Domestication of Dogs
Domestication of Dogs
Domestication of Dogs
2
Peter Savolainen
Introduction
The dog is perhaps the most fascinating of the domestic animals. According to the
available knowledge, it was the first domestic animal, and its wild ancestor, the
wolf, was probably domesticated by mobile hunter-gatherers rather than by
settled farmers, contrary to probably all other domestic animals (Clutton-Brock,
1995). The dog is morphologically very diverse, both in size and shape. In the
range of size it is the most diverse mammal species, with the close to 100-fold dif-
ference in weight between the chihuahua and the great dane being the most
prominent example. Dogs and humans share a number of social and behavioural
signals facilitating communication. Many of these are shared with the wolf and
have undoubtedly had significance in the domestication of the wolf, facilitating
the first contacts, but have also been further selected during the process of domes-
tication (Houpt and Willis, 2001). This gives the dog a behaviour which appeals
to many of us more than that of other domestic animals, explaining its status as
man’s best friend.
Despite this special status, until recently surprisingly little has been known
about the history of the dog. While the origin from the wolf has been the gener-
ally accepted theory, there have been virtually no other details known about the
first origin of the dog, or about how the immense morphological variation,
and the hundreds of dog breeds, have developed. Since there is great interest
in this subject, numerous texts have been written, and there have been many the-
ories about how, where and when the dog originated, but there have been
few facts to build on, making reading about this subject an often frustrating
experience.
© CAB International 2007. The Behavioural Biology of Dogs
(ed. P. Jensen) 21
22 P. Savolainen
There is now overwhelming evidence that the wild ancestor of the domestic dog
is the wolf. An origin primarily from the wolf has been the main theory from the
start of evolutionary theory building, but it has been hypothesized that also other
canids, such as the jackal or coyote, could have contributed to the forming of the
dog, explaining some of the morphological and behavioural variation among
domestic dogs. However, with time, numerous studies of physical and behav-
ioural characteristics, and later also molecular genetic markers, both nuclear
markers and mtDNA, have built an increasingly stronger case for the wolf origin
(Wayne, 1993; Clutton-Brock, 1995). While not disproving some degree of con-
tribution from other species, the total collection of evidence is strong in favour of
the wolf as the only ancestor of the dog. The evidence from mtDNA gives a con-
sistent picture: among more than 1000 analysed dogs, all have mtDNA sequences
which are much closer to those of wolves than of the other wild canids. However,
mtDNA is maternally inherited and can therefore only describe the history of
females, and while studies of nuclear markers consistently support an origin from
the wolf, studies have so far been relatively limited, leaving a possibility for a small
contribution on the male side from, for example, the jackal. Large-scale popula-
tion studies of nuclear markers, for example the Y chromosome, will surely give
the final answer within a few years.
Thus, for a long time research did not bring much new knowledge about the
history of the dog, except the early molecular biological studies which could bring
growing certainty that the dog originates from the wolf rather than from other
canids, and the sporadic addition of new archaeological finds of early dogs,
changing the geographic location for their oldest known remains. Since the clas-
sical instruments for this kind of question, archaeology and palaeontology, failed
to provide any detailed information, population genetic studies was applied rela-
tively soon after the techniques for large-scale DNA-sequencing studies became
available. The studies have, so far, mainly been performed on mtDNA, because
the genetic variation is higher than for nuclear DNA, especially in the noncoding
part, the control region (CR), giving much genetic information from analysis of a
relatively short DNA sequence. mtDNA is maternally inherited and a drawback
is therefore that the mtDNA-studies can only describe the history of the female
24 P. Savolainen
dogs, the history of the males being only indirectly monitored. An advantage on
the other side, shared with the Y chromosome, compared to nuclear autosomal
DNA, is that the mitochondrial genome does not recombine, which implies that
the whole molecule is inherited as an entity through the generations, making the
interpretation of the data straightforward.
The first three major population genetic studies of mtDNA from dogs and
wolves were published in 1996 and 1997: 261 base pairs of the mtDNA CR were
analysed in 140 domestic dogs representing 67 breeds, mostly European but also
a few from Asia, Africa and America, and 162 wolves from throughout Europe,
Asia and North America (Vila et al., 1997), and 970 and 670 base pairs, respec-
tively, of the CR were analysed in a total of 128 dogs, representing mainly
Japanese breeds, and 19 wolves (Okumura et al., 1996; Tsuda et al., 1997). These
studies showed that, in phylogenetic analyses, the domestic dog mtDNA
sequences were distributed into six distinct groups (clades) interspersed by wolf
sequences, showing that the dog originates from several female wolf lines, at least
as many as the number of clades (Fig. 2.1). Importantly, all of the 268 studied
domestic dogs had mtDNA sequences which were closer to those of wolves than
of other canids. Thus, there was no indication of a genetic contribution from any
other canid than the wolf to the domestic dog population. However, the geo-
graphic location of the domestication of the wolf, or any detailed conclusions
about the history of specific dog breeds, could not be determined based on these
relatively limited data sets. The time for the origin of the dog was estimated, based
on the largest genetic distance of the sequences in the largest phylogenetic clade
of dog sequences, giving a possible date of 135,000 years. However, as shown
below, the mtDNA data can be interpreted differently, resulting in a date more
in agreement with that obtained from the archaeological record.
The greatest step so far in the study of the origin of dogs was taken in 2002 with
the publication of two articles, one study of the present-day dog population
throughout the world (Savolainen et al., 2002) and one of pre-European archaeo-
logical remains of American dogs (Leonard et al., 2002). Leonard et al. showed
that domestic dogs on all continents have a common origin from a single gene
pool, and the study by Savolainen et al. suggested that this gene pool originated
just once, somewhere in East Asia.
There are principally two approaches possible for elucidating the place of
domestication of the wolf, using mtDNA analysis. The most straightforward
approach would be to study the mtDNA sequences of today’s domestic dog pop-
ulation and compare these with sequences from a worldwide sample of wolves.
The population of wolves having the most similar, or even identical, sequences
compared to the dogs would then be presumed to be the wolf population from
which the domestic dogs descend, and the geographical region in which these
wolves live would be presumed to be the geographical origin of the dog. However,
Domestication 25
0.005 substitutions/site
coyote
Fig. 2.1. Phylogenetic tree of dog (unlabelled) and wolf (open squares) mtDNA
types (Savolainen et al., 2002). Six clades (A–F) of dog mtDNA types are
indicated. Branch lengths are according to the indicated scale, the branch leading
to the outgroup (coyote) has been reduced by 50%. Clades A–D and F were
imploded, and the relationships between mtDNA types within the clades are
instead shown as minimum-spanning networks in Figs 2.2 and 2.3.
Domestication
Africa 30 (85.7) 14 (5) 4 (11.0) 1 (0) 1 (2.9) 1 (0) 0 0 0 0 0 0 35 16 (5)
Arctic America 25 (100) 5 (1) 0 0 0 0 0 0 0 0 0 0 25 5 (1)
Europe 140 (67.6) 20 (9) 36 (18.0) 4 (1) 12 (5.9) 3 (0) 19 (9.1) 3 (3) 0 0 0 0 207 30 (13)
East Asia 192 (73.8) 44 (30) 39 (15.0) 10 (7) 24 (9.2) 4 (1) 0 0 3 (1.1) 1 (1) 2 (0.8) 2 (2) 260 61 (41)
SW Asia 51 (56.7) 16 (4) 32 (35.0) 4 (2) 5 (5.4) 2 (0) 2 (2.2) 1 (1) 0 0 0 0 90 23 (7)
Siberia 17 (70.8) 9 (1) 2 (8.3) 1 (0) 4 (17.0) 2 (1) 0 0 0 0 1 (4.1) 1 (1) 24 13 (3)
India 11 (84.6) 4 (0) 1 (7.7) 1 (0) 1 (7.7) 1 (0) 0 0 0 0 0 0 13 6 (0)
Total 466 (71.3) 71 114 (17.4) 13 47 (7.2) 5 21 (3.2) 4 3 (0.5) 1 3 (0.5) 3 654 97
27
28 P. Savolainen
70 61
30 16 13
A
UK 2 4
50 17 19 19
P
A
71
12 2 15
B D
58 62
G S 1
UK
01 17 1
14 3 9
10
4 11
UK
33 20 10
18
6 48 3
16 8 56 18
5 7 4
42 6
46
S 1 2 9
3 22 82
8
24 31
75 74
43 5
7 1
34
49
60 29 UK UK
80 14 15
51
72
39 38
73
44
35
P 11
32
54 6
40 6
UK
25
3
59 15 UK
52 14 12
26
C E F
63 5
12
A
13
P
57 1
55 7
S
3
41
28 18 20 65 2
53
8 4 2
27 83 UK
47 64 22
66 1 2
81 68 19 36
45
G GS
67
69 21
introduced to America by the ancestors of today’s Inuits rather than by the ances-
tors of the First Nations tribes. Other parts of America were not represented
because the original First Nations tribes’ dog populations are thought to have
been largely obliterated or mixed with European dogs in connection with the
arrival of Europeans, which would make the interpretation of American dog
sequences complicated. However, in a study of mtDNA in 19 Mexican hairless
dogs, a Mexican breed believed to have remained essentially isolated, the mtDNA
types all belonged to clades A, B and C (Vila et al., 1999b). Furthermore, the study
of archaeological remains of pre-European American domestic dogs (Leonard et
al., 2002) showed that among 13 samples from Mexico, Peru and Bolivia, 800
years or more old, 12 had mtDNA types clustering with clade A and one sample
had a type clustering with clade B. Thus, both clades A and B, and possibly clade
C, according to the data on the Mexican Hairless Dog, were present among the
native American dogs. Perhaps more importantly, none of the American samples
had sequences which did not fall into the same mtDNA-clades as the Old World
samples. Thus, in the available mtDNA data there is no sign of a separate domes-
tication of dogs in America, or of crossbreeding with the wolf. This shows that the
dog populations of the New World have a common origin with the dogs of the
Old World.
Domestication 29
An indication of the geographic origin of the three dog clades, and thereby the
origin of the dog, can be obtained from a comparison of the genetic variation
between geographical regions. If an ancestral population and a derived popula-
tion (formed from a subset of the genetic types of the ancestral population) are
compared, the number of mtDNA types and the nucleotide diversity are expected
30 P. Savolainen
The time of origin for each dog mtDNA clade can be estimated from the mean
genetic distance of the sequences in each clade to the original wolf mtDNA type,
Domestication 31
Clade A
Clade C Clade B
and the mutation rate of the analysed region. Under the assumption that the
clade originates from a single wolf mtDNA type, the age of the clade will ap-
proximately correspond to the origin of the dogs originating from that female
wolf. However, there are two problems with the available data that render
a precise dating of the origin of the dog based on the available mtDNA data
impossible.
The first problem concerns the dating of the palaeontological fix-point
needed to calibrate the mutation rate. The wild canid most closely related to the
wolf is the coyote, and the time for the split into two species from their common
ancestor is the best available calibrating point for the mutation rate of dog
mtDNA. Counting the number of sequence differences between wolf and coyote
in the analysed mtDNA region, and dividing by the divergence time, gives the
mutation rate. However, while a first appearance of wolves ~700,000 years ago
and of coyotes ~1 million years ago (Kurtén, 1968; Kurtén and Anderson, 1980)
indicates a date for the divergence between the two species of approximately
1 million years ago, the time of divergence has not been definitely established and
an earlier date of up to approximately 2 million years cannot be ruled out.
Therefore, without more exact palaeontological evidence, a precise dating of the
32 P. Savolainen
origin of the dog will not be possible, but it can nevertheless give an indication of
the probable range of time, and this information can be used together with other
data from the archaeological record on the first appearance of dogs. To simplify
the following argument, the 1-million-year date for the wolf–coyote divergence
will be used for the calculations based on mtDNA, but bearing in mind the pos-
sibility that the split occurred up to 2 million years ago, in which case the age of
the datings should be doubled. The second problem is that the mutation rate of
the analysed 582 base pairs region does not give resolution between mtDNA types
in the time scale needed to monitor the last 20,000 or 30,000 years. According to
the calculations above, the rate of mutation is approximately one mutation per
24,000 years in a lineage (Savolainen et al., 2002). This implies
that if two wolves, having mtDNA types differing by a single mutation, were
domesticated say 15,000 years ago, the dogs of today originating from those two
lineages would have mtDNA types differing from each other only by the
single original mutation or by just one or two additional mutations. If there
were several wolves having mtDNA types differing by just a few mutations, it
would not be possible to fully resolve the mtDNA types of the dogs originating
from them. Thus, the period around 15,000 years ago, which is suggested to be
the time for the origin of the domestic dog according to the archaeological record,
cannot be fully studied using the 582 base pairs mtDNA region available at
present.
In a domestication event with a subsequent population expansion, a star-like
phylogeny, with the founder mtDNA type in the centre and new mtDNA types
distributed radially, would be expected. The networks of clades B and C are star-
like, indicating an origin from a single wolf mtDNA type (Figs 2.2 and 2.3). In
contrast, clade A has a complicated pattern without an easily identifiable central
node. A distance of up to 11 substitutional steps between mtDNA types would
indicate that clade A is much older than clades B and C, and derives from an
initial domestication of wolves. However, as discussed above, the dog mtDNA
types in clade A do not necessarily originate from a single wolf mtDNA type even
though clade A is an almost completely continuous group of mtDNA types sepa-
rated from each other by single mutational steps. It is possible that clade A origi-
nally was a clade of several closely related wolf mtDNA types, and that several
wolves having a number of these mtDNA types belonging to clade A were domes-
ticated. Looking more closely at clade A it has, instead of a single central node,
several subclusters with star-like shape, suggesting that clade A may have origi-
nated from several wolf mtDNA types (Figs 2.2 and 2.3). The approximate age of
clade A, assuming a single origin from the wolf and a subsequent population
expansion, is estimated from the mean pairwise distance between East Asian dog-
mtDNA-sequences (3.39 substitutions, SD=0.13) and the mutation rate to 41,000
± 4000 years. This calculation may be biased by the population history among
the dogs and wolves. Alternatively, the maximum age of clade A can be estimated
from the number of steps between the most distantly related mtDNA types, 11
substitutions apart which corresponds to ~120,000 years. According to these cal-
culations clade A would have originated 40,000–120,000 years ago, and if it is
Domestication 33
supposed that it was formed in a domestication event from a single wolf mtDNA
type, the domestic dog would have originated 40,000–120,000 years ago. If
instead an origin of clade A from several different wolf mtDNA types is assumed,
several reasonably defined subclusters can be found. To give an alternative dating
of the domestication of the dog, three subclusters of clade A possibly representing
three origins of dog from wolf, marked by lines in Fig. 2.3, can be studied. The
mean genetic distances of the sequences belonging to these subclusters to their
respective nodes (0.45, 0.65 and 1.07 substitutions with SD=0.13, 0.09, 0.27,
respectively) give estimates of 11,000 ± 4000, 16,000 ± 3000 and 26,000 ± 8000
years for their ages, respectively. Thus, clade A has a substructure, suggesting that
it could have been formed by several wolf mtDNA types, possibly ~15,000 years
ago. Assuming single wolf mtDNA types as founders of clades B and C, the mean
distances among East Asian sequences to the nodes (0.54 and 0.71 substitutions,
SD=0.08 and 0.10) give estimated ages of 13,000 ± 3000 and 17,000 ± 3000 years
for clades B and C, respectively. Thus, clade A was formed ~40,000–120,000
years ago but has a substructure indicating later population events, and clades B
and C were formed ~15,000 years ago. Depending on how these data are inter-
preted, it can suggest a first origin of domestic dogs either ~40,000–120,000 years
ago forming only clade A, after which clades B and C were introduced into the
dog gene-pool through crossbreeding between dogs and wolves ~15,000 years
ago, but it can also suggest a single origin ~15,000 years ago involving all the
three clades A, B and C.
The key question for the dating of the origin of the dog is the number of wolf
mtDNA types that formed dog-clade A, but this question cannot be answered
based on the present mtDNA data. However, indirect evidence in the form of the
universal presence of clades A, B and C at similar proportions talks in favour of a
simultaneous origin of the three clades. Assuming that clade A would have origi-
nated from an initial domestication approximately 40,000 years ago, the dog pop-
ulation originating from that domestication event would be expected to have
spread to other parts of the world. An introduction of clades B and C, 15,000
years ago, into the already existing domestic dog gene pool, by regional cross-
breedings with wolves, would require a very thorough mixing to have occurred,
through migration to all parts of Eurasia, to level the frequencies of the three
clades to the very similar proportions now found throughout the world.
Alternatively, if clade A originated from a domestication 40,000 years ago, these
dogs could have remained isolated in the original geographical region, not
spreading to other parts of the world. Clades B and C would then, 15,000 years
ago, have been introduced into the domestic dog gene pool by regional cross-
breedings with wolves, after which the dogs would have spread to other parts of
the world. While the amount of migration and trade 15,000 years ago is not
known, this scenario seems unlikely. It is also contradicted by the age of the oldest
subcluster of clade A in Europe (as determined from the mean genetic distance
from mtDNA types unique to the western part of the world to the nodal mtDNA
type shared with East Asia; 0.39 substitutions, SD=0.09), which is estimated to be
only 9000 ± 3000 years old (Fig. 2.3).
34 P. Savolainen
Thus, a first origin of dog mtDNA clade A, a long time before the origin
of dog clades B and C, does not seem probable considering the very similar
proportions of clades A, B and C around the world. The total sum of circum-
stantial mtDNA evidence therefore indicates a single origin, in both place and
time, for the three clades, approximately 15,000 years ago. Furthermore, in the
absence of a clear result from the mtDNA data, the best evidence for the time of
the first origin of the domestic dog remains the archaeological record, which indi-
cates an origin approximately 15,000 years ago. A synthesis of mtDNA data: the
similar proportions of clades A, B and C, and the larger genetic variation for
clades A and B; and archaeological data: the oldest evidence of domestic dog
dated at approximately 15,000 years ago, therefore point to an origin of the
domestic dog in East Asia ~15,000 BP. In this event, clade A would have had
several origins from wolf mtDNA types, at least around ten, and the first domes-
tication of wolves would not have been an isolated event, but rather a common
practice in the human population in question.
Clades D, E and F have been left out of the discussion so far to simplify the
discussions. These three clades are found only regionally, and clades E and F
also at very low frequency, three dogs for each clade, making population
genetic analyses virtually impossible (Table 2.1). The three clades, which
represent three separate origins from the wolf, could either be the results
of crossbreedings between male dogs and female wolves or separate domestica-
tions of the wolf, or they could have originated together with clades A, B and C
at a single domestication event. Since clades E and F are found in East Asian
dogs, an origin together with clades A, B and C seems possible. In this case they
would have been part of a gene-pool containing the five clades A, B, C, E and F,
but clades E and F, like some parts of clade A, would not have spread to other
parts of the world. For clade D the situation is very different. This clade is found
only in Europe and Turkey, and it has therefore probably an origin outside East
Asia, independent from that of the other clades. There is a clear geographical
division within clade D. mtDNA types D1–D4, which constitute a separate sub-
cluster, were all found in Scandinavian spitz breeds, in total 18 out of 49
Scandinavian dogs, while D5 and D6, which are several mutations away from
D1–D4, were found in two Turkish and one Spanish dog, respectively (Fig. 2.2).
The distance between the subclusters D1–D4 and D5–D6, three substitutions,
corresponds to approximately 36,000 years. It therefore seems probable that
clade D originates from at least two separate origins from the wolf. The diver-
gence within subcluster D1–D4, a single substitution in 18 individuals, indicates
a recent origin, a few thousand years at most. Thus, at least clade D seems to have
originated through crossbreeding between male dogs and female wolves. This
crossbreeding seems to have had some impact on the Scandinavian breeds, con-
sidering the large proportion of dogs having mtDNA types belonging to clade D,
Domestication 35
18 out of 49 dogs (37%). It is probable that, in crosses between male dogs and
female wolves, the chances of the offspring becoming part of the domestic dog
population would be low. The other scenario, that offspring from male
wolf/female dog crossbreedings would ‘become dogs’ seems more probable. It is
believed that in some cultures, especially in the northern parts of Eurasia and
America, dogs, and above all bitches, were crossed with wolves in order to bring
new blood to the dog population. Future studies of the Y chromosome will no
doubt tell us to what degree such crossbreedings have contributed to the devel-
opment of domestic dogs.
An interesting question connected to the history of the domestic dog is the origin
of the Australian dingo. The dingo is a wild canid, morphologically resembling
South Asian domestic dogs, which according to the archaeological record arrived
in Australia between 3500 and 12,000 years ago (Clutton-Brock, 1995). However,
the precise ancestry and time of arrival in Australia of the dingo have not been
known, nor whether, on its arrival, it was domesticated or half-domesticated
before becoming feral, or a truly wild dog. In a study of 211 dingoes, sampled
around Australia, 582 base pairs of the mtDNA CR were analysed and compared
to dog and wolf data, giving very distinct results (Savolainen et al., 2004). All dingo
sequences clustered in dog clade A, forming a single internal cluster of mtDNA
types unique to dingoes, around a central type, A29, found in both dingoes and
dogs (Fig. 2.4). Among domestic dogs, A29 was found only east of the Himalayas,
together with most types in that part of clade A. This indicates that the dingo pop-
ulation originates from a single introduction of a small population of domestic
dogs coming from East Asia, carrying mtDNA type A29. The mean distance
among the dingo sequences to A29 indicates their arrival to Australia ~5000
years ago. The dingo has probably remained isolated since then, and represents
a unique isolate of early undifferentiated dogs.
Summary
A29
Fig. 2.4. Minimum-spanning network of the main dog clade, A (Savolainen et al.,
2004). mtDNA types (circles) and hypothetical intermediates (solid dots) are
separated by one mutational step (substitutions, indels not shown). White circles
are mtDNA types found in dogs only; light grey circles are types found in dingoes
only; dark grey circles are types found in both dingoes and dogs. Squares denote
wolves. Areas of grey circles are proportional to frequencies among dingoes, but
the area of A29 is reduced by 50%.
References
Angleby, H. and Savolainen, P. (2005) A study of the forensic usefulness of the mito-
chondrial DNA variation among and within populations, breeds and types of
domestic dogs. Forensic Science International 154, 99–110.
Clutton-Brock, J. (1995) Origins of the dog: domestication and early history. In:
Serpell, J. (ed.) The Domestic Dog, its Evolution, Behaviour, and Interactions
with People. Cambridge University Press, Cambridge, pp. 7–20.
Clutton-Brock, J. and Noe-Nygaard, N. (1990) New osteological and C-isotope
Domestication 37