Journal of Veterinary Behavior 9 (2014) 228e234

Contents lists available at ScienceDirect

Journal of Veterinary Behavior

journal homepage: www.journalvetbehavior.com

Research

Domestic dog skull diversity across breeds, breed groupings, and

genetic clusters
Dana Georgevsky a, Johanna J. Carrasco a, Michael Valenzuela b,
Paul Damien McGreevy a, *
a
Faculty of Veterinary Science, University of Sydney, Sydney, NSW 2006, Australia
b
Brain and Mind Research Institute, Faculty of Medicine, University of Sydney, Sydney, NSW 2006, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Domestic dogs have become a model organism for studying the extent and consequences of morpho-
Received 4 February 2013 logical diversity, especially in the skull. It has been demonstrated that Cephalic Index (CI, skull width/
Received in revised form skull length) correlates with central concentration of ganglion cells in the retina and with ventral
19 April 2014
rotation of the cerebral hemispheres. These changes may be reflected in the behavior of breeds with
Accepted 22 April 2014
different skulls shapes. This study explored skull variation in the breed groups (n ¼ 7) described by the
Available online 9 May 2014
Australian National Kennel Club to determine if CI differed significantly among the breed groups; groups
were expected to differ not least in behavior. The CI of 12 representative dogs (females, n ¼ 6; males,
Keywords:
dog
n ¼ 6) of the most popular breeds (n ¼ 80; total n ¼ 960 dogs) were measured. Multivariate analysis of
skull variance was performed to determine CI variance among the breed groups and between previously
breed groups reported clusters of breeds with similar DNA, which identifies common ancestry. Although CI differed
genotype significantly among some breed groups, neither the breed groupings nor the DNA clusters satisfactorily
dolichocephalic explained all the variance in CI. The results show that breed groupings and genetic clusters only partially
brachycephalic explain CI differences. They also suggest that CI is on a continuum and that the definition of three
categories of canine skull as dolichocephalic, mesocephalic, and brachycephalic may be overly arbitrary.
Ó 2014 Elsevier Inc. All rights reserved.

Introduction categories used by kennel clubs (KCs) and councils (the umbrella
organizations that govern the rules of dog showing and pedigree
Since domestication, dogs have played many important roles in dog breeding) include herding, retrieving, hunting, guarding, toy,
human life. These traditionally included hunting, guarding, and sporting, and miscellaneous. Of course, the dogs have also been
herding, and the dogs are now a valuable source of companionship used for haulage and indeed for their meat. Different behaviors
in many family homes. Historically, the number of dog breeds grew have been selected for or against during domestication, and these
rapidly from the mid-19th century, and there are now more than are now echoed in the written breed standards against which show
400 recognized breeds. Research on dog skull morphology has been dogs are judged. Yet because there are only limited opportunities to
conducted for almost as long. In particular, Studer (1901) used skull judge temperament and behavior in the show ring, breed selection
morphology to determine five putative original clusters within has tended to focus on appearance rather than behavior. Some have
modern breeds. Purebred dogs are grouped into categories that questioned the logic of this practice because modern purebred dogs
reflect not only their common ancestry and geographical origins are mainly kept as companions rather than for a breed-specific
but also their functional and behavioral attributes. Dog breeds vary purpose (McGreevy and Bennett, 2010).
not only in size but also in their conformation, coat attributes, and Genetic similarities and differences are another approach to
brain case positioning (Drake and Klingenberg, 2010). Common breed classification that avoids problems associated with subjective
value judgments. VonHoldt et al. (2011) classified breeds based on
DNA linkages to suggest 10 genetic clusters, such as toys, spaniels,
* Address for reprint requests and correspondence: Paul D. McGreevy, BVSc, PhD,
scent hounds, working dogs, mastiff-like dogs, small terriers, re-
MACVSc, Faculty of Veterinary Science, University of Sydney, Sydney, NSW 2006,
Australia. trievers, herders, sighthounds, and ancient/spitz breeds. Interest-
E-mail address: [email protected] (P.D. McGreevy). ingly, variability in skull shapes has recently been described as

1558-7878/$ e see front matter Ó 2014 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jveb.2014.04.007
 D. Georgevsky et al. / Journal of Veterinary Behavior 9 (2014) 228e234 229

having multiple genetic loci, particularly in relation to the degree of We also hypothesized that some variance in CI may reflect dif-
brachycephaly (Schoenebeck et al., 2012). ficulty in measuring skull difference among smaller dogs, so we
Morphological diversity has been described in numerous spe- explored relationships between a breed’s CI variance and some
cies, but dogs (Canis lupus familiaris) exhibit the greatest variation, indicators of the average size of dogs within each breed. Another
especially when compared with their ancestral relative, the wolf possible source of variability may be genetic diversity, so we esti-
(Canis lupus) (Drake, 2011). Most breeds of dogs have phylogenet- mated the size of the Australian population for each of the breeds in
ically novel skull shapes when compared with adult wolves (Drake question by examining the number of puppies registered within the
and Klingenberg, 2010). Breed-to-breed variability also applies to breed over the preceding 5-year period. A third possibility is that CI
the canine skull, so it is unsurprising that breed standards often variance reflects a lack of detail in the breed standard about the
devote considerable detail to skull attributes. Of the various preferred conformation of the skull. As a proxy for this detail, we
phenotypic attributes, it appears that skull shape has been sub- examined the role, if any, of the word count for the description of
jected to the most scientific scrutiny. Cephalic Index (CI) is a simple desirable features of the “head and skull” for each breed as specified
method of characterizing skull morphology. In humans, the terms in the ANKC breed standards. Finally, with a large audit of this na-
dolichocephalic, mesocephalic, and brachycephalic are applied to ture, we expected to be able to elucidate how the three categories of
skulls with different shapes based on the CI. The same terms are skull (dolichocephalic, mesocephalic, and brachycephalic) can be
used in canine anatomy, but the thresholds between one category distinguished from one another.
and another have not been defined.
Skull shape is important because it may have a bearing on brain
function and behavior, a question of enduring interest to evolu- Methods
tionary biologists and veterinary scientists (McGreevy et al., 2004).
Reduction in skull length in carnivores correlates with a reduction Cephalic Index
in olfactory lobe size, hypothetically owing to restriction in the
development of frontal brain regions (Gittleman, 1991). This also The method was designed to ensure that the representative dogs
appears to be associated with a truncated behavioral repertoire of each breed were measured for skull attributes. We sampled six
(Goodwin et al., 1997). There is recent evidence that the behavior of females and six males from each breed. To be included, dogs had to
dogs co-varies with skull shape (McGreevy et al., 2013). be aged 2 years or older and of show quality or from show-quality
Variation in skull length is also associated with differences in lines. Littermates of dogs that had already been measured were
retinal ganglion cell distribution (McGreevy et al., 2004), a po- avoided to ensure that the effects of a certain mating were not
tential explanation for an increased ability to focus in the central amplified in the study.
field of view rather than in the periphery and hence to respond to To be included, breeds had to:
human pointing gestures (Gácsi et al., 2009). More recently, by
examining magnetic resonance images (MRIs) of the brains of a 1. Be recognized by the ANKC
range of dog breeds, Roberts et al. (2010) showed that the relative 2. Be owned by the breeders registered with DogsNSW, and
reduction of skull length compared with width (measured by a 3. Have had more than 30 puppies registered nationally with the
higher CI) was significantly correlated with a progressive pitching ANKC in 2009.
of the brain, as well as with a downward shift in the position of
the olfactory lobe. This finding was confirmed by Hussein et al. Dogs were held by an assistant so that the nasal planum was
(2012). horizontal and were then photographed using a dorsoventral view
Heterochrony, a process that requires only simple genetic of the top of the head, which allowed the length and width of the
modifications to develop diversity, relies on changes in the timing skull to be measured. A standardized cloth strap with a rectangular
or rate of ontogenetic pathways (Gould, 1977). Although previous benchmark (2.5 cm  4.9 cm) was placed at the centre of the
research suggested that dogs evolved via heterochrony and are zygomatic arches. A finger placed on the occipital crest was placed
pedomorphic wolves (Coppinger and Coppinger, 1982, 2001; Frank and the photo was taken (Figure 1). The breed, dog’s name, and age
and Frank, 1982), a recent study has challenged the role of het- were all recorded.
erochrony and concluded that dogs are not pedomorphic wolves. It
used geometric morphometric analysis to investigate heterochronic
patterns of the adult dogs (n ¼ 677) representing 106 different
breeds, and compared them with an ontogenetic series of 401
wolves (Drake, 2011). This revealed that none of the modern breeds
of dogs had a cranial shape that resembled the cranial shapes of
wolves. The study confirmed the earlier finding of ventral tilting of
the cerebral hemispheres in association with brachycephaly
(Roberts et al., 2010). Moreover, it reported tilting in the opposite
direction in dolichocephalic and “down-face” breeds, such as the
bull terrier. An additional examination of juvenile wolf skulls
demonstrated that the position of the face and the neurocranium
remain in the same plane throughout maturation (Drake, 2011). In
summary, dogs show a small amount of genetic variation but an
enormous range of phylogenetically novel skull shapes.
In this study, we screened a wide array of common breeds to
assemble the largest report of CIs in dogs. This allowed us to Figure 1. A photograph of a Saluki, a dolichocephalic breed. For the present study, each
establish whether interbreed CI variance was related to Australian photograph was taken with the camera held horizontally, which allowed measure-
ments to be obtained for each dog’s skull length and width. The length was measured
National Kennel Club (ANKC) breed groupings, to known genetic
from the fingertip to the tip of the nose, and the width was measured from each
clusters (VonHoldt et al., 2011) and to groupings used by the zygomatic arch, which was displayed by the tape placed around the widest part of the
Federation Cynologique Internationale (FCI). dog’s head.
230 D. Georgevsky et al. / Journal of Veterinary Behavior 9 (2014) 228e234

Figure 2. Dendrogram of the breed Cephalic Index means. This graph represents similarities across breeds, regardless of the breed groups and genetic clusters. The lengths of the
horizontal lines that form the groups are indicative of how different neighboring groups are.
 D. Georgevsky et al. / Journal of Veterinary Behavior 9 (2014) 228e234 231

The team attended dog shows from November 2011 through Statistical analysis using ANKC classifications
May 2012. Most shows were held at the show grounds at Erskine
Park or Castle Hill. Breeds that had not been completely represented The interaction of breed groups and sex was significant
at shows over this period were then targeted at the Sydney Royal (P ¼ 0.010) in the analysis of CI.
Easter Show. As this is the largest show in NSW, the breeds that There was a highly significant difference (P < 0.001) in mean CI
were still not covered were excluded on the basis that their across breed groups. Of the covariates, breed height was significant
numbers were deemed too small to be representative of the breed (P < 0.001), whereas weight and word counts as specified in the
as a whole. Using this process, we accumulated data on 90 breeds. breed standards and the number of puppies registered were not
(P ¼ 0.68, 0.74, and 0.67, respectively).
The ANKC breed groups all had mean CI values that differed
Heights and weights of the breeds significantly from each other (P < 0.05) apart from groups 1 and 7
(toys and nonsporting), groups 2 and 5 and 6 (terriers and utility
The preferred heights and weights for exhibition purposes were and working), groups 3 and 4 (gun and hounds), and groups 3 and 5
drawn primarily from the ANKC breed standards. Data were avail- (gun and working).
able for the height of 68 breeds and the weight of 31. Where a range Figure 3 plots the ANKC breed group means. It is visually clear
was nominated, the median was calculated. Where either height or that CI does not in itself discriminate the groups, with large areas of
weight was not specified in the ANKC breed standards, the Amer- overlap in CI values: A CI value of say 60% could be a dog in any of
ican Kennel Club (AKC) standards were consulted (and data con- the breed groups.
verted from imperial to metric units). This source provided data on
a further 10 breeds for weight and 3 breeds for height. It was
selected because it provides more detail on heights and weights Statistical analysis using FCI
than other potential sources, such as the UK KC. This process left
only one breed, the bull terrier, for which neither height nor weight The FCI group 9 had mean CI values that were different from all
was specified in the ANKC or AKC breed standards. We therefore others (P < 0.05). The other groups all had mean CI values that were
assigned dogs of this breed to the same size category as the Staf- not significantly different from each other (P > 0.05) apart from
fordshire bull terrier. groups 2 and 7 and 10: pinscher and schnauzerdmolossoid
Measurements were obtained using a GNU image manipulation breedsdSwiss mountain and cattle dogs and other breeds and
program (http://www.gimp.org/) after normalization to the refer- scenthounds and related breeds (P < 0.01) and sighthounds
ence rectangle. The CI was calculated as 100 anterior-posterior (P < 0.01), groups 3 and 7 and 10: terriers and scenthounds and
length divided by skull width. Published genotype groups were related breeds (P < 0.05) and sighthounds (P < 0.05), groups 5 and 7
used to cluster breeds based on VonHoldt et al. (2011) as follows: and 10: spitz and primitive types and scenthounds and related
toys, spaniels, scent hounds, working dogs, mastiff-like, small ter-
riers, retrievers, herding, sighthounds, ancient/spitz breeds. The
ANKC assigns breeds to 7 groups, namely Group 1 (Toys), Group 2
(Terriers), Group 3 (Gundogs), Group 4 (Hounds), Group 5 (Working
Dogs), Group 6 (Utility), and Group 7 (Nonsporting). In contrast, the
FCI assigns breeds to 10 groups. These are based on different
themes such as appearance or use. The 10 groups are sheepdogs
and cattle dogs (except Swiss cattle dogs), pinscher and schnau-
zerdmolossoid breedsdSwiss mountain and cattle dogs and other
breeds, terriers, dachshunds, spitz and primitive types, scent-
hounds and related breeds, pointers and setters, retrieversd-
flushing dogsdwater dogs, companion and toy dogs, and
sighthounds.

Statistical analysis

We used GenStat Version 14 (VSN International, Hemel Hemp-

stead, UK) to compare mean CI across breed groups and across
genetic clusters. We conducted an analysis of variance and a
restricted maximum likelihood (REML) regression with the
threshold set at P-value lower than 0.05. The latter included sex as a
factor and checked whether height, weight, word count of “head
and skull” descriptions and the number of puppies registered were
significant covariates. The reasons for preferring REML over
maximum likelihood for estimation of the variances are that the
estimators are less biased and that it produces an appropriate de-
gree of freedom correction. Figure 3. Range in the cephalic index (CI) means for the 7 classes of the breeds
recognized by the ANKC. Data are based on a total of 960 dogs from 80 breeds. The
breed groups from 1e7 are: 1dToys (mean CI ¼ 72.8; dogs, n ¼ 156; breeds, n ¼ 13);
Results 2dTerriers (mean CI ¼ 51.7; dogs, n ¼ 132; breeds, n ¼ 11); 3dGun dogs (mean
CI ¼ 72.1; dogs, n ¼ 132; breeds, n ¼ 11); 4dWorking dogs (mean CI ¼ 62.8; dogs,
n ¼ 180; breeds, n ¼ 15); 5dHounds (mean CI ¼ 72.9; dogs, n ¼ 108; breeds, n¼9);
A convenient way of graphing so many means is to construct a 6dUtility (mean CI ¼ 61.5; dogs, n ¼ 168; breeds, n ¼ 14); and 7dNonsporting (mean
dendrogram of the means. This is presented in Figure 2. CI ¼ 54.2; dogs, n ¼ 84; breeds, n ¼ 7). ANKC ¼ Australian National Kennel Club.
232 D. Georgevsky et al. / Journal of Veterinary Behavior 9 (2014) 228e234

Figure 5. Distribution of the mean cephalic index (CI) (and standard deviation) for the
Figure 4. Distribution of the mean cephalic index (CI) (and standard deviation) for the 10 genetic clusters of dog breeds. The average number of breeds in each group is 5. The
80 breeds in our series, as grouped by the FCI. Data are based on a total of 960 dogs DNA groupings labeled above correspond to an established arrangement of genetic
from 80 breeds. The breed groups from 1e10 are: 1dSheepdogs and cattle dogs clusters (VonHoldt et al., 2011). The genetic clusters from 1e10 are: 1dToys (mean
(except Swiss cattle dogs; mean CI ¼ 54.7; dogs, n ¼ 120; breeds, n ¼ 10); 2dPinschers CI ¼ 81.3; dogs, n ¼ 84; breeds, n ¼ 7); 2dSpaniels (mean CI ¼ 52.9; dogs, n ¼ 36;
and schnauzersdmolossoid breedsdSwiss mountain and cattle dogs and other breeds breeds, n ¼ 3); 3dScenthounds (mean CI ¼ 55.8; dogs, n ¼ 60; breeds, n ¼ 5);
(mean CI ¼ 62.7; dogs, n ¼ 156; breeds, n ¼ 13); 3dTerriers (mean CI ¼ 59.5; dogs, 4dWorking dogs (mean CI ¼ 66.8; dogs, n ¼ 108; breeds, n ¼ 9); 5dMastiff-like (mean
n ¼ 120; breeds, n ¼ 10); 4dDachshunds (mean CI ¼ 50.7; dogs, n ¼ 12; breeds, n ¼ 1); CI ¼ 75.3; dogs, n ¼ 36; breeds, n ¼ 3); 6dSmall terriers (mean CI ¼ 58.6; dogs, n ¼ 60;
5dSpitz and primitive types (mean CI ¼ 60.4; dogs, n ¼ 120; breeds, n ¼ 10); breeds, n ¼ 5); 7dRetrievers (mean CI ¼ 58.6; dogs, n ¼ 60; breeds, n ¼ 5); 8dHerding
6dScenthounds and related breeds (mean CI ¼ 53.0; dogs, n ¼ 48; breeds, n ¼ 4); (mean CI ¼ 54.6; dogs, n ¼ 72; breeds, n ¼ 6); 9dSighthounds (mean CI ¼ 47.6; dogs,
7dPointers and setters (mean CI ¼ 47.7; dogs, n ¼ 72; breeds, n ¼ 6); n ¼ 48; breeds, n ¼ 4); and 10dAncient/Spitz (mean CI ¼ 54.5; dogs, n ¼ 84; breeds,
8dRetrieversdflushing dogsdwater dogs (mean CI ¼ 54.9; dogs, n ¼ 72; breeds, n ¼ 7).
n ¼ 6); 9dCompanion and toy dogs (mean CI ¼ 77.6; dogs, n ¼ 168; breeds, n ¼ 14);
and 10dSighthounds (mean CI ¼ 47.6; dogs, n ¼ 72; breeds, n ¼ 6). FCI ¼ Federation
Hussein et al., 2012) may be redundant. We propose that readers
Cynologique Internationale.
can use CI and the dendrogram to decide how well different breeds
align using this metric. We note that the dendrogram shows that
breeds (P < 0.05) and sighthounds (P < 0.05). The means are plotted the papillon and the chihuahua are at what would be termed the
in Figure 4. brachycephalic end of the spectrum. These petite breeds lack the
strong but crowded dentition of typically brachycephalic breeds
such as the pug and French bulldog. The CI fails to discriminate
Statistical analysis using genetic clusters
adequately among these four breeds and so any analyses that align
skull morphology with behavior are likely to suffer from lack of CI
The interaction of breed groups and sex was almost significant
variation at this end of the spectrum. At the other end of the
(P ¼ 0.052) in the analysis of CI. There was more overlap of CI
spectrum, however, we note that Borzois have a particularly long
among the 10 genetic clusters. The genetic clusters all had mean CI
skull and that representative numbers of dogs of this breed will be
values that were not significantly different from each other
especially important in any analyses that align skull morphology
(P > 0.05) apart from clusters 1 and 2, 3, 4, 6, 7, 8, 9, and 10: toys and
with behavior.
spaniels (P < 0.001), scenthounds (P < 0.001), working dogs
Our results show that the CI of dogs is fully explained by neither
(P < 0.001), small terriers (P < 0.001), retrievers (P < 0.001),
the breed groups (ANKC and FCI) nor the genetic clusters. That said,
herding (P < 0.001), sighthounds (P < 0.001), and ancient/spitz
they show that the mean CIs of some breed groups are more similar
(P < 0.001); clusters 2 and 5: spaniels and mastiff-like (P ¼ 0.016);
than others. This is evident, for example, in the gun dogs and
clusters 2 and 5: spaniels and scenthounds (P ¼ 0.018); clusters 4
working dogs and may suggest that these breeds align as more
and 8 and 9: working dogs, herding (P ¼ 0.039) and sighthounds
closely similar in terms of conformation than other more diverse
(P ¼ 0.005); and clusters 5 and 6, 7, 8, 9, 10: mastiff-like and small
groups. We know that breed groupings are largely based on the
terriers (P ¼ 0.042); retrievers (P ¼ 0.042); herding (P ¼ 0.01);
original purpose of the breeds within a group, an attribute that
sighthounds (P ¼ 0.002); and ancient/spitz (P ¼ 0.008). The means
generally reflects behavior more than conformation (McGreevy,
are plotted in Figure 5.
2009). So, it may be that behavioral differences among breed
groups may align with, or even be predicted by, skull attributes. It is
Discussion something of a surprise to find that the FCI clustering of scent-
hounds separate from sighthounds revealed very little difference in
The dendrogram of CI shows how breeds with different skull skull morphology. It may be that the hounds from both groups have
morphology can be clustered together. To an extent, it shows how CI in common a tendency to chase, regardless of the modality of prey
is on a continuum and why the imposition of three categories of stimulus. Further analysis of breed-specific behavioral tendencies
skull (dolichocephalic, mesocephalic, and brachycephalic; e.g., see as they relate (or otherwise) with CI is required.
 D. Georgevsky et al. / Journal of Veterinary Behavior 9 (2014) 228e234 233

Breed groups used for showing purposes are not entirely Conclusions
consistent. For example, breed groupings vary among countries.
From one country to another, there are a different number of breed This study has contributed to the knowledge of morphological
groups, as well as different breeds included within each group diversity in domestic dogs, as it has identified skull variation among
(Turcsán et al., 2011). The ANKC has seven groups, namely gundogs, dogs of different breed groups and genetic clusters. Understanding
working, toys, terriers, hounds, utility, and nonsporting. Similarly, morphological diversity among breed groups may contribute to an
the UK KC has 7 slightly different groups, namely gundogs, hounds, appreciation of the specific behavioral tendencies dogs show and
pastoral, terrier, toy, utility, and working (Turcsan et al., 2011). may help to explain behaviors that are expected in dogs with a
Meanwhile, the FCI has 10 groups. Despite these differences, in given skull morphology. Further work should determine the effect
general, breed-specific behavioral tendencies remain a funda- of skull variation and behaviors of breeds.
mental attribute in these categorizing systems.
The ANKC breeds that exhibit the most within-group skull
Acknowledgments
variation in the present study were in the toy group and the
nonsporting group. Similarly, the FCI breeds that exhibit the most
The authors are most grateful to all participating dog breeders
within-group skull variation in the present study were the com-
who allowed us to take photographs of their dogs throughout the
panion and toy dogs. This may simply reflect the considerable di-
data collection period of this study. Dr. Navneet Dhand and Dr. Mick
versity of these breeds, which generally had no working origins and
O’Neill are warmly thanked for their guidance in the statistical
therefore were, and still are, valued chiefly for their appearance.
analysis of this study. Associate Professor Peter Thomson is thanked
This corresponds with the suggestion that skull variation is largely a
for his work on the dendogram that appears as Figure 2.
result of the selection criteria humans are placing on breeds
(Schoenebeck et al., 2012). The head shapes of these breeds, and the
degree of skull variation is a result of many quantitative trait loci Conflict of interest
(QTL) interactions, in particular bone morphogenetic protein 3,
which alters the conformation of the head of dogs (Schoenebeck The authors declare no conflict of interest. The idea for the
et al., 2012). article was conceived by Paul McGreevy. The experiments were
A further significant finding from the present study is that the designed by Paul McGreevy. The experiments were performed by
clusters of breeds with similar genotype profile show more skull Johanna Carrasco and Dana Georgevsky. The data were analyzed by
variation than the breed groups. In summary, there are twice as Dr. Mick O’Neill and Dr. Navneet Dhand. The article was written by
many significant differences among the genetic clusters than Paul McGreevy, Dana Georgevsky, and Michael Valenzuela. The
among the breed groups. This suggests that, regardless of their unfunded study was entirely noninvasive.
genetic similarity, show dogs may be selected to show distinct
phenotypic traits that possibly reflect their original function. It may
be that this phenotypic attribute varies more than genotypes alone References
owing to the selective forces that operate in pedigree dog breeding.
Carrasco, J.J., Georgevsky, D., Valenzuela, M., McGreevy, P.D., 2013. Sexual dimor-
It may even be that judges who focus on a particular group may phism in the skull morphology of domestic dogs. J. Vet. Behav.: Clin. Appl. Res 9,
have a preferred range of CI that they favor across the breeds within 43e46.
that group. Coppinger, L., Coppinger, R.P., 1982. Livestock-guarding dogs that wear sheep’s
clothing. Smithsonian 13, 64e73.
The interactions between sex, breed group, and CI were signif- Coppinger, R.P., Coppinger, L., 2001. Dogs: A Startling New Understanding of Canine
icant. This sexual dimorphism has been reported elsewhere Origin. Behavior and Evolution Scribner, New York, NY.
(Carrasco et al., 2013). A significant influence on CI was also the size Drake, A.G., Klingenberg, C.P., 2010. Large-scale diversification of skull shape in
domestic dogs: disparity and modularity. Am. Nat. 175, 289e301.
of the breed, as stated in the breed standards. However, some
Drake, A.G., 2011. Dispelling dog dogma: an investigation of heterochrony in dogs
caution is required in interpreting this finding. In the breed stan- using 3D geometric morphometric analysis of skull shape. Evol. Dev. 13, 204e213.
dards of the smaller breeds, the average height range is the same for Frank, H., Frank, M.G., 1982. On the effects of domestication on canine social
both sexes, suggesting either no dimorphism or differences that are development and behavior. Appl. Anim. Ethol. 8, 507e525.
Gácsi, M., McGreevy, P., Kara, E., Miklósi, Á., 2009. Effects of selection for coopera-
too small to standardize or report. In contrast, for the larger breeds, tion and attention in dogs. Behav. Brain. Func. 5, 31.
the preferred heights are larger for male dogs. It is possible that the Gittleman, J.L., 1991. Carnivore olfactory bulb size: allometry, phylogeny and ecol-
apparent influence of size on CI also reflects a possible shortfall in ogy. J. Zool. 225, 253e272.
Goodwin, D., Bradshaw, J.W.S., Wickens, S.M., 1997. Paedomorphosis affects visual
our method. Its reliance on linear measurements of skull width and signals of domestic dogs. Anim. Behav. 53, 297e304.
length when viewed from above fails to capture potentially Gould, S.J., 1977. Ontogeny and Phylogeny. Belknap Press, Harvard, MA.
important differences in the lateral profile of dogs’ skulls. Breeds Hussein, A.K., Sullivan, M., Penderis, J., 2012. Effect of brachycephalic, mesatice-
phalic, and dolichocephalic head conformations on olfactory bulb angle and
such as the chihuahua, with very small heads, may cause the level of orientation in dogs as determined by use of in vivo magnetic resonance im-
within-breed variation of the width-to-length ratio to fall. Dogs aging. Am. J. Vet. Res. 73, 946e951.
with dramatically different head morphologies can have virtually McGreevy, P.D., Bennett, P.B., 2010. Challenges and paradoxes in the companion
animal niche. Darwinian selection, selective breeding and the welfare of ani-
identical CI values. So it is, perhaps, not so surprising that the dif- mals. Anim. Welf. 19, 11e16.
ferences among breeds are often small and uninformative. Perhaps McGreevy, P.D., Grassi, T.D., Harman, A.M., 2004. A strong correlation exists between
analysis of individual lengths and widths taken on skulls and not the distribution of retinal ganglion cells and nose length in the dog. Brain.
Behav. Evol. 63, 13e22.
living dogs would shed additional light on this issue. There are
McGreevy, P., 2009. A Modern Dog’s Life. UNSW Press, Randwick, NSW, Australia.
several methods in common use in phylogenetic and morphometric McGreevy, P.D., Georgevsky, D., Carrasco, J., Valenzuela, M., Duffy, D.L., Serpell, J.A.,
analysis that would allow a finer analysis of the effects of CI. Using 2013. Dog behavior co-varies with height, bodyweight and skull shape. PLoS
skulls rather than photographs of live dogs would both improve the ONE 8 (12), e80529. http://dx.doi.org/10.1371/journal.pone.0080529.
Roberts, T., McGreevy, P., Valenzuela, M., 2010. Human induced rotation and reor-
accuracy of measurements and also make considerably more ganization of the brain of domestic dogs. PLoS One, e11946.
anatomical variables available for investigation. Catalogued collec- Schoenebeck, J.J., Hutchinson, S.A., Byers, A., Beale, H.C., Carrington, B., Faden, D.L.,
tions of dog skulls such as those held by The Albert Heim Foun- Rimbault, M., Decker, B., Kidd, J.M., Sood, R., Boyko, A.R., Fondon 3rd, J.W.,
Wayne, R.K., Bustamante, C.D., Ciruna, B., Ostrander, E.A., 2012. Variation of
dation for Canine Research in Switzerland could be used for this BMP3 contributes to dog breed skull diversity. PLoS Genet 8, e1002849. http://
sort of study in future. dx.doi.org/10.1371/journal.pgen.1002849.
234 D. Georgevsky et al. / Journal of Veterinary Behavior 9 (2014) 228e234

Studer, T., 1901. Prehistoric dogs and their relation to modern breeds. Abhandlungen Vonholdt, B.M., Pollinger, J.P., Earl, D.A., Knowles, J.C., Boyko, A.R., Parker, H.,
der Schweizerischen palontologischen Gesellschaft (Memoires de la Socit Pal- Geffen, E., Pilot, M., Jedrzejewski, W., Jedrzejewska, B., Sidorovich, V.,
ontologie Suis) 28, 1e127. Greco, C., Randi, E., Musiani, M., Kays, R., Bustamante, C.D., Ostrander, E.A.,
Turcsán, B., Kubinyi, E., Miklósi, Á., 2011. Trainability and boldness traits differ be- Novembre, J., Wayne, R.K., 2011. A genome-wide perspective on the evolu-
tween dog breed clusters based on conventional breed categories and genetic tionary history of enigmatic wolf-like canids. Genome Res. 21 (8), 1294e1305.
relatedness. Appl. Anim. Behav. Sci. 132, 61e70. PMID: 21566151.