Positive allometry for exaggerated structures in the ceratopsian dinosaur Protoceratops andrewsi supports socio-sexual signaling

Full text

Palaeontologia Electronica
palaeo-electronica.org
Hone, David W. E., Wood, Dylan, and Knell, Robert J. 2016. Positive allometry for exaggerated structures in the ceratopsian dinosaur
Protoceratops andrewsi supports socio-sexual signaling. Palaeontologia Electronica 19.1.5A: 1-13
palaeo-electronica.org/content/2016/1369-sexual-selection-in-ceratopsia
Copyright Palaeontological Association, January 2016
Positive allometry for exaggerated structures in the
ceratopsian dinosaur Protoceratops andrewsi
supports socio-sexual signaling
David W. E. Hone, Dylan Wood, and Robert J. Knell
ABSTRACT
The allometry in the frill and jugal bosses of the small ornithischian dinosaur Pro-
toceratops andrewsi are assessed. An analysis of 37 specimens, encompassing four
distinct size classes of animal, shows that the frill (in both length and width) under-
goes positive allometry during ontogeny, while the jugals also show a trend towards an
increase in relative size. In conjunction with other data, this provides support that these
features were under selection as socio-sexual dominance signals.
David W. E. Hone. School of Biological and Chemical Sciences, Queen Mary University of London, Mile
End Road, E1 4NS. d.hone@qmul.ac.uk
Dylan Wood. School of Biological and Chemical Sciences, Queen Mary University of London, Mile End
Road, E1 4NS. d.wood@se12.qmul.ac.uk
Robert J. Knell. School of Biological and Chemical Sciences, Queen Mary University of London, Mile End
Road, E1 4NS. r.knell@qmul.ac.uk
Keywords: Ceratopsian; sexual selection; behavior; gregariousness
Submission: 7 August 2015. Acceptance: 10 November 2015
INTRODUCTION
Sexual selection is a powerful and near-ubiq-
uitous evolutionary pressure (Andersson, 1994)
that is responsible for much of the morphological
and behavioural diversity of extant animals, which
was presumably also a major evolutionary driver in
the past (Knell et al., 2013a). It has been sug-
gested by several authors (e.g., Galton, 1971;
Sampson, 2001; Hone et al., 2012; Knell et al.,
2013a) that many of the wide variety of ornaments
and exaggerated structures on extinct animals
were sexually selected, either through acting as
signals in mate choice scenarios, as signals or
weapons used in intrasexual contests or a combi-
nation of these. Some modern signaling traits are
also used in social signaling and which can be
more or less related to reproduction (West-Eber-
hard, 1983; Kraaijeveld et al., 2004) and for some
of the exaggerated traits of extinct organisms it
might be better to think of explanations in terms of
“socio-sexual signaling” rather than trying to be
definitive about exactly how they were used (e.g.,
see Hone et al., 2012).
These ideas remain controversial, however,
with misunderstandings over mutual sexual selec-
tion and the evolutionary biology of costly signals
leading to recent exchanges in the literature
regarding the validity of sexual selection as a

HONE, WOOD, & KNELL: SEXUAL SELECTION IN CERATOPSIA
2
hypothesis to explain exaggerated structures pres-
ent in extinct animals (e.g., Padian and Horner,
2011a, b, 2013, 2014; Knell and Sampson, 2011;
Hone et al., 2012; Hone and Naish, 2013; Knell et
al., 2013a, b; Borkovic and Russell, 2014). The
fundamental problem with trying to understand the
biology of extinct organisms is that we cannot
observe their behaviour directly and thus it is diffi-
cult to assess (Hone and Faulkes, 2014). As a con-
sequence we have to fall back on more indirect
assessments of whether a trait evolved through
sexual selection, one of which is the rate of change
of the traits in question during ontogeny (e.g., Tom-
kins et al., 2010; Knell et al., 2013a).
It has been acknowledged for some years
now that one feature of many sexually selected
traits is positive allometry — a slope greater than 1
when the log trait size is regressed against log
body size (e.g., Petrie, 1992; Simmons and Tom-
kins, 1996; Bonduriansky and Day, 2003; From-
hage and Kokko, 2014). This translates to an
exponent with a value greater than 1 in the basic
allometric equation y = axb where y is the trait size
and x is the body size of the organism in question.
This positive allometry means that the trait in ques-
tion increases more than proportionally with body
size, so that large animals have proportionally
larger traits than smaller ones. Not all traits that are
sexually selected are positively allometric (Bondu-
riansky, 2007) and there are a few examples of
positively allometric traits that are not under sexual
selection. For example, Simmons and Tomkins
(1996) reported that earwig elytra show some
degree of positive allometry, and Green (1992)
described positive allometry for tail length in
smooth newts Triturus vulgaris, but these are rare
and probably mostly associated with locomotory
traits. Sexually selected display traits and weap-
ons, by contrast, are certainly very commonly posi-
tively allometric (Kodric-Brown et al., 2006).
Allometric relations can be extracted from the fossil
record when sufficient specimens are available,
meaning that we can test hypotheses regarding the
evolution of putative sexually selected traits by esti-
mating their allometry. If a trait is shown to be posi-
tively allometric then, in the absence of a plausible
alternative regarding its evolution, we can take this
as evidence supporting a hypothesis of evolution
via sexual selection or socio-sexual dominance
signals (e.g., Knell and Fortey, 2005; Tomkins et
al., 2010; Knell et al., 2013a).
The herbivorous ornithischian dinosaurs pro-
vide an important case for allometry linked to sex-
ual or socio-sexual selection in the fossil record.
Numerous members of this diverse group of extinct
archosaurs possess large cranial crests or various
other ornaments on their bodies including horns,
frills, plates and ‘helmets’. Members of the ceratop-
sian lineage (the ‘horned dinosaurs’) are excellent
candidates for analysis for a number of reasons.
There are numerous, well-preserved specimens
including both juveniles of various sizes and adults
(Figure 1), and all known members show a variety
of cranial structures including horns, bosses and in
particular a large frill that extends from the back of
the head over the neck (Dodson et al., 2004; You
and Dodson, 2004) (Figure 1). The small ceratop-
sian Protoceratops from the Late Cretaceous of
China and Asia is especially well suited for study
as it is known from a large number of specimens
that include young juveniles through to large adults
(Brown and Schlaikjer, 1940; Maryańska and
Osmólska, 1975; You and Dodson, 2004).
FIGURE 1. Size categories of specimens of Protoceratops andrewsi used in this study. Right to left: young juvenile,
juvenile, subadult, adult. Scale bar is 1 m. Image modified from Hone et al. (2014a), original illustration by David
Maas.

PALAEO-ELECTRONICA.ORG
3
The shapes of ceratopsian skulls do change
during ontogeny (Brown and Schlaikjer, 1940;
Maryańska and Osmólska, 1975; Chapman, 1990;
Sampson et al., 1997) but studies have been lim-
ited by the available data. Maryańska and Osmól-
ska (1975), for example, reached five main
conclusions about the ontogeny of the cranium of
Bagaceratops and suggested that these also likely
applied to Protoceratops (a close relative). These
changes were “the relative length of the orbit
decreases, the length of the snout increases
slightly, the length of the frill at first increases, than
[sic] appears to stop growing and its length
becomes relatively shorter during the successive
stages, the width of the frill increases, the width
across the quadrates and jugals increases”
(Maryańska and Osmólska, 1975). However,
although these conclusions echoed some similar
patterns observed in Protoceratops by Brown and
Schlaikjer (1940), they were limited by a lack of
complete specimens that spanned a broad range
of ages. Similarly, Dodson (1976) assessed sexual
dimorphism in Protoceratops and his analyses
included allometric regressions of various mea-
surements, but his study focused on adult or near-
adult animals with only a very limited sample or
younger specimens.
A large number of hypotheses have been
advocated for the functions of the cranial features
of ceratopsians including signaling and/or sexual
selection (Farlow and Dodson, 1975), though this
has in the past often been incorrectly ruled out
because of a lack of apparent sexual dimorphism
in many taxa (see Hone et al., 2012 for review). In
at least some ceratopsians the horns were used in
intraspecific combat (Farke et al., 2009) and may
have been involved in interspecific combat (Happ,
2008), though they may have served additional
purposes (e.g., as aposematic signals). A particu-
larly wide range of explanations has been pro-
posed for the functions of the ceratopsian frill
including: as a temperature regulating device, as
an aposematic signal, as a defence against preda-
tors and for socio-sexual dominance signals. How-
ever, in most taxa many of these can be ruled out
(see Hone et al., 2012 for a review), leaving sexual
or socio-sexual signaling as the most plausible cur-
rent explanation for the evolution of these struc-
tures.
Here we assess the growth of the frill in Proto-
ceratops based on a larger sample of specimens
and across a range of body sizes in excess of an
order of magnitude, including animals that can be
assigned to four distinct size classes from young
juveniles to large adults (Hone et al., 2014a). We
show that both the length and width of the frill are
positively allometric, suggesting a signaling func-
tion.
METHODS
Data on Protoceratops andrewsi skulls were
taken from the literature and measured from pub-
lished photographs of specimens. Data on larger
specimens of P. andrewsi were taken from Dodson
(1976) that focused on animals of ‘subadult’ and
adult sizes. In his analysis of skull dimorphism,
Dodson (1976) illustrated a number of major met-
rics of skull and frill dimensions based around land-
marks of the cranium. His measure of basal skull
length (variable 1 of Dodson, 1976) was taken as
the standard unit of size of the skull (Dodson 1976)
measured from the tip of the snout to the base of
the articular in line with the ventralmost point of the
lower temporal fenestra (also corresponding to the
position of the foramen magnum). The length of the
frill was taken as Dodson’s (1976) total skull length
less the basal skull length, however, where the
anterior anchor could not be determined, a mini-
mum was measured to the anterior margin of the
frill fenestrae and a maximum to the posterior mar-
gin of the orbit. The width of the frill was measured
as the maximum width of the frill (variable 9 of Dod-
son, 1976), and the width across the jugal bosses
(variable 8 of Dodson, 1976) was recorded, as
were the orbit length and height (variables 13 and
14 of Dodson, 1976).
Additional measurements from younger speci-
mens were taken from (smallest to largest)
Fastovsky et al. (2011), Hone et al. (2014a) and
Handa et al. (2012) corresponding to young juve-
niles, juveniles, and subadults (Figure 1). These
non-adult specimens at least all correspond to sim-
ilar localities within the Djadochta Formation and
can be considered part of a single population (see
Hone et al., 2014a), and the material measured by
Dodson, (1976) also herald from the same forma-
tion (Brown and Schlaikjer, 1940). Measurements
were taken from figures in the papers and unpub-
lished photographs by the authors. The measure-
ments taken from these non-adult animals were
made using the program ImageJ (Schneider et al.,
2012) and correspond with the variables described
above to allow integration of the data. However, the
inclusion of very young animals with very different
proportions and shapes to the skull (Figure 2)
makes it difficult to ensure that some other mea-
surements correspond properly with Dodson’s
(1976) landmarks.
Where images of the small specimens pre-
sented the skull only in dorsal view (e.g., very
young specimens of Fastovsky et al., 2011), the

HONE, WOOD, & KNELL: SEXUAL SELECTION IN CERATOPSIA
4
basal skull length was measured in the midline of
the skull from the tip of the rostrum to a point mid-
way between the posterior margin of the orbit and
the anterior margin of the fenestra in the frill (Fig-
ure 3). Similarly, additional measurements were
taken for the frill of the smallest specimens. The
length of the frill was taken to a point midway
between the dorsoposterior margin of the orbit and
the anteriomedial margin of the fenestra, but a
maximum frill length and minimum frill length were
also recorded, terminating at these respective
points on the orbit and fenestra (Figure 3). Mea-
surements of specimens taken from photographs
may be subject to parallax error; however, as far as
possible photographs were taken parallel to the
long axis of the respective total skull lengths (cf
Figure 3) to minimise the effects of these errors.
A total of 37 specimens were included in the
analysis, although not all measurements were
available for each specimen. Specimens ranged in
size from a basal skull length of 23.5 to 357 mm (or
including the frill, from 35 to 515 mm) — see data
in Appendix. Assessment was based on the search
for positive allometry of the crests with respect to
isometry of other elements (see Tomkins et al.,
2010), here using basal skull length. This measure
has been shown by Dodson (1976) to change iso-
metrically with other major features of the skull
suggesting it is an appropriate metric for this pur-
pose. Although there are obvious limitations with
small datasets (Brown and Vavrek, 2015), the data
set used here is larger than a number of previous
assessments of allometry in fossil taxa and covers
a greater range of animal sizes, and thus is suffi-
cient given the limitations of the fossil record.
Allometric slopes for the log-transformed val-
ues for frill length and width, the width across the
jugal bosses and also orbit length and width were
calculated using Standardised Major Axis (SMA)
regressions (Warton et al., 2006; Westoby, 2006)
using the smatr package (Warton et al., 2012) run-
ning in R v. 3.1.2 (R Core Team, 2014). SMA is
sometimes referred to as Reduced Major Axis but it
has been argued that this latter term can be mis-
leading and is, in fact, based on a mistranslation
from the French (Warton et al., 2006, appendix A).
Tests for deviation of the slopes from a value of 1
were carried out by calculating an r test statistic in
the same way as for a conventional linear regres-
sion as recommended in Warton et al. (2006).
Sexual dimorphism has been proposed to
explain the frill and other cranial characteristics of
P. andrewsi (see Dodson 1976). However, sexual
dimorphism has recently been reassessed and
rejected as a hypothesis for these characteristics
(Maiorino et al., 2015) and so we analysed all avail-
able data and did not split the adults into the puta-
tive male and female morphs. This would also have
reduced the amount of available data for each
analysis if the sexes were treated separately.
RESULTS
Both frill length and frill width showed positive
allometry in Protoceratops (Table 1, Figure 4). The
width across the jugal bosses gave an allometric
FIGURE 2. Changes in skull shape in Protoceratops andrewsi. All skulls are drawn to the same total length and are
seen in dorsal view (upper row) and right lateral view (lower row). Left to right (with sources in parentheses) small
juveniles (Fastovsky et al., 2011), juveniles (MPC-D 100/526), subadults (MPC-D 100534), putative ‘female’ morph,
putative ‘male’ morph (both Dodson, 1976). The large fenestrae seen in the smallest animals are supratemporal
fenestra and are not homologous with the frills of the fenestra in the larger animals.

PALAEO-ELECTRONICA.ORG
5
slope that was not significantly different from 1,
although the slope was suggestive of positive
allometry with confidence intervals which only just
overlapped with 1 (0.983–1.34) and a p-value of
0.078 for the test comparing the slope to isometry.
In the case of frill length, this result was qualita-
tively unchanged when the minimum and maxi-
mum frill lengths were used as alternate
measurements from the smaller specimens (maxi-
mum frill length: slope = 1.21, minimum frill length:
slope = 1.27). As skull basal length increases,
therefore, both the length and the width of the frill
become larger relative to the rest of the skull.
By contrast with the increase in frill length and
width, both orbit length and orbit height showed
negative allometry, although height increased with
basal skull length faster than did width, indicating
that the shape of the orbits changed as the animals
got larger. As can be seen from Figure 2, small
Protoceratops have orbits that are longer than they
are high, whereas the larger animals have orbits
that are closer to being circular, with orbit length
and orbit height being roughly equal in the largest
specimens.
DISCUSSION
The results here provide strong evidence for
the growth of the frill in Protoceratops andrewsi
being greater than the overall growth of the animal
by demonstrating positive allometry (Table 1). The
different dimensions of the frill show different allo-
metric slopes, demonstrating that the frill changed
in shape as well as size (as the changes in frill
width are slightly greater than length). The
increase in length and width fits with the sugges-
tions of both Brown and Schlaikjer (1940) and
Maryańska and Osmólska (1975) although the lat-
ter noted that, in Bagaceratops at least, the frill at
some point stopped increasing in length before
expanding again. That cessation of growth is not
seen here, though the limited number of speci-
mens in their analysis means that this pattern could
have been caused by one or a few unusual individ-
uals. Chapman (1990) suggested that the length of
the frill in Bagaceratops was close to isometry, but
this is not the case here. Dodson (1976) used a dif-
ferent allometric regression technique to that used
here, but the results should still be qualitatively
comparable. Across his data, Dodson (1976)
recorded a slight decrease in frill length with a
slight increase in frill width for Protoceratops
although in both cases the slope was below signifi-
cance at P = 0.05. However, his dataset contained
only one small individual (basal skull length under
100 mm) with most specimens representing sub-
adults or adults, though it is still a rather different
result to our own that shows consistent positive
allometry for the frill across all skull sizes.
Furthermore, the metrics used here are a sim-
plification of the changes that Protoceratops expe-
riences as other patterns are clearly visible (though
not directly assessed here). For example, as well
as increasing in both length and width, the crest
also appears to deflects upwards in adults to be
steeply inclined relative to the skull roof whereas it
is largely in line with the skull roof in younger ani-
mals (Figures 1, 2). Although there is not a com-
plete ontogenetic sequence available for
Protoceratops, this deflection is not seen in
FIGURE 3. Measurements taken from skulls of Protoc-
eratops based on an idealised adult in dorsal view
(above) and lateral view (below). Black lines and num-
bers indicate the measurements taken according to the
variable of Dodson (1976). These are: 1, basal skull
length; 2, total length (frill length is variable 2 subtracted
from variable 1); 8, jugal width; 9, frill width; 13, orbit
length; 14, orbit height. The grey lines indicate the max-
imum and minimum lengths of the frill as measured in
juvenile animals. See text for further details.

HONE, WOOD, & KNELL: SEXUAL SELECTION IN CERATOPSIA
6
younger specimens suggesting that it does not
occur until the animals are relatively old. This
change also means that the length of the frill is
effectively foreshortened when measured in lateral
or dorsal view (as done here cf. Dodson, 1976)
rather than measuring along the actual elements of
the frill themselves. Thus as the frill deflects in
older individuals this measurement will underesti-
mate the true length of the frill and the proportional
growth rate is therefore likely to be greater than
that calculated here. Maiorino et al. (2015) also
noted the dorsal deflection of the frill during ontog-
eny as well as additional changes not commented
on here such as the elongation of the rostrum and
development of a nasal bump in larger specimens.
This latter increase in the nasal bump was also
noted by Brown and Schlaikjer (1940) and is effec-
tively the same point raised by Lull and Gray
(1949) who noted an increase in depth in the nasal
region.
Finally, the smallest specimens have large
supratemporal fenestrae that dominate the frill and
almost contact the orbits, whereas in adults these
fenestrae are closed and the fenestrae in the frill
are limited to the posterior part of the frill (Figure 2).
Other apparent changes include the elongation of
the premaxillary teeth that are proportionally short
in juveniles and enlarged in adults (Figures 1, 2)
and in the postcranium, the enlarged neural spines
on the midpoint of the caudal vertebral series
appear to grow taller later in ontogeny.
Interestingly the width of the skull across the
jugals is here shown not to differ from isometry.
This differs from what was suggested by
Maryańska and Osmólska (1975), although Chap-
man (1990) suggested that the jugal width is iso-
metric in Bagaceratops at least, and Dodson
(1976) recorded near isometry for the jugal width
among larger individuals. The result we have
returned may be confounded by an overall change
in the shape of the skull as we and others have
measured the entire width of the skull including the
jugal bosses, rather than just the bosses them-
selves. Additional data may confirm if this dimen-
sion is truly isometric or does demonstrate positive
allometry, though if they are isometric, it would sug-
FIGURE 4. Allometric relationships for frill length (1), frill width (2) and the width across the jugal bosses (3). Solid
lines show the fitted lines from SMA regression, dashed grey lines show the line of isometry (slope = 1 and intercept
= 0). All measurements were originally in mm prior to log transformation.
TABLE 1. SMA slopes and associated statistics for frill and other skull traits from Protoceratops andrewsi.
Measure Intercept Slope Slope CIs Test statistic P slope ≠ 1
Frill length -1.46 1.23 1.14–1.34 r = 0.719, 26df <0.0001
Frill width -1.21 1.29 1.19–1.41 r = 0.827, 18df <0.0001
Jugal Boss width -0561 1.15 0.983–1.34 r = 0.403, 18df 0.078
Orbit width 0.272 0.699 0.650–0.751 r = -0.877, 30df <0.0001
Orbit height -0.583 0.843 0.760–0.935 r = -0.526, 30df 0.002

PALAEO-ELECTRONICA.ORG
7
gest a functional role that does not change with
ontogeny.
Although orbit shape is not directly related to
crest morphology, this has been observed to
change in Protoceratops (Brown and Schlaikjer,
1940; Maryańska and Osmólska, 1975; Dodson,
1976; Maiorino et al., 2015). Orbit size has been
used as a metric against which other values can be
scaled in assessments of ontogenetic allometry in
extinct amniotes and may be isometric in at least
some taxa (Tomkins et al., 2010). Maryańska and
Osmólska (1975) observed a decrease in relative
orbit width for Bagaceratops with Dodson (1976)
demonstrating a sharp decrease in orbit height and
an even greater decrease in width (his ‘orbit
length’). Both of these observations are consistent
with the results recovered here showing a strong
negative allometry for the orbit but also a change in
shape given the differences in slope recovered for
the two different dimensions.
The positive allometry seen in the frill demon-
strates that the frill grows faster than the rest of the
animal as individuals increase in size. Definitions
of maturity in non-avian dinosaurs are complicated
as at least some were reproductively mature
before they were fully osteologically mature (e.g.,
Lee and Werning, 2008) and thus it is difficult to
determine exact timing of changes between ‘juve-
niles’ and ‘adults’ (see Handa et al., 2012 and
Hone et al., 2014a for discussions of age catego-
ries in Protoceratops). However, it is clear that the
frill was disproportionally small in younger individu-
als, grew faster than the rest of the animal during
ontogeny, and reached full size only in large adult
specimens. Such a pattern of rapid growth com-
bined with ontogenetic change makes socio-sexual
dominance displays a strong hypothesis to explain
these results (Knell et al., 2013a) as the relative
sizes of the crests suggests a function in adults but
not in juveniles.
If the frill acted as a socially or sexually
selected signal, this would explain other observed
patterns in Protoceratops. It is common for animals
to exhibit multiple sexually selected traits (Omland,
1996) and this would match the potential use of the
premaxillary teeth and the neural spines of the tail
as additional signals. The former might have a role
in signaling or even intraspecific combat, and it has
been suggested by Tereschenko and Singer (2013)
that the tail of Protoceratops might be well suited to
being raised as a signal. As dinosaurs apparently
reached sexual maturity before full adult size this
may also potentially explain the isometric growth of
the frill recorded by Dodson (1976) for his study of
subadult and adult animals. If the subadults were
sexually mature then this may correlate with a
growth of the frill to reach proportional peak size at
this point in order to act as a viable signal of matu-
rity.
Species recognition has been advocated as a
selective pressure to explain the presence of
crests such as ceratopsian frills in various dinosau-
rian taxa (e.g., see Padian and Horner, 2011a, b,
2013). This has been strongly criticised as it fails to
explain the observed patterns in crest distribution
on taxa, or the presence of such costly signals in
the fossil record that are unknown in extant taxa
(e.g. see Knell and Sampson, 2011; Knell et al.,
2013a, b; Hone and Naish, 2013). It is possible that
the presence of these crests in adults did help indi-
viduals recognise conspecifics, but this would be a
co-opted function, not the primary origin and drive
behind its evolution (Hone and Naish, 2013).
As a concept, ‘species recognition’ may relate
to either correct mate identification or the more
general ability of individuals to recognise conspe-
cifics such as in relation to forming and maintaining
a group (see Hone and Naish, 2013). In the case of
the latter aspect, this is not supported for P.
andrewsi based on the known behavior of these
animals. Although many specimens of P. andrewsi
are known from isolated remains, there are now
size-segregated aggregations of this species at
multiple sizes, which likely represent different age
classes (Hone et al., 2014a). As such, if the frill
was important to help identify conspecifics in
aggregations, then it would be selected to be pres-
ent in even young animals, or alternatively if the frill
as seen in juveniles was a sufficient signal, there
would be no need for it to increase in size and
change shape in adults. In terms of mate recogni-
tion, as noted for other ceratopsians (Hone and
Naish, 2013), the fact that the frill continued to
change shape after sexual maturity would poten-
tially confound signals. In addition, neither aspect
of species recognition explains the strong similarity
of frill (and indeed general cranial morphology) of
P. andrewsi to other sympatric species of Protocer-
atops or other protoceratopsids. Selection for rec-
ognition would drive species to produce more
disparate, rather than highly similar forms (Hone
and Naish, 2013).
Similar patterns of positive allometric growth
of cranial crests during ontogeny are also seen in
other ornithischian dinosaurs. Sampson et al.
(1997) stated that both the horns and frills of the
centrosaurine ceratopsian dinosaurs showed posi-
tive allometry during growth (and these features

HONE, WOOD, & KNELL: SEXUAL SELECTION IN CERATOPSIA
8
FIGURE 5. Life restoration of adult Protoceratops andrewsi (foreground) engaging in speculative display postures, an
activity in which non-mature animals (background) do not take part. Artwork by Rebecca Gelernter, who retains the
copyright on this image — used with permission.

PALAEO-ELECTRONICA.ORG
9
also change in shape as they grow), though this
was stated without a formal analysis owing to a
lack of complete cranial material. Similarly, Mallon
et al. (in press) described a specimen of Arrhinoc-
eratops at circa 70% of adult size but where the frill
was less than 50% of that of the adult, and the
horns 30% of that of the adult (their figure 10),
implying relatively rapid growth late in ontogeny.
Evans (2010) also found strong positive allometry
in the development of the crests of a number of
hadrosaur genera with increases in size and
changes in shape relatively late in ontogeny and
juveniles exhibiting no, or only incipient, crests.
Although not discussed in terms of possible func-
tions by Evans (2010), the implications are similar
to those discussed here, and it is possible to infer
that these were selected for socio-sexual domi-
nance.
Some hadrosaur and ceratopsian mass mor-
tality sites do show size-sorted groups and include
the presence of adult-only groups (e.g., see Hone
et al., 2014b) or non-adult only groups (e.g., see
Lehman, 2007; Lauters et al., 2008). However, oth-
ers show groups of multiple different sizes and
include both juveniles (or at least subadults) and
adults together (e.g., see Sampson et al., 1997;
Brinkman, 2014), a pattern also seen in some
trackways (Lockley, 1991). For Protoceratops,
however, the presence of mixed-age groups, as
well as clusters of adults only and juveniles only,
suggests that the enlarged cranial crests of various
ornithischians were not linked to herd coherency
(in the sense of correct identification of conspecif-
ics). If they were, then such features would be
present in both juveniles and adults and to the
same degree, as this would be beneficial to both
age groups in mixed herds, or to groups of all juve-
niles. Instead, again the positive allometry and the
presence in adults but not younger animals
strongly implies this is a function that benefits only
adult animals, and thus socio-sexual signals are a
strong candidate to explain this pattern of growth.
Our results do not rule out additional functions
of such crests they might have had beyond sexual
and social dominance signals, although some
hypotheses can be excluded. The ceratopsian frill
would have provided little protection as armour
against animals like large tyrannosaurs that lived
alongside P. andrewsi (see Jerzykiewicz et al.,
1993) and that could bite into or through much
thicker and stronger bones (see Hone and Rauhut,
2010). A large frill may act as an aposematic signal
to heterospecifics (especially predators), though as
juvenile dinosaurs were generally more vulnerable
to predators than adults (Hone and Rauhut, 2010)
then the small frill size in juveniles suggests limited
value. Large bodied (multi-ton) tyrannosaurs may
have been little deterred by such signals and unlike
many other ceratopsians, Protoceratops lacked
large horns with which it could engage potential
predators (cf. Happ, 2008). This would make a sig-
nal effectively a false one, and such signals typi-
cally only operate if the bearer is rare (Mallett and
Joron, 1999) when in fact Protoceratops is gener-
ally the most common large tetrapod in its environ-
ment (based on the huge numbers of recovered
specimens). If ceratopsians generally used their
frills as aposematic signals, then we might predict
convergence of form through Mullerian mimicry
among those that were sympatric, and perhaps
even convergent Batesian mimicry from rarer taxa.
CONCLUSIONS
Based on the available data it seems likely
that the frill, and perhaps the jugal bosses, of Pro-
toceratops andrewsi were features that functioned
in signaling (Figure 5). The absence of support for
alternate hypotheses combined with the positive
allometry shown here across a wide range of indi-
viduals suggests a feature that functioned only in
adults and a socio-sexual dominance signal is the
best available candidate.
Despite extensive debate in recent years as to
the functions of these and similar structures in vari-
ous archosaur and other lineages, relatively little
hypothesis testing has been carried out (e.g., see
Tomkins et al., 2010). These results therefore pro-
vide an important step in assessing the function of
such features in the fossil record and provide con-
firmation that these likely evolved as a result of
pressures favouring socio-sexual dominance sig-
nals.
ACKNOWLEDGEMENTS
We thank P. Dodson for making his dataset
accessible and N. Handa for providing supporting
photographs. Our thanks to two anonymous refer-
ees and the editor for the suggestions and assis-
tance in taking this paper to publication.
REFERENCES
Andersson, M. 1994. Sexual Selection. Princeton Uni-
versity Press, Princeton.
Bonduriansky, R. 2007. Sexual selection and allometry:
a critical reappraisal of the evidence and ideas. Evo-
lution, 61:838-849.

HONE, WOOD, & KNELL: SEXUAL SELECTION IN CERATOPSIA
10
Bonduriansky R. and Day, T. 2003. The evolution of
static allometry in sexually selected traits. Evolution,
57:2450-2458.
Borkovic, B. and Russell, A. 2014. Sexual selection
according to Darwin: a response to Padian and
Horner’s interpretation. Comptes Rendus Palevol,
13:701-707.
Brinkman, D.B. 2014. The size-frequency distribution of
hadrosaurs from the Dinosaur Park Formation of
Alberta, Canada, p. 416-421. In Eberth, D.A. and
Evans, D.C. (eds.), Hadrosaurs. Indiana University
Press, Bloomington.
Brown, B. and Schlaikjer, E.M. 1940. The structure and
relationships of Protoceratops. Annals of the New
York Academy of Sciences, 40:133-266.
Brown, C.M. and Vavrek, M.J. 2015. Small sample sizes
in the study of ontogenetic allometry; implications for
palaeobiology. PeerJ, 3:e818.
Chapman, R.E. 1990. Shape analysis in the study of
dinosaur morphology, p. 21-42. In Carpenter, K. and
Currie, P. (eds.), Dinosaur Systematics: Approaches
and Perspectives. Cambridge University Press, New
York.
Dodson, P. 1976. Quantitative aspects of relative growth
and sexual dimorphism in Protoceratops. Journal of
Paleontology, 50:929-940.
Dodson, P., Forster, C.A., and Sampson, S.D. 2004. Cer-
atopsidae, p. 494-513. In Weishampel, D.B., Dodson,
P., and Osmólska, H. (eds.), The Dinosauria. Univer-
sity of California Press, Berkley.
Evans, D.C. 2010. Cranial anatomy and systematics of
Hypacrosaurus altispinus, and a comparative analy-
sis of skull growth in lambeosaurine hadrosaurids
(Dinosauria: Ornithischia). Zoological Journal of the
Linnean Society, 159:398-434.
Farke, A.A., Wolff, E.D.S., and Tanke, D.H. 2009. Evi-
dence of combat in Triceratops. PLoS ONE, 4:e4252.
Farlow, J.O. and Dodson, P. 1975. The behavioural sig-
nificance of frill and horn morphology in ceratopsian
dinosaurs. Evolution, 29:353-361.
Fastovsky, D., Weishampel, D., Watabe, M., Barsbold,
R., Tsogtbaatar, K., and Narmandakh, P. 2011. A nest
of Protoceratops andrewsi (Dinosauria, Ornithischia).
Journal of Paleontology, 85:1035-1041.
Fromhage, L. and Kokko, H. 2014. Sexually selected
traits evolve positive allometry when some matings
occur irrespective of the trait. Evolution,
68:1332-1338.
Galton, P.M. 1971. A primitive dome-headed dinosaur
(Ornithischia: Pachycephalosauridae) from the lower
Cretaceous of England, and the function of the dome
in pachycehalosaurids. Journal of Paleontology,
45:40-47.
Green, A.J. 1992. Positive allometry is likely with mate
choice, competitive display and other functions. Ani-
mal Behaviour, 43:170-172.
Handa, N., Watabe, M., and Tsogtbaatar, K. 2012. New
specimens of Protoceratops (Dinosauria: Neocera-
topsia) from the Upper Cretaceous in Udyn Sayr,
Southern Gobi Area, Mongolia. Paleontological
Research, 16:179-198.
Happ, J. 2008. An analysis of predator-prey behavior in
a head-to-head encounter between Tyrannosaurus
rex and Triceratops, p. 355-370. In Larson, P. and
Carpenter, K (eds.). Tyrannosaurus rex: the Tyrant
King. Indiana University Press, Blommington.
Hone, D.W.E., Farke, A.A., Watabe, M., Shigeru, S., and
Tsogtbaatar, K. 2014a. A new mass mortality of juve-
nile Protoceratops and size-segregated aggregation
behaviour in juvenile non-avian dinosaurs. PLoS
ONE, 9:e113306.
Hone, D.W.E. and Faulkes, C.J. 2014. A proposed
framework for establishing and evaluating hypothe-
ses about the behaviour of extinct organisms. Jour-
nal of Zoology, 292:260-267.
Hone, D.W.E. and Naish, D. 2013. The ‘species recogni-
tion hypothesis’ does not explain the presence and
evolution of exaggerated structures in non-avialan
dinosaurs. Journal of Zoology, 290:172-180.
Hone, D.W.E., Naish, D., and Cuthill, I.C. 2012. Does
mutual sexual selection explain the evolution of head
crests in pterosaurs and dinosaurs? Lethaia,
45:139-156.
Hone, D.W.E. and Rauhut, O.W.M. 2010. Feeding
behaviour and bone utilization by theropod dino-
saurs. Lethaia, 43:232-244.
Hone, D.W.E., Sullivan, C., Zhao, Q., Wang, K., and Xu,
X. 2014b. Body size distribution in a colossal hadro-
saurid death assemblage from the Upper Cretaceous
of Zhucheng, Shandong Province, China, p. 524-
531. In Eberth, D.A. and Evans, D.C. (eds.), Hadro-
saurs. Indiana University Press, Bloomington.
Jerzykiewicz, T., Currie, P.J., Eberth, D.A., Johnston,
P.A., and Zheng, Z.-Z. 1993. Djadokhta Formation
correlative strata in Chinese Inner Mongolia: an over-
view of the stratigraphy, sedimentary geology, and
paleontology and comparisons with the type locality
in the pre-Altai Gobi. Canadian Journal of Earth Sci-
ences, 30:2180-2190.
Knell, R.J. and Fortey, R.A. 2005. Trilobite spines and
beetle horns: sexual selection in the Palaeozoic?
Biology letters, 1:196-199.
Knell, R., Naish, D., Tompkins, J.L., and Hone, D.W.E.
2013a. Sexual selection in prehistoric animals:
detection and implications. Trends in Ecology and
Evolution, 28:38-47.
Knell, R., Naish, D., Tompkins, J.L., and Hone, D.W.E.
2013b. Is sexual selection defined by dimorphism
alone? A reply to Padian and Horner. Trends in Ecol-
ogy and Evolution, 28:250-251.
Knell, R.J. and Sampson, S. 2011. Bizarre structures in
dinosaurs: species recognition or sexual selection? A
response to Padian and Horner. Journal of Zoology,
83:18-22.
Kodric-Brown, A., Sibly, R.A., and Brown, J.H. 2006. The
allometry of ornaments and weapons. Proceedings
of the National Academy of Sciences, 103:8733-
8738.

PALAEO-ELECTRONICA.ORG
11
Kraaijeveld, K., Gregurke, J., Hall, C., Komdeur, J., and
Mulder, R.A. 2004. Mutual ornamentation, sexual
selection, and social dominance in the black swan.
Behavioural Ecology, 15:380-389.
Lauters, P., Bolotsky, Y.L., Van Itterbeeck, J., and Gode-
froit, P. 2008. Taphonomy and age profile of a Latest
Cretaceous dinosaur bone bed in Far Eastern Rus-
sia. Palaios, 23:153-162.
Lehman, T.M. 2007. Growth and population age struc-
ture in the horned dinosaur Chasmosaurus, p. 259-
317. In Carpenter, K. (ed.) Horns and Beaks: Cera-
topsian and Ornithopod Dinosaurs. Indiana Univer-
sity Press, Bloomington.
Lee, A.H. and Werning, S. 2008 Sexual maturity in grow-
ing dinosaurs does not fit reptilian growth models.
Proceedings of the National Academy of Sciences,
105:582-587.
Lockley, M.G. 1991. Tracking dinosaurs: a new look at an
ancient world. Cambridge University Press, Cam-
bridge.
Lull, R.S. and Gray, S.W. 1949. Growth patterns in the
Ceratopsia. American Journal of Science, 247:492-
503.
Maiorino, L., Farke, A.A., Kotsakis, T., and Piras, P.
2015. Males Resemble Females: Re-Evaluating Sex-
ual Dimorphism in Protoceratops andrewsi (Neocera-
topsia, Protoceratopsidae). PLoS ONE,
10:e0126464.
Mallet, J. and Joron, M. 1999. Evolution of diversity in
warning color and mimicry: polymorphisms, shifting
balance, and speciation. Annual Review of Ecology
and Systematics, 30:201-233.
Mallon, J.C., Ryan, M.J., and Campbell, J.A. In press.
Skull ontogeny in Arrhinoceratops brachyops (Ornith-
ischia: Ceratopsiade) and other horned dinosaurs.
Zoological Journal of the Linnean Society.
Maryańska, T. and Osmólska, H. 1975. Protoceratopsi-
dae (Dinosauria) of Asia. Palaeontologia Polonica,
33:133-181.
Omland, K.E. 1996. Female mallard mating preferences
for multiple male ornaments. Behavioural Ecology
and Sociobiology, 39:353- 360.
Padian, K. and Horner, J.R. 2011a. The evolution of
‘bizarre structures’ in dinosaurs: biomechanics, sex-
ual selection, social selection or species recognition?
Journal of Zoology, 283:3-17.
Padian, K. and Horner, J.R. 2011b. The definition of sex-
ual selection and its implications for dinosaurian biol-
ogy. Journal of Zoology, 283:23-27.
Padian, K. and Horner, J.R. 2013. Misconceptions of
sexual selection and species recognition: a response
to Knell et al. and to Mendelson and Shaw. Trends in
Ecology and Evolution, 28:249-250.
Padian, K. and Horner, J.R. 2014. The species recogni-
tion hypothesis explains exaggerated structures in
non-avialan dinosaurs better than sexual selection
does. Comptus Rendus Palevol, 13:97-107.
Petrie, M. 1992. Are all secondary sexual display struc-
tures positively allometric, and, if so, why? Animal
Behaviour, 43:173-175.
R Core Team, 2014. R: A language and environment for
statistical computing. R Foundation for Statistical
Computing, Vienna, Austria. http://www.R-proj-
ect.org/
Sampson, S.D. 2001. Speculations on the socioecology
of ceratopsid dinosaurs (Ornithischia: Neoceratop-
sia), p. 263-276. In Tanke, D.H. and Carpenter, K.
(eds.), Mesozoic Vertebrate Life. Indiana University
Press, Bloomington.
Sampson, S.D., Ryan, M.J., and Tanke, D.H. 1997. Cra-
niofacial ontogeny in centrosaurine dinosaurs
(Ornithischia: Ceratopsidae): taxonomic and behav-
ioral implications. Zoological Journal of the Linnean
Society, 121:293-337.
Schneider, C.A., Rasband, W.S., and Eliceiri, K.W. 2012.
NIH Image to ImageJ: 25 years of image analysis.
Nature Methods, 9:671-675.
Simmons, L.W. and Tomkins, J.L. 1996. Sexual selection
and the allometry of earwig forceps. Evolutionary
ecology, 10:97-104.
Tereschenko, V.S. and Singer, T. 2013. Structural fea-
tures of neural spines of the caudal vertebrae of pro-
toceratopoids (Ornithischia: Neoceratopsia).
Paleontological Journal, 47:618-630.
Tomkins, J.L., Lebas, N.R., Witton, M.P., Martill, D.M.,
and Humphries, S. 2010. Positive allometry and the
prehistory of sexual selection. American Naturalist,
176:141-148.
Warton, D.I., Wright, I.J., Falster, D.S., and Westoby, M.
2006. Bivariate line-fitting methods for allometry. Bio-
logical Reviews, 81:259-291.
Warton, D.I., Duursma, R.A., Falster, D.S., and Taskinen,
S. 2012. smatr 3 - an R package for estimation and
inference about allometric lines. Methods in Ecology
and Evolution, 3:257-259.
West-Eberhard, M.J. 1983. Sexual selection, social com-
petition, and speciation. Quarterly Review of Biology,
58:155-183.
Westoby, M. 2006. Bivariate line-fitting methods for
allometry. Biological Reviews, 81:259-291.
You, H. and Dodson, P. 2004. Basal Ceratopsia, p.
478-493 In Weishampel, D.B., Dodson, P., and
Osmólska, H. (eds.), The Dinosauria. University of
California Press, Berkley.

HONE, WOOD, & KNELL: SEXUAL SELECTION IN CERATOPSIA
12
APPENDIX.
Table of all data used in the analyses. All measurements are in mm. AMNH = American Museum
of Natural History, New York; MPC/D = Mongolian Paleontological Centre, Ulan Baator.
Source Specimen Number
Skull
length
total
Skull
length
basal
Frill
length
Frill
width
Jugal
boss
width
Orbit
length
Orbit
height
Frill
minimum
Frill
maximum
Dodson,
1976
AMNH 6419 115 76 51.5 67 71 26.5 26.2 NA NA
Dodson,
1976
AMNH 6434 190 123 90.9 NA NA 34.1 20 NA NA
Dodson,
1976
AMNH 6430 NA 137 NA NA NA 37.1 30 NA NA
Dodson,
1976
AMNH 6251 NA 140 NA NA NA 43.7 42.4 NA NA
Dodson,
1976
AMNH 6431 259 150 146 185 162 44.4 31.9 NA NA
Dodson,
1976
AMNH 6486 281 150 125 NA 170 50.7 35.3 NA NA
Dodson,
1976
AMNH 6432 168 92.3 80.8 NA 200 43.5 NA NA NA
Dodson,
1976
AMNH 6428 170 90 57.6 NA NA 32.4 30.3 NA NA
Dodson,
1976
AMNH 6409 304 191 193 387 295 47.6 55 NA NA
Dodson,
1976
AMNH 6480 NA 200 NA NA 225 NA 54 NA NA
Dodson,
1976
AMNH 6444 340 210 191 374 325 51.1 40.6 NA NA
Dodson,
1976
AMNH 6485 NA 229 NA NA NA 52.5 52.7 NA NA
Dodson,
1976
AMNH 6408 314 235 152 242 233 54.4 54.6 NA NA
Dodson,
1976
AMNH 6433 410 261 178 NA NA 62.5 57.8 NA NA
Dodson,
1976
AMNH 6429 408 169 208 333 399 70.2 70.8 NA NA
Dodson,
1976
AMNH 6439 348 271 202 NA NA 62.5 61 NA NA
Dodson,
1976
AMNH 6441 NA 272 NA NA NA 58.3 41.5 NA NA
Dodson,
1976
AMNH 6477 490 303 265 NA NA 63.8 59.3 NA NA
Dodson,
1976
AMNH 6417 NA NA 182 375.5 373 NA 40.4 NA NA
Dodson,
1976
AMNH 6425 469 313 264 471 360 70.2 79 NA NA
Dodson,
1976
AMNH 6413 421 314 234 525 360 75.7 58.2 NA NA
Dodson,
1976
AMNH 6414 461 341 249 490 400 77.6 72.6 NA NA
Dodson,
1976
AMNH 6438 NA 352 NA NA NA 87.5 94 NA NA

PALAEO-ELECTRONICA.ORG
13
Dodson,
1976
AMNH 6466 491 357 262 465 458 78.3 75 NA NA
Dodson,
1976
AMNH 6467 515 285 257.5 445 358 61.1 61.3 NA NA
Handa et
al., 2012
MPC/D 100/539 NA 185.5 NA NA 191.5 53 48 NA NA
Hone et al.,
2014a
MPC/D 100/534 285.5 145.5 138.5 240 159 38.5 47 NA NA
Hone et al.,
2014a
MPC/D 100/526 B 128 79 49 52.5 44 31 23 23 53.5
Hone et al.,
2014a
MPC/D 100/526 C 125 91.5 33.5 NA NA 25.5 28.5 24 53
Fastovsky
et al., 2011
MPC/D 100/530 a 41.4 26 15.5 22 22 14.5 NA 13 16
Fastovsky
et al., 2011
MPC/D 100/530 b 41 23.5 17 25 26 11 NA 14 17.5
Fastovsky
et al., 2011
MPC/D 100/530 c 35 NA 15.5 19 NA 10.5 NA 14 18
Fastovsky
et al., 2011
MPC/D 100/530 d 46.5 29.5 16 24 NA 14 10.5 14 17
Fastovsky
et al., 2011
MPC/D 100/530 e 39.5 28 12 23.5 NA 12.5 9.5 13 10
Fastovsky
et al., 2011
MPC/D 100/530 f 51 33.5 16.5 25.5 41 17 11 16.5 13.5
Fastovsky
et al., 2011
MPC/D 100/530 g 40.5 32 12 21 NA 15 10 13.5 15
Fastovsky
et al., 2011
MPC/D 100/530 h 35 27 12 22.5 NA 12 8.5 12.5 13.5
Source Specimen Number
Skull
length
total
Skull
length
basal
Frill
length
Frill
width
Jugal
boss
width
Orbit
length
Orbit
height
Frill
minimum
Frill
maximum
APPENDIX (continued).