Rethinking dog domestication by integrating genetics,

archeology, and biogeography

Greger Larson a,1, Elinor K. Karlsson b,c, Angela Perri a, Matthew T. Webster d, Simon Y. W. Ho e, Joris Peters f,
Peter W. Stahl g, Philip J. Piperh,i, Frode Lingaasj, Merete Fredholmk, Kenine E. Comstockl, Jaime F. Modianom,n,
Claude Schellingo, Alexander I. Agoulnikp, Peter A. Leegwaterq, Keith Dobneyr, Jean-Denis Vignes, Carles Vilàt,
Leif Anderssond,u, and Kerstin Lindblad-Tohb,d
a
Durham Evolution and Ancient DNA, Department of Archaeology, University of Durham, Durham DH1 3LE, United Kingdom; bBroad Institute of MIT
and Harvard, Cambridge MA 02142; cFaculty of Arts and Sciences Center for Systems Biology, Harvard University, Cambridge MA 02138; dScience for Life
Laboratory Uppsala, Department of Medical Biochemistry and Microbiology, Uppsala University, SE-751 23 Uppsala, Sweden; eSchool of Biological Sciences,
University of Sydney, Sydney NSW 2006, Australia; fVeterinary Sciences Department, Institute of Palaeoanatomy, Domestication Research and the History of
Veterinary Medicine, Ludwig-Maximilian University, 80539 Munich, Germany; gDepartment of Anthropology, University of Victoria, Victoria, BC, Canada V8W
2Y2; hSchool of Archaeology and Anthropology, Australian National University, Canberra, Australian Capital Territory 200, Australia; iArchaeological Studies
Program, University of the Philippines, Diliman, 1101, Quezon City, Philippines; jDepartment of Basic Sciences and Aquatic Medicine, Division of Genetics,
Norwegian School of Veterinary Science, 0033 Oslo, Norway; kFaculty of Life Sciences, Division of Genetics and Bioinformatics, Department of Basic Animal
and Veterinary Sciences, University of Copenhagen, 7-1870 Frederiksberg C, Denmark; lUniversity of Michigan-Dearborn, Dearborn, MI 48128; mDepartment
of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, St. Paul, MN 55108; nMasonic Cancer Center, University of Minnesota,
Minneapolis, MN 55455; oDepartment of Animal Sciences, Swiss Federal Institute of Technology Zurich and Vetsuisse Faculty Zurich, University of Zurich, 8092
Zurich, Switzerland; pDepartment of Human and Molecular Genetics, Herbert Wertheim College of Medicine, Florida International University, Miami, FL
33199; qDepartment of Clinical Sciences of Companion Animals, Faculty of Veterinary Medicine, Utrecht University, 3508 TD Utrecht, The Netherlands;
r
Department of Archaeology, University of Aberdeen, Aberdeen AB24 3UF, Scotland, United Kingdom; sDépartement “Ecologie et Gestion de la Biodiversité”,
Muséum National d’Histoire Naturelle, “Archéozoologie, Archéobotanique: Sociétés, Pratiques et Environnements,” Centre National de la Recherche
Scientifique-Institut Ecologie et Environement, F-75005 Paris, France; tConservation and Evolutionary Genetics Group, Doñana Biological Station, 41092
Seville, Spain; and uDepartment of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, SE-75123 Uppsala, Sweden

Edited by Joachim Burger, Johannes Gutenberg-University, Mainz, Germany, and accepted by the Editorial Board April 17, 2012 (received for review February
20, 2012)

The dog was the first domesticated animal but it remains uncertain key morphological characters established by zooarcheologists to
when the domestication process began and whether it occurred just differentiate domestic animals from their wild wolf ancestors (e.g.,
once or multiple times across the Northern Hemisphere. To ascertain size and position of teeth, dental pathologies, and size and pro-
the value of modern genetic data to elucidate the origins of dog portion of cranial and postcranial elements) were not yet fixed
domestication, we analyzed 49,024 autosomal SNPs in 1,375 dogs during the initial phases of the domestication process. Further-
(representing 35 breeds) and 19 wolves. After combining our data more, the range of natural variation among these characters in
with previously published data, we contrasted the genetic signatures ancient wolf populations and the time it took for these traits to
of 121 breeds with a worldwide archeological assessment of the appear in dogs are unknown. Free-ranging wolves attracted to the
earliest dog remains. Correlating the earliest archeological dogs with refuse generated by human camps most likely followed a com-
the geographic locations of 14 so-called “ancient” breeds (defined by mensal pathway to domestication that was neither deliberate nor
their genetic differentiation) resulted in a counterintuitive pattern. directed (3). Because the process was not unidirectional, the
First, none of the ancient breeds derive from regions where the oldest telltale traits archeologists use to differentiate wolves and dogs
archeological remains have been found. Second, three of the ancient probably took numerous generations to become apparent in the
breeds (Basenjis, Dingoes, and New Guinea Singing Dogs) come from
archeological record.
Despite the difficulties associated with the use of archeological
regions outside the natural range of Canis lupus (the dog’s wild an-
evidence to pinpoint the timing of domestication, there is a general
cestor) and where dogs were introduced more than 10,000 y after
consensus that domestic dogs were present in the Levant (including
domestication. These results demonstrate that the unifying charac-
Cyprus), Iraq, Northern China, and the Kamchatka peninsula in
teristic among all genetically distinct so-called ancient breeds is a lack Far Eastern Russia by ∼12,000 y ago, and in western Europe a few
of recent admixture with other breeds likely facilitated by geographic millennia before that. Recent studies have made claims that do-
and cultural isolation. Furthermore, these genetically distinct ancient mestic (or incipient) dogs were present even earlier during the Late
breeds only appear so because of their relative isolation, suggesting Pleistocene in Belgium (4), the Czech Republic (5), and south-
that studies of modern breeds have yet to shed light on dog origins. western Siberia (6). Morphological analyses suggest that although
We conclude by assessing the limitations of past studies and how some of the early canid remains possess characteristics broadly
next-generation sequencing of modern and ancient individuals may similar to those found in modern dogs, it remains possible that the
unravel the history of dog domestication. bones represent either wolves going through the initial phases of an
incomplete domestication process (6) or a morphologically distinct
genomics | phylogeography local, now-extinct population of wolves.
The use of more advanced morphometric analyses is allowing
zooarcheologists to have greater confidence in identifying early
D arwin speculated about the origins of several domestic ani-
mals and suggested that, given the vast morphological vari-
ation across numerous breeds, dogs must have had more than
one wild ancestor (1). Recent genetic studies, however, support Author contributions: G.L., E.K., and K.L.-T. designed research; G.L., E.K., A.P., F.L., M.F., K.E.C.,
the notion that dogs are descended exclusively from the gray wolf J.F.M., C.S., A.I.A., P.L., C.V., and K.L.-T. performed research; G.L., E.K., M.T.W., S.Y.W.H., J.P.,
P.W.S., P.J.P., J.-D.V., C.V., L.A., and K.L.-T. analyzed data; and G.L., E.K., A.P., M.T.W., S.Y.W.H.,
(Canis lupus) (2). J.P., P.W.S., P.J.P., J.F.M., K.D., J.-D.V., C.V., L.A., and K.L.-T. wrote the paper.
Beyond questions regarding wild ancestry, geneticists and
The authors declare no conflict of interest.
generations of archeologists have investigated not only how and
why dogs were domesticated, but also when, where, and how many This article is a PNAS Direct Submission. J.B. is a guest editor invited by the Editorial Board.

times it may have occurred. Unique among all domestic animals, Freely available online through the PNAS open access option.
the first unambiguous domestic dogs precede the appearance of 1
To whom correspondence should be addressed. E-mail: greger.larson@durham.ac.uk.
settled agriculture in the archeological record by several thousand This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
years. Identifying the earliest dogs is difficult, however, because 1073/pnas.1203005109/-/DCSupplemental.

8878–8883 | PNAS | June 5, 2012 | vol. 109 | no. 23 www.pnas.org/cgi/doi/10.1073/pnas.1203005109

dogs (7). Given the geographical breadth of these finds, arche- recent historical records clearly demonstrate that most modern
ologists have (generally) been reluctant to postulate exact loca- breeds experienced significant population fluctuations within the
tions where dogs may have been domesticated. Instead, they past 100 y (Table S1). Here, we only use the term “breed” when
have broadly accepted the plausibility of the existence of nu- referring to modern dog breeds recognized by kennel clubs.
merous, independent centers of dog domestication beginning in To test the branching pattern of the previously identified basal
the Late Pleistocene (8). breeds and to assess the status of unstudied breeds (Table 1 and
Many genetic studies of modern dogs and wolves have been less Table S1), we used 49,024 SNPs typed in 19 wolves and 1,375
circumspect. Armed first with fragments of mitochondrial DNA dogs from 35 breeds. In addition, we compiled a broad temporal
and molecular clocks, the authors of one study concluded that and geographic survey of dog domestication by undertaking
dogs were domesticated 135,000 y ago (9). A separate study later a global examination of the archeological record (Tables S2 and
analyzed a similar mitochondrial fragment sequenced from 654 S3). By comparing the zooarcheological evidence with the geo-
dogs and, on the basis of regional patterns of modern dog diversity,
graphical origins of the total set of modern breeds, we estab-
deduced that dogs were domesticated just once in East Asia (10).
Both of these claims have since been challenged. First, it is lished a framework for understanding why some breeds have
highly likely that the use of deep fossil calibrations for molecular retained basal signatures and why most have not.
clocks has led to a significant overestimation of the timing of dog
Results and Discussion
domestication (11). Second, analyses of African street dogs sug-
gested that a single East Asian origin was too simplistic (12). A Genetic History of Modern Breeds. A neighbor-joining phylogenetic
study of 48,000 SNPs in 912 dogs and 225 gray wolves concluded tree inferred using our data (Fig. 1) was broadly similar to those
that both East Asian and Near Eastern wolf populations con- described previously (13, 15). A deep genetic split is evident be-
tributed DNA to modern dog breeds (13). Other studies that have tween Old World and New World wolves (Table S4) at the base of
incorporated nuclear markers also suggest diverse geographic the tree. From there, high bootstrap values (>95%) support the
origins of dogs (14), and with the application of a broader, more basal position and genetic distinctiveness of the so-called ancient
integrated approach, the genetic and archeological perspectives (basal) breeds: the Akita, Basenji, Eurasier, Finnish Spitz, Saluki,
have become more closely aligned. However, despite the volume

ANTHROPOLOGY
and Shar-Pei (Fig. 1 and Table 1). Although the relationships
of new data, the estimates of when, where, and how many times between numerous breeds that have been crossed recently (e.g.,
dogs were domesticated remain disconcertingly imprecise. Dachshunds) are well supported, and although each of nonbasal
One significant insight that genetic studies have yielded, using breeds is strongly monophyletic, the relationships between them
both microsatellites (15) and SNPs (13), is the identification of are poorly resolved (Fig. 1).
several genetically divergent modern dog breeds in well-sup- When our results are combined with those from the two previous
ported basal positions on phylogenetic trees. This early-branching studies (13, 15), the total number of basal breeds increases to 16.
pattern has been used to designate these breeds as “ancient” (13). Two of these basal breeds have shallow histories. The American
To avoid conflating genetic differentiation with presumed ancient
Eskimo breed was deliberately created by crossing Keeshonds,
heritage (16), we will instead refer to these lineages as “basal.”
The term “breed” is also problematic. The focus on general Volpinos, and Pomeranians, and after World War II, Japanese
classes of dogs (e.g., sight hounds, scent hunters, shepherd dogs, Spitzes may also have been incorporated (18). The name “Amer-
and giant dogs) likely has prehistoric roots and led to the de- ican Eskimo” was derived from the kennel that originally began
velopment of broadly distinct forms of dogs. For example, three breeding them, despite the fact that the breed never had an asso-
differently sized dog types have been recorded at the 8,000-y-old ciation with Inuits. The highly mixed heritage of the breed is evi-
Svaerdborg site in Denmark (17). Modern breeding practices, dent from its position on the phylogenetic tree, which is depen-
focused on distinct breeds with strict aesthetic requirements and dent on the choice of analytical technique. The American Eskimo
closed bloodlines, only emerged in the 19th century, and claims appears alongside the basal Samoyed in trees estimated using 10-
for the antiquity (and long-term continuity) of modern breeds are SNP windows; however, it is positioned next to Pomeranians on
based upon little or no historical or empirical evidence. In fact, a tree inferred using individual SNPs (13).

Table 1. A list of 16 breeds that were either labeled “ancient” in previous publications or were
identified as basal in this study
Breed Parker et al. (15) Vonholdt et al. (13) Present study
1
Afghan Hound y y
Akita2 y y y
Alaskan Malamute3 y y
American Eskimo (recent) y*
Basenji4 y y y
Canaan5 y
Chow Chow6 y y
Dingo7 y
Eurasier (recent) y
Finnish Spitz8 y
New Guinea singing dog9 y
Saluki10 y y y
Samoyed11 n y
Shar-Pei12 y y y
Shiba Inu13 y
Siberian Husky14 y y

The letters “y”, “n”, and “y*” indicate basal breeds, nonbasal breeds, and an inconclusive result, respectively.
The absence of a letter indicates the breed was not a part of the study in question. Superscripted numbers
following breed names correlate with the numbers under the dog symbols in Fig. 2. Detailed descriptions of
these breeds are provided in Table S1.

Larson et al. PNAS | June 5, 2012 | vol. 109 | no. 23 | 8879

 Chinese Shar-Pei (50)
1000
Akita (4)
Basenji (10)
Saluki (10)
Afghan (1)
Eurasier (49)
Finnish Spitz (68)
Tibetan Terrier (17)
Pekingese (8)
Pug (10)
Chinese Crested (39)
Schipperke (24)
White Shepherd (15)
Leonberger (34)
Doberman Pinscher (203) Fig. 1. A neighbor-joining tree depicting the rela-
Rottweiller (21)
Boxer tionships between 35 breeds (with sample sizes) and
(94) rooted with New and Old World Wolves. All clades
English Bulldog (2)
0 Mastiff (9) have been collapsed. Gray branches are poorly
Neopolitan Mastiff (11) supported, whereas black branches and black circles
Shetland Sheepdog (49) indicate bootstrap values >95%. Clade colors depict
Pembroke Welsh Corgi (56) breeds that retain a basal signature (red), non-Eu-
Greyhound (38)
Australian Cattle Dog (10)
ropean breeds that are not basal (blue), and Euro-
Kerry Blue Terrier (3) pean breeds that are rumored to have deep histo-
Poodle (49) ries but are not basal (brown). The well-supported
Lagotto Romagnolo (24) English Cocker Spaniel relationships between Rottweilers and Doberman
(109) Pinschers, Neapolitan Mastiffs, Mastiffs, English Bull-
German Shorthaired Pointer (10)
dogs, Boxers, Shetland Sheepdogs, and Pembroke
Basset Hound (10)
English Setter (10)
Welsh Corgis are the result of known or suspected
Dachshund (24) recent admixture between these breeds. The well-
Golden Retriever support relationship between Dachshunds and En-
(303)
Labrador Retriever (1) glish Setters reflects a recent interbreeding between
Wolves (New World) (6) the Dachshund individuals used in this study with
Wolves (Old World) (14) English Setters.

The Eurasier is also a recently created breed, developed de- (18) and Italian Greyhounds (22) vanished completely and many
liberately and fixed in the 1960s by mixing Chow Chows with other breeds suffered significant bottlenecks (Table S1). Bol-
Keeshonds and a single Samoyed (18). Because the majority of stering or recreating these breeds was accomplished by crossing
the breeds used to create Eurasiers possess basal signatures (13, numerous other breeds, a practice that obscured whatever ge-
15), Eurasiers also appear basal, although they are the only breed netic signatures of their early heritage that existed before the
whose monophyly is weakly supported (33% bootstrap value). World Wars, and ultimately led to highly inbred modern popula-
The remaining 14 basal breeds [including Samoyeds, which do tions (23). Interestingly, the recent genetic homogenization has
not appear basal on the phylogenetic tree inferred from micro- occurred despite the increase in phenotypic disparity as breeders
satellite data (15), but are basal when using SNPs (13)] have have simultaneously closed breeding lines and selected for ex-
generally avoided admixture with other breeds (Table S1). This treme morphological traits (24).
avoidance is probably the only reason why they retain a genetic Even the basal breeds identified in this and other studies expe-
legacy that extends beyond the age of modern breeding and the rienced recent and significant demographic change. The Shiba Inu
establishment of kennel clubs during the second half of the 19th faced extinction in World War II and the modern breed is an
century (19). amalgamation of three isolated and distinct Japanese lineages
Despite the long history of human selection for specific dog (18). The Finnish Spitz, supposedly used for millennia by Finno-
forms, there is a major disconnect between truly ancient dogs and Ugric people, was nearly extinct by 1880. A single breeder, Hugo
modern breeds. For example, unsubstantiated claims have been Roos, set out to rescue the type by traveling to remote villages and
made for the antiquity of the modern Irish Wolfhound. Wolf- collecting the few remaining individuals least likely to have been
hound-type dogs were used to hunt wolves across Europe. In crossed (accidentally or purposely) with other breeds (18). The fact
Ireland, wolves were exterminated by 1786 (20), after which the that Finnish Spitzes retain a basal genetic signature is testament to
demand for Wolfhounds plummeted, and by 1840 the type was the success of Roos’s efforts to obtain uncrossed individuals.
either extinct or all but extinct. George Augustus Graham re- With the exception of the Alaskan Malamute, all 14 basal
vitalized (or recreated) the form by breeding one possible wolf- breeds have geographic origins in the Old World (Table S1); this
hound to Scottish Deerhounds, and then incorporated Borzois is despite the fact that dogs were an integral part of the human
and Great Danes to create the modern breed that retained the occupation of the New World and that several modern breeds,
aesthetic of the original form, but not the genetic ancestry (18). including the Chihuahua, are thought to have been at least partly
The story of the Irish Wolfhound is not unusual. Although the derived from domestic dogs native to the New World. The general
origin myths of the Cardigan and Pembroke Welsh Corgis state lack of basal lineages in the Americas is likely because of the fact
that their respective introductions to England differed by 2,000 y that European breeds, initially introduced only 500 y ago, have
(21), both types were allowed to interbreed for centuries before overwhelmed the native lineages. This finding was demonstrated
being split into two modern breeds in the 1920s (18). Whatever by a recent study of mitochondrial variation among street dogs in
their deeper history, these breeds form strongly supported sister South America, which concluded native maternal lineages were
clades on phylogenetic trees (13), meaning that their pre- almost entirely absent in New World dogs (25).
admixture heritage is invisible even with the resolving power of Finally, numerous widely geographically distributed dog pop-
tens of thousands of SNPs. ulations share identical mutations responsible for specific pheno-
Both World Wars had a major impact on the genetic diversity types. Chinese and Mexican breeds both possess the same hair-
of the domestic dog. In the United Kingdom, English Mastiffs less gene (26), sub-Saharan African and Thai breeds possess a
were reduced to 14 individuals (18), Sussex Spaniels to 10 (22), ridged line of hair on their backs caused by the same genetic
and Manchester Terriers to 11 (18). Bernese Mountain Dogs mutation (27), and at least 19 different breeds possess the

8880 | www.pnas.org/cgi/doi/10.1073/pnas.1203005109 Larson et al.

identical mutation for foreshortened limbs (28). These mutations relative to other prey and domestic animals in the fossil record.
are unlikely to have arisen multiple times independently, im- In addition, the absence of archeological excavations in many
plying a significant degree of gene flow between breeds. This parts of the world biases our interpretation of domestication
evidence, combined with known demographic fluctuations in history. The universal human propensity to bury dogs either on
numerous breeds, suggests that throughout history global dog their own or within human burials (32), however, has significantly
populations experienced numerous episodes of diversification enhanced the archeological visibility of dogs.
and homogenization. Each successive round further reduced the Finally, even when zooarcheologists can confidently attribute
power of genetic data derived from modern breeds to infer the remains to Canis familiaris, dating can prove problematic. The
early history of dog domestication. earliest dogs in North America were originally reported from the
Jaguar Cave site in Idaho with an associated date of 10,400 y cal
Dogs in the Archeological Record. Identifying dog remains in the
archeological record is not always straightforward. First, it can be B.P. (33). Subsequent direct dating of the bones revealed that
difficult to discriminate between dogs and wolves, because dogs two Jaguar Cave dogs are ∼3,500 and ∼1,000 y old (34).
were still morphologically wolf-like at the earliest stages of do- An interesting pattern emerges when directly dated and con-
mestication. In addition, and in contrast to their modern patchy fidently identified dog specimens (Table S3) are mapped onto
distribution, wolves were once dispersed across the Northern the historic distribution of wolves across the Old and New
Hemisphere (29) (Fig. 2). As a result, zooarcheologists cannot Worlds (Fig. 2). First, remains ∼12,000 y or older are present in
establish the wild or domestic status of dog remains based solely numerous sites in Europe, the Levant, Iraq, Northern China, and
on geographic location as they can for sheep and goats, the na- the Kamchatka peninsula in the Russian Far East. Dogs appear
tive wild ranges of which were much more restricted. in contexts older than 8,000 y everywhere else within the maxi-
Second, identifying dogs can be confounded by the presence of mal distribution of wolves, suggesting independent domes-
several other extant and extinct species of similar-sized canids, tications of local populations of wolves, migration of humans
including foxes (Vulpes spp.) and maned wolves (Chrysocyon possessing dogs, or the secondary acquisition of dogs by groups
brachyurus) in South America, dholes (Cuon spp.) in Europe and that were not involved in the domestication process.

ANTHROPOLOGY
Asia, jackals in Africa and Asia (Canis aureus, Canis adustus, and Dogs appear south of the original wolf distribution in the Old
Canis mesomelas), and African wild dogs (Lycaon pictus) (30). and New Worlds almost always with the arrival of agriculture.
Recent efforts have been made to differentiate dogs from these For example, despite the fact that human remains are present in
canid species using shape analyses (7), and numerous early claims
much older contexts at Coxcatlan Cave in Mexico, dogs first
for domestic dogs have since been rejected because reanalyses
appear only ∼5,200 B.P. alongside the appearance of agricultural
have revealed contradictory designations (Table S2). This is often
the case when preserved specimens are relatively scarce or frag- communities (35). The same is true in sub-Saharan Africa, where
mented, reducing the presence of specific distinguishing features dogs appear after the advent of the Sudanese Neolithic ∼5600
necessary to discriminate between closely related forms. B.P. (36), in Peninsular Southeast Asia ∼4,200 B.P. (37), and in
Third, a variety of factors can introduce biases against the Island Southeast Asia ∼3,500 B.P. (38). Dogs only arrived in
preservation of certain vertebrate taxa in the archeological re- South Africa ∼1,400 y ago following the arrival of cows, sheep,
cord. These include taphonomic processes [particularly in humid and goats a few hundred years before (39), and in southern South
tropical settings (31)], and the general paucity of canid remains America ∼1,000 y ago with the arrival of sedentary societies (40).

3 8
11 14

13
10 6
1
5

12

4
9

Fig. 2. A world map in which the approximate maximal range of gray wolves (Canis lupus) is shaded in gray (based on ref. 29). Green circles represent regions
where confidently dated remains of domestic dogs have been described in at least one archeological site (Table S3). Circles are divided into eight segments,
each of which represents 1,500 y, visually depicting the age of the oldest remains at sites in the region over which the circle sits. Filled circles represent remains
older than 10,500 y. Each red dog represents a basal breed. The number under each dog refers to the breeds in Table 1; their locations are based upon their
suspected geographic origins, described in Table S1.

Larson et al. PNAS | June 5, 2012 | vol. 109 | no. 23 | 8881

Biogeographical Perspective. Mapping the geographic location of continent until they reached South Africa and southern South
the 14 basal dog lineages onto the maximal wolf distribution and America <1,400 y ago. The number of differentiated, isolated dog
the archeological data reveals several counterintuitive patterns. populations has since been reduced through human movement and
First, although domestic dogs were present in numerous Euro- trade that subsequently led to increased gene flow and population
pean archeological sites ∼15,000 y ago, and despite the fact that homogenization, and through warfare, which often resulted in ex-
textual references or depictions superficially suggest temporally treme demographic fluctuations (including extinction). Each time
deep origins for 13 European breeds including the Pharaoh and a lineage that had been evolving in isolation came into contact with
Ibizan Hounds (Table S1), only the Finnish Spitz retains a basal introduced dogs, the resulting descendant admixture blurred the
signature. Second, although dogs reached Island Southeast Asia genetic signature, making it more difficult to deduce their origins
∼3,500 y ago and southern Africa ∼1,400 y ago, the branches before the assimilation.
leading to three breeds from these regions (Basenjis, New This pattern is not unique to dogs. When human populations
Guinea Singing Dogs, and Dingoes) are located in basal posi- transported domesticates into new regions, the most common
tions on the tree (Fig. 1). This pattern confounds the expectation result has been an admixed population of introduced and local
that basal breeds should originate from the regions that possess varieties, many of which arrived during previous expansion epi-
the oldest archeological dog remains, or at least the regions that sodes. Examples of this phenomenon include European domestic
possess the deepest historical records of types recognizable in grapes (48), Central American maize (49), and Western Eurasian
modern breeds. sheep (50).
The two breeds closest to central Europe that retain basal Basal dog lineages fall outside the large, poorly supported clade
signatures (the Finnish Spitz and the Israeli Canaan Dog), are that includes most modern dog breeds (Fig. 1). This result is not
both known to have been isolated from their European coun- because they more closely approximate the earliest domestic
terparts. Efforts to create modern breed standards included dogs, but because they have mostly avoided recent admixture with
a policy of avoiding those individuals that had been bred with other breeds that themselves possess a merged genetic heritage
foreign, recently introduced breeds (18). Most basal breeds have from dogs that evolved in a wide variety of geographic regions. It
hybridized with other lineages. If those breeds have either been is far easier to avoid introgression by existing at the periphery,
crossed with other basal breeds (e.g., the Shiba Inu) or if a few of beyond landscape and cultural barriers. This theory explains why
the least introgressed individuals are retained and bred [e.g., the numerous basal lineages are from those regions where dogs only
Finnish Spitz or the Dingo; though at least 80% of wild dingoes recently arrived, outside the natural range of wolves, and why no
have interbred with European breeds (41)], then a basal genetic central European breeds retain an ancient signature despite the
signal is retained. ∼15,000-y history of domestic dogs. The vast majority of modern
As discussed above, many basal breeds have also experienced breeds were only created in the past 150 y, emerging from what
severe bottlenecks that have exaggerated their unique genetic sig- was a relatively homogeneous gene pool formed as a result of
natures. The extant captive population of the New Guinea Singing millennia of human migration and the subsequent merging of
Dog is descended from only eight individuals (42), European multiple, previously independently evolving dog lineages. This
Afghans went extinct during the World Wars and were re-estab- history, along with the closed gene pools and small effective
lished using just three imported dogs, and the modern European population sizes associated with recent breed formation, also
Basenji stock was initiated with just a handful of individuals col- explains the strongly supported genetic monophyly of individual
lected in 1936 and supplemented with dogs acquired from central breeds and the lack of resolved relationships between them.
Africa in 1988 (21). The combination of introgression and bottle- The shallow history of breed formation has eased the process of
necks suggests that basal breeds have little or no genetic con- correlating known breed-specific phenotypes with, in some cases,
nections to their ancestral populations, and that genetic their causal mutations (51). Unfortunately, our understanding of
distinctiveness alone cannot be used as a proxy to signify an ancient dog origins has been hampered by our reliance on limited marker
heritage. sets that type a small portion of the 2.4 billion DNA bases that
The most predictive factor in determining whether a breed make up the dog genome (2). Even in datasets that type numerous
retains a basal signature is a lack of gene flow, or at least a lack of individuals, methods that use mitochondrial sequences or even
introgression with breeds that do not possess basal signatures. tens of thousands of SNPs are only capable of recovering sig-
Thus, the unifying characteristic among the 14 basal dog lineages natures that have resulted from the effects of bottlenecks and
(Table 1) is geographic or cultural isolation from the primary reticulate evolution that took place during 19th and 20th century
center of dog breeding in Europe that began in the 19th century. breed formation. As a result, our ability to investigate the deeper
If geography alone determined basal status, however, then the history of dog domestication has been severely hampered.
Africanis, Chihuahua, Chinese Crested, Lhasa Apso, Pekingese, The advent of rapid and inexpensive DNA sequencing tech-
Pug, Rhodesian Ridgeback, Shih Tzu, and Tibetan Terrier should nology has made it possible to significantly increase the volume
also be basal. In these cases, however, a significant degree of in- and commensurate resolving power of genetic data, thus allow-
trogression with European breeds is recorded or strongly sus- ing a greater time depth to be accessed. In humans, dense gen-
pected (Table S1). Although there is pictorial, written (43), and otyping (millions of SNPs) and complete genomes of both
zooarcheological (44) evidence for toy dogs spanning at least the ancient and modern individuals have revealed a far more com-
last 2,000 y, no toy breeds possess a basal signature, probably a plex history (including inter- and intraspecies admixture) than
result of the ease with which they can be transported and in- was previously available using sparser datasets (52). Comparable
terbred with local dogs. genetic analyses of modern and ancient domestic dog genomes
Populations of numerous taxa that live at isolated peripheries, and the resolving power they possess will soon yield equally
including the Falkland Islands Wolf (45), Homo floresiensis (46), complex insights into their domestication and subsequent evo-
and woolly mammoths (47), often either outlived or appear lution, thus revealing our deep, shared history with dogs.
different from their continental relatives. Island populations of
dogs (both real and metaphorical) are more likely to retain their Materials and Methods
genetic integrity not because related populations on the main- Genetics. DNA was isolated from 1,375 domestic dogs (Table S1) and 19 wolves
land have gone extinct, however, but because peripheral pop- (Table S4) and genotyped for 49,664 SNPs on the Affymetrix canine v2 arrays
ulations have avoided amalgamation into a larger group that, as using the snp5-geno-qc software package, with subsequent QC done using
a consequence, has lost its genetic distinctiveness. PLINK (53). SNPs on chromosome X and SNPs with genotyping rates <95%,
were removed, yielding a dataset of 49,024 SNPs. Duplicate samples were
Conclusion identified and merged based on genome-wide average identity-by-state
Though clear signs of the dog domestication process are visible pairwise identity higher than 98%. Breed assignment was confirmed using
15,000 y ago, dogs were not present across every habitable principal component analysis with smartpca (part of the EIGENSOFT software

8882 | www.pnas.org/cgi/doi/10.1073/pnas.1203005109 Larson et al.

package) (54). All dogs included in the analysis had genotyping rates > 75% lists all of the locations, sites, and elements used in Fig. 2. We applied a con-
(median of 98% in dogs and 96% in wolves). servative approach when deciding whether or not to accept individual claims
To construct phylogenetic trees, pairwise identity-by-state genetic dis- for remains that were identified as domestic dogs. The specific criteria and
tances between samples were first estimated across all SNPs that passed rationales are discussed in the SI Results and Discussion.
quality filters using PLINK (53). The distances were then used to construct
a neighbor-joining tree using Phylip (55). The dataset was bootstrapped ACKNOWLEDGMENTS. We thank April McMahon, Alan de Quieroz, Matthew
1,000 times to obtain support values for each node. Breen, Gary Johnson, and Hannes Lohi for comments on the manuscript.
G.L. is currently a Research Councils United Kingdom Academic Fellow and
Archeology. The survey of the archeological literature revealed numerous was supported by a European Molecular Biology Organization postdoctoral
reports of remains, the details of which (species designation, status de- fellowship; K.L.-T. is a European Young Investigator award recipient funded
termination, and dating) the authors were confident. Many other claims by the European Science Foundation, and was supported by grants from the
were contentious. We created two tables. The first (Table S2) lists reports of Swedish Research Council; and A.P. was supported by the British Association
domestic dogs and the rationales for not including them in Table S3, which for Japanese Studies.

1. Darwin C (1859) On the Origin of Species by Means of Natural Selection, or the 30. Macdonald DW, Sillero-Zubiri C, eds (2004) The Biology and Conservation of Wild
Preservation of Favoured Races in the Struggle for Life (John Murray, London). Canids (Oxford Univ Press, Oxford).
2. Lindblad-Toh K, et al. (2005) Genome sequence, comparative analysis and haplotype 31. Van Neer W (2000) Domestic animals from archaeological sites in Central and West-
structure of the domestic dog. Nature 438:803–819. Central Africa. The Origins and Development of African Livestock: Archaeology, Ge-
3. Zeder MA (2012) Pathways to animal domestication. Biodiversity in Agriculture: Do- netics, Linguistics and Ethnography, eds Blench R, MacDonald K (Routledge, New
mestication, Evolution and Sustainability, eds Gepts P, et al. (Cambridge Univ Press, York), pp 163–190.
Cambridge). 32. Morey DF (2006) Burying key evidence: The social bond between dogs and people.
4. Germonpré M, et al. (2009) Fossil dogs and wolves from Palaeolithic sites in Belgium, J Archaeol Sci 33(2):158–175.
the Ukraine and Russia: Osteometry, ancient DNA and stable isotopes. J Archaeol Sci 33. Lawrence B (1968) Antiquity of large dogs in North America. Tebiwa. The Journal the
36:473–490. Idaho State University Museum 11(2):43–48.
5. Germonpré M, Láznicková-Galetová M, Sablin MV (2011) Palaeolithic dog skulls at the 34. Gowlett JAJ, Hedges R, Law I, Perry C (1987) Radiocarbon dates from the Oxford AMS
Gravettian Predmostí site, the Czech Republic. J Archaeol Sci 39:184–202.
system: Archaeometry datelist 5. Archaeometry 29(1):125–155.
6. Ovodov ND, et al. (2011) A 33,000-year-old incipient dog from the Altai Mountains of

ANTHROPOLOGY
35. Flannery KV (1967) Vertebrate fauna and hunting patterns. The Prehistory of the
Siberia: Evidence of the earliest domestication disrupted by the Last Glacial Maxi-
Tehuacan Valley, ed Byers DS (Univ of Texas Press, Austin), pp 132–177.
mum. PLoS ONE 6:e22821.
36. Gautier A (2002) The evidence for the earliest livestock in North Africa: Or adventures
7. Pionnier-Capitan M, et al. (2011) New evidence for Upper Palaeolithic small domestic
with large bovids, ovicaptrids, dogs and pigs. Droughts, Food and Culture: Ecological
dogs in South-Western Europe. J Archaeol Sci 38:2123–2140.
Change and Food Security in Africa’s Later Prehistory, ed Hassan FA (Kluwer Aca-
8. Crockford SJ (2000) A commentary on dog evolution: Regional variation, breed de-
demic/Plenum, New York), pp 195–207.
velopment and hybridisation with wolves. Dogs Through Time: An Archaeological
Perspective, ed Crockford SJ (Archaeopress, Oxford, UK), pp 295–312. 37. Higham CFW (1998) Archaeology, linguistics and the expansion of the East and
9. Vilà C, et al. (1997) Multiple and ancient origins of the domestic dog. Science 276: Southeast Asian Neolithic. Expansion of East and Southeast Asian Neolithic in Ar-
1687–1689. chaeology and Language II: Archaeological Data and Linguistic Hypotheses, eds
10. Savolainen P, Zhang YP, Luo J, Lundeberg J, Leitner T (2002) Genetic evidence for an Blench R, Spriggs M (Routledge, New York), pp 103–114.
East Asian origin of domestic dogs. Science 298:1610–1613. 38. Bellwood P (1998) The archaeology of Papuan and Austronesian prehistory in the
11. Ho SYW, Larson G (2006) Molecular clocks: When times are a-changin’. Trends Genet Northern Moluccas, Eastern Indonesia. Expansion of East and Southeast Asian Neo-
22:79–83. lithic in Archaeology and Language II: Archaeological Data and Linguistic Hypotheses,
12. Boyko AR, et al. (2009) Complex population structure in African village dogs and its eds Blench R, Spriggs M (Routledge, New York), pp 128–140.
implications for inferring dog domestication history. Proc Natl Acad Sci USA 106: 39. Plug I, Voigt EA (1985) Archaeozoological studies of Iron Age communities in
13903–13908. southern Africa. Advances in World Archaeology 4:189–238.
13. Vonholdt BM, et al. (2010) Genome-wide SNP and haplotype analyses reveal a rich 40. Prates L, Prevosti FJ, Berón M (2010) First records of prehispanic dogs in southern
history underlying dog domestication. Nature 464:898–902. South America (Pampa-Patagonia, Argentina). Curr Anthropol 51:273–280.
14. Gray MM, Sutter NB, Ostrander EA, Wayne RK (2010) The IGF1 small dog haplotype is 41. Corbett LK (1995) The Dingo in Australia and Asia (Comstock/Cornell, Ithaca, NY).
derived from Middle Eastern grey wolves. BMC Biol 8:16. 42. Koler-Matznick J, Yates BC, Bulmer S, Brisbin IL (2007) The New Guinea singing dog:
15. Parker HG, et al. (2004) Genetic structure of the purebred domestic dog. Science 304: Its status and scientific importance. Aust Mammal 29:47.
1160–1164. 43. Toynbee JMC, Oost SI (1973) Animals in Roman life and art. History: Reviews of New
16. Miklósi Á (2007) Dog Behaviour, Evolution, and Cognition (Oxford Univ Press, New Books 2(2):34–34.
York). 44. Peters J (1998) Roman Animal Keeping and Breeding: A Synthesis Based on
17. Henriksen BB, Sørensen KA, Sørensen I (1976) Sværdborg I: Excavations 1943–44: A Zooarchaeological Analysis and the Written and Pictorial Record (in German).
Settlement of the Maglemose Culture (Akademisk, Köbenhavn). (Leidorf, Rahden, Germany).
18. Morris D (2002) Dogs: The Ultimate Dictionary of Over 1,000 Dog Breeds (Trafalgar 45. Slater GJ, et al. (2009) Evolutionary history of the Falklands wolf. Curr Biol 19:
Square, North Pomfret, VT). R937–R938.
19. Arman K (2007) A new direction for kennel club regulations and breed standards. Can 46. Brown P, et al. (2004) A new small-bodied hominin from the Late Pleistocene of
Vet J 48:953–965.
Flores, Indonesia. Nature 431:1055–1061.
20. Hickey KR (2000) A geographical perspective on the decline and extermination of the
47. Vartanyan SL, Arslanov KA, Karhu JA, Possnert G, Sulerzhitsky LD (2008) Collection of
Irish wolf Canis lupus: An initial assessment. Ir Geogr 33:185–198.
radiocarbon dates on the mammoths (Mammuthus primigenius) and other genera of
21. American Kennel Club (2006) The Complete Dog Book (Ballantine, New York), 20th Ed.
Wrangel Island, northeast Siberia, Russia. Quat Res 70(1):51–59.
22. Wilcox B, Walkowicz C (1993) Atlas of Dog Breeds of the World (TFH Publications,
48. Myles S, et al. (2011) Genetic structure and domestication history of the grape. Proc
Neptune, NJ), 4th Ed.
Natl Acad Sci USA 108:3530–3535.
23. Calboli FCF, Sampson J, Fretwell N, Balding DJ (2008) Population structure and in-
49. van Heerwaarden J, et al. (2011) Genetic signals of origin, spread, and introgression in
breeding from pedigree analysis of purebred dogs. Genetics 179:593–601.
24. Drake AG, Klingenberg CP (2010) Large-scale diversification of skull shape in domestic a large sample of maize landraces. Proc Natl Acad Sci USA 108:1088–1092.
dogs: Disparity and modularity. Am Nat 175:289–301. 50. Chessa B, et al. (2009) Revealing the history of sheep domestication using retrovirus
25. Castroviejo-Fisher S, Skoglund P, Valadez R, Vilà C, Leonard JA (2011) Vanishing na- integrations. Science 324:532–536.
tive American dog lineages. BMC Evol Biol 11:73. 51. Karlsson EK, et al. (2007) Efficient mapping of mendelian traits in dogs through ge-
26. Drögemüller C, et al. (2008) A mutation in hairless dogs implicates FOXI3 in ecto- nome-wide association. Nat Genet 39:1321–1328.
dermal development. Science 321:1462. 52. Stoneking M, Krause J (2011) Learning about human population history from ancient
27. Salmon Hillbertz NH, et al. (2007) Duplication of FGF3, FGF4, FGF19 and ORAOV1 and modern genomes. Nat Rev Genet 12:603–614.
causes hair ridge and predisposition to dermoid sinus in Ridgeback dogs. Nat Genet 53. Purcell S, et al. (2007) PLINK: A tool set for whole-genome association and pop-
39:1318–1320. ulation-based linkage analyses. Am J Hum Genet 81:559–575.
28. Parker HG, et al. (2009) An expressed fgf4 retrogene is associated with breed-defining 54. Patterson N, Price AL, Reich D (2006) Population structure and eigenanalysis. PLoS
chondrodysplasia in domestic dogs. Science 325:995–998. Genet 2:e190.
29. Nowak RM (2003) Wolf evolution and taxonomy. Wolves: Behavior, Ecology, and 55. Felsenstein J (1993) PHYLIP (phylogeny inference package), version 3.5 c (Joseph
Conservation, eds Mech L, Boitani L (Univ of Chicago Press, Chicago), pp 239–258. Felsenstein, University of Washington, Seattle).

Larson et al. PNAS | June 5, 2012 | vol. 109 | no. 23 | 8883