Phylogenetic analysis of the evolution of lactose digestion in adults. 1997.

Full text

Page 1

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers,
academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.
Phylogenetic Analysis of the Evolution of Lactose Digestion in
Adults
Author(s): Clare Holden and Ruth Mace
Source: Human Biology, 81(5/6):597-619. 2009.
Published By: Wayne State University Press
DOI: http://dx.doi.org/10.3378/027.081.0609
URL: http://www.bioone.org/doi/full/10.3378/027.081.0609
BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological,
ecological, and environmental sciences. BioOne provides a sustainable online platform for over
170 journals and books published by nonprofit societies, associations, museums, institutions,
and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated content indicates
your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use.
Usage of BioOne content is strictly limited to personal, educational, and non-commercial use.
Commercial inquiries or rights and permissions requests should be directed to the individual
publisher as copyright holder.

Page 2

key words: lactose digestion capacity, cultural evolution, pasto-
ralism, phylogeny.
Human Biology, October 1997, v. 69, no. 5, pp. 605–628.
Copyright © 1997 Wayne State University Press, Detroit, Michigan 48201-1309
Phylogenetic Analysis of the Evolution of Lactose Digestion
in Adults
Clare Holden1 and Ruth Mace1
Abstract In most of the world’s population the ability to digest lactose de-
clines sharply after infancy. High lactose digestion capacity in adults is com-
mon only in populations of European and circum- Mediterranean origin and
is thought to be an evolutionary adaptation to millennia of drinking milk from
domestic livestock. Milk can also be consumed in a processed form, such as
cheese or soured milk, which has a reduced lactose content. Two other selec-
tive pressures for drinking fresh milk with a high lactose content have been
proposed: promotion of calcium uptake in high- latitude populations prone to
vitamin- D deficiency and maintenance of water and electrolytes in the body
in highly arid environments. These three hypotheses are all supported by the
geographic distribution of high lactose digestion capacity in adults. However,
the relationships between environmental variables and adult lactose digestion
capacity are highly confounded by the shared ancestry of many populations
whose lactose digestion capacity has been tested. The three hypotheses for
the evolution of high adult lactose digestion capacity are tested here using a
comparative method of analysis that takes the problem of phylogenetic con-
founding into account. This analysis supports the hypothesis that high adult
lactose digestion capacity is an adaptation to dairying but does not support the
hypotheses that lactose digestion capacity is additionally selected for either
at high latitudes or in highly arid environments. Furthermore, methods using
maximum likelihood are used to show that the evolution of milking preceded
the evolution of high lactose digestion.
The ability to digest lactose in adults is a genetic polymorphism inherited as a
dominant genetic trait (Sahi et al. 1973; Johnson et al. 1977; Ransome- Kuti et al.
1975; Metneki et al. 1984). This trait is common in a few of the world’s popula-
tions. The physiological cause of high lactose digestion capacity (LDC) in adult-
hood is the retention of high levels of lactase in the small intestine beyond infancy
(lactase persistence), which contrasts with the standard mammalian developmental
pattern of a steep decline in small intestine lactose levels after infancy (Flatz 1987).
The LDC of over 20,000 individuals worldwide has been tested. High fre-
quencies (?70%) of adults with high LDC are found in northern Europeans and
1 Department of Anthropology, University College London, Gower Street, London WC1E 6BT, England.
HB_5-6_FINAL.indb 597 5/3/2010 12:28:34 PM

Page 3

598 / holden and mace
their descendants in North America and Australia and among some African pas-
toralist groups thought to have originated in North or East Africa. Intermediate
frequencies (30–70%) of adults with high LDC are found around the Mediterra-
nean, the Middle East, and in central and south Asia. Regions whose adult popula-
tions predominantly have low LDC include much of sub- Saharan Africa, east and
southeast Asia, and the native populations of the Americas and Oceania. Patchy
sampling in much of sub- Saharan Africa and in southern and central Asia neces-
sitates caution in our overview of these areas, particularly because the distribution
of adult LDC in these regions appears to be locally variable and complex.
Selective Pressures for Drinking Milk
There are three major hypotheses for the evolution of high adult LDC. The
first hypothesis was independently proposed by Simoons (1969, 1970) and Mc-
Cracken (1971). They observed that in regions where milk was not normally con-
sumed until recently, adults mostly have a low capacity to digest lactose. Therefore
it was hypothesized that lactase persistence in adulthood is an adaptation to millen-
nia of pastoralism and milk consumption, a theory known as the culture- historical
hypothesis (Simoons 1970a). This is a coevolutionary theory in which selection
of a genetic trait is influenced by the cultural environment, the herding and milk-
ing of livestock. According to this hypothesis, the capacity to digest lactose has
a selective advantage in adults in pastoralist populations. Individuals with a high
LDC are able to derive a nutritional benefit from the lactose in milk, which is not
available to individuals with low LDC. Individuals with low LDC may also suffer
from symptoms of lactose intolerance when they consume fresh milk, including
abdominal discomfort, flatulence, and diarrhea. It has been suggested that, because
of these symptoms, milk could be nutritionally detrimental to individuals with low
LDC, although this suggestion has been much debated (Scrimshaw and Murray
1988). Populations that keep livestock but do not milk them, for example, popula-
tions in China and southeast Asia and parts of sub- Saharan Africa (Murdock 1967;
Simoons 1970), would not be expected to have evolved high adult LDC according
to the culture- historical hypothesis (Simoons 1979).
The other two hypotheses refer specifically to the consumption of fresh
milk. Milk is often processed into dairy products, such as cheese and yogurt,
which have a reduced lactose content. Two specific selective advantages to drink-
ing fresh milk, with a high lactose content, have been proposed. Flatz and Rot-
thauwe (1973) suggested that in high- latitude environments, where sunshine is
limited, humans are at risk of vitamin D deficiency and rickets. The lactose in
fresh milk, like vitamin D, promotes the uptake of calcium, also present in milk.
This hypothesis could explain the high frequency of lactose digesters in northern
European populations. Durham (1991) has used this hypothesis to explain the dif-
ference between northern Europe, where fresh milk is consumed in quantity and
most adults have high LDC, and the Mediterranean, where milk is mostly eaten as
cheese and the population has predominantly low LDC.
HB_5-6_FINAL.indb 5985/3/2010 12:28:34 PM

Page 4

Evolution of Lactose Digestion / 599
The third hypothesis, proposed by Cook and Al- Torki (1975) and Cook
(1978), states that in highly arid environments the water content of fresh milk
increases the survival chances of lactose- digesting milk drinkers among desert-
dwelling nomadic pastoralists and also that diarrhea and consequent water
depletion in lactose- intolerant members of the group cause selection against mal-
digesters. This hypothesis is supported by the high frequencies of adults with the
LDC observed in pastoralist groups in hot areas, for example, the Middle East and
North Africa, including the Bedouin, the Tuareg, and the Fulani (Table 1).
The evolution of adult lactase persistence has been modeled several times.
Bodmer and Cavalli- Sforza (1976) estimated that a selection coefficient of 0.04
would be necessary for high LDC to increase from an initial prevalence of 0.001%
to the levels observed today in northern European populations (estimated fre-
quency of 0.5) within 290 generations (9000 years). If the initial frequency of the
lactase persistence gene were 1.0%, a selection coefficient of only 0.015 would
be required. This time scale is realistic for the Middle East, where livestock were
first domesticated around 8000–7000 B.C. (Clutton- Brock 1987). Flatz (1987)
estimated that for the gene to reach contemporary European levels in the 3500
years or less since the first known domestic livestock in northern Europe, starting
from an initial frequency of 0.005%, a higher selection coefficient of between 3%
and 7% would be required.
More recently, attempts have been made to model the coevolution of a gene
for lactase persistence and the cultural trait of milk drinking. Aoki (1986) esti-
mated that for the selection of the gene for lactase persistence to increase from an
initial prevalence of 0.05% to the prevalence observed in northern Europe today
(estimated gene frequency of 0.7) within the time available since the advent of
dairying (6000 years ago) and with an effective population size of 500, the se-
lection coefficient must have been greater than 5%. Feldman and Cavalli- Sforza
(1989) also found that a selection coefficient of greater than 5% was necessary for
a gene frequency of 0.70 to be reached in 6000 years.
However, in these dual- inheritance coevolutionary models milk drinking is
a cultural trait with a low initial frequency whose selection coefficient depends on
the prevalence of the lactase persistence gene. The ethnographic evidence does
not support this assumption, insofar as milk consumption apparently has been
universally adopted by populations with predominantly low LDC, for example,
the Mongols, the Herero, the Nuer, and the Dinka. Milk- based pastoralism may
be the best means of subsistence, particularly in dry, marginal environments, even
for lactose nondigesters. Milk processing and the consumption of fresh milk in
only small quantities are cultural and behavioral means by which many lactose
malabsorbers manage to consume milk products without suffering the symptoms
of lactose intolerance. After milk- based pastoralism had been adopted as a means
of subsistence, high LDC would have enabled adults to consume more fresh milk
and to derive a nutritional benefit from the lactose component of fresh milk, and so
be selected for. The initial frequency of the cultural trait of milk consumption may
therefore be virtually 100%, which could reduce the selection coefficient required
HB_5-6_FINAL.indb 599 5/3/2010 12:28:34 PM

Page 5

600 / holden and mace
Table 1. Data Used in the Analysis



Pastoralism as



Proportion
Total Number

Number of
Global Solar
of Total
of Individuals
Frequency of
Population and
Dry Months
Radiation
Subsistence
Tested in Each
Low LDC
Ethnographic Atlas Code
per Year
(kcal/cm/yr)
Activity (%)
Population
(%)
Reference
Apache (Nh17)
12
162
0
22
100
Johnson et al. (1978)
Australian Aboriginesa
12
167
0
45
84
Brand et al. (1983)
Baggara (Habbania) (Cb13)
10
174
50.5
19
53
Bayoumi et al. (1981)
Baggara (Messiria) (Cb15)
10
174
70.5
20
60
Bayoumi et al. (1981)
Bedouin (Jordanian (Cj2)
10
162
80.5
162
24
Hijaki et al. (1983)
Bedouin (Saudi) (Cj5)
12
192
93
35
17
Cook and Al- Torki (1975), Dissanyake et






al. (1990)
Beja (Amarar) (Ca35)
12
192
93
82
13
Bayoumi et al. (1982)
Beja (Beni Amir) (Ca36)
12
174
80.5
40
13
Bayoumi et al. (1982)
Beja (Haddendoa) (Ca43)


60.6
137
20
Bayoumi et al. (1982)
Beja (Bisharin) (Ca5)
12
192
80.5
22
14
Bayoumi et al. (1982)
Chippewa (Na36)
7
124
0
33
97
Newcomer et al. (1977)
Czechs (Ch3)
9
124
30.5
200
13
Madzarovova- Nohejlova (1974)
Dinka (Aj11)
5
144
50.5
213
76
Bayoumi et al. (1982), Elliott et al. (1973)
Egyptians (Cd2)
12
192
30.5
742
64
Hussein et al. (1982), Hussein and






Ezzilarab (1994)
Eskimo (Greenland) (Na25)
8
79
0
119
85
Gudmand- Hoyer and Jarnum (1969),






Gudmand- Hoyer et al. (1973), Asp et






al. (1975)
Fijians (Ih4)
0
167
0b
12
100
Masarei et al. (1972)
Fulani (pastoralist) (Cb8)
7
174
80.5
9
22
Kretchmer et al. (1971)
Fulani (sedentary) (Cb22)
7
174
40.5
24
71
Kretchmer et al. (1971)
HB_5-6_FINAL.indb 6005/3/2010 12:28:35 PM

Page 6

Evolution of Lactose Digestion / 601
Ganda (Ad7)
0
144
10.5
27
96
Cook and Kajubi (1966), Cook and






Dahlquist (1968)
Greeks (Ce7)
8
162
30.5
800
52
Kanaghinis et al. (1974), Ladas et al.






(1982)
Hausa (Cb26)
7
174
30.5
17
76
Kretchmer et al. (1971)
Hazara Tajiki (Ea3)
10
162
50.5
79
82
Rahimi et al. (1976)
Herero (Ab2)
9
178
60.5
37
97
Currie et al. (1978)
Hopi (Nh18)
10
162
10.5
21
100
Johnson et al. (1978)
Hungarian (Ch8)
7
124
20.5
535
37
Czeizel et al. (1983)
Hutu (Ae10)
2
145
30.5
51
51
Cook and Kajubi (1966), Cox and Elliott






(1974)
Igbo (Af10)
4
144
10.5
16
81
Elliott et al. (1973), Olatumbosun and






Adadevoh (1971), Ransome- Kuti et al.






(1975)
Iranian (Ie9)
12
162
30.5
40 (>12 yrs)
83
Sadre and Karbasi (1979)
Irish (Cg3)
0
94
40.5
50
4
Fielding et al. (1981)
Italians (South) (Ce5)
4
124
10.5
197
67
De Ritis et al. (1970), Burgio et al.






(1984), Rinaldi (1984), Cavalli- Sforza






et al. (1987)
Japanese (Ed5)
1
162
10.5
66
81
Nose et al. (1979), Shibuya et al. (1970),






Yoshida et al. (1975)
Javanese (Indonesia) (Ib2)
1
144
20.5
53
91
Surjono et al. (1973)
Jordanians (Cj6)
10
162
30.5
204
75
Hijaki et al. (1983), Snook et al. (1976)
Lapps (Cg4)
8
79
60.5
519
41
Isokoshi et al. (1981)
Lebanese (Cj7)
5
162
20.5
225
79
Loiselet and Jarjouhi (1974), Nasrallah






(1979)
Mongols (Eb7)
10
124
80.5
198
88
Wang et al. (1984)
Nama (Aa3)
9
178
50.5
18
50
Jenkins and Nurse (1976)
Northern Chinese (Han)c
8
162
20.5
314
88
Wang et al. (1984), Zheng et al. (1988)
Nubians (Midobi) (Cb11)
12
174
93
21
67
Bayoumi et al. (1981)
Nuer (Aj3)
5
144
50.5
23
78
Bayoumi et al. (1982)
HB_5-6_FINAL.indb 6015/3/2010 12:28:35 PM

Page 7

602 / holden and mace
Table 1. (continued)



Pastoralism as



Proportion
Total Number

Number of
Global Solar
of Total
of Individuals
Frequency of
Population and
Dry Months
Radiation
Subsistence
Tested in Each
Low LDC
Ethnographic Atlas Code
per Year
(kcal/cm/yr)
Activity (%)
Population
(%)
Reference
Papago (Ni2)
10
162
0
14
93
Johnson et al. (1978)
Pathans/Pushtu (Ea2)
10
162
30.5
86
65
Rab and Baseer (1976), Rahimi et al.






(1976)
Pima (Ni6)
10
162
0
62
95
Johnson et al. (1977), Johnson et al.





(>4 yrs)
(1978)
Punjabis (Ea13)
5
162
20.5
384
56
Rab and Baseer (1976), Ahmad and






Flatz (1984), Abbas and Ahmad (1983)
Russians (Ch11)
6
94
30.5
103
57
Lember et al. (1991)
San (!Kung and #hua)d
9
178
0
65
95
Jenkins et al. (1974), Nurse and Jenkins






(1974)
Shilluk (Ai6)
10
144
20.5
8
63
Bayoumi et al. (1982)
Sindhi (Ea1)
11
192
30.5
45
42
Rab and Baseer (1976), Ahmad and






Flatz (1984)
Sinhalese (Eh6)e
0
144
30.5
158
73
Senewiratne et al. (1977)
Sotho (Ab8)
5
178
30.5
23
65
Segal et al. (1983)
Spanish (Ce6)
5
162
30.5
265
15
Pena Yanez et al. (1971, 1972)
Swazi (Ab2)
4
192
20.5
12
75
Segal et al. (1983)
HB_5-6_FINAL.indb 6025/3/2010 12:28:35 PM

Page 8

Evolution of Lactose Digestion / 603
Tamils (Eg2)e
0
144
20.5
31
71
Senewiratne et al. (1977)
Thai (Ej9)
4
174
10.5
339
98
Flatz and Saengudom (1969), Flatz et al.






(1969), Keusch et al. P(1969),






Rotthauwe et al. (1971)
Tswana (Ab13)
6
178
40.5
24
83
Segal et al. (1983)
Tuareg (Aulliminden) (Cc8)
8
144
60.5
118
13
Flatz et al. (1986)
Tunisians (Cd16)
8
162
20.5
43
83
Filiali et al. (1987)
Turks (Ci5)
6
162
40.5
470
71
Flatz et al. (1986)
Tutsi (Ae10)
2
145
40.5
59
7
Cook and Dahlquist (1968), Cook and






Kajubi (1966), Cox and Elliott (1974),






Elliott et al. (1973)
Xhosa (Ab11)
3
178
30.5
17
82
Segal et al. (1983)
Yoruba (Af6)
4
144
10.5
100
91
Kretchmer et al. (1971), Olatunbosun






and Adadevoh (1971), Ransome- Kuti






et al. (1975)
Zulu (Ab12)
3
178
40.5
32
81
Segal et al. (1983)
a. Ethnographic Atlas not used. Subsistence practices and longitude and latitude taken from Brand et al. (1983).
b. 6–15% dependence on livestock, but pig- based livestock economy.
c. Shantung Chinese, Murdock (1967) ref. Ed10, cluster 163, used.
d. Nyae Nyae Kung, Ethnographic Atlas, ref. Aa1, used.
e. Longitude and latitude of Sri Lanka used, instead of using Ethnographic Atlas.
HB_5-6_FINAL.indb 6035/3/2010 12:28:35 PM

Page 9

604 / holden and mace
for the lactase persistence gene to reach observed frequencies in the time avail-
able. In this case these traits would not be truly coevolutionary, because selection
for milk consumption would not depend on the gene for lactase persistence.
Materials and Methods
One way to test the three hypotheses would be to regress the relevant en-
vironmental variables (pastoralism, sunshine intensity, and aridity) against the
prevalence of adult LDC in different populations worldwide. However, the rela-
tionships between environmental variables and LDC are highly confounded by
the shared ancestry of many human groups. Shared ancestry of several variables
can either produce a spurious, apparently functional relationship between vari-
ables or obscure a real association between variables (Mace and Pagel 1994).
For example, the association between high latitude and high LDC in Europe
could be parsimoniously interpreted as the result of northern European popula-
tions sharing a relatively recent common ancestral population in which high LDC
was prevalent. It is not necessarily correct to count each case of high prevalence
of LDC in northern European populations (e.g., in the Irish, British, Danes, Nor-
wegians, and Finns) (Sahi 1994) as an independent evolutionary adaptation, pro-
viding independent statistical evidence for the correlated evolution of high LDC
capacity and living at high latitudes.
Alternatively, many sedentary Arab populations in North Africa and the
Middle East have predominantly low LDC, although livestock were originally
domesticated in this region about 10,000 years ago (Table 1). These populations
present a challenge to the hypothesis that adult LDC is an evolutionary adaptation
to keeping domestic livestock. However, if these populations are viewed histori-
cally, it is clear that they share a recent common ancestor (Figures 1 and 2). A
parsimonious interpretation of the prevalence of low LDC in sedentary Arabs is
that low LDC was prevalent in a common ancestor of these populations. This re-
duces the challenge that is presented by these populations to the culture- historical
hypothesis if they are all counted separately.
Phylogenetic comparative methods overcome these problems by placing all
the populations on a phylogeny and by measuring the amount of change along
the branches of the tree (Felsenstein 1985). Regression analysis is done on inde-
pendent occurrences of change in the relevant variables along the branches of the
phylogeny, known as contrasts, rather than just on the branch tips, or populations
as they are seen today (Pagel 1992).
Both genetic and cultural phylogenies were used as models of the historical
relationships between populations, because both genetic and cultural transmission
might be implicated in the evolution of high LDC since lactase persistence is a ge-
netic trait, whereas pastoralism and milk drinking are cultural traits. The phyloge-
nies used here represent only estimates of past relationships between populations,
but they are assumed to be better approximations of the past than would be ob-
tained without a phylogenetic model. Not explicitly using a phylogeny implicitly
HB_5-6_FINAL.indb 604 5/3/2010 12:28:35 PM

Page 10

Evolution of Lactose Digestion / 605
Figure 1. First genetic tree of populations used in analysis, using Cavalli- Sforza’s FST linkage tree
(Cavalli- Sforza et al. 1994, p. 78), showing frequencies of lactose malabsorption in these
populations. Lactose malabsorption frequencies are grouped into three levels here for il-
lustrative purposes only: black, 0–30%; shaded, 30–70%; white, ?70%.
HB_5-6_FINAL.indb 6055/3/2010 12:28:35 PM

Page 11

606 / holden and mace
Figure 2. Second genetic tree of populations used in analysis, a modified Nei linkage tree (Cavalli-
Sforza et al. 1994, p. 78), showing percentage of dependence on pastoralism in these pop-
ulations. Dependence on pastoralism is grouped into three levels for illustrative purposes
only: black, ?50% dependence; shaded, 30–50% dependence; white, 0–30% dependence.
HB_5-6_FINAL.indb 606 5/3/2010 12:28:35 PM

Page 12

Evolution of Lactose Digestion / 607
assumes that all populations are equally related to one another, which is certainly
a less accurate representation of the past than the models used here.
An analysis using a maximum- likelihood method (Pagel 1994) was also
performed. This method gives information about the direction of evolutionary
change, that is, about which variable changed first in the course of evolution.
Data. The data used in this analysis were selected from comprehensive re-
views by Simoons (1978), Flatz (1987), and Sahi (1994) and from a literature
search of the Science Citation Index from 1981 to 1996 using the terms “lactose
absorption” and “lactose malabsorption.” Except where stated in Table 1, only
samples from adults were included.
The comparative method employed in this analysis requires that all the pop-
ulations be placed on a genetic and linguistic tree to model the historical relation-
ships between populations. All populations included are found in Cavalli- Sforza
et al.’s History and Geography of Human Genes (1994) and therefore can be
placed on a world genetic tree. Cavalli- Sforza et al. (1994) include only aboriginal
populations, defined as populations present in their current locations before 1492.
Recent migrant populations were excluded, including all nonnative Americans
and non- Aboriginal Australians. All samples of emigrant populations were also
omitted here to decrease the probability of recent genetic admixture in the data
set. Samples from populations recognized to have mixed ancestry in the original
studies were excluded. Because a linguistic tree was also used in this analysis, all
samples are also from populations whose language or language group is listed
in Ruhlen’s Guide to the World’s Languages (1991). A number of language syn-
onyms were found in Voeglin and Voeglin’s (1977) book.
All the samples used here come from populations represented in the Eth-
nographic Atlas, a cross- cultural database originally written by Murdock (1967)
and currently in the process of being revised and computerized by P. Gray (per-
sonal communication). Data on pastoralism and geographic location were taken
from this source. The Ethnographic Atlas code for each population is included in
Table 1. The Australian Aborigines are an exception. They are not a single Eth-
nographic Atlas culture, but they are a genetically monophyletic group [accord-
ing to Cavalli- Sforza et al. (1994, p. 78)] who were traditionally hunter- gatherers
without domesticated livestock.
In total, 7905 individuals from 62 distinct cultures (as recognized in the
Ethnographic Atlas) were included in the data set. The greatest loss of individual
samples resulting from the selection criteria was from the nonaboriginal popula-
tions of the Americas and Australia. More important for the aims of this study,
a number of samples from anthropological populations were unable to be used
because these populations are not known genetically or are not included in the
Ethnographic Atlas. Anthropologically interesting populations that could not
be included were the hunter- gatherer Khants from western Siberia (94% lactose
maldigesters; Lember et al. 1995) and several other groups from the former So-
viet Union (Sahi 1994), the pastoralist Kasakhs (Wang et al. 1984), the Roma
HB_5-6_FINAL.indb 6075/3/2010 12:28:35 PM

Page 13

608 / holden and mace
(Gypsies) (56% lactose maldigesters; Czeizel et al. 1983), and various Indian
groups (45% lactose maldigesters in North India, 67% in South India; Tandon et
al. 1981). The inclusion criteria applied here, however, had the advantage of al-
lowing all populations to be compared using the same source of variables, such as
pastoralism, which had been previously quantified by an independent researcher.
It also allowed the use of different trees as models of human evolution, with the
same data set of populations in each tree.
The longitude and latitude of each population were taken from the Ethno-
graphic Atlas.
The Ethnographic Atlas estimates the percentage of dependence on live-
stock in each population’s total subsistence activities. Pastoralism in this analy-
sis is a quantitative cultural trait, not a qualitative category. The midpoint of the
Ethnographic Atlas estimate was used as a measure of pastoralism here (e.g.,
where Murdock coded a culture’s reliance on livestock as 3, that is, 26–35% of
total subsistence activities, this was counted as 30.5% dependence here). Pasto-
ralism is used here only in reference to livestock capable of being milked, thus
permitting the evolution of LDC if the culture- historical hypothesis is correct. The
Ethnographic Atlas also states the main type of livestock kept and whether or not
milking was traditionally practiced. If the main type of animals kept was recorded
as pigs or small domestic animals (e.g., dogs), the society is recorded as nonpas-
toralist here (0% dependence). The analysis was repeated twice, firstly including
and then excluding pastoralists who traditionally did not milk their animals. Popu-
lations that kept livestock but did not milk them are found in China and southeast
Asia and in parts of sub- Saharan Africa (Murdock 1967; Simoons 1970).
No distinction is made in this analysis between populations that consume
predominantly processed, low- lactase forms of milk and populations that con-
sume significant amounts of fresh milk, with a high lactose content. Ethnographic
evidence suggests that milk- processing technologies are present in all dairying or
pastoralist groups today. In hot climates such as Africa and the Middle East milk
is soured naturally if it is left to stand. It seems probable that milk- processing
technologies were present early in the history of milking domestic animals. It
is therefore assumed here that all pastoralists have had equally effective milk-
processing technology, whether they lived 6000 years ago or more recently. This
assumption contrasts with Durham (1991), who interprets the high frequencies of
high LDC in some present- day North African pastoralists as the outcome of their
adopting a pastoralist mode of subsistence early on, before the full development
of milk- processing technologies.
An estimate of aridity was taken as the number of months per year with less
than 50 ml of rainfall in the area inhabited by each population (Pearce and Smith
1993). Other measures of aridity were tried, including average annual rainfall
and average rainfall in environments above 30°C, but this made no difference to
the outcome of the analysis. The sunshine experienced by each population was
estimated from the global solar radiation for land at that longitude and latitude
(Kessler 1985).
HB_5-6_FINAL.indb 6085/3/2010 12:28:35 PM

Page 14

Evolution of Lactose Digestion / 609
Statistical Methods. Correlated evolution of high LDC was tested with the
following quantitative traits: dependence on pastoralism (%), solar radiation, and
(alternatively) dry months per year and average rainfall. Independent contrasts
(which are measures of change at each branch of the tree) were generated by fol-
lowing the method of Pagel (1992), which is implemented in the computer pro-
gram CAIC (Purvis and Rambaut 1995). Regressions on the contrasts are through
the origin (Pagel 1992). We assume that all the branch lengths on the trees are
equal. LDC was used as the special variable to resolve unresolved nodes.
Two alternative genetic trees were tested, both taken from Cavalli- Sforza et
al. (1994). These two trees were constructed using different methods of calculat-
ing genetic distance. The first is an FST linkage tree by Cavalli- Sforza (Figure 1).
The second was constructed using Nei’s modified method of calculating genetic
distance (Figure 2). The cultural descent of populations was modeled using a lan-
guage tree based on the classifications of Ruhlen (1991) and assuming a mono-
phyletic origin of language (Figure 3). On a worldwide scale language groups
show a broad correspondence with genetic relationships (Cavalli- Sforza et al.
1988). The world language phylogeny used here is less resolved than the genetic
trees because of the lack of resolution above the language phylum level. Because
language evolves much more rapidly than genes do, similarities resulting from
deep historical relationships are generally considered completely obscured after
10,000 years of divergence [although see Cavalli- Sforza et al. (1988) and Ruhlen
(1991) for an opposing argument]. Because of the lack of resolution back through
time, using the language tree was rather similar to using a test involving only the
tips of the genetic trees. The genetic and linguistic trees in Figures 1 through 3
were drawn using the software MacClade. The variables were traced onto the
branches of each tree using parsimony (Maddison and Maddison 1992).
Statistical tests using independent contrasts depend on inferring, by par-
simony, a single set of values at the internal nodes of the genetic tree. Where
characters evolve rapidly and repeatedly, as is likely to be the case with cultural
traits, parsimony methods may give unreliable answers. Pagel (1994) describes a
comparative method for analyzing binary discrete characters that does not require
inferring a particular pattern of changes at internal nodes. This method finds evi-
dence for correlated evolution in two discrete characters by considering all the
possible transitions among traits on a phylogeny. The method can test hypotheses
about the extent to which the evolution of one trait is likely to be correlated with
the evolution of another and can also test hypotheses about the direction of evo-
lutionary change.
The test begins by fitting two alternative statistical models to a data set.
The model of independent change allows the two binary variables to evolve in-
dependently of each other along each of the branches of the tree. The model of
dependent change makes the probability of change in one variable dependent on
the state of the other variable. We are interested in whether the dependent vari-
able LDC is more likely to change from low to high if the dependent variable
milking is present. The independent and dependent models of change are fitted
HB_5-6_FINAL.indb 609 5/3/2010 12:28:35 PM

Page 15

610 / holden and mace
Figure 3. Linguistic tree for populations used in analysis, based on Ruhlen (1991), assuming a
monophyletic origin of language.
HB_5-6_FINAL.indb 610 5/3/2010 12:28:36 PM