency for
ornithopods with a ‘blunt’ prefix to group closer together, and those with a ‘pointed’ prefix
to group closer, still with moderate overlap between these two supra-groupings, and also
within the secondary groups. This suggests that it is the geometry of the distal portion of
the snout acting primarily to distinguish groups, and that the lateral geometry acts
secondarily to pull these groups apart, but still retains marginal overlap between all groups.
When the CV model axes for both of the shape groupings are back-projected into PCA
space, they both display similar patterns of shape variation along the first axis. For the first
shape-defined groups, the modelled CV axis defines a transect cutting through both the
major groupings, with the mean shape nestled firmly in the zone where the two overlap
(Fig. 42). The two extreme points of this axis are both outside the shape envelopes defined
by the data set, with respect to the first two principal components, and as such represent
only hypothetical extremes of either group. This axis represents the transformation in
shape that acts to maximise discrimination between these two groups. What is instantly
evident when these models are visualised in a strobe plot, is the insufficiency of the terms
‘blunt’ and ‘pointed’ in describing their geometry.
Figure 41 – Two-dimensional CVA plot for ornithopods when classified according to a secondary
shape descriptor. Note the size of the space occupied by those classified as ‘blunt-concave’, a
group dominated by hadrosaurs.
Page 56
55
The two extreme plots represent a warping of specimens such as the fourth morphotype of
Lambeosaurus lambei along PC-1, and at the other extreme a warping of Zalmoxes robustus
the opposite direction along PC-1 (Fig. 43). Overall, the plot describes deformation from a
broad-tipped and laterally compressed shape, through the mean shape, typically
representing iguanodontians, towards a shape that is broad at the maxilla-premaxilla
junction, and tapers distally into a rounded tip, representative of specimens such as
Hypsilophodon foxii. The overlay plot here is similar to that produced for the equivalent
analysis for the ruminant data set (Fig. 28), except the deformations are more pronounced
in terms of the lateral expansion, and there is considerably less compression or extension of
the distal tip.
Figure 42 – Single modelled CV axis (black points) back-projected into the PC space defined by the
first two principal components when ornithopods are classed as either ‘blunt’ or ‘pointed’.
Legend and axes as Figure 37.
Figure 43 – Strobe plot representing the single modelled CV discriminant axis in a space defined by
PC-1 and PC-2.
Page 57
56
The specimens defined by broadening of the distal tip are confined to the hadrosaurs
occupying the higher end of PC-2, and the proximally broad snout is confined to
hypsilophodonts and heterodontosaurs, with the iguanodonts and occasional members of
the other families tending to congregate around the central shape. This is a strong pattern
of variation, exhibited in conjunction with phylogenetic affinity, and also mimics the pattern
that both CV-2 and CV-3 show in Figure 36 (exclusive of the anteroposterior expansion or
compression of the distal tip of the snout), when ornithopods are classified according to
their chronostratigraphic age.
This pattern is similarly recovered when the model CV axes representing the secondary
shape categorisation are back-projected and visualised in PC space. Although given the
number of axes, it is somewhat difficult to assign them to particular species or spaces (a
problem which is also exacerbated by the overlapping nature of many of the groups), a
near-identical transformational sequence is described by CV-1 (Fig. 44).
Figure 44 – Strobe plot representing the first 4 modelled CV axes in PC space when ornithopods are
classed in a manner that more completely describes their snout profile geometry. Note that the
fifth modelled axis has been excluded here, as it showed no deformation between model points.
Page 58
57
The first CV axis in Figure 44, although reversed in direction, is nearly identical to the
modelled axis shown in Figure 43, with the exception that the transformations are more
gently curving, and slightly less pronounced in the distal region. The same pattern of
deformation is evident however, with a hypsilophodont- or heterodontosaur-grade profile
occupying one end, and an advanced hadrosaur-grade profile occupying the other, relating
to ‘pointed-convex’ and ‘blunt-concave’ geometries, respectively.
The second and third CV axes represent transects through all of the group spaces, with the
exception of the ‘pointed-linear’ group. The patterns of transformation they reveal are
fairly similar, given their near-corresponding orientations in PC space. Broadly comparable
to CV-1, they represent a gradation from a laterally compressed shape into a distinct distal
laterally expanded and flattened profile, with this expansion propagating down the length
of the snout. There is also substantial distal compression accompanying this, a
transformation not picked out by CV-1. The fourth CV axis is spatially constrained in 2D PC
space, describing a near-vertical transect, with lambeosaurine hadrosaurs occupying one
end of the axis, and the second morph of Iguanodon bernissartensis occupying the other.
The variation this axis describes is deformation from a bulbous-tipped profile to one that
retains a similar distal width, but becomes moderately compressed in the mid-premaxilla
and laterally expanded proximally adjacent to the maxilla-premaxilla suture into a club-like
geometry.
The patterns indicate that initially, early ornithopods predominantly had distally-pointed
premaxilla, plausibly the plesiomorphic condition, the geometry of which was retained in
lineages such as Zalmoxes, possibly relating to its extensive isolation in the European island
archipelago (Weishampel et al., 2003). Following this, as ornithopods grew in body size and
became more morphologically complex adopting the typical iguanodont-grade bauplan, it
appears that they experimented with various snout shapes, possibly reflecting adaptation
to different diets and nutritional pathways reflecting the need for increased forage
consumption required for increasing body size. Contemporary hypsilophodonts retained
more gracile bauplans, and a more pointed snout; this may relate to increased selectivity, in
Page 59
58
the manner proposed for ruminants, as they attempted to compete against the more
diverse iguanodonts.
Hadrosaurs exhibit a substantially wider and flattened distal snout compared to the other
ornithopod groups. There is currently a lack of consensus as to whether hadrosaurs were
more analogous to ground-level grazers or higher level browsers. What is apparent here is
that hadrosaurs exhibit the ‘blunt’ distal snout shape that has previously been considered
indicative of a grazing feeding strategy in ruminants, but rejected as a defining trait in the
current study. However, if the role of snout shape indeed lies in selectivity, in spite of the
fact that browsing and grazing ruminants cannot be distinguished based on snout profile
shape, it can still be proposed that hadrosaurs were non-selective feeders, with a relatively
higher intake rate.
Page 60
59
5.0 Discussion
The results obtained by this study suggest a preferential method of analysis for future
investigation of functional ecology. For individual, but not necessarily independent,
morphophysiological traits, multivariate discriminant analyses with all known parameters,
while independently correcting for allometry in body mass and phylogeny, are required to
identify the traits that exhibit a deterministic influence overall on feeding strategy.
Following this, traits that operate in conjunction with each other the most strongly, once
elucidated, should be assigned functional guilds representing the ecology of distinct groups
of organisms. Pérez-Barbería et al. (2004) reached a similar conclusion in that the current
boundaries are somewhat arbitrary between feeding strategies in ruminants. A viable
approach to resolve the problem should employ a covariate or group of covariates as
continuous variables, with thresholds being based on the identification of functionally
significant and discrete clusters. However, using this methodology, the authors who have
investigated this issue so far have found no morphological discrepancies that can explain
variation in ruminant digestive efficiency based on digestive, not ingestive, morphology (see
also Janis, 1995; Clauss et al., 2003; Sponheimer et al., 2003; Clauss et al., 2008a). This
perplexing result may be, in part, due to treating species as static entities, when realistically
thresholds should be constructed on a sliding scale accounting for spatiotemporal
variations where appropriate (Owen-Smith, 1997). It also seems that general laws must be
flexible enough to account for singular exceptions (e.g., frugivores) and are insufficient to
encapsulate the full diversity of ruminant feeding habits. The real problem, however, may
stem from the fact that previous work has attempted to arbitrarily sub-divide and
categorise species that, in reality, form a continuum, with ‘browsers’ and ‘grazers’
occupying terminal points on the continuum, representing the most stationary or inflexible
feeding types. This scenario is most likely the one explained by the results of the current
investigation.
Whitlock (2011) in an analysis of diplodocoid feeding habits uses snout (i.e., premaxilla)
shape as a proxy for selectivity, and combines it with dental microwear as a proxy for
relative browse height. He followed the same methods as Solounias et al. (1988) and
Page 61
60
Solounias and Moelleken (1993) to discriminate snout varieties, adding two additional
comparative metrics based on the quality of the specimens. This study is intriguing, as
obviously most dinosaurs cannot be fitted into the regular classes of browsers and grazers
due to the general lack of contemporary grasslands. Nevertheless, the additional dimension
of browsing height allows comparisons to be drawn with extant organisms (i.e., ungulates),
but applied to different ecological hypotheses. Whitlock (2011) draws comparisons with
Carrano et al. (1999), in that different groups of sauropods and hadrosaurs are analogous in
their feeding styles, despite their phylogenetic and ecological distance. Carrano et al. (1999)
in an analysis of muzzle shape (using width as a proxy) recovered a functional dichotomy
between typically dimorphic lambeosaurine and monomorphic hadrosaurine hadrosaurs, in
that the former have narrower snouts (premaxillae) indicative of selective browsing, and
the latter having broader snouts for non-selective browsing. This hypothesized shape-based
dichotomy was not recovered in this study, despite the use of a similar range of hadrosaurs.
The results obtained in Whitlock (2011) are more applicable to the patterns detected by
this study. Within sauropods, specifically the clade Diplodocoidea, a functional dichotomy
was resolved with respect to snout shape, relating to both selectivity and browsing height,
with the simple use of ‘square’ and ‘round’ shapes defining these feeding strategies.
Diplodocoids with a ‘round’ snout shape are considered by Whitlock (2011) as mid-height
canopy browsers with a selective diet. Those with ‘square’ snouts were interpreted as
ground-level non-selective browsers. Application of these inferences to ornithopods is
somewhat problematic, given the disparity in body size between the two groups
(ornithopods were typically considerably smaller). However, selectivity and browsing height
are largely independent factors – just because an organism browses at a mid-level height, it
does not necessarily mean that they were selective. The significant aspect of this analysis is
that functional groups, relating to two independent dimensions of feeding style, are
strongly associated with snout shape. Applying these inferences to the current
investigation, it is plausible that ornithopods with more pointed snouts (i.e., the
heterodontosaur- and hypsilophodont-grade species) had selective diets, but at a browsing
height closer to ground level befitting their typically small stature. Furthermore, hadrosaurs
can be considered to be ground-level non-selective browsers, a somewhat counter-intuitive
Page 62
61
deduction given their increased height. This could further support the notion that
hadrosaurs were facultatively bipedal, adopting a quadrupedal stance for ground-level
foraging (see Sellers et al. 2009).
The relation of dinosaurian feeding mechanisms to contemporary floral diversity has been a
focus of study for many years (see Butler et al. 2009). Butler et al. (2009), using a large
occurrence database of all known Cretaceous plants and herbivorous dinosaurs, found that
there is no correlation between the relative diversity of one particular dinosaurian clade
(e.g., Ornithopoda, Sauropoda) to that of contemporary floras, specifically focussing on
angiosperms. This is consistent with the underlying co-radiation hypothesis between
dinosaurs and angiosperms (the proposition that a ‘faunal changeover’ from high-browsers
to low-browsers at the Jurassic-Cretaceous boundary drove the radiation of early
angiosperms) within a phylogenetic framework; i.e., can the relative phylogenetic diversity
of any herbivorous dinosaur clade be associated with a concurrent variation in the diversity
of early flowering plants. Nevertheless, this interesting result does not constitute a test of
“co-evolution” (see Sander et al. (2010), p. 352); it is a test of co-radiation, based on
analysis of relative diversity, of which co-evolution is a more intricate detail within,
consisting of the synchronized evolution of adaptive traits in, for example, competitive
organisms or those with predator-prey interactions (i.e., a strong ecological coupling).
Results by Butler et al. (2009) were non-significant for all dinosaurian clades. However,
Leckey (2004) found a strong correlation between end-Cretaceous North American
ornithischians (including ornithopods) and angiosperms. Butler and Barrett (2008) found
that both hadrosaurids and non-hadrosaurid ornithopods are significantly correlated with
Cretaceous marine sediments (albeit within coarse sedimentological and taxonomic
groupings), representing environments such as coastal plain. However, ornithopods also
occur within terrestrial sediments (e.g., Tenontosaurus is associated with Early Cretaceous
lacustrine and fluvial sediments). This supports the view advocated by Sander et al. (2010)
that ornithopods deliberately segregated themselves spatially from direct competition with
contemporary sauropodomorphs. But there is no evidence to indicate this is a secondary
consequence of an ornithopod association with angiosperms, as the authors speculate.
Page 63
62
The earliest angiosperm macrofossils are rare, and occur in Lower Aptian to Lower
Cenomanian (125-96Ma) sequences from the United States mid-Atlantic coast (Hochuli et
al., 2006). Within this sequence, the angiosperms become increasingly abundant through
time, in facies representing highly-disturbed riparian environments (Royer et al., 2010).
These earliest angiosperms are interpreted as rapidly growing weedy shrubs and herbs
(Royer et al., 2010). Contiguous ornithopods, as revealed by this investigation, typically
possessed distally rounded or pointed snouts, belonging to iguanodontid- or
hypsilophodont-grade species respectively. These forms were succeeded in most cases by
hadrosaurs during the Upper Cretaceous, representing a turnover in body size and feeding
strategies. In agreement with Williams et al., (2009), it is suggested here that hadrosaurs
may have been ground-level feeders. If this is indeed the case, then what this observation
represents is an ecological shift from smaller, ‘browsing’ ornithopods, with a transitional
and experimental phase in iguanodontid feeding strategies, the timing of which coincides
with initial angiosperm diversification, followed by hadrosaurs with a high consumption
rate of ground-level forage.
According to Butler et al. (2009), angiosperm diversification increased significantly during
the Santonian-Maastrichtian to a similar degree as the initial Albian-Cenomanian radiation.
It is possible that this second radiation, which coincides with increasing hadrosaur diversity,
is related to this changeover in feeding strategy. During the latter half of the Upper
Cretaceous hadrosaurs were dominant terrestrial herbivores, with a fairly cosmopolitan
distribution. As angiosperm diversity has often been linked theoretically with increased
environmental disturbance (see Butler et al. 2009), which plausibly could result from
increased ‘grazing’ by hadrosaurs, it is possible that increasing angiosperm diversity during
the Upper Cretaceous may be closely coupled with increasing hadrosaur diversity
(conforming to Leckey, 2004), and concurrent changes in feeding strategy accompanying
this. However, there is a distinct paucity of macrofossils for both hadrosaurs and
angiosperms during the Turonian-Santonian periods (Butler et al., 2009). As a result, the
exact timing of this turnover and diversification of both hadrosaurs and angiosperms is
poorly understood on a spatiotemporal scale. What would be worthwhile is the mapping of
Page 64
63
snout shapes on to all known ornithopod occurrences, following the methods of Butler et
al. (2009), and observing not just the changes in relative diversification patterns, but seeing
if angiosperm diversity patterns can be spatiotemporally linked with variations in
ornithopodan feeding strategies, as revealed from snout profile shape.
Ornithopods are unique among all other Ornithischians in that it is likely they favoured a
grinding mechanism using complex ‘dental batteries’ as opposed to slicing with pointed
‘beaks’, except for perhaps some plesiomorphic species and ceratopsians (Bell et al., 2009).
This complexity of the masticatory system is conceivably a mechanical adaptation relating
to increased consumption of low-quality forage, which at ground level will have included
minor volumes of grit, and possible silicophytoliths if early grasses were consumed too
(Williams et al., 2009). Future analysis of all ornithischian clades should further shed light
on the feeding strategies of contemporary and ecologically significant organisms, and how
they interacted in a more subtle manner with their surrounding ecosystems.
Page 65
64
6.0 Conclusions
Using a two-dimensional representation of the ruminant snout in ventral aspect, it is found
in this study there was no relation between snout shape and feeding strategy within a
highly diverse sample of the major ruminant clades (i.e., rejection of the primary hypothesis
and acceptance of both the null and alternative hypotheses, conforming largely to Pérez-
Barbería and Gordon, 1999). The lack of between-group discrimination was statistically
significant in most cases as assessed by a likelihood ratio test. These results are not
believed to be a phylogenetic artefact, contra Raia et al. (2010) among others. It is also
found that previously used categorisations relating snout shape to feeding strategy are
inadequate to in their descriptions of the full range of exhibited morphological variation
(i.e., ‘browsers’ do not specifically have ‘pointed’ snouts, and ‘browsers’ do not have ‘blunt’
snouts as previously postulated). The geometric complexity is more extensive than this, and
forms a continuum of shape variation (figs 13, 28 and 29). These results suggest that
attempts to place arbitrary thresholds on other related factors involved in feeding are
problematic and should be discontinued (following the recommendations of Gordon and
Illius, 1994). For example, attempting to classify ruminants based on various subjective
thresholds for their percentage of grass consumption seems highly suspect (see
Introduction). In light of this analysis, results obtained by Janis et al. (2010) - that intake
rate forces covariation in the anatomical strength of the mandible - should be reanalysed to
determine whether grazing ruminants have a more robust masticatory apparatus than
browsing ruminants. It is proposed here, in an manner analogous to that of Codron et al.
(2007), that ruminant diets represent a continuum with variation explicitly occurring on a
spatiotemporal scale for all feeding strategies, with individual strategies varying in a
homoplastic manner that is inconsistent with phylogenetic relation.
The fact that feeding strategies were not demonstrated to be associated with snout shape
in this investigation should be a point of concern for future ecological studies involving
ruminants. Accordingly, it is recommended snout shape should be subjected to further
study to determine whether it constitutes a morphological feature capable of delimiting
browsing and grazing ruminants (e.g., Janis and Ehrhardt, 1998). Reconstruction of
Page 66
65
palaeodiets based on snout shape should similarly be questioned (Solounias et al., 1988;
Solounias and Moelleken, 1993), given that this appears to be an inadequate proxy for
extant ruminants. It is conceivable that the results here a product of the lack of consistency
in defining functional feeding groups for ruminants with respect to other
morphophysiological traits. The functional significance of snout shape in both intake rate
and selectivity is not questioned. However, the results of this study imply that closer
inspection of this relation is required, with intake rate and selectivity being related to snout
shape within an analytically robust statistical framework.
Following the ecological classifications proposed in Whitlock (2011) regarding diplodocoid
sauropods, the results for the ornithopod data set reveal stronger patterns relating snout
morphology to feeding style than those found in ruminants. Hypsilophodont- and
heterodontosaur-grade species are largely consistent with a ‘pointed’ snout geometry,
reflecting a selective diet, close to ground-level given their typical size. Patterns in several
ornithopod clades (e.g., Dryosauridae, Camptosauridae) are difficult to discern due to low
sampling frequency. Iguanodonts appear to have gone through a transitional and
experimental phase, with a variety of snout morphologies reflecting adaptation to
numerous feeding roles. Hadrosaurs took on the role of non-selective ground-level
‘grazers’, reflected in their widened and distally flattened snout morphologies, an assertion
corroborated by analysis of tooth microwear in hadrosaurs (Williams et al., 2009). Previous
assertions that hadrosaurs and ruminants are ecological parallels are challenged based on
the disparate snout shapes exhibited by the two groups (Fig. 30). However, there are two
issues with these results. It is likely that taphonomic deformation anisotropically modified
the shape of several of the specimens (Fig. 36), creating lateral distortions in the snout
profile shape. It may be feasible to remove these potentially confounding artefacts to
reconstruct the true specimen profile for future analyses. Also, little is known about the
ontogenetic stage of majority of the specimens. If premaxilla shape varies in an allometric
manner, this could account for some aspects of the patterns observed here. Resolution of
this issue will require careful consideration in future analyses of morphological shape
variation. Nevertheless, it appears in the dataset investigated here that genera were
Page 67
66
taxonomically well constrained. This suggests ontogeny is not affecting the results to a
significant degree (except perhaps in the overlapping hadrosaur-iguanodont shape
domain).
Furthermore, the phylogenetic and ecological turnover between iguanodontians and
hadrosaurians may be associated with a secondary burst of angiosperm diversification
during the Upper Cretaceous (approximately Santonian age). The theoretical backing for
this putative linkage is that it is believed angiosperms flourished initially in a highly
disturbed environment, with which the transition in feeding strategy for hadrosaurs may be
associated (i.e., hadrosaurs were megaherbivores with a low-selectivity and high intake rate
to maintain their increasing body size). This secondary hypothesis requires an empirical
synthesis between previous studies (e.g., Butler and Barrett, 2008; Butler et al., 2009) to
determine whether this trait in hadrosaurs can be correlated within both environmental
and spatiotemporal parameters.
Following this study, future investigations could follow many routes to answer a number of
important ecological questions regarding the evolutionary histories of both ruminants and
ornithopods. Snout shape was found to be a poor basis for discriminating between feeding
strategies in ruminants. Subsequently, study is needed regarding the extent to which snout
shape can be associated with other digestive and ingestive parameters, and to other
morphophysiological factors. If other parameters are found to be related to feeding
strategy, the issue then becomes whether these factors covary equally, or force
modification upon each other?
The phylogenetic comparisons drawn in this study were predominantly qualitative. What is
undoubtedly required is the dissection of recovered signals to determine what proportion
of trait covariation can be explained by phylogenetic relationships (i.e., morphological
similarity based on common ancestry; e.g., Figuerido et al., 2010; Raia et al., 2010).
Applicable methods include the phylogenetic GLS (general least squares) model, which has
gained increasing interest in the fields of macroecology and macroevolution (e.g., Cardillo
et al., 2010). Furthermore, if singular or multiple functional traits are found to be
Page 68
67
phylogenetic artefacts, it could be possible to track the sequence of acquisition, and
therefore trace the functional and ecological histories of ruminants. In addition to
phylogenetic correlation, other factors such as ontogeny and sexual dimorphism should be
scrutinised within a statistical framework to detect potential allometric variation, and
possible synchronisation of trait acquisition and evolution patterns between sexes.
Page 69
68
7.0 Acknowledgements
First and foremost, I would like to extend my gratitude to Norman MacLeod (Natural
History Museum, London), for helping me develop and improve this project from day one,
constantly providing technical assistance on many of the more difficult aspects of geometric
morphometrics, and also providing the initial ornithopod profile outlines, self-made
analytical software, and a laboratory with which to work in. I am also indebted to Roberto
Portela Miguez (Natural History Museum, London), for allowing me access to specimens on
numerous occasions, in spite of my terrible organisational skills. For additional ornithopod
material, I would also like to extend my appreciations to Clint Boyd (University of Texas),
Andrew McDonald (University of Pennsylvania), and Richard Butler (Bayerische
Staatssammlung für Paläontologie und Geologie). Finally, I would like to thank Marcus
Clauss (University of Zurich) for providing several difficult to obtain documents used in this
study.
Page 70
69
8.0 References
Adams, D. C., Rohlf, F. J. and Slice, D. E. (2004) Geometric morphometrics: ten years of
progress following the ‘revolution’, Italian Journal of Zoology, 71, 5-16
Barden, H. E. and Maidment, S. C. R. (2011) Evidence for sexual dimorphism in the
stegosaurian dinosaur Kentrosaurus aethiopicus from the Upper Triassic of Tanzania,
Journal of Vertebrate Palaeontology, 31(3), 641-651
Barrett, P. M. (2000) Prosauropod dinosaurs and iguanas: speculations on the diet of extinct
reptiles, In: (Sues, H. D. (ed.), Evolution of Herbivory in Terrestrial Vertebrates: Perspectives
from the Fossil Record, Cambridge University Press, Massachusetts, pp. 42-78
Barrett, P. M. and Rayfield, E. J. (2006) Ecological and evolutionary implications of dinosaur
feeding behaviour, Trends in Ecology and Evolution, 21(4), 217-224
Barrett, P. M. and Han, F.-L. (2009) Cranial anatomy of Jeholosaurus shangyuanensis
(Dinosauria: Ornithischia) from the Early Cretaceous of China, Zootaxa, 2072, 31-55
Bell, P. R., Snively, E. and Shychoski, L. (2009) A comparison of the jaw mechanics in
hadrosaurid and ceratopsid dinosaurs using finite element analysis, The Anatomical Record,
292, 1338-1351
Brill, K. and Carpenter, K. (2007) A description of a new ornithopod from the Lytle Member
of the Purgatoire Formation (Lower Cretaceous) and a reassessment of the skull of
Camptosaurus, In: Carpenter, K. (ed.), Horns and Beaks: Ceratopsian Ornithopod Dinosaurs,
Indiana University Press, Bloomington, pp. 49-67
Bodmer, R. E. (1990) Ungulate frugivores and the browser-grazer continuum, Oikos, 57,
319-325
Bourliére, F. and Hadley, M. (1970) The ecology of tropical savannahs, Annual Review of
Ecology and Systematics, 1, 125-152
Page 71
70
Boyd, C. A., Brown, C. M., Scheetz, R. D. and Clarke, J. A. (2009) Taxonomic revision of the
basal neornithischian taxa Thescelosaurus and Bugenasaura, Journal of Vertebrate
Palaeontology, 29(3), 758-770
Butler, R. J. and Barrett, P. M. (2008) Palaeoenvironmental controls on the distribution of
Cretaceous herbivorous dinosaurs, Naturwissenschaften, 95, 1027-1032
Butler, R. J., Upchurch, P. and Norman, D. B. (2008) The phylogeny of the Ornithischian
dinosaurs, Journal of Systematic Palaeontology, 6(1), 1-40
Butler, R. J., Barrett, P. M., Kenrick, P. and Penn, M. G. (2009) Diversity patterns amongst
herbivorous dinosaurs and plants during the Cretaceous: implications for hypotheses of
dinosaur/angiosperm evolution, Journal of Evolutionary Biology, 22, 446–459
Campbell, N. A. and Atchley, W. R. (1981) The geometry of canonical variates analysis,
Systematic Zoology, 30(3), 268-280
Cardillo, M., Gittleman, J. L. and Purvis, A. (2008) Global patterns in the phylogenetic
structure of island mammal assemblages, Proceedings of the Royal Society B, 275, 1549-
1556
Carrano, M. T., Janis, C. M. and Sepkoski Jr., J. J. (1999) Hadrosaurs as ungulate parallels:
lost lifestyles and deficient data, Acta Palaeontologica Polonica, 44(3), 237-261
Chapman, R. E., Weishampel, D. B., Hunt, G. and Rasskin-Gutman D. (1997) Sexual
dimorphism in dinosaurs, In: Wolberg, D. L., Stump, E. and Rosenberg, G. (eds.), Dinofest
International, Proceedings of a Symposium Sponsored by Arizona State University, 83–93,
Philadelphia, Philadelphia Academy of Natural Sciences
Chin, K. (2007) The palaeobiological implications of herbivorous dinosaur coprolites from
the Upper Cretaceous Two Medicine formation of Montana: why eat wood?, Palaios, 22,
554-566
Page 72
71
Chin, K., Eberth, D. A., Schweitzer, M. H., Rando, T. A., Sloboda, W. J. and Horner, J. H.
(2003) Remarkable preservation of undigested tissue within a Late Cretaceous
tyrannosaurid coprolite from Alberta, Canada, Palaios, 18, 286-294
Clauss, M., Lechner-Doll, M. and Streich, W. J. (2002) Faecal particle size distribution in
captive wild ruminants: an approach to the browser/grazer dichotomy from the other end,
Oecologia, 131, 343-349
Clauss, M., Lechner-Doll, M. and Streich, W. J. (2003) Ruminant diversification as an
adaptation to the physiochemical characteristics of forage. A re-evaluation of an old debate
and a new hypothesis, Oikos, 102, 532-562
Clauss, M. and Hummel, J. (2005) The digestive performance of mammalian herbivores:
why big may not be that much better, Mammal Review, 35, 174-187
Clauss, M., Schwarm, A., Ortmann, S., Streich, W. J. and Hummel, J. (2007) A case of non-
scaling in mammalian physiology? Body size, digestive capacity, food intake, and ingesta
passage in mammalian herbivores, Comparative Biochemistry and Physiology, 148, 249-265
Clauss, M., Hofmann, R. R., Streich, W. J., Fickel, J. and Hummel, J. (2008a) Higher masseter
muscle mass in grazing than browsing ruminants, Oecologia, 157, 377-385
Clauss, M., Kaiser, T. and Hummel, J. (2008b) The morphophysiological adaptations of
browsing and grazing mammals, In: Gordon, I. J. et. al. (eds.), The ecology of browsing and
grazing, Berlin, pp. 47-88
Clauss, M., Fritz, J., Bayer, D., Nygren, K., Hammer, S., Hatt, J.-M., Südekum, K.-H. and
Hummel, J. (2009a) Physical characteristics of rumen contents in four large ruminants of
different feeding type, the addax (Addax nasomaculatus), bison (Bison bison), red deer
(Cervus elaphus) and moose (Alces alces), Comparative Biochemistry and Physiology Part A,
152, 398-406
Page 73
72
Clauss, M., Hofmann, R. R., Fickel, J., Streich, W. J. and Hummel, J. (2009b) The intraruminal
papillation of different feeding types: implications for rumen physiology, Journal of
Morphology, 270, 929-942
Clauss, M., Hofmann, R. R., Streich, W. J., Fickel, J. and Hummel, J. (2009c) Convergence in
the macroscopic anatomy of the reticulum in wild ruminant species of different feeding
types and a new resulting hypothesis on reticular function, Journal of Zoology, 281, 26-38
Clauss, M., Hume, I. D. and Hummel, J. (2010) Evolutionary adaptations of ruminants and
their potential relevance for modern production systems, Animal, 4(7), 979-992
Codron, D., Lee-Thorp, J. A., Sponheimer, M. and Codron, J. (2007) Nutritional content of
savannah plant foods: implications for browser/grazer models of ungulate diversification,
European Journal of Wildlife Research, 53, 100-111
Codron, D. and Clauss, M. (2010) Rumen physiology constrains diet niche: linking digestive
physiology and food selection across wild ruminant species, Canadian Journal of Zoology,
88, 1129-1138
Cooper. N and Purvis, A. (2010) Body size evolution in mammals: complexity in tempo and
mode, The American Naturalist, 175(6), 727-738
Demment, M. W. and Van Soest, P. J. (1985) A nutritional explanation for the body-size
patterns of ruminant and nonruminant herbivores, The American Naturalist, 125(5), 641-
672
Dodson, P. (1975) Taxonomic implications of relative growth in lambeosaurine hadrosaurs,
Systematic Zoology, 24, 37-54
Du Toit, J. T. (2003) Large herbivores and savannah heterogeneity, In: du Toit et al. (eds.),
The Kruger Experience, Island Press, pp. 292-309
Ezard, T. H. G., Pearson,. P. N. and Purvis, A. (2010) Algorithmic approaches to aid species’
delimitation in multidimensional morphospace, BMC Evolutionary Biology, 10, 1-11
Page 74
73
Farlow, J. O. (1987) Speculations about the diet and digestive physiology of herbivorous
dinosaurs, Palaeobiology, 13(1), 6-72
Fernández, M. H. and Vrba, E. S. (2005) A complete estimate of the phylogenetic
relationships in Ruminantia: a dated species-level supertree of the extant ruminants,
Biological Reviews, 80, 269-302
Fiorillo, A. R. (1998) Dental microwear patterns of the sauropod dinosaurs Camarasaurus
and Diplodocus: evidence for resource partitioning in the Late Jurassic of North America,
Historical Biology, 13, 1-16
Fletcher, T. M., Janis, C. M. and Rayfield, E. J. (2010) Finite element analysis of ungulate
jaws: can mode of digestive physiology be determined?, Palaeontologica Electronica, 13(3),
1-15
Figuerido, B., Serrano-Alarcón., Slater, G. C. and Palmqvist, P. (2010) Shape at the
crossroads: homoplasy and history in the evolution of the carnivoran skull towards
herbivory, Journal of Evolutionary Biology, 23, 2579-2594
Fritz, J., Hummel, J., Kienzle, E., Arnold, C., Nunn, C. and Clauss, M. (2009) Comparative
chewing efficiency in mammalian herbivores, Oikos, 118, 1623-1632
Fryxell, J. M. (1991) Forage quality and aggregation by large herbivores, The American
Naturalist, 138(2), 478-498
Galis, F. (1996) The application of functional morphology to evolutionary studies, Trends in
Ecology and Evolution, 11(3), 124-129
Galton P.M. (1974) The Ornithischian dinosaur Hypsilophodon from the Wealden of the Isle
of Wight, Bulletin of the British Museum (Natural History) Geology, 25, 1–152c
Galton, P. M. (1983) The cranial anatomy of Dryosaurus, a hypsilophodontid dinosaur from
the Upper Jurassic of North America and East Africa, with a review of hypsilophodontids
from the Upper Jurassic of North America, Geologica et Palaeontologica, 17, 207-243
Page 75
74
Galton, P. M. (1984) Cranial anatomy of the prosauropod dinosaur Plateosaurus from the
Knollenmergel (Middle Keuper, Upper Triassic) of Germany, I. Two complete skulls from the
Trossingen/Württemberg with comments on the diet, Geologica et Palaeontologica, 18,
139-171
Gordon, I. J. (2003) Browsing and grazing ruminants: are they different beasts?, Forest
Ecology and Management, 181, 13-21
Gordon, I. J. and Illius, A. W. (1988) Incisor arcade structure and diet selection in ruminants,
Functional Ecology, 2, 15-22
Gordon, I. J. and Illius, A. W. (1994) The functional significance of the browser-grazer
dichotomy in African ruminants, Oecologia, 98, 167-175
Gordon, I. J. and Illius, A. W. (1996) The nutritional ecology of African ruminants: a
reinterpretation, Journal of Animal Ecology, 65, 18-28
Gordon, I. J., Illius, A. W. and Milne, D. (1996) Sources of variation in the foraging efficiency
of grazing ruminants, Functional Ecology, 10, 219-226
Goswami, A., Flynn, J. J., Ranivoharimanana, L. and Wyss, A. R. (2005) Dental microwear in
Triassic amniotes: implications for paleoecology and masticatory mechanics, Journal of
Vertebrate Palaeontology, 25(2), 320-329
Goswami, A., Milne, N. and Wroe, S. (2010) Biting through constraints: cranial morphology,
disparity and convergence across living and fossil carnivorous mammals, Proceedings of the
Royal Society B, doi:10.1098/rspb.2010.2031
Gwynne, M. D. and Bell, R. H. V. (1968) Selection of vegetation components by grazing
ungulates in the Serengeti National Park, Nature, 220, 390-393
Haas, G. (1963) A proposed reconstruction of the jaw musculature of Diplodocus, Annals of
the Carnegie Museum, 36, 139-171
Page 76
75
Hadley, C., Milne, N. and Schmitt, L. H. (2009) A three-dimensional geometric
morphometric analysis of variation in cranial size and shape in tammar wallaby (Macropus
eugenii) populations, Australian Journal of Zoology, 57, 337-345
Herring, S. W. (1985) The ontogeny of mammalian mastication, American Zoology, 25, 339-
349
Hoculi, P. A., Heimhofer, U. and Weissert, H. (2006) Timing of early angiosperm radiation:
recalibrating the classical succession, Journal of the Geological Society, 163, 587-594
Hofmann, R. R. (1973) The Ruminant Stomach, Stomach Structure and Feeding Habits of
East African Game Ruminants, East African Monographs in Biology, Volume II, East African
Literature Bureau, Nairobi, Kenya
Hofmann, R. R. (1988) Morphophysiological evolutionary adaptations of the ruminant
digestive system, In: Dobson, A. and Dobson, M. J. (eds.), Aspects of Digestive Physiology in
Ruminants, Cornell University Press, Ithaca, New York, pp. 1-19
Hofmann, R. R. (1989) Evolutionary steps of ecophysiological adaptation and diversification
of ruminants: a comparative view of their digestive system, Oecologia, 78, 443-457
Hofmann, R. R. (1991) Endangered tropical herbivores – their nutritional requirements and
habitat demands, In: Ho, Y. W., Wong, H. K., Abdullah, N. and Tajuddin, Z. A. (eds.), Recent
Advances on The Nutrition of Herbivores, Malaysia Society of Animal Production, UPM
Serdang, pp. 27-34
Hofmann, R. R. (1999) Functional and comparative digestive system anatomy of Arctic
ungulates, Rangifer, 20, 71-81
Hofmann, R. R. and Stewart, D. R. M. (1972) Grazer or browser: a classification based on the
stomach-structure and feeding habits of East African ruminants, Extrait de Mammalia,
36(2), 226-240
Page 77
76
Hyde, B. (2010) The stomach contents of Lower Jurassic ichthyosaurs from the South Coast
of England; their taphonomic significance, Unpublished MEarthSci Thesis, University of
Manchester, UK
Janis, C. M. (1990) Correlation of cranial and dental variables in mammals: a comparison of
macropodoids and ungulates, Memoirs of Queensland Museum, 281, 349-366
Janis, C. M. (1995) Correlations between craniodental morphology and feeding behaviour in
ungulates: reciprocal illumination between living and fossil taxa, In: Thomason, J. J. (ed.),
Functional Morphology in Vertebrate Palaeontology, Cambridge University Press, New York,
pp. 76-98
Janis, C. M. and Ehrhardt, D. (1988) Correlation of relative muzzle width and relative incisor
width with dietary preference in ungulates, Zoological Journal of the Linnean Society, 92,
267-284
Janis, C. M., Damuth, J. and Theodor, J. M. (2000) Miocene ungulates and the terrestrial
primary productivity: where have all the browsers gone?, Proceedings of the National
Academy of Science, 97(14), 7899-7904
Janis, C. M., Constable, E. C., Houpt, K. A., Streich, W. J. and Clauss, M. (2010) Comparative
ingestive mastication in domestic horses and cattle: a pilot investigation, Journal of Animal
Physiology and Animal Nutrition, 94, 402-409
Jarman, P. J. (1974) The social organisation of antelope in relation to their ecology,
Behaviour, 48, 215-267
Jones, K. E. and Goswami, A. (2010) Morphometric analysis of cranial morphology in
pinnipeds (Mammalia, Carnivora): convergence, ecology, ontogeny and dimorphism, In:
Goswami, A. and Friscia, A. (eds.), Carnivoran Evolution: New Views on Phylogeny, Form,
and Function, Cambridge University Press, pp. 342-373
Kar, R. K., Sharma, N. and Kar, R. (2000) Occurrence of fossil fungi in dinosaur ding and its
implication on food habit, Current Science, 87, 1053-1056
Page 78
77
Klingenberg, C. P. (2010) Evolution and development of shape: integrating quantitative
approaches, Nature Reviews: Genetics, 11, 623-635
Kobayashi, Y. and Azuma, Y. (2003) A new iguanodontian (Dinosauria: Ornithopoda) from
the Lower Cretaceous Kitadani Formation of Fukui Prefecture, Japan, Journal of Vertebrate
Palaeontology, 23(1), 166-175
Krauss, D. (2001) An analysis of feeding habits of herbivorous dinosaurs through the
examination of phytoliths trapped on tooth grinding surfaces, Journal of Vertebrate
Palaeontology, 21, 69A
Leckey, E. H. (2004) Co-evolution in herbivorous dinosaurs and land plants: the use of fossil
locality data in an examination of spatial co-occurrence, Unpublished MSc Thesis, University
of California, Santa Barbara, USA
Liyong, J., Jun, C., Shuqin, Z., Butler, R. J. and Godefroit, P. (2010) Cranial anatomy of the
small ornithischian dinosaur Changchunsaurus parvus from the Quantou Formation
(Cretaceous: Aptian-Cenomanian) of Kilin Province, northeastern China, Journal of
Vertebrate Palaeontology, 30(1), 196-214
Lohmann, G. P. and Schweitzer, P.N. (1990) On eigenshape analysis, In: Rohlf, F. J. and
Bookstein, F. L. (eds.), Proceedings of the Michigan Morphometrics Workshop, Ann Arbor:
University of Michigan Museum of Zoology, pp. 147–166.
MacFadden, B. J. (2000) Cenozoic mammalian herbivores from the Americas:
reconstruction ancient diets and terrestrial communities, Annual Reviews of Ecology and
Systematics, 31, 33-59
MacLeod, N. (2005) Principal components analysis (eigenanalysis & regression 5),
Palaeontological Association Newsletter, 59, 44-57
MacLeod, N. (2007a) Groups I, Palaeontological Association Newsletter, 64, 35-45
MacLeod, N. (2007b) Groups II, Palaeontological Association Newsletter, 65, 36-49
Page 79
78
MacLeod, N. (2009) Form and shape models, Palaeontological Association Newsletter, 72,
14-27
Makovicky, P. J., Kilbourne, B. M. Sadleir, R. W. and Norell, M. A. (2011) A new basal
ornithopod (Dinosauria, Ornithischia) from the Late Cretaceous of Mongolia, Journal of
Vertebrate Palaeontology, 31(3), 626-640
Manchester, S. R., Lehman, T. M. and Wheeler, E. A. (2010) Fossil palms (Arecaceae,
Coryphoideae) associated with juvenile herbivorous dinosaurs in the Upper Cretaceous
Aguja Formation, Big Bend National Park, Texas, International Journal of Plant Sciences,
171(6), 679-689
McDonald, A. T. (2011) The taxonomy of species assigned to Camptosaurus, (Dinosauria:
Ornithopoda), Zootaxa, 2783, 52-68
Meijaard, E. and Groves, C. P. (2004) A taxonomic revision of the Tragulus mouse-deer
(Artiodactyla), Zoological Journal of the Linnean Society, 140, 63-102
Mendoza, M., Janis, C. M. and Palmqvist, P. (2002) Characterizing complex craniodental
patterns related to feeding behaviour in ungulates: a multivariate approach, Journal of the
Zoological Society of London, 258, 223-246
Mendoza, M. and Palmqvist, P. (2006) Characterising adaptive morphological patterns
related to diet in Bovidae (Mammalia: Artiodactyla), Acta Zoologica Sinica, 52(6), 988-1008
Mitteroecker, P. and Gunz, P. (2009) Advances in geometric morphometrics, Evolutionary
Biology, 36, 235-247
Molnar, R. E. (1977) Analogies in the evolution of combat and display structures in
ornithopods and ungulates, Evolutionary Theory, 3, 165-190
Molnar, R. E. (1991) The cranial morphology of Tyrannosaurus rex, Palaeontographica
Abteilung, 217, 137-176
Page 80
79
Murray, M. G. and Brown, D. (1993) Niche separation of grazing ungulates in the Serengeti:
an experimental test, Journal of Animal Ecology, 62, 380-389
Norman, D. B. (1984) On the cranial morphology and evolution of ornithopod dinosaurs,
Symposium of the Zoological Society of London, 52, 521-547
Norman, D. B. and Weishampel, D. B. (1985) Ornithopod feeding mechanisms: their bearing
on the evolution of herbivory, The American Naturalist, 126(2), 151-164
Ostrom, J. H. (1961) Cranial morphology of the hadrosaurian dinosaurs of North America,
Bulletin of the American Museum of Natural History, 122, 33-186
Ostrom, J. H. (1964) A functional analysis of jaw mechanics in the dinosaur Triceratops,
Postilla, Peabody Museum of Natural History, Yale University, 88, 1-35
Ostrom, J. H. (1970) Stratigraphy and palaeontology of the Cloverly Formation (lower
Cretaceous) of the Bighorn Basin area, Wyoming and Montana, Bulletin of the Peabody
Museum of Natural History, 35, 234pp.
Owen-Smith, N. (1997) Distinctive features of the nutritional ecology of browsing versus
grazing ruminants, Proceedings of the First International Symposium on Physiology and
Ethology of Wild and Zoo Animals, 11, 176-191
Pérez-Barbería, F. J. and Gordon, I. J. (1988) Factors affecting food comminution during
chewing in ruminants: a review, Biological Journal of The Linnean Society, 63, 233-256
Pérez-Barbería, F. J. and Gordon, I. J. (1998) The influence of molar occlusal surface area on
the voluntary intake, digestion, chewing behaviour and diet selection of red deer (Cervus
elaphus), Journal of the Zoological Society of London, 245, 307-316
Pérez-Barbería, F. J. and Gordon, I. J. (1999) The functional relationship between feeding
type and jaw and cranial morphology in ungulates, Oecologia, 118, 157-165
Page 81
80
Pérez- Barbería, F. J. and Gordon, I. J. (2000) Differences in body mass and oral morphology
between the sexes in the Artiodactyla: evolutionary relationships with sexual segregation,
Evolutionary Ecology Research, 2, 667-684
Pérez- Barbería, F. J. and Gordon, I. J. (2001) Relationships between oral morphology and
feeding style in the Ungulata: a phylogenetically controlled evaluation, Proceedings of the
Royal Society of London B, 268, 1023-1032
Pérez- Barbería, F. J., Gordon, I. J. and Nores, C. (2001a) Evolutionary transitions among
feeding styles and habitats in ungulates, Evolutionary Biology Research, 3, 221-230
Pérez- Barbería, F. J., Gordon, I. J. and Illius, A. (2001b) Phylogenetic analysis of stomach
adaptation in digestive strategies in African ruminants, Oecologia, 129, 498-508
Pérez- Barbería, F. J., Elston, D. A., Gordon, I. J. and Illius, A. W. (2004) The evolution of
phylogenetic differences in the efficiency of digestion in ruminants, Proceedings of the
Royal Society of London B, 271, 1081-1090
Prieto-Márquez, A. (2010) Global phylogeny of Hadrosauridae (Dinosauria: Ornithopoda)
using parsimony and Bayesian methods, Zoological Journal of the Linnean Society, 159, 435-
502
Raia, P., Carotenuto, F., Meloro, C., Piras, P. and Pushkina, D. (2009) The shape of
contention: adaptation, history, and contingency in ungulate mandibles, Evolution, 64(5),
1489-1503
Royer, D. L., Miller, I. M., Peppe, D. J. and Hickey, L. J. (2010) Leaf economic traits from
fossils support a weedy habit for early angiosperms, American Journal of Botany, 97(3),
438-445
Robbins, T. C., Spalinger, D. E. and van Hoven, W. (1995) Adaptation of ruminants to browse
and grass diets: are anatomical-based browser-grazer interpretations valid?, Oecologia,
103, 208-213
Page 82
81
Rohlf, F. J. and Slice, D. (1990) Extensions of the Procrustes method for the optimal
superimposition of landmarks, Systematic Biology, 39(1), 40-59
Rybczynski, N., Tirabasso, A., Bloskie, P., Cuthbertson, R. and Hollidat, C. (2008) A three-
dimensional model of Edmontosaurus (Hadrosauridae) for testing chewing hypotheses,
Palaeontologica Electronica, 11(2), 1-14
Sanson, G. (2006) The biomechanics of browsing and grazing, American Journal of Botany,
93(10), 1531-1545
Satorra, A. and Saris, W. E. (1985) Power of the likelihood ratio test in covariance structure
analysis, Psychometrika, 50, 83-90
Sellers, W. I., Manning, P. L., Lyson, T., Stevens, K. and Margetts, L. (2009) Virtual
palaeontology: gait reconstruction of extinct vertebrates using high performance
computing, Palaeontologica Electronica, 12(3), 11A, 26p.
Semprebon, G. M. and Rivals, F. (2010) Trends in the paleodietary habits of fossil camels
from the Tertiary and Quaternary of North America, Palaeogeography, Palaeoclimatology,
Palaeoecology, 295, 131-145
Shipley, L. A. (1994) Grazers and browsers: how digestive morphology affects diet selection,
In: Grazing Behaviour of Livestock and Wildlife, eds. K. L. Launchbaugh, K. D. Sanders and J.
C. Mosley, Idaho Forest, Wildlife and Range Expeditions Station Bulletin, University of
Idaho, Moscow, pp. 20-27
Shipley, L. A., Gross, J. E., Spalinger, D. E., Hobbs, N. T. and Wunder, B. A. (1994) The scaling
of intake rate in mammalian herbivores, The American Naturalist, 143(6), 1055-1082
Solounias, N. and Dawson-Saunders, B. (1988) Dietary adaptations and paleoecology of the
Late Miocene ruminants from Pikermi and Samos in Greece, Palaeogeography,
Palaeoclimatology, Palaeoecology, 65, 149-172
Solounias, N. and Moelleken, S. M. C. (1993) Dietary adaptation of some extinct ruminants
determined by premaxillary shape, Journal of Mammalogy, 74, 1059-1071
Page 83
82
Solounias, N., Teaford, M. and Walker, A. (1988) Interpreting the diet of extinct ruminants:
the case of a non-browsing giraffid, Palaeobiology, 14(3), 287-300
Solounias, N., Moelleken, M. C. and Plavcan, J. M. (1995) Predicting the diet of extinct
bovids using masseteric morphology, Journal of Vertebrate Palaeontology, 15, 795-805
Sponheimer, M., Lee-Thorp, J. A., DeRuiter, D., Smith, J. M., Van der Merwe, N. J., Reed, K.,
Grant, C. C., Ayliffe, L. K., Robinson, T. F., Heidelberger, C. and Marcus, W. (2003) Diets of
Southern African Bovidae: stable isotopic evidence, Journal of Mammalogy, 84, 471-479
Tennant, J. P. (2010) Osteology of a near-complete specimen of Tenontosaurus tilletti
(Dinosauria: Ornithopoda) from the Cloverly Formation, Montana, USA, Unpublished
MEarthSci Thesis, University of Manchester, UK
Therrien, F., Henderson, D. M. and Ruff, C. B. (2006) Bite me: biomechanical models of
theropod mandibles and implications for feeding behaviour, In: Carpenter, K. (ed.), The
Carnivorous Dinosaurs, Indiana University Press, pp. 179-237
Tweet, J. S., Chin, K., Braman, D. R. and Murphy, N. (2008) Probable gut contents within a
specimen of Brachylophosaurus canadensis (Dinosauria: Hadrosauridae) from the Upper
Cretaceous Judith River Formation of Montana, Palaios, 23, 624-635
Upchurch, P. and Barrett, P. M. (2000) The evolution of sauropod feeding mechanisms, In:
Sues, H. D. (ed.), Evolution of Herbivory in Terrestrial Vertebrates: Perspectives from the
Fossil Record, Cambridge University Press, Massachusetts, pp. 79-122
Van Bocxlaer, B. and Schultheiß, R. (2010) Comparison of morphometric techniques for
shapes with homologous landmarks based on machine learning approaches to biological
discrimination, Palaeobiology, 36(3), 497-515
Van Zyl, J. H. M. (1965) The vegetation of S. A. Lombard Nature Reserve and its utilisation
by certain antelopes, Zoologica Africana, 1, 55-71
Varicchio, D. J. (2001) Gut contents from a Cretaceous tyrannosaurid: implications for
theropod digestive tracts, Journal of Palaeontology, 75, 401-406
Page 84
83
Voeten, M. M. and Prins, H. H. T. (1999) Resource partitioning between sympatric wild and
domestic herbivores in the Tarangire region of Tanzania, Oecologia, 120, 287-294
Walker, A. C. (1984) Extinction in hominid evolution, In: Nitecki, M. H. (ed.), Extinctions,
University of Chicago Press, Chicago, pp. 119-152
Weishampel, D. B. (1984) Evolution of jaw mechanisms in ornithopod dinosaurs, Advances
in Anatomy, Embryology and Cell Biology, 87, 1-109
Weishampel, D. B. (1985) An approach to jaw mechanics and diversity: the case of
ornithopod dinosaurs, In: Duncker, H. R. and Fliescher, G. (eds.), Vertebrate Morphology,
Gustav Fischer Verlag, Stuttgart, pp. 261-263
Weishampel, D. B. and Norman, D. B. (1989) Vertebrate herbivory in the Mesozoic: jaws,
plants, and evolutionary metrics, Geological Society of America Special Paper, 238, 87-100
Weishampel, D. B. and Jianu, C-M. (2000) Plant-eaters and ghost lineages: dinosaurian
herbivory revisited, In: Sues, H. D. (ed.), Evolution of Herbivory in Terrestrial Vertebrates:
Perspectives from the Fossil Record, Cambridge University Press, Massachusetts, 123-143
Weishampel, D. B., Norman, D. B. and Grigorescu, D. (1993) Telmatosaurus transsylvanicus
from the Lower Cretaceous of Romania: the most basal hadrosaurid dinosaur,
Palaeontology, 36(2), 361-385
Weishampel, D. B., Jianu, C-M., Csiki, Z. and Norman, D.B. (2003) Osteology and phylogeny
of Zalmoxes (N. G.), an unusual euornithopod dinosaur from the latest Cretaceous of
Romania, Journal of Systematic Palaeontology, 1(2), 65-123
Whitlock, J. A. (2011) Inferences of diplodocoid (Sauropoda: Dinosauria) feeding behaviour
from snout shape and microwear analyses, PLoS One, 6(4), 1-20
Williams, S. H. and Kay, R. F. (2001) A comparative test for adaptive explanations for
hypsodonty in ungulates and rodents, Journal of Mammalian Evolution, 8, 207-229
Page 85
84
Williams, V. S., Barrett, P. M. and Purnell, M. A. (2009) Quantitative analysis of dental
microwear in hadrosaurid dinosaurs, and the implications for hypotheses of jaw mechanics
and feeding, Proceedings of the National Academy of Sciences, 106(27), 11194-11199
Wilson, D. E. and Cole, F. R. (2000) Order Artiodactyla, In: Common Names of Mammals of
the World, Smithsonian Institution Press, pp. 92-100
Winkler, D. A., Murry, P. A. and Jacobs, L. L. (1997) A new species of Tenontosaurus
(Dinosauria: Ornithopoda) from the Early Cretaceous of Texas, Journal of Vertebrate
Palaeontology, 17(2), 330-348
Page 86
85
9.0 Appendices
Fam
ily
Subfa
mily
Genus
Speci
es
Sub-s
peci
es
Eco
logy
Cri
teri
on
Refe
rence
Shape I
Shape I
I
Ant
iloca
pridae
-A
nti
loca
pra
am
eric
ana
-In
term
edia
teU
nkno
wn
Cla
uss
et
al.
(2002)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Aep
ycer
otin
aeA
epyc
eros
mel
am
pus
-In
term
edia
teH
ypso
dont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Alc
elap
hina
eA
lcel
aphus
buse
laphus
majo
rG
raze
r>
75%
gra
sses
Gord
on
and I
llius
(1996)
Blu
ntB
lunt
-conv
ex
Bovi
dae
Alc
elap
hina
eB
eatr
agus
hunte
ri-
Gra
zer
Var
ious
Men
doza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conv
ex
Bovi
dae
Alc
elap
hina
eC
onnoch
aet
esgnou
-G
raze
r>
75%
gra
sses
Gord
on
and I
llius
(1996)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Alc
elap
hina
eC
onnoch
aet
esta
uri
nus
johnst
oni
Gra
zer
90%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Alc
elap
hina
eD
am
alisc
us
alb
ifro
ns
-G
raze
rG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Bovi
dae
Alc
elap
hina
eD
am
alisc
us
dorc
as
-G
raze
rV
ario
usM
endoza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-lin
ear
Bovi
dae
Alc
elap
hina
eD
am
alisc
us
korr
igum
-G
raze
r>
75%
gra
sses
Gord
on
and I
llius
(1996)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Alc
elap
hina
eD
am
alisc
us
liec
hte
nst
einii
-G
raze
rG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onv
ex
Bovi
dae
Alc
elap
hina
eD
am
alisc
us
lunatu
s-
Gra
zer
Hyp
sodont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conv
ex
Bovi
dae
Alc
elap
hina
eD
am
alisc
us
pyg
arg
us
-G
raze
r>
80%
gra
sses
Codro
n an
d C
laus
s (2
010)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeA
mm
odorc
as
clark
ei-
Bro
wse
rV
ario
usM
endoza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeA
nti
dorc
as
mars
upia
lis
anglo
ensi
sIn
term
edia
te30%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onc
ave
Bovi
dae
Ant
ilopin
aeA
nti
lope
cerv
icapra
-In
term
edia
teM
esodont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeD
orc
atr
agus
meg
alo
tis
-In
term
edia
teV
ario
usM
endoza
and
Pal
mqvi
st (
2006)
Poin
ted
Poin
ted-c
onc
ave
Bovi
dae
Ant
ilopin
aeE
udorc
as
thom
soni
-In
term
edia
teH
ypso
dont
yS
olo
unia
s et
al.
(2010)
Poin
ted
Poin
ted-c
onv
ex
Bovi
dae
Ant
ilopin
aeG
aze
lla
spek
ei-
Inte
rmed
iate
50%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
ben
net
tii
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
cuvie
ri-
Inte
rmed
iate
Gen
eric
affin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
dam
aru
fico
llis
Inte
rmed
iate
47.5
% g
rass
esC
laus
s et
al.
(2009c)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
dorc
as
Inte
rmed
iate
Var
ious
Men
doza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
gaze
lla
ara
bic
aIn
term
edia
teM
esodont
yS
olo
unia
s et
al.
(2010)
Poin
ted
Poin
ted-l
inea
r
Bovi
dae
Ant
ilopin
aeG
aze
lla
lepto
cero
s-
Inte
rmed
iate
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Poin
ted-l
inea
r
Bovi
dae
Ant
ilopin
aeG
aze
lla
rufi
frons
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
saudiy
a-
Inte
rmed
iate
Gen
eric
affin
ityIn
ferr
edB
lunt
Blu
nt-l
inea
r
Bovi
dae
Ant
ilopin
aeG
aze
lla
soem
mer
ingi
-In
term
edia
te50%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeG
aze
lla
subgutt
uro
sa-
Inte
rmed
iate
50%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-lin
ear
Bovi
dae
Ant
ilopin
aeL
itocr
aniu
sw
aller
i-
Bro
wse
r>
75%
bro
wse
Gord
on
and I
llius
(1996)
Poin
ted
Poin
ted-c
onc
ave
Bovi
dae
Ant
ilopin
aeM
adoqua
cord
eauxi
-B
row
ser
Conc
entr
ate
sele
ctor
Hoffm
an (
1989)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeM
adoqua
phillipsi
-B
row
ser
Conc
entr
ate
sele
ctor
Hoffm
an (
1989)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeM
adoqua
salt
iana
erla
nger
iB
row
ser
10%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Ant
ilopin
aeM
adoqua
swayn
eipia
centi
nii
Bro
wse
rC
onc
entr
ate
sele
ctor
Hoffm
an (
1989)
Blu
ntB
lunt
-conc
ave
Page 87
86
Fa
mily
Subfa
mily
Genus
Speci
es
Sub-s
peci
es
Eco
log
yC
rite
rio
nR
efe
rence
Sha
pe I
Sha
pe I
I
Bo
vid
aeA
ntilo
pin
aeN
an
ger
gra
nti
-In
term
edia
teH
ypso
dont
yS
olo
unia
s et
al.
(2
010
)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeA
ntilo
pin
aeN
eotr
ag
us
ba
tesi
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeA
ntilo
pin
aeN
eotr
ag
us
mo
sch
atu
s-
Bro
wse
r>
75
% b
row
seG
ord
on
and
Illi
us (
19
96
)B
lunt
Po
inte
d-c
onv
ex
Bo
vid
aeA
ntilo
pin
aeN
eotr
ag
us
pyg
ma
eus
-B
row
ser
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeA
ntilo
pin
aeO
reo
tra
gus
ore
otr
ag
us
-B
row
ser
5%
gra
sses
Cla
uss
et a
l. (
20
09
c)B
lunt
Po
inte
d-c
onv
ex
Bo
vid
aeA
ntilo
pin
aeO
ure
bia
ou
reb
i-
Inte
rmed
iate
Hyp
sod
ont
yS
olo
unia
s et
al.
(2
010
)P
oin
ted
Po
inte
d-c
onc
ave
Bo
vid
aeA
ntilo
pin
aeP
roca
pra
gu
ttu
rosa
-In
term
edia
te2
8%
gra
sses
Cla
uss
et a
l. (
20
09
c)B
lunt
Blu
nt-l
inea
r
Bo
vid
aeA
ntilo
pin
aeP
roca
pra
pic
tica
uda
ta-
Inte
rmed
iate
Gen
eric
affin
ityIn
ferr
edB
lunt
Blu
nt-l
inea
r
Bo
vid
aeA
ntilo
pin
aeP
roca
pra
prz
ewa
lskii
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-l
inea
r
Bo
vid
aeA
ntilo
pin
aeR
ap
hic
eru
sca
mp
estr
is-
Bro
wse
r>
75
% b
row
seG
ord
on
and
Illi
us (
19
96
)P
oin
ted
Po
inte
d-l
inea
r
Bo
vid
aeA
ntilo
pin
aeR
ap
hic
eru
sm
ela
noti
s-
Inte
rmed
iate
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)P
oin
ted
Po
inte
d-c
onc
ave
Bo
vid
aeA
ntilo
pin
aeR
ap
hic
eru
ssh
arp
eico
lon
icu
sIn
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edP
oin
ted
Po
inte
d-c
onc
ave
Bo
vid
aeA
ntilo
pin
aeR
hyn
cho
tra
gus
do
mo
ren
sis
va
ria
ni
Bro
wse
rC
onc
entr
ate
sele
cto
rH
offm
an (
19
89
)B
lunt
Blu
nt-c
onc
ave
Bo
vid
aeA
ntilo
pin
aeR
hyn
cho
tra
gus
kir
kii
min
or
Bro
wse
r>
75
% b
row
seG
ord
on
and
Illi
us (
19
96
)B
lunt
Blu
nt-c
onc
ave
Bo
vid
aeA
ntilo
pin
aeR
hyn
cho
tra
gus
gu
enth
eri
ha
dso
ni
Bro
wse
rV
ario
usM
end
oza
and
Pal
mq
vist
(2
006
)B
lunt
Blu
nt-c
onc
ave
Bo
vid
aeB
ovi
nae
Bo
sela
phu
str
ag
oca
mel
us
-In
term
edia
teM
eso
dont
yS
olo
unia
s et
al.
(2
010
)B
lunt
Blu
nt-l
inea
r
Bo
vid
aeB
ovi
nae
Ta
uro
tra
gus
ory
x-
Inte
rmed
iate
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)B
lunt
Blu
nt-l
inea
r
Bo
vid
aeB
ovi
nae
Tet
race
rus
qu
ad
rico
rnis
-G
raze
rH
ypso
dont
yS
olo
unia
s et
al.
(2
010
)P
oin
ted
Po
inte
d-c
onc
ave
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
sa
nga
si-
Inte
rmed
iate
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)B
lunt
Blu
nt-c
onv
ex
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
sb
uxto
ni
-B
row
ser
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)B
lunt
Blu
nt-l
inea
r
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
seu
ryce
rus
-B
row
ser
Bra
chyd
ont
yS
olo
unia
s et
al.
(2
010
)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
sim
ber
bis
-B
row
ser
Bra
chyd
ont
yS
olo
unia
s et
al.
(2
010
)B
lunt
Blu
nt-l
inea
r
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
ssc
rip
tus
-B
row
ser
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
ssp
ekii
-In
term
edia
teM
eso
dont
yS
olo
unia
s et
al.
(2
010
)B
lunt
Blu
nt-c
onv
ex
Bo
vid
aeB
ovi
nae
Tra
gel
ap
hu
sst
rep
tisc
ero
s-
Bro
wse
rV
ario
usM
end
oza
and
Pal
mq
vist
(2
006
)B
lunt
Blu
nt-c
onv
ex
Bo
vid
aeC
aprina
eP
seu
dois
na
yau
r-
Inte
rmed
iate
Var
ious
Men
do
za a
nd P
alm
qvi
st (
20
06
)B
lunt
Blu
nt-c
onc
ave
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sd
ors
alis
-B
row
ser
Bra
chyd
ont
yS
olo
unia
s et
al.
(2
010
)P
oin
ted
Po
inte
d-l
inea
r
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sh
arv
eyi
ign
ifer
Bro
wse
r1
% g
rass
esC
laus
s et
al.
(2
009
c)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sm
axw
elli
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Po
inte
d-l
inea
r
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sm
on
tico
lasc
hu
ltze
iF
rugi
vore
Unk
now
nC
laus
et
al.
(2
002
)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sn
ate
len
sis
na
tele
nsi
sB
row
ser
1%
gra
sses
Cla
uss
et a
l. (
20
09
c)B
lunt
Blu
nt-c
onc
ave
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sn
iger
-B
row
ser
Bra
chyd
ont
yS
olo
unia
s et
al.
(2
010
)P
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sn
igri
fro
ns
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Po
inte
d-c
onv
ex
Bo
vid
aeC
epha
lop
hina
eC
eph
alo
phu
sru
fila
tus
-B
row
ser
Co
ncen
trat
e se
lect
or
Ho
ffm
an (
19
89
)P
oin
ted
Po
inte
d-c
onv
ex
Page 88
87
Fam
ily
Subfa
mily
Genus
Speci
es
Sub-s
peci
es
Eco
logy
Cri
teri
on
Refe
rence
Shape I
Shape I
I
Bovi
dae
Cep
halo
phi
nae
Cep
halo
phus
silv
icult
or
rufi
cris
tus
Bro
wse
rG
ener
ic a
ffin
ityIn
ferr
edP
oin
ted
Poin
ted-c
onv
ex
Bovi
dae
Cep
halo
phi
nae
Cep
halo
phus
zebra
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Poin
ted-c
onc
ave
Bovi
dae
Cep
halo
phi
nae
Syl
vic
apra
gri
mm
ia-
Fru
givo
reV
ario
usM
endoza
and
Pal
mqvi
st (
2006)
Poin
ted
Poin
ted-l
inea
r
Bovi
dae
Hip
potr
agin
aeA
ddax
naso
macu
latu
s-
Gra
zer
80%
gra
ssC
laus
s et
al.
(2009c)
Poin
ted
Poin
ted-c
onv
ex
Bovi
dae
Hip
potr
agin
aeH
ippotr
agus
equin
us
-G
raze
rH
ypso
dont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Hip
potr
agin
aeH
ippotr
agus
nig
er-
Gra
zer
Var
ious
Men
doza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Hip
potr
agin
aeO
ryx
bei
sabei
saG
raze
r>
75%
gra
sses
Gord
on
and I
llius
(1996)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Hip
potr
agin
aeO
ryx
gaze
lla
-G
raze
r82%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Hip
potr
agin
aeO
ryx
leuco
ryx
-G
raze
rG
rass
/roug
hage
eat
ers
Hoffm
an (
1989)
Blu
ntB
lunt
-conv
ex
Bovi
dae
Red
unci
nae
Kobus
def
ass
a-
Gra
zer
Hyp
sodont
yS
olo
unia
s et
al.
(2010)
Poin
ted
Poin
ted-c
onc
ave
Bovi
dae
Red
unci
nae
Kobus
ellipsi
pry
mnus
-G
raze
rH
ypso
dont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conv
ex
Bovi
dae
Red
unci
nae
Kobus
kob
leuco
tis
Gra
zer
Var
ious
Men
doza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Red
unci
nae
Kobus
lech
e-
Gra
zer
Hyp
sodont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Red
unci
nae
Kobus
meg
ace
ros
-G
raze
rG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Bovi
dae
Red
unci
nae
Kobus
vard
onii
-G
raze
rV
ario
usM
endoza
and
Pal
mqvi
st (
2006)
Poin
ted
Poin
ted-c
onc
ave
Bovi
dae
Red
unci
nae
Pel
eaca
pre
olu
s-
Bro
wse
r7%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onc
ave
Bovi
dae
Red
unci
nae
Red
unca
aru
ndin
um
-F
rugi
vore
Var
ious
Men
doza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Red
unci
nae
Red
unca
fulv
uro
fula
-In
term
edia
teV
ario
usM
endoza
and
Pal
mqvi
st (
2006)
Blu
ntB
lunt
-conc
ave
Bovi
dae
Red
unci
nae
Red
unca
redunca
-G
raze
rH
ypso
dont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cap
reolin
aeA
lces
alc
es-
Bro
wse
r<
20%
gra
sses
Codro
n an
d C
laus
s (2
010)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cap
reolin
aeA
lces
pulm
atu
s-
Bro
wse
rC
onc
entr
ate
sele
ctor
Hoffm
an (
1989)
Bun
tB
lunt
-conc
ave
Cer
vidae
Cap
reolin
aeB
last
oce
ras
bez
oart
icus
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Cer
vidae
Cap
reolin
aeB
last
oce
ras
dic
hoto
mus
-In
term
edia
te24%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-lin
ear
Cer
vidae
Cap
reolin
aeC
apre
olu
sca
pre
olu
s-
Bro
wse
r<
20%
gra
sses
Codro
n an
d C
laus
s (2
010)
Poin
ted
Poin
ted-l
inea
r
Cer
vidae
Cap
reolin
aeH
ippoca
mel
us
anti
sensi
s-
Inte
rmed
iate
Gen
eric
affin
ityIn
ferr
edB
lunt
Blu
nt-c
onv
ex
Cer
vidae
Cap
reolin
aeM
aza
ma
am
eric
ana
-B
row
ser
1%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-lin
ear
Cer
vidae
Cap
reolin
aeM
aza
ma
gouazo
ubia
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Poin
ted-l
inea
r
Cer
vidae
Cap
reolin
aeO
doco
ileu
shem
ionus
-B
row
ser
Unk
now
nC
laus
s et
al.
(2002)
Blu
ntB
lunt
-lin
ear
Cer
vidae
Cap
reolin
aeO
doco
ileu
svir
gin
ianus
-B
row
ser
9%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conv
ex
Cer
vidae
Cap
reolin
aeO
zoto
cera
sbez
oart
icus
-In
term
edia
teU
nkno
wn
Cla
uss
et a
l. (
2002)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cap
reolin
aeP
udu
puda
-B
row
ser
3%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onv
ex
Cer
vidae
Cap
reolin
aeR
angif
erta
randus
-In
term
edia
te36%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onc
ave
Cer
vidae
Cer
vina
eA
xis
axis
-In
term
edia
te70%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onc
ave
Page 89
88
Fam
ily
Subfa
mily
Genus
Speci
es
Sub-s
peci
es
Eco
logy
Cri
teri
on
Refe
rence
Shape I
Shape I
I
Cer
vidae
Cer
vina
eC
ervus
axis
ceyl
onen
sis
Inte
rmed
iate
Gen
eric
affin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Cer
vidae
Cer
vina
eC
ervus
duvauce
li-
Inte
rmed
iate
Mes
odont
yS
olo
unia
s et
al.
(2010)
Poin
ted
Poin
ted-c
onc
ave
Cer
vidae
Cer
vina
eC
ervus
elaphus
-In
term
edia
te20-8
0%
gra
sses
Codro
n an
d C
laus
s (2
010)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cer
vina
eC
ervus
eldii
eldii
Inte
rmed
iate
Unk
now
nC
laus
s et
al.
(2002)
Poin
ted
Poin
ted-l
inea
r
Cer
vidae
Cer
vina
eC
ervus
kuhli
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Cer
vidae
Cer
vina
eC
ervus
nip
pon
-In
term
edia
te50%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cer
vina
eC
ervus
schom
burg
ki
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-c
onc
ave
Cer
vidae
Cer
vina
eC
ervus
tim
ore
nsi
s-
Inte
rmed
iate
Unk
now
nC
laus
s e
t al.
(2002)
Poin
ted
Poin
ted-c
onv
ex
Cer
vidae
Cer
vina
eC
ervus
unic
olo
rbro
okei
Inte
rmed
iate
Mes
odont
yS
olo
unia
s et
al.
(2010)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cer
vina
eD
am
adam
a-
Inte
rmed
iate
Unk
now
nF
letc
her
et a
l. (
2010)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Cer
vina
eE
laphuru
sdavid
ianus
-G
raze
r75%
gra
sses
Cla
uss
et a
l. (
2009c)
Blu
ntB
lunt
-conc
ave
Cer
vidae
Hyd
ropotin
aeH
ydro
podus
iner
mis
-In
term
edia
te50%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onv
ex
Cer
vidae
Mun
tiaci
nae
Ela
phodus
cephalo
phus
-B
row
ser
Unk
now
nC
laus
s et
al.
(2002)
Poin
ted
Poin
ted-c
onv
ex
Cer
vidae
Mun
tiaci
nae
Munti
acu
scr
inif
rons
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Poin
ted-l
inea
r
Cer
vidae
Mun
tiaci
nae
Munti
acu
sm
untj
ak
vagin
alis
Bro
wse
r10%
gra
sses
Cla
uss
et a
l. (
2009c)
Poin
ted
Poin
ted-c
onv
ex
Cer
vidae
Mun
tiaci
nae
Munti
acu
sre
eves
ire
eves
iB
row
ser
Gen
eric
affin
ityIn
ferr
edB
lunt
Blu
nt-l
inea
r
Giraf
fidae
-G
iraff
aca
mel
opard
alis
-B
row
ser
>75%
bro
wse
Gord
on
and I
llius
(1996)
Poin
ted
Poin
ted-c
onc
ave
Mosc
hidae
-M
osc
hus
mosc
hif
erous
-In
term
edia
teG
ener
ic a
ffin
ityIn
ferr
edB
lunt
Blu
nt-l
inea
r
Tra
gulid
ae-
Tra
gulu
sja
vanic
us
insu
lari
sB
row
ser
Unk
now
nF
letc
her
et a
l. (
2010)
Poin
ted
Poin
ted-c
onv
ex
Tra
gulid
ae-
Tra
gulu
skanch
ilum
bri
nus
Bro
wse
rG
ener
ic a
ffin
ityIn
ferr
edP
oin
ted
Poin
ted-c
onv
ex
Tra
gulid
ae-
Tra
gulu
snapu
-B
row
ser
Gen
eric
affin
ityIn
ferr
edP
oin
ted
Poin
ted-l
inea
r
Ap
pen
dix
1 –
Ru
min
ant
sam
ple
lis
t fo
r th
is s
tud
y. T
axo
no
mic
det
ails
wer
e ex
trac
ted
fro
m W
ilso
n a
nd
Co
le (
20
00
). T
he
eco
logi
cal
det
ails
we
re e
xtra
cted
fro
m a
ran
ge o
f so
urc
es;
gen
eric
aff
init
y as
a c
rite
rio
n s
ign
ifie
s as
sign
men
t to
a p
arti
cula
r ec
olo
gy b
ased
up
on
det
ails
re
gard
ing
spec
ies
wit
hin
th
e sa
me
gen
us.
Ass
ign
ed
shap
e gr
ou
pin
gs a
re d
iscu
sse
d in
th
e te
xt.
Page 90
89
Speci
men I
DT
axonom
ic G
roup
Genus
Speci
es
Refe
rence
Age
Shape I
Shape I
I
YP
M 1
880
Cam
pto
saur
idae
Cam
pto
sauru
sdis
par
A. M
cDona
ldU
pper
Jur
assi
cB
lunt
Blu
nt-l
inea
r
CM
3392
Dry
osa
urid
aeD
ryosa
uru
s alt
us
A. M
cDona
ldU
pper
Jur
assi
cB
lunt
Blu
nt-c
onc
ave
N/A
Dry
osa
urid
aeD
ryosa
uru
s le
ttow
vorb
ecki
Gal
ton
(1983)
Upper
Jur
assi
cB
lunt
Blu
nt-c
onc
ave
AM
NH
5730
Had
rosa
urid
aeA
nato
tita
nco
pei
N. M
acle
od; A
. M
cDona
ldU
pper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 8
893
Had
rosa
urid
aeB
rach
ylophosa
uru
sca
naden
sis
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
AM
NH
5338
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
AM
NH
5382
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
CM
11375
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 8
433
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
RO
M 1
933
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
RO
M 1
947
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
RO
M 7
59
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
AM
NH
5359
Had
rosa
urid
aeC
ory
thosa
uru
sca
suari
us
N. M
acle
od; A
. M
cDona
ldU
pper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
AM
NH
5060
Had
rosa
urid
aeE
dm
onto
sauru
sre
galis
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
Chi
cago
I15
Had
rosa
urid
aeE
dm
onto
sauru
sre
galis
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 2
288
Had
rosa
urid
aeE
dm
onto
sauru
s re
galis
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NS
R4036
Had
rosa
urid
aeE
dm
onto
sauru
s re
galis
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
AM
NH
5465
Had
rosa
urid
aeG
ryposa
uru
s la
tiden
sN
. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 2
278
Had
rosa
urid
aeP
rosa
uro
lophus
nota
bilis
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 2
246
Had
rosa
urid
aeH
ypacr
osa
uru
salt
ispin
us
N. M
acle
od
Upper
Cre
tace
ous
Poin
ted
Poin
ted-c
onc
ave
NM
C 2
247
Had
rosa
urid
aeH
ypacr
osa
uru
salt
ispin
us
N. M
acle
od
Upper
Cre
tace
ous
Poin
ted
Blu
nt-c
onc
ave
RO
M 7
82
Had
rosa
urid
aeL
am
beo
sauri
ne
indet
.N
. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
US
NM
16600
Had
rosa
urid
aeL
am
beo
sauri
ne
indet
.N
. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
AM
NH
5340
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 8
703
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
P67 9
166
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-lin
ear
RO
M 1
218
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Poin
ted
Poin
ted-c
onc
ave
RO
M 7
58
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
RO
M 7
94
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
TM
P P
78.1
6.1
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
US
NM
IV
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
YP
M 3
222
Had
rosa
urid
aeL
am
beo
sauru
sla
mbei
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
Page 91
90
Appendix 2 – Ornithopod sample list for this study (including ‘Heterodontosauridae’). Names refer
to people who personally provided images for analysis. Taxonomic groups are discussed in the
text. Shape categories were assigned in the same manner as for the ruminants.
Speci
men I
DT
axonom
ic G
roup
Genus
Speci
es
Refe
rence
Age
Shape I
Shape I
I
PU
22405
Had
rosa
urid
aeM
aia
saura
pee
ble
soru
mN
. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
MN
HM
GD
F 3
00
Had
rosa
urid
aeO
ura
nosa
uru
snig
erie
nsi
sN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conc
ave
IVM
C 2
277
Had
rosa
urid
aeP
rosa
uro
lophus
maxim
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
NM
C 2
970
Had
rosa
urid
aeP
rosa
uro
lophus
maxim
us
N. M
acle
od
Upper
Cre
tace
ous
Poin
ted
Poin
ted-c
onv
ex
PU
14970
Had
rosa
urid
aeP
rosa
uro
lophus
maxim
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conv
ex
R3386/7
Had
rosa
urid
aeT
elm
ato
sauru
s tr
anss
ylvanic
us
Wei
sham
pel
et
al.
(1993)
Upper
Cre
tace
ous
Blu
ntB
lunt
-conc
ave
-H
eter
odont
osa
urid
aeA
gilis
auru
slo
uder
back
iR
. B
utle
rM
iddle
Jur
assi
cP
oin
ted
Poin
ted-c
onc
ave
SA
M K
337
Het
erodont
osa
urid
aeH
eter
odonto
sauru
stu
cki
N. M
acle
od
Low
er J
uras
sic
Poin
ted
Poin
ted-c
onc
ave
SA
M K
1332
Het
erodont
osa
urid
aeH
eter
odonto
sauru
s tu
cki
N. M
acle
od
Low
er J
uras
sic
Poin
ted
Poin
ted-c
onc
ave
JLU
M L
0204-Y
-23
Hyp
silo
pho
dont
idae
Changch
unsa
uru
sparv
us
Liy
ong
et
al.
(2010)
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
IGM
100/2
017
Hyp
silo
pho
dont
idae
Haya
gri
va
Mak
ovi
cky
et a
l. (
2011)
Upper
Cre
tace
ous
Poin
ted
Poin
ted-l
inea
r
R2477
Hyp
silo
pho
dont
idae
Hyp
silo
phodon
foxii
Gal
ton
(1974)
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
R196
Hyp
silo
pho
dont
idae
Hyp
silo
phodon
foxii
N. M
acle
od
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
R197
Hyp
silo
pho
dont
idae
Hyp
silo
phodon
foxii
N. M
acle
od
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
IVP
P V
12529
Hyp
silo
pho
dont
idae
Jeholo
sauru
ssh
angyu
anen
sis
Bar
rett a
nd H
an (
2009)
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
MO
R 1
636a
Hyp
silo
pho
dont
idae
Ory
ctodro
meu
scu
bic
ula
ris
C. B
oyd
Upper
Cre
tace
ous
Poin
ted
Poin
ted-l
inea
r
4 H
AI
7A
Hyp
silo
pho
dont
idae
Thes
celo
sauru
sneg
lect
us
N. M
acle
od
Upper
Cre
tace
ous
Blu
ntB
lunt
-conv
ex
NC
SM
15728
Hyp
silo
pho
dont
idae
Thes
celo
sauru
ssp
.C
. B
oyd
Upper
Cre
tace
ous
Poin
ted
Poin
ted-c
onv
ex
MC
Z 4
392
Hyp
silo
pho
dont
idae
Zep
hyr
osa
uru
ssc
haff
iC
. B
oyd
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
SD
SM
8691
Igua
nodont
iaD
akata
don
lakota
ensi
sN
. M
acle
od; A
. M
cDona
ldL
ow
er C
reta
ceous
Blu
ntB
lunt
-conv
ex
FP
DM
-V-4
0-3
Igua
nodont
iaF
ukuis
auru
ste
tori
ensi
sK
obay
ashi
and
Azu
ma
(2003)
Low
er C
reta
ceous
Blu
ntB
lunt
-conv
ex
BM
NH
9764
Igua
nodont
iaM
ante
llis
auru
sath
erfi
elden
sis
N. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conv
ex
IRS
N B
II
30A
Igua
nodont
iaIg
uanodon
ber
nis
sart
ensi
sN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-lin
ear
IRS
NB
1457
Igua
nodont
iaIg
uanodon
ber
nis
sart
ensi
sN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conc
ave
IRS
NB
1536
Igua
nodont
iaIg
uanodon
ber
nis
sart
ensi
sN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conc
ave
IRS
NB
II
35A
Igua
nodont
iaIg
uanodon
ber
nis
sart
ensi
sN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conc
ave
FW
MS
H 9
3B
1Ig
uano
dont
iaT
enonto
sauru
sdoss
iW
inkle
r e
t al.
(1997)
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onc
ave
YP
M 5
456
Igua
nodont
iaT
enonto
sauru
sti
llet
oru
mO
stro
m (
1970);
A. M
cDona
ldL
ow
er C
reta
ceous
Blu
ntB
lunt
-conc
ave
AM
NH
3031
Igua
nodont
iaT
enonto
sauru
sti
llet
oru
mN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conc
ave
MC
Z 4
205
Igua
nodont
iaT
enonto
sauru
sti
llet
oru
mN
. M
acle
od
Low
er C
reta
ceous
Blu
ntB
lunt
-conv
ex
TM
MC
A C
UB
0.5
3Ig
uano
dont
iaT
enonto
sauru
sti
llet
oru
mN
. M
acle
od
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
LL
.12275
Igua
nodont
iaT
enonto
sauru
sti
llet
oru
mT
enna
nt (
2010)
Low
er C
reta
ceous
Poin
ted
Poin
ted-c
onc
ave
YP
M 1
887
Igua
nodont
iaT
hei
ophyt
alia
ker
riA
. M
cDona
ldL
ow
er C
reta
ceous
Poin
ted
Poin
ted-c
onv
ex
R3398
Rha
bdodont
idae
Zalm
oxes
robust
us
Wei
sham
pel
et
al.
(2003)
Upper
Cre
tace
ous
Poin
ted
Poin
ted-c
onc
ave