ArticlePDF Available

The Comparative Method in Anthropology [and Comments and Reply]

January 1994
Current Anthropology 35:549-564

Authors:

Ruth Mace

University College London

Show all 8 authorsHide

Phylogenie homology and (bottom) analogy.

…

Figures - uploaded by Ruth Mace

Content may be subject to copyright.

Content uploaded by Ruth Mace

Content may be subject to copyright.

The Comparative Method in Anthropology [and Comments and Reply]

Author(s): Ruth Mace, Mark Pagel, John R. Bowen, Biman Kumar Das Gupta, Keith F.

Otterbein, Mark Ridley, Thomas Schweizer and Eckart Voland

Source:

Current Anthropology,

Vol. 35, No. 5 (Dec., 1994), pp. 549-564

Published by: The University of Chicago Press on behalf of Wenner-Gren Foundation for

Anthropological Research

Stable URL: http://www.jstor.org/stable/2744082

Accessed: 13-02-2018 06:55 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide

range of content in a trusted digital archive. We use information technology and tools to increase productivity and

facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

http://about.jstor.org/terms

Wenner-Gren Foundation for Anthropological Research, The University of Chicago

Press

are collaborating with JSTOR to digitize, preserve and extend access to

Current

Anthropology

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

CURRENT ANTHROPOLOGY Volume 35, Number 5, December I994

The Comparative

Method in

Anthropology

by Ruth Mace

and Mark Pagel

Cross-cultural comparison is a common method of testing

hypotheses regarding the co-evolution of elements of cultures or

of the adaptiveness of a cultural practice to some aspect of the

environment. It has long been recognized, however, that cultures

are not independent but rather may share many cultural ele-

ments by virtue of common ancestry and proximity. Attempts to

address this issue, known as Galton's problem, range from statis-

tically removing confounding variables to using a standard sam-

ple of "independent cultures." We show here that when testing

any hypothesis of co-evolution one should not attempt to iden-

tify independent cultures or to create them statistically. Rather,

cross-cultural comparative studies must be based upon the identi-

fication of independent events of cultural change. Once this prin-

ciple is applied, it becomes apparent that it is in fact groups of

closely related cultures that are potentially the most informative

for testing cross-cultural hypotheses. Constructing phylogenies

of cultures and placing upon them independent instances of cul-

tural elements' arising or changing is an essential part of this

task.

RUTH MACE iS a Royal Society University Research Fellow in

the Department of Anthropology of University College London

(Gower St., London WCiE 6BT, England). Born in i96i, she was

educated at Oxford University (B.A., i983; D.Phil., i987) and has

held research positions in the Centre for Environmental Technol-

ogy of Imperial College (i987-89) and the School of Develop-

ment Studies of the University of East Anglia (i989-9i). Her re-

search interests are subsistence economics, pastoralism, human

ecology, and conservation. She has published "Pastoralist Strate-

gies for Survival in an Unpredictable Environment: A Model of

Herd Composition" (Agricultural Systems 3I:i85-204), "Over-

grazing Overstated" (Nature 349: 280-gI), "Nomadic Pastoralists

Adopt Strategies That Maximise Long-term Household Survival"

(Behavioural Ecology and Sociobiology 33:329-34), and "Transi-

tions between Cultivation and Pastoralism in Sub-Saharan Af-

rica" (CA 34:363-82).

MARK PAGEL directs a research unit on ecology and behaviour in

the Department of Zoology of Oxford University. He was born in

I954 and educated at the University of Washington (B.A., I976;

Ph.D., ig80). He has taught evolutionary theory and statistics at

Harvard University (i990-92) as well as at Oxford. His publica-

tions include "Detecting Correlated Evolution on Phylogenies: A

General Method for the Comparative Analysis of Discrete Char-

acters" (Proceedings of the Royal Society B 255:37-45), "The

i. This research was funded by the Royal Society (RM) the Science

and Engineering Research Council, the National Environment Re-

search Council, and the Biotechnology and Biological Sciences Re-

search Council (MP).

Adaptationist Wager," in Phylogenetics and Ecology, edited by P.

Eggleton and D. Vane-Wright (London: Academic Press, in press),

"The Evolution of Conspicuous Oestrous Advertisement in Old

World Monkeys" (Animal Behaviour 47:I333-4I), and, with

P. H. Harvey, The Comparative Method in Evolutionary Biology

(Oxford: Oxford University Press, i99i).

The present paper was submitted in final form 6 XII 93.

For well over a century, anthropologists have used infor-

mation obtained by comparisons across cultures to test

ideas about cultural evolution. Since the i 950S many

of these comparisons have made use of tabulations of

cultural information compiled by G. P. Murdock and

others, who codified over I,250 cultures extant at that

time. Shortly after the first studies making use of Mur-

dock's Atlas of World Cultures were published, a large

branch of anthropology (mostly located in Britain) re-

acted to the use of facts about culture in this way. The

idea that the facts themselves should be considered as

social constructs came into the ascendancy, and much

of anthropology redefined itself as an interpretive hu-

manity rather than a science, concerned with cultural

specificity rather than comparison across cultures. Holy

(i987) remarked that "the line between comparativists

and non-comparativists is probably more sharply drawn

than ever before," the latter being in the numerical ma-

jority.

Meanwhile, the past decade has seen a great expan-

sion in other branches of anthropology in which anthro-

pologists study humanity, including cultural diversity,

as a scientific endeavour. Here the comparative method

is widely used both by those who are interested in the

evolution of culture and in other disciplines of the hu-

man sciences, such as demography, in which, for exam-

ple, fertility is considered to be at least in part cultural.

Issues that have recently been addressed using cross-

cultural comparison as empirical evidence include ques-

tions concerning the co-evolution of rules of inheritance

and marriage (Hartung I982, i985), pathogen risk and

polygyny (Low I988, i990), incest taboos and exogamy

(Durham i99i, Thornhill i99I ), reproduction and social

organization (Lesthaeghe i989), and modes of subsis-

tence and fertility (Campbell and Wood I988, Bentley,

Goldberg, and Jasienska I993).

Most models of cultural evolution include the as-

sumption that culture evolves by "descent with modi-

fication" (Cavalli-Sforza and Feldman I98I, Lumsden

and Wilson I98I, Boyd and Richerson I985, Durham

I 99 1), probably from a single origin. Most models of cul-

tural evolution also acknowledge the existence of bun-

dles of characters that can be transmitted as indepen-

dent units (i.e., even when other aspects of the culture

are not passed on); we shall call these "elements" of

culture.

Perhaps the most common way of analysing compara-

tive data is to correlate two or more elements across a

group of cultures. However, Francis Galton realized as

early as I889 that cultures cannot be treated as indepen-

dent for purposes of investigating cross-cultural trends.

549

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

550 I CURRENT ANTHROPOLOGY Volume 35, Number 5, December 1994

The historical patterns of relatedness among societies

mean that they cannot be assumed to have evolved or

acquired their particular characteristics independently.

Therefore a cross-cultural association between two or

more elements of culture does not necessarily indicate

that the two elements tend to evolve or to be acquired

together. They may both have been acquired for unre-

lated reasons in an older or ancestral culture and then

subsequently transmitted to many later descendant cul-

tures. Since Galton's time, anthropologists have devel-

oped two types of methodology to confront the problem

of non-independence in comparative studies.

"Solutions" to Galton's Problem

Murdock attempted to solve Galton's problem by cate-

gorizing the I,250 cultures in his famous Atlas into a

smaller number of clusters. He assumed that the clus-

ters represented independent groupings but argued that

the cultures within a cluster would display many simi-

larities, presumably owing to both proximity and com-

mon descent. Murdock and White (i969) published a

sample of cultures in which each cluster is represented

only once. This "standard cross-cultural sample" of I86

cultures is still widely used today. Other investigators

have chosen different samples of cultures depending

upon the availability of certain kinds of data.

Both Murdock's standard sample and other samples

drawn from his clusters correctly identify the problem

of non-independence and therefore improve upon simple

cross-cultural correlations that ignore history. However,

for two reasons this approach fails to provide a princi-

pled solution to Galton's problem. One is that it merely

moves back in time the issue of historical relatedness:

the clusters may not be independent for all the same

logical reasons that cultures within a cluster are not.

Sets of cultures will be related to varying degrees. If we

believe in cultural monogenesis, then there are no such

things as "independent cultures," only cultures with dif-

fering degrees of relatedness, and therefore Murdock's

clusters must also be related, even if distantly. This

means that clusters will not be independent on variables

that evolved or were acquired and retained in common

ancestors to those clusters. An additional limitation of

"standard samples" is that such samples prevent re-

searchers from making comparisons among closely re-

lated societies. Murdock's approach recognizes the im-

portance of history but confronts it with a very broad

brush. There may be ample variation among the closely

related cultures within a cluster that is ignored. Further,

cultures within a cluster are treated as equally related

and therefore as equally non-independent. But, if histori-

cal relationships are identified on a more recent level,

subclusters or even subsubclusters may emerge. This is

a serious limitation because we may have the best

chance of understanding historical patterns of relat-

edness among closely related societies (e.g., Eggan I975

[I9541) and therefore of understanding patterns of cross-

cultural correlation.

The second set of techniques for confronting Galton's

problem relies upon statistical procedures to control for

or remove from the data points those portions of the

variation that are thought not to be independent among

cultures. Although statistical approaches vary in detail,

they are based on a shared premise: that variation among

cultural groups can be divided into a portion caused by

the historical patterns of descent and of proximity and

a portion representing unique independent cultural evo-

lution. Only the latter portion-a statistical residual-is

used to investigate comparative trends. The statistical

techniques, including various regression models (or,

equivalently, analyses of variance) and methods of spa-

tial autocorrelation analysis (Dow et al. I984, Dow

i99i), can in principle produce sets of statistically inde-

pendent data points. Like Murdock's standard sample,

then, they are an improvement upon ignoring common

ancestry entirely. However, they also fail to provide a

principled solution to Galton's problem. These meth-

ods, in effect, are victims of their own zeal. By control-

ling for the effects of ancestry and/or proximity they

remove that which in Murdock's approach would

roughly correspond to the differences among clusters.

We say "roughly," however, because the methods are

not limited to controlling only for those major differ-

ences associated with clusters. Depending upon how

precisely one specifies the "network" of connections

(the shared ancestry and/or proximity) among cultures,

much more variation may be removed. The variation

that remains once the "shared" variation is removed is,

again roughly, equivalent to that which would be found

from comparing only very closely related samples.

Thus, the distinguishing comparison between the

standard sample and statistical approaches is, ironically,

that each technique makes use of that which the other

one either discards or must ignore: statistical ap-

proaches discard the variation that Murdock's cluster

approach retains, whereas Murdock's approach cannot

make use of the variation among closely related cultures

that the statistical techniques retain. The dilemma is

that both types of information are relevant to testing

cross-cultural comparative trends.

The challenge confronting comparative anthropolo-

gists, then, is managing to make use of both differences

among major cultural clusters and differences among

more closely related societies. The answer lies in identi-

fying what we shall call independent instances of cul-

tural change. Darwin recognized implicitly that the va-

lidity of comparative biological trends depended upon

counting independent instances of evolutionary change

in species (see Ridley i992). Galton later recognized the

same problem in cross-cultural analyses. Ridley (i983)

formalized this idea for discrete or categorical traits, and

later work (Felsenstein I985 and others, reviewed in Pa-

gel and Harvey I 98 8; Harvey and Pagel i 9 9 i) has gener-

alized the principle to traits that vary on a continuous

scale.

For purposes of cross-cultural comparative studies, we

shall call any instance of the de novo invention or acqui-

sition of a new element by copying from another culture

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method I 55

AB AB AB AB AB A'B' A'B' A'B' A'B'

A'B' AB AB A'B' AB A'B' AB AB A'B'

B \ I / B' B'

FIG. i. Phylogenies of culture showing correlations between elements A' and B' that result from (top)

homology and (bottom) analogy.

or the change or loss of an element as an independent

instance of cultural change. The critical point of this

essay is that the validity of comparative methods for

anthropology depends upon correctly counting indepen-

dent instances of cultural change. Independent instances

of cultural change, in turn, cannot be identified without

the construction of a phylogeny (or cladogram) showing

the patterns of hierarchical descent of the cultures being

studied.

The Importance of History

Consider that we find a number of cultures that contain

the elements A and B. That B arises in the presence of

A in some cultures does not mean that all instances of

B were caused by A (Boas I940 [i8961), but if the ele-

ments are causally or functionally related, then they will

appear together more often than would be expected by

chance. Whether this is the case is what comparative

studies are designed to test. Central to our approach is

the idea of a cultural phylogeny. Evidence for correlated

cultural change is obtained from the phylogeny if pairs

of cultural elements change together on repeated, inde-

pendent occasions. Crucially, then, it is not the number

of cultures that currently have the elements that mat-

ters for testing cross-cultural hypotheses. Rather, what

counts as evidence for correlated evolutionary change

is the number of times those elements have changed

together.

If cultures evolve by descent with modification, then

when one culture is descended from another the two

will have many common elements. Elements present in

the daughter culture that were inherited from the

mother culture are described as homologous. Figure i

(top) shows a phylogeny of hypothetical cultures in

which we assume that only vertical transmission of cul-

tural information has occurred. If one were to examine

the cultures that are represented by the tips of the phy-

logeny, one might be tempted to imagine that A' was

co-evolving with B' and that they were somehow func-

tionally linked: five cultures have the elements (AB) and

four have (A'B'), with no exception to the rule that A'

occurs with B'.

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

552 CURRENT ANTHROPOLOGY Volume 35, Number 5, December 1994

Now, consider that a historical analysis reveals that

the ancestral culture from which all these other cultures

evolved was (AB). A evolved into A' and then B evolved

into B' only once, as shown in figure i (top). The associa-

tion of A' with B' at the tips can be fully understood on

the basis of one instance of B' evolving in the presence

of A'. There is no evidence for the repeated co-evolution

of these two elements. The fact that the branch of the

tree in which this occurred has since evolved into many

separate cultures, all of which still share A' and B', is of

no statistical significance to our hypothesis. A' and B'

may each be functional, which explains why both ele-

ments are maintained. However, they may have no con-

nection with each other whatsoever. A correlation be-

tween the two elements owing to homology is not

evidence for repeated co-evolution.

In contrast, if an element evolves on many separate

occasions in separate cultures, then the cultural ele-

ments are described as analogous. Figure i (bottom) il-

lustrates an example with an equal degree of associa-

tion between A' and B' in extant cultures as figure i

(top), but in this case the instances of A' and B' are anal-

ogous. A correlation across a range of cultures between

two or more analogous elements is evidence for the

repeated and independent co-evolution of those ele-

ments.

We are now in a position to restate in terms of a phylo-

genetic approach why the previous attempts to solve

Galton's problem fail. Murdock's standard-sample ap-

proach is analogous to drawing one culture from each of

the main branches of figure i. These clusters would be

independent for the characters A, B, and A'B'. However,

any cultural element that evolved prior to the root of

the tree in figure i will be shared by descent among

the clusters of cultures: correlations calculated across

clusters may still be artefacts of shared ancestry. Fur-

thermore, using any standard sample such as Murdock's

we could not study cultural variation within the closely

related cultures of figure i. The elements of interest may

have been gained and lost several times within one of

the clusters (such as in fig. i [bottom]), but such relevant

details could not be included in an analysis based on the

standard sample. Conversely, the statistical techniques

would tend to remove the variation represented by the

main and secondary branches of figure i. These meth-

ods, in effect, would analyze only the variation repre-

sented by pairs of cultures at the tips of the phylogeny

and discard variation that originated earlier.

Reducing the statistical analysis to independent in-

stances of cultural change as we advocate might remove

correlations found in past studies that seem intuitively

rather unlikely to indicate any functional link, such as

that between circumcision and climate (Whiting I964,

Saucier I972) or between postpartum sexual abstinence

and ritual art (Lesthaeghe i989). It may also help to re-

solve cases in which two studies testing the same hy-

pothesis have reached different conclusions, such as

Campbell and Wood's (i988) finding that fertility is not

associated with subsistence practice and Bentley, Gold-

berg, and Jasienska's (I993) finding that it is.

Reconstructing Phylogenies and Identifying

Ancestral Character States

Once it is appreciated that a comparative hypothesis is

actually a hypothesis about the pattern of origination

and modification of cultural elements in different soci-

eties, much debate will ensue about whether a particular

phylogeny is the correct phylogeny for a particular set

of cultures. We give here only a very brief overview of

the key issues of reconstructing phylogenies and of iden-

tifying ancestral character states. Discussions of many

of the issues of reconstructing ancestral character states

can be found in the biological literature (Felsenstein

I979, I983; Maddison, Donoghue, and Maddison I984;

Ridley I986; Harvey and Pagel i99i).

Unlike biological phylogenies, in which a true pattern

of vertical descent exists, cultural phylogenies, even if

predominantly vertical, will contain many instances of

populations interbreeding and the horizontal transmis-

sion (e.g., between neighbours) of cultural elements.

(Horizontal transmission has been variously described

as diffusion, copying, borrowing, or "synology" [Dur-

ham I990]; we use the term horizontal transmission be-

cause it implies nothing about the mechanism of or

motivation for the adoption of the element.) We shall

have to accept, then, that a cultural phylogeny rep-

resents only broadly the cultural path that most of the

ancestors of the majority of members of that culture

followed.

On the evidence of Cavalli-Sforza et al. (i988), the

cultural and genetic phylogenies may not differ greatly.

For this reason genetic evidence will often provide a use-

ful starting point for reconstructing phylogenies. How-

ever, a phylogeny should probably be based on all the

evidence available, be it genetic, linguistic, archaelogi-

cal, historical, or cultural (assuming that the elements

relevant to the hypothesis being tested are not also used

to construct the phylogeny, which would be circular).

In this way, there is the greatest chance that one will

converge upon the true phylogeny.

Language has the advantage of evolving in what lin-

guists have described as a genetic way (Ruhlen i987).

Languages also diverge from each other in fairly consis-

tent ways, and therefore the extent of difference between

two related language groups can be used to estimate

when those two languages separated. Genes and lan-

guage do not evolve in identical ways, so the construc-

tion of linguistic phylogenies is difficult and controver-

sial (Ruvolo I987, Bateman et al. i990), especially for

establishing relationships between groups of more dis-

tantly related cultures. For more closely related cul-

tures, however, languages may be very useful. Language

differences typically evolve much more rapidly than ge-

netic differences and therefore can be used to separate

groups that may be difficult to distinguish on common

genetic indices. If it is true that the loss of a native lan-

guage in favour of another language is generally associ-

ated with the adoption of most or all of a whole range of

elements from that culture, then linguistic phylogenies

may be even more appropriate than genetic ones. Lan-

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method 553

90;, <,0 c:fi<>/ qg>OOs C; Gx C;G :o

Nilotic < Cushitic

C-

FIG. 2. A phylogeny of nine Kenyan pastoralist cultures based on linguistic similarity, showing (C +)

camel-keeping cultures and (C-) cultures without camels, (black bars) some possible positions of independent

events of the adoption of camel herding, and (arrow) a possible route for the horizontal transmission of camel

herding across cultures. (Branch lengths do not necessarily represent time.)

guage is also not likely to be one of the elements in

our co-evolutionary hypothesis. Whilst being far from

perfect, it may offer the best general method of recon-

structing cultural phylogenies that we have, and Ruh-

len's (i987) classification of the world's languages (build-

ing on the work of Joseph Greenberg and others)

provides the basis for such a phylogeny.

Once a phylogeny is available, the process of identi-

fying ancestral character states can take place. Direct

historical or archaeological evidence may sometimes be

available. Where evidence of ancestral states is lacking,

it is necessary to infer them. One of the best-known

techniques for inferring ancestral character states is par-

simony. Parsimony methods find the minimum number

of evolutionary events on a tree that is required to ex-

plain the distribution of elements among the contempo-

rary cultures. Given the many potential methods of cul-

tural transmission, parsimony may often underestimate

the true number of evolutionary events. However, any

evidence of repeated co-evolution based on parsimony

is therefore likely to be real (statistical type-I errors will

be minimized), although real cases of repeated co-

evolution may be overlooked (statistical type-2 errors

may be increased). A useful review of computer pro-

grams that can use a variety of specified criteria to con-

struct phylogenetic trees (including PHYLIP, PAUP, and

Hennig86) and ancestral character states (MacClade) can

be found in Maddison and Maddison (I989, I992).

The Adoption of Camel Herding by East

African Pastoralists

To illustrate the processes of phylogeny reconstruction

and the identification of instances of cultural change,

we shall consider some pastoralist cultures in Kenya.

We are interested in testing the idea that camel herding

is adopted in dry climates. We have constructed a phy-

logeny of these groups based on linguistic evidence (fig.

2). This cultural phylogeny portrays the broad division

between the Nilotes and the Cushites that is also sup-

ported by genetic differences (Cavalli-Sforza et al. I988).

Among the Cushitic groups, four extant cultures are

considered. Classified on the basis of linguistic similar-

ity, Gabbra and Borana (who speak essentially the same

language) fall together on one branch and Rendille and

Somali fall together on another. Among the Nilotes,

Samburu and Maasai speak the same language and Tur-

kana and Pokot are more distantly related.

To illustrate the reconstruction of ancestral character

states, we may look at these pastoralist groups and see

which of them are camel-keeping. (Technically we are

interested in the meme [Dawkins I976] or ideational

unit [Durham i99i] of aspiration to camel keeping

rather than the behaviour itself, because not all individ-

uals in any culture will own camels; the poor may not

be able to acquire camels, nor would it necessarily be in

their interest to do so [Mace and Houston i989, Mace

I990], but most would aspire to own them if they were

rich enough.) Those cultures that are camel-keeping are

shown in figure 2. (All the cultures that are not camel-

keeping specialize in cattle.) Archaeological evidence

suggests that camels may have arrived in the region

around 4,000 years ago, which may also have been a

period of increased climatic drying (Gowlett I988). Ca-

valli-Sforza et al.'s (I988) genetic analysis suggests that

the separation of the Nilosaharan and Afroasiatic (Ethio-

pian) culture clusters pre-dates this by some time. Thus

their last common ancestral culture at the root of the

phylogeny did not herd camels (and may not have herded

anything).

When the splits among the Cushitic groups occurred

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

554 CURRENT ANTHROPOLOGY Volume 35, Number 5, December I994

is unknown, but it was probably after the arrival of cam-

els in the region (Ruhlen's rule of thumb is that separa-

tion for 500 years makes languages mutually difficult to

comprehend and separation for i,ooo years makes them

mutually incomprehensible). One of the most parsimo-

nious patterns of the adoption of camel herding by differ-

ent cultures that is consistent with the historical and

archaeological evidence is that shown in figure 2. Other

equally or less parsimonious trees are possible. An alter-

native pattern of events is that the common ancestor of

the four Eastern Cushitic cultures herded camels and

the Borana then lost this element. Further historical,

archaeological, genetic, and linguistic analyses will no

doubt throw more light on the evolution of these cul-

tures.

Whichever of these reconstructions is correct, the ob-

served pattern of camel herding in these nine Kenyan

pastoralist groups can be explained by a minimum of

four independent instances of cultural change. The Maa-

sai, southern Samburu, Pokot, and Borana all live in wet-

ter areas than any of the camel-keeping groups. The

northern Samburu have taken up camel herding only

very recently, especially since a severe drought in I984.

Thus in all four cases the adoption of camel herding

and a dry environment occur together. Including a wider

range of pastoralist cultures would further strengthen

our confidence that this association is statistically un-

likely to be due to chance and that the behaviour is

functionally linked to a dry environment.

The same phylogeny can be used to test additional

ideas. Circumcision practices also vary among these pas-

toralist groups but differently from the practice of camel

herding. The Gabbra, Borana, and Somali all practise a

drastic form of circumcision on very young girls. The

Rendille, on the other hand, practise a different type of

circumcision on young women at their marriage. The

similarities among the Gabbra, Borana, and Somali prac-

tices are such that it is implausible that each culture

invented these practices in isolation; some horizontal

transmission (such as observing and copying) must have

occurred. Nor is it possible to construct a phylogeny in

which camel herding and circumcision practices evolve

together. If we consider the linguistic phylogeny to be

the true phylogeny of the cultural groups, then there

must have been more than one independent instance of

a gain or a loss of both camel herding and circumcision,

even within this cluster of four closely related cultures.

This therefore provides an example of elements of cul-

ture evolving independently of each other.

Horizontal Transmission and Comparative

Studies

Many elements of culture are transmitted horizontally

among societies. As we have seen, camel herding and

some circumcision practices are very likely to have

spread by horizontal transmission among the Kenyan

pastoralist groups. Although vertical transmission (i.e.,

from the parental culture) may be the most common

form of transmission for most elements of culture, there

is a range of possible mechanisms by which cultural ele-

ments can be transmitted horizontally (Boyd and Richer-

son I985), whether or not groups are interbreeding, and

horizontal transmission should not be considered rare.

Many comparative analyses have statistically re-

moved effects of geographical proximity (Naroll I970,

Dow et al. I984) on the grounds that they are not con-

cerned with cultural similarities that result from "diffu-

sion." Implicit in such analyses is that acquisition of an

element of culture by horizontal transmission should

not be considered as an independent event of cultural

change. But horizontal transmission should not be ig-

nored in comparative studies of the evolution of culture

which seek to test adaptive hypotheses. Indeed, horizon-

tal transmission can be used to identify additional in-

stances of the correlated evolution of cultural elements.

In the case of the Kenyan pastoralists, camel herding

may have been transmitted horizontally along the path

shown in figure 2. The northern Samburu got the idea

of camel herding (and the camels) from neighbouring

cultures, especially the Turkana, but this does not make

their adoption of camels any less relevant to our hypoth-

esis. Indeed, contrasting northern and southern popu-

lations of Samburu that intermarry and do not differ

greatly in language or other aspects of their culture but

do differ in the relevant environmental parameters and

in camel keeping could provide some of the strongest

evidence in favour of the hypothesis.

Nevertheless, not all the circumstances of horizontal

transmission should be considered as independent

events. Consider that we wished to test another hypoth-

esis on figure 2: that female circumcision co-evolves

with pastoralism. Historical sources (Worthy I959,

Cronk I989) report that earlier this century the Muko-

godo changed from living as hunter-gatherers to living

as pastoralists and started circumcising their daughters.

We might take this as evidence of an event of co-

evolution of these two cultural elements. But the history

also reveals that the changes coincided with the onset

of intermarriage between the Mukogodo and Maasai

groups and the Mukogodo were to a great extent assimi-

lated into Maasai culture. Dress, religion, marriage

rules, methods of housing, and even language all

changed to the Maasai form. The phylogeny of these

cultures contains what is known as an anastomosis-a

fusion of two branches of the tree. The culture emergent

from the anastomosis could be different from either of

the two parent cultures or (as is probably true in this

case) just a continuation of one of them (the Maasai

branch). If all the Mukogodo were assimilated into Maa-

sai culture, then the Mukogodo branch of the tree would

cease to be represented by any tip on the phylogeny-

that is, that culture would be extinct. It will generally

be incorrect to consider any cultural elements found in

the tree immediately after the anastomosis to have

arisen independently if that element was present in ei-

ther of the parent cultures. Thus the identification of

anastomoses could be important when counting inde-

pendent events of cultural change.

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method | 555

Another example of an anastomosis is the formation

of the Ariaal by the intermarriage of Samburu and Ren-

dille (Spencer I973). Cultural practices from both of the

parent cultures are practised by the Ariaal. It may be

that they are simply a transitional group housing mainly

migrants from the Rendille to the Samburu culture; al-

tematively, they may develop their own cultural iden-

tity, even their own language, over time. Many argue,

however, that one culture and language will generally

win out over the other in cases of fusion (Renfrew I987).

This view implies that splitting, separation, and barriers

to cultural mixing rather than cultural fusion have been

the more prevalent mechanisms for the origination of

cultural diversity as they are for biological diversity. The

fact that it is possible to identify, individual cultures,

rather than human cultural diversity's appearing as an

amorphous soup of continuous variation, also supports

this hypothesis (see Naroll I97I on how to treat lan-

guage ambiguities).

Another type of horizontal transmission that should

probably not be counted as an independent instance of

cultural change is imposed cultural change (Durham

I99I). Warfare and domination of one cultural group by

another, slavery, colonialism, and statehood can all

cause such cultural change through imposition. Fre-

quently cultural elements of the subordinate culture are

actively exterminated, as in religious persecution. There

is also the less easily defined area of groups adopting

cultural elements from wealthier or more powerful cul-

tures because the association with the higher-status

group, rather than the element itself, has some perceived

or actual benefit (Boyd and Richerson [I985] call this

"indirect transmission bias"). Female circumcision may

currently be spreading through the Sahelian zone in this

way because of its association with the wealthier Is-

lamic cultures and the association of the uncircumcised

state with slavery (Lightfoot-Klein I989). Long-distance

warfare, colonialism, and large intercultural disparities

in wealth may be relatively recent phenomena in the

evolutionary history of human culture, but they are now

widespread and may be responsible for the loss of much

of the world's cultural diversity. These phenomena are

likely to complicate the interpretation of comparative

studies of cultural evolution, and special care should be

taken to recognize such phenomena as causal agents in

apparent instances of the repeated co-evolution of the

cultural elements that may be spread in this way.

Horizontal transmission, then, can often be treated as

an independent instance of cultural change in the cul-

ture in which it arises and thus a useful data point for

comparative study. Copying and learning by one culture

of another culture's practices can provide evidence for

the functional links among various elements of culture.

However, great care must be taken in the interpretation

of large blocks of cultural elements appearing together,

which may be due to anastomoses. Imposition of cul-

tural elements and acquisition by status association may

also mislead tests of functional associations.

If a comparative hypothesis is formulated that con-

cerns only instances of de novo invention or modifica-

tion of an element within a culture, then efforts will

have to be made to separate effects of horizontal trans-

mission from local effects. For example, although camel

herding was transmitted horizontally among pastoralists

in northern Kenya, the breeds now used by the different

groups are physiologically distinct (except that the

Samburu acquired their camels very recently from the

Turkana and the breeds used by these two groups are

still similar). Thus an element of de novo adaptation to

the new culture and environment has occurred since the

acquisition of the broader idea of camel husbandry.

Comparative Studies Using Continuous

Variables

We have based all of our examples so far on elements of

culture that can be classified as taking a small number

of discrete values: camel herding or not, a particular cir-

cumcision practice or not, dry climate or not. However,

many elements of culture or environmental variables

are more naturally thought of as varying along a contin-

uous scale. Fertility, bride-price, degree of polygyny,

temperature, and rainfall are examples of continuous

variables. In the example of figure 2 we might instead

have studied the (average) number of camels per house-

hold as a function of the average amount of rainfall, and

this might have given us more information relevant to

our hypothesis.

When the relationship between two quantitative cul-

tural elements is sought, the phylogeny is used to define

a series of comparisons or "contrasts" between pairs of

"sister-cultures" on the phylogeny. Sister-cultures have

an immediate common ancestor (e.g., cultures 3 and 4,

6 and 7, and 8 and 9 in figure 3). The differences between

two such cultures are, by definition, differences that

have been acquired since the two cultures separated and

thus are events of cultural change. For the example

given here one would compare each sister-culture pair

on the average rainfall and the average size of a herd. If

the hypothesis is correct, greater differences in rainfall

between cultures should be associated with greater dif-

ferences in camel herd size.

This form of analysis is not limited to comparisons

among sister-cultures, although they will usually be the

simplest and most straightforward to define. It is also

possible to define contrasts between the ancestral nodes

on the phylogeny or between a culture at the tip and a

node. Figure 3 (top) shows the sister-culture compari-

sons and the culture-node and node-node comparisons

defined by that phylogeny, which comprises all the pos-

sible contrasts. Thus, in addition to the three sister-

culture comparisons, the comparisons between nodes

X2 and X4 and between X7 and X5 and the culture-node

comparisons between culture i and X6, between culture

2 and X3, and between culture 5 and Xi provide five

additional independent pieces of evidence that bear on

the presumed functional relationship. The difficulty

with these additional comparisons is that one needs a

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

556) CURRENT ANTHROPOLOGY Volume 35, Number y, December I994

1 2 3 4 5 6 7 8 9

3 ~~~~~4

1 2 3 4 5 6 7 8 9

3 ~~~~4

\~~Xy

FIG. 3. Contrasts counted when testing for an association (top) between two quantitative variables and

(bottom) between a categorical element (black bars) and a quantitative variable. Numbers, extant cultures;

X's, nodes.

way to estimate the values of the characters at the

nodes. The simplest approach is to estimate every higher

node as the mean or average of the points immediately

above it on the phylogeny; thus Xi would be estimated

as (culture 3 + culture 4)/2. The values of X2 and X4

can be estimated by the same procedure. The value for

X5 would then be found as the mean of the values found

for X2 and X4, and so on throughout the tree. This form

of averaging makes implicit assumptions about rates of

cultural evolution in the various branches of the tree.

However, it has been shown to have reasonable statisti-

cal properties in several computer simulation studies

(Martins and Garland I99I, Gittleman and Luh I 992,

Purvis, Gittleman, and Luh I994). It should be empha-

sized that the averaging process is not a way of estimat-

ing what was going on in the past but merely a way of

partitioning the variability among the cultures of the

phylogeny into independent parts. Other ways of esti-

mating the nodes are possible, however (see, e.g.,

Felsenstein I985, Harvey and Pagel I99I, Martins and

Garland I99I, and Pagel I992).

Frequently one will want to test for the association

between a quantitative variable and a categorical vari-

able (such as the amount of rainfall and the presence or

absence of camel husbandry; fig. 3 [bottom]). Here again,

the phylogeny makes clear how to test such a hypothe-

sis. Wherever a transition occurs from one state of the

categorical variable to the other, we compare the average

value of the continuous variable in the two conditions.

This can often be achieved with a sister-culture compar-

ison, such as those between cultures 3 and 4 and cul-

tures 6 and 7. Cultures 2 and 5 also provide such a com-

parison, if not as close a one as the others. The

remaining transition to camel herding requires that one

develop estimates of X4 and X5. One could in this case

estimate X4 = (culture 8 + culture 9)/2. The estimate

of X5 is made difficult by the fact that cultures 6 and 7

have already been used. To avoid any such overlap one

could compare the estimate of X4 with the value for

culture i, but such "long-distance" comparisons are

generally not as well controlled as sister-culture com-

parisons. Many other elements of culture will have

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method J 557

changed along the branches leading from X8 to cultures

i and 9.

Statistical Tests

The phylogenetic approach that we have outlined pro-

vides a framework within which to conduct statistical

tests of comparative hypotheses. The common feature of

all the analyses we have described is that the phylogeny

identifies a set of independent contrasts or comparisons

that can therefore be used as units of analysis in appro-

priate statistical tests. It is not necessary to limit one's

analyses to two variables at a time. Comparisons can

simultaneously be formed on a large number of vari-

ables, some of which may be used as controlling vari-

ables in later statistical analyses.

To test hypotheses about transitions in two categori-

cal variables, one first reconstructs the probable ances-

tral character states at each of the nodes of the phylog-

eny, using, for example, parsimony or related methods.

Then one notes each branch of the tree in which one,

the other, or both of the variables changes state between

the beginning and end points of the branch. For these

branches only (that is, ignoring branches in which nei-

ther variable changes), one writes down their end-states.

Four end-states are possible, and these define a two-by-

two contingency table in which one merely tallies the

observed end-states of those branches in which some

change occurred. If changes in the two variables are posi-

tively correlated, there will be a preponderance of tallies

in the X = o, Y = o and X = i, Y = i boxes of the

contingency table, where X and Y are the two variables

under investigation. A negative relationship will yield a

preponderance of tallies in the X = o, Y = i, and X =

I, Y = O boxes. If there is no relationship, each of the

boxes will tend to have the same number of tallies. The

association can then be tested by a chi-squared statistic

or a Fisher's exact test. Ridley (I983), Pagel and Harvey

(I988), and Harvey and Pagel (i99i) discuss this method

in greater detail.

A limitation of the contingency table method is that

it fails to distinguish among some kinds of transition.

For example, a branch in which X changed from o to i

but Y remained unchanged in state i would be tallied

in the same box as a branch in which Y changed from o

to i and X remained unchanged in state i and the same

as a branch in which both X and Y changed from o to i

(see Harvey and Pagel i 99 fI). Maddison (1 990) provides a

method that can overcome this limitation. This method

assesses whether an observed pattern of transitions on

the phylogeny can be used to support the hypothesis

that changes in one of the variables make changes in

the other more or less likely. Pagel (I994) reports a maxi-

mum-likelihood technique based on Markov transition

models that also can test whether certain directions or

patterns of change occur more often than expected by

chance. This method tests for correlated cultural change

on a phylogeny without relying upon reconstruction of

ancestral character states.

To test for a relationship between two quantitative

variables one merely correlates the two or more sets of

pairwise comparisons or contrasts. Here partial correla-

tion can be used to control statistically for the effects of

one variable whilst examining the relationship between

two others. Tests between a continuous and a categori-

cal variable make use of the fact that each contrast de-

fines a difference on the continuous variable between a

group (or node of the phylogeny) in which the categorical

variable is present and a group (or node) in which it is

absent. This set of pairwise difference scores can then

be tested by means of a paired t statistic (or equivalent

non-parametric statistic).

Conclusion

It may be tempting to ignore phylogeny in comparative

methods. Constructing phylogenies requires extra work,

they will often be contentious, their use often reduces

the degrees of freedom in statistical tests, and some of

the assumptions made in "contrast" comparative meth-

ods may seem unrealistic. But it has to be remembered

that simple cross-cultural correlations, analyses of vari-

ance, regressions, and chi-squared tests of the type that

are currently being used in comparative studies of hu-

man cultures also make implicit assumptions about the

relationships among cultures and about rates of evolu-

tion, whether the investigators realize it or not. Such

analyses are generally only justified if the elements un-

der study have evolved or been adopted independently

in every culture. This will be true only if all the cultures

in the analysis separated from some common ancestor

at the same time, in some presumed explosive radiation

at one point in their history, or all the elements of inter-

est evolved independently since they separated from

common ancestors (Felsenstein I985, Harvey and Pagel

I99I). There are few cases in which we can be sure that

no degree of similarity is brought about by descent or

some coincidence in history. If there are no other clues

to phylogeny, language at least is always known. This

leaves few excuses for omitting phylogeny from cross-

cultural comparative studies.

Comments

JOHN R. BOWEN

Department of Anthropology, Washington University,

St. Louis, Mo. 63I30-4899, U.S.A. 20 V 94

Mace and Pagel argue that if one wishes to test univer-

salistic (here, "comparative") hypotheses then one

ought to adopt a particular form of controlled compara-

tive method. By mapping the adoption of new cultural

elements onto an evolutionary tree diagram, they argue,

one can limit comparison to elements that became

available for adoption by two or more populations after

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

558 | CURRENT ANTHROPOLOGY Volume 35, Number 5, December I994

those populations had diverged. One can then see if, over

large numbers of cases, similar conditions lead to simi-

lar adopt/not-adopt decisions (or the continuous equiva-

lents thereof ).

This argument is welcome in that it draws attention

to the practice of controlled comparison, which, though

indicated here by a lone footnote to Fred Eggan, is the

comparative method of choice for most sociocultural an-

thropologists. Indeed, the relative isolation of those who

do HRAF-style work could best be overcome by integrat-

ing regional studies into a more general comparative

framework. To attract a large percentage of the field,

such a general framework would best not be limited to

Murdock-style studies. It should, for example, include

the small-n studies of large-scale events-revolutions

being perhaps the type case-that predominate in histor-

ical sociology and comparative political science. Al-

though the article does not imply such a convergence, I

would offer as a friendly amendment a step-wise analy-

sis that proceeds from studies of parallel processes to

general claims about what kinds of changes seem to re-

cur in different and unrelated world regions. We would

be concerned no longer with + camels but with the

general ability and proclivity of humans to switch

modes of production. There is nothing particularly new

in this idea; indeed, it seems very close to the focus on

parallel processes of evolution that animated the work

of Julian Steward and his students. (One problem in this

article is that, coming as it does-from outside the field

of sociocultural anthropology, it knows not its own fore-

bears.)

A puzzle remains about what "cultural phylogeny"

means. Mace and Pagel use "culture" in two very differ-

ent ways. Sometimes "culture" means a congeries of

cultural elements; at other times it means a population

or a society. This double usage stems from our idea of

people as "culture carriers" on an analogy with genetic

material.

Mace and Pagel use linguistic reconstructions as the

basis for constructing "cultural phylogeny." Linguistics

acts as a proxy for genetic relationships. Yet linguistic

relationships among small-scale societies are more dif-

ferent in shape from genetic relationships than is appar-

ent from their article. I happen to work in highland Su-

matra, where many of today's inhabitants come from

South and West Asia. I am not sure what an aggregate

genetic "tree" would look like for any one Sumatran

population, but it would surely look very different from

a linguistic reconstruction. And, of course, a linguistic

reconstruction looks different from a reconstruction of

religious (in this case, Islamic) or agricultural ideas.

Even were we to settle on some general idea of "de-

scent" as the basis for the tree, the link between the

biological and the cultural lines of transmission is not

specified here. Presumably it is through a process of

teaching ideas and norms to one's children. But why

should this mode of transmission be the "tree" (the

"true phylogeny") against which all other modes of

transmission are merely noise, or "horizontal trans-

mission" ?

BIMAN KUMAR DAS GUPTA

34, Malakarpara Rd., Flat 3A, Calcutta 38, India.

29 IV 94

The major difficulty with the phylogenetic approach

highlighted here is that the identification of "mother"

and "daughter" cultures and the explanation of the "co-

evolution of culture" remain matters of conjecture. An-

thropological models must be based on empiricism, and

it is the sharing of culture traits, the rationales for shar-

ing, and the nature of the clustering of culture traits that

can be examined empirically. Moreover, determining

the association of traits such as polyandry and high alti-

tude in the case of India and female circumcision and

pastoralism in the case of East Africa through phylogen-

esis will yield regional rather than universal patterns.

Phylogenetic studies will not help to reveal any global

pattern of co-evolution and descent of cultural traits,

and therefore their application remains limited.

Mace and Pagel's contention that constructing phy-

logenies of culture is an important aspect of the testing

of cross-cultural comparative hypotheses is useful, but

one must be cautious in treating cultural and genetic

elements in virtually the same way; biological and ge-

netic elements are more stable and permanent than cul-

tural markers, some of which are more dependent upon

the influence of geography. The material aspects of cul-

ture do change rapidly because of exposure to neighbour-

ing cultures, regional geographical conditions, and so on,

but the aspects of social structure and cultural values

which form the core elements of culture are less amena-

ble to change and tend to retain their distinctiveness.

For example, the Gallong and Minyong of north-east In-

dia share many material traits because of geographical

proximity, but in polyandry, phratry organisation, nam-

ing patterns, and dormitory system they retain individ-

ual identities. This may make cross-cultural comparison

somewhat difficult.

Cross-cultural studies help us to understand the shar-

ing of cultural traits, beliefs, and practices by groups

living in a particular territory. This sharing has to be

explained by a number of factors, viz., a common politi-

cal-historical tradition and the sharing of cultural ele-

ments from a regional cultural pool. For example, when

a particular erstwhile feudal estate of West Bengal and

Bihar was anthropologically studied from I957 to 196I

it was observed that all 64 groups in the area shared

material aspects of life (food habits, dress styles, house

types, household implements), economic organisation,

regional fairs and festivals, and local ritual order. Apart

from the factor of proximity, this was the result of the

occupational compartmentalisation of artisan groups,

the monopoly of ritual specialists, adherence to tradi-

tional regional festivals and shrines, and the use of a

local dialect that people considered their own. Similar

regional cultures have been identified in other parts of

India through cross-cultural studies. Cross-cultural

studies may also have potential for revealing the nature

of the flow of cultural traits in acculturative situations.

With regard to the origin of culture traits, phylogeny

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method 1 559

may prove useful, though its multi-disciplinary applica-

tion is complicated. For studies of clustering and change

in culture traits, however, in addition to phylogeny the

migration of groups and the transmission of traits from

one region to another must be given due consideration.

Traits such as the market can be viewed phylogeneti-

cally. The origin of daily markets, especially in India,

may be seen in terms of an evolution from annual fairs

where economic transactions were made to regular gath-

erings for the sale of livestock and craft products organi-

sed under the patronage of kings to markets held weekly

in the territorial jurisdiction of one feudal lord or an-

other to daily markets in urban centres. When we try to

understand the distribution of types of consanguineous

marriage or surname or village exogamy, however, we

may have to look to various geo-cultural factors. In the

Indian context, the caste hierarchy can be understood

only in terms of its functional utility and the concepts

of purity and pollution. Similarly, the specific distribu-

tion of net operators and trap operators along the coast

of the Chilka Lake in eastern India has to be seen in

terms of its operational and functional utility rather

than in terms of a phylogenetic model.

Phylogeny may be a useful tool for analysing varia-

tions in a trait among subgroups. Thus in north-eastern

India, both the Khasi and the Garo of Meghalaya are

matrilineal, but the Garo matrilineal structure centres

round the roles of the nokna and the nokrom whereas

that of the Khasi does not. Again, the type of bilocal

residence evident among the Pnar Khasi of Meghalaya

is unique among Khasi groups. This has to be explained

not simply through phylogeny but also through in-depth

study of migration, the impact of local geography, and

the proximity of other groups from which a specific cul-

tural trait or bundle of traits has been borrowed. Phylog-

eny may provide some essential clues for explaining the

persistence of traits such as polyandry only in the moun-

tains of north-western, north-eastern, and southern In-

dia or, say, the concentration of generalised exchange

in the alliance system of certain tribal groups in the

north-east whereas in the south and west of central India

but not elsewhere FSD, MBD, and SD marriage is

prescriptive or preferential. Though keeping property

within the kin group is offered as a reason for the genesis

of this trait, there are other reasons too; for example,

sister's daughter marriage can persist only when the

family of birth and the family of marriage are not differ-

entiated. Phylogeny can help us to reconstruct a culture

or its descent but cannot explain a historical process

empirically.

KEITH F. OTTERBEIN

Department of Anthropology, State University of New

York at Buffalo, Buffalo, N.Y. 1426I, U.S.A. 5 v 94

Mace and Pagel have argued that their phylogenetic ap-

proach is a replacement for the worldwide cross-cultural

study, an approach which is limited by the nonindepen-

dence of cases (i.e., Galton's problem). Critics of cross-

cultural studies have often seized upon Galton's prob-

lem as a reason for rejecting both the approach and the

findings. Thus some cross-cultural researchers have

gone to elaborate efforts to develop solutions to the prob-

lem (e.g., Naroll developed five separate solutions

[Otterbein I987:I37]). Other researchers, for example,

Ember and Otterbein (I99I:222-25), recognize, as do

Mace and Pagel, that neighboring societies are not com-

plete duplicates of each other and hence can be retained

in cross-cultural samples. If one is concerned about du-

plicate cases in a sample, we suggest, "a method we like

is to drop a 'duplicate' case after the sample has been

chosen randomly (e.g., when two cultures have been

chosen which share the same language, one culture is

chosen randomly to be dropped)" (p. 223). Another

method is to drop a case only if the points on the vari-

ables used in a hypothesis test are identical-of course,

this would only be done if the societies shared a lan-

guage, were neighbors, or gave evidence of direct culture

contact. Thus I certainly would not cease using world-

wide samples in comparative research because of the

alleged difficulties that arise from the nonindependence

of cases.

In addition to employing random sampling or proba-

bility sampling, a cross-cultural study should derive

hypotheses from theories where possible (i.e., adopting

a deductive rather than an inductive approach) and code

the data for one's own variables in preference to using

precoded data (Otterbein I990). A variation on random

sampling which tends to "spread out" the sample soci-

eties and thus presumably reduce the number of dupli-

cate cases is to number the societies in the sampling

universe, determine the sample size desired, divide this

number into the total number of societies in the sam-

pling universe, call this number n, and select every nth

society in the sampling universe (Otterbein I989). Vari-

ous sampling universes are described in Ember and Ot-

terbein (I99I:228-3i) and Otterbein (I989).

The phylogenetic approach resembles the regional

analysis advocated by Radcliffe-Brown and his students.

Except for one brief reference to Eggan (I954), a student

of Radcliffe-Brown, Mace and Pagel fail to anchor their

approach in that respected tradition. Eggan, in the article

they cite, called it the "method of controlled compari-

son," but his article provided no instruction in this

method. Mace and Pagel do provide us with the informa-

tion we need to apply the phylogenetic approach, but

their method is not controlled comparison. I will point

out two of its limitations.

First, only those regions for which there is a great

amount of information-historical (including archaeo-

logical reports), linguistic, and ethnographic-can be ex-

amined. It was the recognition of the need for this infor-

mation if comparative studies were to be undertaken

that led Radcliffe-Brown to urge detailed ethnographic

studies. Earlier Boas had urged his students to collect

information in the field so that the regional approach he

advocated, called the historical method, could eventu-

ally be employed. Mace and Pagel do not tell us how

many regions are sufficiently well-described to permit

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

560 I CURRENT ANTHROPOLOGY Volume 35, Number 5, December 1994

the use of the phylogenetic approach. I doubt that there

are many. Even nearly two centuries of research on the

Iroquoian peoples of the Eastern Woodlands of North

America have probably not yielded sufficient informa-

tion to allow the successful use of the phylogenetic

method.

Second, the single example given (namely, that camel

herding will be adopted by peoples living in the arid

areas of a region) suggests that the hypotheses which

can be tested are narrow in scope-that is, restricted to

a particular region. It might also be argued that the ex-

ample given is trivial, but I will not so argue. I would

like to see Mace and Pagel demonstrate their approach

with a hypothesis or hypotheses applicable to any re-

gion. Since the purpose of a worldwide cross-cultural

study is just that, they must show that the phylogenetic

approach can test hypotheses of universal applicability

if it is to replace the worldwide cross-cultural study.

The example given is not universally applicable-camel

herding has not been adopted in the dry-climate regions

of the New World. It can perhaps be generalized if Mace

and Pagel are willing to consider herd animals other

than camels, such as sheep and goats, as one variable in

the hypothesis. Moreover, not all herd animals are

adapted to dry climates.

I believe that there is a place for both methods in

comparative research. In my own teaching and research

I advocate a research progression wherein case studies

are followed by small-scale comparative studies (and I

would include both the traditional controlled compari-

son and the phylogenetic approach in my methodologi-

cal tool kit) or longitudinal studies and then by world-

wide cross-cultural studies. Finally, cases that deviate

from predictions can be scrutinized in a new round of

case studies (see Otterbein 199I:xxiv).

MARK RIDLEY

Departments of Anthropology and Biology, Emory

University, Atlanta, Ga. 30322, U.S.A. 2 v 94

It is one of the established orthodoxies of the compara-

tive method that the characters under study must not

have been built into the phylogeny used to study them.

I have probably said this myself on some occasion, and

it is not my purpose to criticize Mace and Pagel for say-

ing it. However, in a recent study of the relation be-

tween the gene frequencies of two blood groups among

human populations, I met the problem in its pressing,

real form rather than its polite rhetorical one. The phy-

logenies I used were those of Cavalli-Sforza, Menozzi,

and Piazza (I994) and Nei and Roychoudhury (I993), and

they were constructeJ from classical nuclear genetic

characters. Cavalli-Sforza et al., for example, used 42

nuclear gene loci, including the two blood-group loci I

was interested in. Maybe I should have rerun the data

for the 40 other loci and recalculated the phylogeny, but

that is the kind of advice that is easier to give than to

take. There were 40 loci, about 25 populations, and

some of the loci (such as Rhesus) have many alleles: I

reckoned that I would have had to rekey 5,ooo or so

numbers. I also reckoned that the whole exercise would

not have made the least difference to the result: what I

needed was an argument. I duly reconsidered that or-

thodoxy.

Circles can be virtuous as well as vicious. If the phy-

logeny was constructed using the characters under in-

vestigation, far from building the premiss into the con-

clusion it must make it more difficult to find a

significant result. In an extreme case, such as mammary

lactation, which is used to define a major group (mam-

mals), it becomes impossible to find any comparative

relationships because the sample size is necessarily i.

When the characters under investigation are 2/42 char-

acters used to construct the phylogeny, the same conser-

vative process operates in weaker form. The "circular"

reasoning actually makes any significant conclusion

more robust. Maybe more relaxed significance levels

should be demanded in investigations of such virtue and

rigor that they find a significant result even when the

target variables are used to build the phylogeny! The

same procedure is also a way of answering critics who

say that comparative results depend on the phylogeny

that is chosen, for it shows that the result is robust even

under a rigorous phylogeny. I expect that, if my point is

right, it could also be expressed in terms of type-I and

type-2 errors.

The obvious snag with using phylogenetic methods

for intraspecific comparative research is that there is no

one phylogeny. Different species such as ring-tailed le-

mur, common chimpanzee, human, and rhesus macaque

have a unique branching relation between them; all (or

almost all) the DNA of a species has been forced through

the same history. But there is no unique branching rela-

tion between the various attributes (cultural, morpho-

logical genetic, etc.) of samples of individuals within a

modern species. One character may have experienced a

different "phylogeny" from another (Hennig i966); in-

deed, if enough characters are studied it is inevitable

that they will not share a phylogeny. The intraspecific

human phylogenies of genes and languages, for instance,

disagree or agree depending on the populations one looks

at (Cavalli-Sforza, Menozzi, and Piazza I994). This kind

of "incongruence" in phylogenies of species is usually

due to error, but in intraspecific phylogenies it can be

real.

The correct procedure, I suspect, is to aim to use the

most accurate phylogeny for the characters under study.

Then the effects of shared history will be removed ad

hoc for each character. For cultural characters like the

ones mainly discussed by Mace and Pagel, a linguistic

or cultural phylogeny may be best, whereas for the blood

groups I was studying the genetic phylogeny may be

more appropriate. It may even be sensible, when looking

at the relation between two characters, to use a different

phylogeny for each. But however that may be, my point

(and I am not saying that Mace and Pagel suggest other-

wise) is that intraspecific phylogenies are not the same

as specific phylogenies, and when faced with particular

problems we should return to the fundamental logic of

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method | 56I

comparative inference rather than simply carrying over

techniques of biological origin into anthropology.

THOMAS SCHWEIZER

Institut ffir V6lkerkunde, University of Cologne,

Albertus-Magnus-Platz, D-o0923 Kiln, Germany.

9 V 94

Mace and Pagel's paper addresses important aspects of

the comparative method in anthropology. It should force

the discipline to reconsider the established practice and

methodology of cross-cultural studies in a more dy-

namic framework. Comparative research in anthropol-

ogy has often gained in methodological rigor and sub-

stantive insight by borrowing ideas and procedures from

more advanced scientific fields (for example, network

autocorrelation methods [see Dow et al. I984, White et

al. 198I] and sampling theory [Dow i99i]). In a similar

spirit, Mace and Pagel introduce methods of compara-

tive historical reconstruction from evolutionary biology.

Their new method of disentangling independent events

of cultural change in anthropological comparisons

seems to me at least thought-provoking. Since it handles

qualitative as well as quantitative variables and contains

an additional statistical component, it promises to be

a valuable supplement to the tool kit of comparative

anthropology. As is to be expected with a novel method,

not all facets of it are yet well-understood, and its

strengths and weaknesses remain to be explored through

further applications, but I am fully convinced by the

arguments and the empirical example of camel-breeding,

circumcision, and climate among Kenyan pastoralists

that it ought to be tried. It would be most instructive

to have a more comprehensive example and an explicit

comparison between this method and network autocor-

relation analysis applied to the same problem and data.

So far it seems to me that the phylogenetic method

proposed by Mace and Pagel works best at the level of

comparisons within a region. It is a novel contribution

to a tradition extending from Kroeber's culture area

studies to the statistical comparisons of ethnic groups

within a region or continent developed by Driver and

others (Jorgensen I974). As is well-established (e.g.,

White, Burton, and Dow i98i:845), regions are more ho-

mogeneous in ecological, historical, and cultural vari-

ables than worldwide samples. Generally, in regional

comparisons the command of ethnographic sources, in-

cluding data on the historical context and trajectories of

the cases studied, is richer and better than in worldwide

cross-cultural studies. Thus regional comparativists are

in a better position to establish causal relations. Given

Mace and Pagel's new method of distinguishing indepen-

dent instances of cultural elements from duplicates due

to common descent or "horizontal transmission," as

they aptly call it, we should be in an even better position

to decompose the causal nexus apparent at the regional

level. This is clearly the case in the interesting regional

comparison of Kenyan ethnic groups. However, the

weak point of regional comparisons in general also ex-

tends to Mace and Pagel's analysis: Every region is lim-

ited in the range of variables and possibly also in the

relationship among variables that it exhibits (see, for in-

stance, White's I988 finding of different types of polyg-

yny in the New vs. the Old World). When testing univer-

sal hypotheses it is not sufficient to restrict the test to

particular regions (although it may be wise to do so in

the stage of development and exploration of hypotheses).

It is also necessary to vary the regional context in order

to test the temporal-spatial limits of hypotheses. On a

practical level I do not see how one could do a worldwide

cross-cultural study with the Mace and Pagel approach,

which requires considerable in-depth data on the cul-

tural phylogeny and the historical background of the

cultures studied. For the time being, then, I would still

use large cross-cultural samples (and network autocorre-

lation methods) to test universal hypotheses in addition

to attempting regional comparisons with Mace and Pa-

gel's method.

More specifically, I would like to know how Mace and

Pagel might tackle some of the standard problems of

cross-cultural comparisons. How would they cope with

the problem of shifting boundaries among the units

studied? (Schlee [I989] shows for some of the Kenyan

ethnic groups they compare that their members inter-

marry and even shift ethnic affiliations opportunis-

tically.) What about the problem of time-lag among

variables? In most cross-cultural comparisons the obser-

vational data are cross-sectional snapshots in time, and

therefore it could be the case that the independent vari-

ables have changed but the dependent variables will ad-

just only later, at a time not covered by the "ethno-

graphic present" of the field study. Finally, comparative

data are highly aggregated and notoriously error-prone;

how do factual errors in the data and flaws in the recon-

struction of the cultural phylogeny affect the results? Is

there empirical evidence or even simulation data from

biology that might provide us an estimate of the ro-

bustness of their method?

ECKART VOLAND

Institut fur Anthropologie, Universitat Gottingen,

Biirgerstrasse 50, D-37073 Gottingen, Germany. 4 v 94

Introducing the historical dimension into the methodol-

ogy of cross-cultural comparison as Mace and Pagel sug-

gest cannot but be beneficial in that it deepens our un-

derstanding of the origins of human cultural diversity

and, furthermore, sensitizes us to the snares known as

Galton's problem. However, their key conclusion, that

"cross-cultural comparative studies must be based upon

the identification of independent events of cultural

change," needs a bit more clarification at least with re-

gard to two aspects: (i) What actually are "events of

cultural change"? and (2) Is the mutual independence

of cultural changes really a conditio sine qua non for

meaningful cross-cultural analysis?

Human cultural behaviour is often the outcome of

conditional behavioural strategies. When this is the

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

562 CURRENT ANTHROPOLOGY Volume 35, Number 5, December 1994

case, environmental characteristics determine which

specific behaviours of the broad pool of behavioural op-

tions available to an individual actually become mani-

fest (Dunbar I993, Voland I994). Thus, under conditions

of environmental fluctuation, human cultural behaviour

is likely to fluctuate as well, but strategic flexibility of

this kind has absolutely nothing to do with "de novo

inventions" of new cultural elements. Cultural changes

then are simply tactical switches within the scope of a

broader behavioural strategy. They are reflections of a

wide range of behavioural possibilities and capabilities

all of which are known and available to the population

in principle but which have different probabilities of be-

ing adopted in different socio-ecological circumstances.

Camel-herding by East Africans may be a good example

of an environmentally induced switch in cultural pat-

terns, since the amount of rainfall decides whether

camel-herding or some other subsistence strategy yields

the most profit. Thus, ethnic and historical variation in

camel-keeping may be regarded as a reflection of purely

environmental variation with little or no cultural

change on the innovation level involved. In giving up

camel-herding on economic grounds, the option of

camel-herding is not lost but simply not chosen for good

economic reasons. In spite of an adaptive behavioural

change, the population's "meme pool" remains un-

changed. With respect to conditional behaviour it is dif-

ficult to speak of events of cultural change with the de-

gree of strictness that is required by Mace and Pagel's

model of "cultural phylogenies."

If, however, instead of cultural behaviour, "memes"

become the focus of interest in the camel-keeping exam-

ple, then the East African people possibly experienced

an event of cultural change only once, namely, when

approximately 4,000 years ago camels arrived in the re-

gion and techniques of herding them were added to the

existing meme pool. In any case, this example nicely

demonstrates what an enormous difference it makes

whether one is concerned with behaviour or cognitive

units, be they called "memes," "ideas," "concepts,"

"culturgens," or something else, in constructing cul-

tural phylogenies and performing comparative studies in

anthropology.

Human behaviour is based on psychological machin-

ery that has evolved in the course of hominisation (Bar-

kow, Cosmides, and Tooby i992). On these grounds

alone we may question whether independence of behav-

ioural changes is ever achieved. For example, differing

paternal investment tendencies are not independent of

one another in that they are based on what seems to

be an evolved "universal idea of paternity" implicitly

holding that high paternal confidence predicts high pa-

temal investment and low patemal confidence low pa-

ternal investment (Alexander I979, Borgerhoff Mulder

i992). This is part of our vertebrate heritage (Clutton-

Brock i99i) and is necessarily integrated into all the

differing cultures of fatherhood (Flinn I98I, Hartung

i985). In Mace and Pagel's terms, differing paternal in-

vestment would be difficult to study using the compara-

tive method, since cross-cultural variation in it is ho-

mologous variation and their aim hinges on the

independent variation of analogous traits.

Natural selection continuously checks the adaptive-

ness of cultural elements and regulates their degree of

occurrence. If cultural traits are maintained for a long

enough time, this may be biologically functional. When

studying the functions of cultural behaviour, whether

the traits in question are actually homologous (i.e., de-

pendent) or analogous (i.e., independent) in terms of

their cultural origin seems to me a secondary problem.

In the long term, their probability of occurrence is deter-

mined solely by their adaptive function, not by the

mechanisms involved in their adoption. As in "real biol-

ogy," the analysis of both homologous and analogous

traits can contribute to an understanding of functional

significance.

To sum up: Mace and Pagel's contribution is valuable

for its plea for methodological rigour that is not always

achieved in comparative anthropology. However, that

comparative analyses "are generally only justified if the

elements under study have evolved or been adopted in-

dependently in every culture" (my emphasis) seems to

me an overstatement in that it unnecessarily limits the

number of phenomena whose functional significance

can be studied using the comparative method.

RUTH MACE AND MARK PAGEL

London, England. 30 VI 94

We would like to take the opportunity, in this reply,

simply to reiterate some of the main points of our argu-

ment. If a cross-cultural comparison is to be used to test

a hypothesis about the correlated evolution of elements

of culture, then some assumptions have to be made

about the phylogenetic relationships among those cul-

tures. We use "phylogeny" here in its broadest sense

to mean the "origin of groups" and thus include in its

meaning both horizontal and vertical relationships

among cultures. To demonstrate statistically that two

or more elements of culture (or elements of culture and

environmental factors) have co-evolved, one must show

that those elements have arisen, changed, or been modi-

fied in many places on the phylogeny. This is what we

have termed the analysis of independent instances of

cultural change. No other approach to establishing

cross-cultural associations is, in principle, valid.

Our approach is not restricted to investigating in-

stances of de novo invention (as Voland suggests). We

attach great significance to the horizontal transmission

of cultural traits (not, as Bowen suggests, treating it as

"merely noise"), because if two elements are repeatedly

adopted together from other cultures rather than inher-

ited together from a mother culture, then this can be

used as evidence for their functional association. Often

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

MACE AND PAGEL The Comparative Method 1 563

the cultural elements under investigation are function-

ally unrelated but nonetheless appear together many

times. This occurs when each of many daughter cultures

inherits the elements from the mother culture. The

maintenance of the elements in a large number of cul-

tures is not evidence for their functional association,

even if the elements are functional in their own right.

Methods that do not use a phylogenetic approach do not

control for this possibility. Cross-cultural comparison

may not be appropriate for testing functional hypotheses

if there is great variation in an element within cultures

and not between or if there is virtually no variation at

all across cultures. Optimality or other analyses that de-

scribe the conditions that maintain variation within cul-

tures may be useful for the former case; careful func-

tional analysis of cultural elements may be helpful in

the latter.

The phylogenetic approach that we have described is

general. It does not require the use of a worldwide or a

regional sample; any sample of cultures can be used,

chosen according to whether the hypothesis being tested

is worldwide or regional in its significance. Cultures do

not have to be discarded arbitrarily (as in methods sug-

gested by Otterbein). Phylogenies can be constructed on

the basis of genetic, linguistic, archaeological, or histori-

cal information but should not be constructed from the

elements to be tested for association. Linguistic phylog-

enies have the desirable feature of indicating what lin-

guists refer to as "genetic" relationships and may often

be preferable to genetic phylogenies as maps of the pat-

tem of inheritance of most of the elements of a culture.

We illustrated this with the Mukogodo case, in which a

unit with a different genetic history-the Mukogodo-

loses its culture but becomes part of another culture-

that of the Maa-speakers-and adopts the latter's lan-

guage.

Our phylogenetic approach assumes that cultures are

real, that they persist through time, and that they occa-

sionally give rise to daughter cultures. In spite of this,

some elements of culture will have been acquired hori-

zontally from other cultures. This means that some in-

dividual elements of culture may have phylogenies

(maps of their origins and alterations) that differ from

the cultural phylogeny. In the extreme, one might hold

the view that cultures represent only ephemeral collec-

tions of cultural elements that have come together (hori-

zontally) from a variety of sources and will at some fu-

ture date come together in other collections to define a

new set of cultures. Were this true, the phylogenetic

approach that we describe would properly simplify to a

correlation across cultures because each culture would

represent an instance either of the acquisition or of the

rejection of the cultural elements under study. We

doubt, however, that this scenario is much more than a

whimsical view of the evolution of cultures. The recog-

nition that cultures are real and that they often give rise

to broadly similar daughter cultures is fundamental to

the field of cross-cultural comparison. Given this, the

identification of instances of cultural change is essential

to comparative studies in anthropology.

References Cited

ALEXANDER, RICHARD D. Darwinism and human affairs. Seat-

tle: University of Washington Press. [EV]

BARKOW, JEROME H., LEDA COSMIDES, AND JOHN TOOBY.

Editors. I 992. The adapted mind: Evolutionary psychology

and the generation of culture. New York: Oxford University

Press. [EV]

BATEMAN, R., I. GODDARD, R. O'GRADY, V. A. FUNK, R.

MOOI, W. J. KRESS, AND P. CANNELL. I990. Speaking of

forked tongues: The feasibility of reconciling human phylog-

eny and the history of language. CURRENT ANTHROPOLOGY 3I:

I-24.

BENTLEY, G., T. GOLDBERG, AND G. JASIENSKA. I993. The

fertility of agricultural and non-agricultural traditional societ-

ies. Population Studies 47:269-8i.

BOAS, F. I940 (i896). "The limitations of the comparative

method of anthropology," in Race, language, and culture. New

York: Macmillan.

BORGERHOFF MULDER, MONIQUE. i992. "Reproductive deci-

sions," in Evolutionary ecology and human behavior. Edited

by E. A. Smith and B. Winterhalder, pp. 339-74. Hawthorne,

N.Y.: Aldine de Gruyter. [EV]

BOYD, R., AND P. RICHERSON. I985. Culture and the evolu-

tionary process. Chicago: University of Chicago Press.

CAMPBELL, K., AND J. WOOD. I988. "Fertility in traditional so-

cieties," in Natural human fertility: Social and biological

mechanisms. Edited by P. Diggory et al., pp. 39-69. London.

CAVALLI-SFORZA, L. L., AND M. FELDMAN. I98I. Cultural

transmission and evolution: A quantitative approach.

Princeton: Princeton University Press.

CAVALLI-SFORZA, L. L., P. MENOZZI, AND A. PIAZZA. I994.

The history and geography of human genes. Princeton:

Princeton University Press. [MR]

CAVALLI-SFORZA, L. L., A. PIAZZA, P. MENOZZI, AND J.

MOUNTAIN. I988. Reconstruction of human evolution: Bring-

ing together genetic, archaeological, and linguistic data. Pro-

ceedings of the National Academy of Sciences (U.S.A.) 85:

6002-6.

CLUTTON-BROCK, T. H. I99I. The evolution of parental care.

Princeton: Princeton University Press. [EV]

CRONK, L. I989. From hunters to herders: Subsistence change as

a reproductive strategy among the Mukogodo. CURRENT AN-

THROPOLOGY 30:224-34.

DAWKINS, R. I976. The selfish gene. Oxford: Oxford University

Press.

DOW, M. 1991. Statistical inference in comparative research:

New directions. Behaviour Science Research 25S:235-58.

DOW, M., M. BURTON, D. WHITE, AND K. REITZ. I984. Gal-

ton's problem as a network autocorrelation. American Ethnolo-

gist II:754-70.

DUNBAR, ROBIN I. M. I993. "Behavioural adaptation," in Hu-

man adaptation. Edited by G. A. Harrison, pp. 73-98. Oxford:

Oxford University Press. [EV]

DURHAM, W. I990. Advances in evolutionary culture theory. An-

nual Review of Anthropology I9:I87-2I0.

.I99I. Co-evolution: Genes, culture, and human diver-

sity. Stanford: Stanford University Press.

EGGAN, F. I975 (I9S4). "Social anthropology and the method of

controlled comparison," in Essays in social anthropology and

ethnology. Chicago: University of Chicago Press.

EMBER, MELVIN, AND KEITH F. OTTERBEIN. I99I. Sampling

in cross-cultural research. Behavior Science Research 25:

2I7-33. [KFO]

FELSENSTEIN, J. I979. Alternative methods of phylogenetic in-

ference and the interrelationships. Systematic Zoology 28:

49-62.

.I983. Parsimony in systematics: Biological and statistical

issues. Annual Review of Ecology and Systematics I4:3I3-33.

.I985. Phylogenies and the comparative method. Ameri-

can Naturalist i.2:i-iS.

F LI N N, MAR K V. 198i. "Uterine versus agnatic kinship variabil-

ity and associated cousin marriage preferences: An evolution-

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

564 CURRENT ANTHROPOLOGY Volume 35, Number 5, December 1994

ary biological analysis," in Natural selection and social

behavior: Recent research and new theory. Edited by R. D.

Alexander and D. W. Tinkle, pp. 439-75. New York: Chiron

Press. [EV]

GITTLEMAN, J. L., AND H-K. LUH. i992. On comparing com-

parative methods. Annual Review of Ecological Systematics

23:383-404.

GOWLETT, J. I988. "Human adaptation and long-term climatic

change in Northeast Africa: An archaeological perspective." in

Ecology of survival: Case studies from Northeast African his-

tory. Edited by D. Johnson and D. Anderson, pp. 27-46. Boul-

der: Westview Press.

HARTUNG, J. I982. Polygyny and the inheritance of wealth. CUR-

RENT ANTHROPOLOGY 23:I-I2.

. I985. Matrilineal inheritance: New theory and analysis.

Behavioral and Brain Sciences 8:66i-88.

HARVEY, P., AND M. PAGEL. I99I. The comparative method

in evolutionary biology. Oxford: Oxford University Press.

HENNIG, w. I966. Phylogenetic systematics. Urbana: University

of Illinois Press. [MR]

HOLY, L. Editor. I987. Comparative anthropology. Oxford: Black-

well.

JORGENSEN, J. G. Editor. I974. Comparative studies by Harold

E. Driver and essays in his honor. New Haven: HRAF Press.

[TS]

LESTHAEGHE, R. j. Editor. I989. Reproduction and social orga-

nization in sub-Saharan Africa. Berkeley: University of Califor-

nia Press.

LIGHTFOOT-KLEIN, H. I989. Prisoners of ritual: An odyssey

into female genital circumcision in Africa. New York: Harring-

ton Park Press.

LOW, B. I988. "Pathogen stress and polygyny in humans," in

Human reproductive behaviour: A Darwinian perspective. Ed-

ited by L. Betzig et al. Cambridge: Cambridge University Press.

. I990. "Human responses to environmental extremes and

uncertainty: A cross-cultural perspective," in Risk and uncer-

tainty in tribal and peasant economies. Edited by E. Cashdan.

Boulder: Westview Press.

LUMSDEN, C. J., AND E. 0. WILSON. I98I. Genes, mind, and

culture: The co-evolutionary process. Cambridge: Harvard Uni-

versity Press.

MACE, R. iggo. Pastoralist herd compositions in unpredictable

environments: A comparison of model predictions and data

from camel-keeping groups. Agricultural Systems 33: I - I I .

MACE, R., AND A. I. HOUSTON. I989. Pastoralist strategies for

survival in an unpredictable environment: A model of herd

composition which maximises household viability. Agricul-

tural Systems 3 I: I 8 5 -204.

MADDISON, W. P. I990. A method for testing the correlated evo-

lution of two binary characters: Are gains or losses concen-

trated on certain branches of a phylogenetic tree? Evolution

44:539-57.

MADDISON, W. P., M. J. DONOGHUE, AND D. R. MADDISON.

I984. Outgroup analysis and parsimony. Systematic Zoology

33:83-I03.

MADDISON, W. P. AND D. R. MADDISON. I989. Interactive

analysis of phylogeny and character evolution using the com-

puter program MacClade. Folia Primatologica 53: I 90-202.

. I992. MacClade: Analysis of phylogeny and character evo-

lution. Version 3.0. Sunderland, Mass.: Sinauer.

MARTINS, E. P., AND T. H. GARLAND. I99I. Phylogenetic anal-

yses of the correlated evolution of continuous characters: A

simulation study. Evolution 45:534-57.

MURDOCK, G. P., AND D. R. WHITE. I969. Standard cross-

cultural sample. Ethnology 9:329-69.

NAROLL, R. I970. "Galton's problem," in A handbook of

method in cultural anthropology. Edited by R. Naroll and R.

Cohen, pp. 974-89. New York: Natural History Press.

. I97I. The double language boundary in cross-cultural sur-

veys. Behavior Science Research 6:95-I02.

NEI, M., AND A. K. ROYCHOUDHURY. I993. Evolutionary rela-

tionships of human populations on a global scale. Molecular Bi-

ology and Evolution IO:927-43. [MR]

OTTERBEIN, KEITH F. I987. Raoul Naroll (i920-i985). Ameri-

can Anthropologist 89:I36-432. [KFO]

I 989. Sampling and samples-an update. Cultural An-

thropology Methods Newsletter I (2):4-5. [KFO]

. I990. Two styles of cross-cultural research. Cultural An-

thropology Methods Newsletter 2(j2):6-7. [KFO]

. Editor. I993. Feuding and warfare: Selected works of

Keith F. Otterbein. New York: Gordon and Breach. [KFO]

PAGEL, M. i992. A method for the analysis of comparative data.

Journal of Theoretical Biology I56:43I-42.

. I994. Detecting correlated evolution on phylogenies: A

new method for the analysis of discrete categorical data. Pro-

ceedings of the Royal Society B 255. In press.

PAGEL, M., AND P. HARVEY. I988. Recent developments in the

analysis of comparative data. Quarterly Review of Biology 63:

4I3-40.

PURVIS, A., J. L. GITTLEMAN, AND H-K. LUH. I994. Truth or

consequences: Effects of phylogenetic accuracy on two compar-

ative methods. Journal of Theoretical Biology. In press.

RENFREW, C. I987. Archaeology and linguistics. Cambridge:

Cambridge University Press.

RIDLEY, M. I983. The explanation of organic diversity: The

comparative method and adaptations for mating. Oxford: Ox-

ford University Press.

1 986. Evolution and classification: The reformation of

cladism. London: Longman.

. i992. Darwin sound on comparative method. Trends in

Ecology and Evolution 7: I 7.

RUHLEN, M. I987. A guide to the world's languages. Vol. I.

Classification. Stanford: Stanford University Press.

RUVOLO, M. I987. "Reconstructing genetic and linguistic trees:

Phenetic and cladistic approaches," in Biological metaphor

and cladistic classification. Edited by H. Hoenigswald and L.

Wiener. Philadelphia: University of Pennsylvania Press.

SAUCIER, J. I972. Correlates of long postpartum taboo: A cross-

cultural study. CURRENT ANTHROPOLOGY I3:238-49.

S C H L E E, G. I 989. Identities on the move: Clanship and pastoral-

ism in northern Kenya. Manchester: Manchester University

Press. [TS]

SPENCER, P. I973. Nomads in alliance: Symbiosis and growth

among the Rendille and Samburu of Kenya. Oxford: Oxford

University Press.

THORNHILL, N. W. I99I. An evolutionary analysis of rules regu-

lating human inbreeding and marriage. Behavioral and Brain

Sciences I4:247-93.

VOLAND, ECKART. I994. "Reproductive decisions-viewed

from an evolutionarily informed historical demography," in Hu-

man reproductive decisions. Edited by R. I. M. Dunbar. Lon-

don: Macmillan. In press. [EV]

WHITE, D. R. I988. Rethinking polygyny: Co-wives, codes, and

cultural systems. CURRENT ANTHROPOLOGY 29:529-7I. [TS]

WHITE, D. R., M. L. BURTON, AND MALCOLM M. DOW. I98I.

Sexual division of labor in African agriculture: A network auto-

correlation analysis. American Anthropologist 83:824-49. [TS]

WHITING, j. W. M. I964. "Effects of climate on certain cultural

practices," in Explorations in cultural anthropology. Edited by

W. H. Goodenough. New York: McGraw-Hill.

WORTHY, G. G. I959. A historical and anthropological back-

ground to the Dorobo of Mukogodo. Kenya National Archives

file no. DC/NKI/3/I2.

This content downloaded from 144.82.114.114 on Tue, 13 Feb 2018 06:55:55 UTC

All use subject to http://about.jstor.org/terms

The biogeography and evolution of land ownership

Article

Full-text available

Mar 2023

Aim Land ownership norms are well documented and play a central role in social–ecological systems. Yet only recently has the spatial and temporal distribution of land ownership been examined using biogeographical and evolutionary approaches. We incorporate biogeographical and evolutionary modelling to test associations between land ownership and environmental, subsistence and cultural contact predictors. Location Africa. Taxon Bantu and Bantoid ethnolinguistic groups (73 societies). Methods Based on ethnographies for 73 societies, we coded land ownership norms as none, group, kin or individual. We paired these data with language phylogenies, and measured phylogenetic and geographical signal and modelled alternative evolutionary trajectories using maximum likelihood methods. We tested the influence of environmental, subsistence and cultural predictors on spatial variation in land ownership, using a multi‐model inference approach based on logistic regression. Results Bantu land ownership norms likely evolved on a unilinear trajectory (i.e. societies progress or regress along a series of ownership types), but not one requiring consistent increase in exclusivity (i.e. restrictions towards ownership by smaller groups) as suggested by prior theory. Our biogeographical analyses suggest land ownership is more likely where neighbours also own land and resource productivity is predictable. Reliance on agriculture has relatively small effect sizes and low importance in the model. Main Conclusions We find support for multiple evolutionary pathways. Lack of resolution may be due to localized horizontal transfer of norms consistent with the influence of neighbours we find from biogeographical analyses. We cannot rule out other untested mechanisms. Although long‐standing theories propose links between subsistence practices and land ownership, our results suggest subsistence plays only a modest role. Our results also support resource defensibility theory (i.e. land ownership is more likely where environmental productivity is predictable). Overall, we demonstrate the value of combining analytical approaches from evolution and biogeography to test hypotheses on the spatial and temporal variation of human cultural traits.

Filogenia, innovación y recambio tecnológico en las puntas de proyectil del Holoceno medio de las Tierras Altas Surandinas

Article

Full-text available

Oct 2021

En este trabajo abordamos el estudio del tempo y modo de la evolución de las puntas de proyectil lanceoladas del Holoceno medio de las Tierras Altas Surandinas, analizando la continuidad formal de algunos diseños y la superposición temporal de clases y atributos. Mediante la construcción de cronofilogramas mostramos diferencias en la tasa de innovación de las puntas líticas y sus caracteres a lo largo del tiempo, el aumento de la longevidad de las clases, y una estructura de evolución en mosaico, con ciertos rasgos con dinámicas direccionales y otros con patrones de movimiento aleatorio Browniano. Concluimos que desde comienzos del Holoceno medio, una población de cazadores recolectores más grande junto con la especialización en la caza de camélidos, propiciaron la evolución selectiva de una mayor diversidad de clases de puntas líticas, que con el tiempo dio lugar a la evolución de clases de mayor longevidad por competencia. Además, la coexistencia de distintos sistemas de armas y estrategias de caza con diferente performance permitió tanto una mayor variabilidad, como la evolución no direccional de atributos en la filogenia de estos artefactos.

The cultural evolution of witchcraft beliefs

Article

Sep 2024
EVOL HUM BEHAV

Identifying culture as cause: Challenges and opportunities

Article

Full-text available

Jan 2024

Causal inference lies at the core of many scientific endeavours. Yet, answering causal questions is challenging, especially when studying culture as a causal force. Against this backdrop, this paper reviews research designs and statistical tools that can be used—together with strong theory and knowledge about the context of study—to identify the causal impact of culture on outcomes of interest. We especially discuss how overlooked strategies in cultural evolutionary studies can allow one to approximate an ideal experiment wherein culture is randomly assigned to individuals or entire groups (instrumental variables, regression discontinuity design, and epidemiological approach). In doing so, we also review the potential outcome framework as a tool to engage in causal reasoning in the cultural evolutionary field.

Nutritional condition and nutrient intake predict moral condemnation of food wasting

Article

Full-text available

May 2024
FOOD QUAL PREFER

Ecology, culture and leadership: Theoretical integration and review

Article

Oct 2023
LEADERSHIP QUART

Universal interpretations of vocal music

Article

Full-text available

Sep 2023

Despite the variability of music across cultures, some types of human songs share acoustic characteristics. For example, dance songs tend to be loud and rhythmic, and lullabies tend to be quiet and melodious. Human perceptual sensitivity to the behavioral contexts of songs, based on these musical features, suggests that basic properties of music are mutually intelligible, independent of linguistic or cultural content. Whether these effects reflect universal interpretations of vocal music, however, is unclear because prior studies focus almost exclusively on English-speaking participants, a group that is not representative of humans. Here, we report shared intuitions concerning the behavioral contexts of unfamiliar songs produced in unfamiliar languages, in participants living in Internet-connected industrialized societies (n = 5,516 native speakers of 28 languages) or smaller-scale societies with limited access to global media (n = 116 native speakers of three non-English languages). Participants listened to songs randomly selected from a representative sample of human vocal music, originally used in four behavioral contexts, and rated the degree to which they believed the song was used for each context. Listeners in both industrialized and smaller-scale societies inferred the contexts of dance songs, lullabies, and healing songs, but not love songs. Within and across cohorts, inferences were mutually consistent. Further, increased linguistic or geographical proximity between listeners and singers only minimally increased the accuracy of the inferences. These results demonstrate that the behavioral contexts of three common forms of music are mutually intelligible cross-culturally and imply that musical diversity, shaped by cultural evolution, is nonetheless grounded in some universal perceptual phenomena.

Phylogenetic insight into the origin of tones

Article

Full-text available

Jul 2023

The origin of tone, also known as tonogenesis, has long been a topic of great interest in language evolution and human cognition studies. Several linguistic studies of tonal languages have proposed various hypotheses that tonal origin may be related to different changes of phonological structures. However, such hypotheses have not been quantitatively tested in an evolutionary framework. Here, we conducted phylogenetic comparative analyses to assess the likelihood of different hypotheses of tonogenetic mechanisms across 106 Sino-Tibetan languages, of which approximately 70% are tonal. Our results showed that the presence of tones has a strong phylogenetic pattern and that Proto-Sino-Tibetan languages were most likely non-tonal. Our findings identified that tonal origin was strongly associated with the evolution of specific phonological structures, such as the loss of syllable-final consonants and voice quality on vowels. Furthermore, we found that tonal origin probably did not influence the diversification rates of Sino-Tibetan languages. These findings enabled us to better understand that tone arose as a compensatory mechanism for the structural organization and evolution of languages.

A quantitative global test of the complexity trade-off hypothesis: the case of nominal and verbal grammatical marking

Article

Full-text available

Dec 2022

Nouns and verbs are known to differ in the types of grammatical information they encode. What is less well known is the relationship between verbal and nominal coding within and across languages. The equi-complexity hypothesis holds that all languages are equally complex overall, which entails trade-offs between coding in different domains. From a diachronic point of view, this hypothesis implies that the loss and gain of coding in different domains can be expected to balance each other out. In this study, we test to what extent such inverse coevolution can be observed in a sample of 244 languages, using data from a comprehensive cross-linguistic database (Grambank) and applying computational phylogenetic modelling to control for genealogical relatedness. We find evidence for coevolutionary relationships between specific features within nominal and verbal domains on a global scale, but not for overall degrees of grammatical coding between languages. Instead, these amounts of nominal and verbal coding are positively correlated in Sino-Tibetan languages and inversely correlated in Indo-European languages. Our findings indicate that accretion and loss of grammatical information in nominal words and verbs are lineage-specific.

Challenges of sampling and how phylogenetic comparative methods help: With a case study of the Pama-Nyungan laminal contrast

Article

Full-text available

Feb 2022

Phylogenetic comparative methods are new in our field and are shrouded, for most linguists, in at least a little mystery. Yet the path that led to their discovery in comparative biology is so similar to the methodological history of balanced sampling, that it is only an accident of history that they were not discovered by a linguistic typologist. Here we clarify the essential logic behind phylogenetic comparative methods and their fundamental relatedness to a deep intellectual tradition focussed on sampling. Then we introduce concepts, methods and tools which will enable typologists to use these methods in everyday typological research. The key commonality of phylogenetic comparative methods and balanced sampling is that they attempt to deal with statistical non-independence due to genealogy. Whereas sampling can never achieve independence and requires most comparative data to be discarded, phylogenetic comparative methods achieve independence while retaining and using all comparative data. We discuss the essential notions of phylogenetic signal; uncertainty about trees; typological averages and proportions that are sensitive to genealogy; comparison across language families; and the effects of areality. Extensive supplementary materials illustrate computational tools for practical analysis and we illustrate the methods discussed with a typological case study of the laminal contrast in Pama-Nyungan.

PHYLOGENETIC ANALYSES OF THE CORRELATED EVOLUTION OF CONTINUOUS CHARACTERS: A SIMULATION STUDY

Article

Full-text available

May 1991
EVOLUTION

We use computer simulation to compare the statistical properties of several methods that have been proposed for estimating the evolutionary correlation between two continuous traits, and define alternative evolutionary correlations that may be of interest. We focus on Felsenstein's (1985) method and some variations of it and on several "minimum evolution" methods (of which the procedure of Huey and Bennett [1987] is a special case), as compared with a nonphylogenetic correlation. The last, a simple correlation of trait values across the tips of a phylogeny, virtually always yields inflated Type I error rates, relatively low power, and relatively poor estimates of evolutionary correlations. We therefore cannot recommend its use. In contrast, Felsenstein's (1985) method yields acceptable significance tests, high power, and good estimates of what we term the input correlation and the standardized realized evolutionary correlation, given complete phylogenetic information and knowledge of the rate and mode of character change (e.g., gradual and proportional to time ["Brownian motion"] or punctuational, with change only at speciation events). Inaccurate branch length information may affect any method adversely, but only rarely does it cause Felsenstein's (1985) method to perform worse than do the others tested. Other proposed methods generally yield inflated Type I error rates and have lower power. However, certain minimum evolution methods (although not the specific procedure used by Huey and Bennett [1987]) often provide more accurate estimates of what we term the unstandardized realized evolutionary correlation, and their use is recommended when estimation of this correlation is desired. We also demonstrate how correct Type I error rates can be obtained for any method by reference to an empirical null distribution derived from computer simulations, and provide practical suggestions on choosing an analytical method, based both on the evolutionary correlation of interest and on the availability of branch lengths and knowledge of the model of evolutionary change appropriate for the characters being analyzed. Computer programs that implement the various methods and that will simulate (correlated) character evolution along a known phylogeny are available from the authors on request. These programs can be used to test the effectiveness of any new methods that might be proposed, and to check the generality of our conclusions with regard to other phylogenies.

The Evolution of Parental Care

Book

Dec 1991

Tim Clutton-Brock

Nomads in Alliance: Symbiosis and Growth Among the Rendille and Samburu of Kenya.

Article

Sep 1974
Man

Feuding and Warfare: Selected Works of Keith F. Otterbein.

Article

Sep 1995

Reproductive decisions

Article

Jan 1992

M. Borgerhoff Mulder

Standard Cross-Cultural Sample

Article

Jan 1980

Reproductive Decisions Viewed from an Evolutionary Informed Historical Demography

Chapter

Jan 1995

Eckart Voland

The double determination of human behaviour and reproduction by biologically evolved reproductive interests on the one hand and socio-ecologically constrained opportunities for living and breeding on the other has become the focus of both the sister disciplines of sociobiology and behavioural ecology in recent years. Human sociobiological research aims at uncovering the genetic aspects of human behavioural patterns by investigating the ultimate causes, mechanisms and adaptive consequences of kin selection, social reciprocity, differential parental investment, sexual selection and other basic elements of social evolution (Trivers, 1985). By focusing on genetical concepts, sociobiological approaches achieve a high degree of generalisability when explaining human social performance (Winterhalder and Smith, 1992); however, only rarely do they achieve a satisfactory specificity when the concern is the explanation of the contextual variability of human social behaviour. For example, the evolved preferences in male and female mate choice and reproductive strategies are well-known and easily explained as an outcome of sexual selection; but the causes of their phenotypic plasticity remain obscure, Under what circumstances do male and female interests converge in monogamous mating systems as opposed to polygamous ones? Questions of this kind fuel behavioural ecology.

Phylogenetic Systematics

Article

Mar 1968
COPEIA

Darwinism and Human Affairs.

Article

May 1981

Alternative Methods of Phylogenetic Inference and Their Interrelationship

Article

Mar 1979
Syst Zool

Joseph Felsenstein

A probabilistic model of evolution in a character is presented. It involves two character states, 0 and 1. The population may have a third state, 01, in which there is polymorphism for both character states. There are three evolutionary events in the model: origination of state 1, reversion from state 1, and loss of polymorphism, plus an event corresponding to total misinterpretation of the character by the taxonomist. The maximum likelihood method of estimating the phylogeny is described. When the probabilities of the four events are taken to be extreme, then depending on their relative sizes under different circumstances four different phylogenetic inference methods emerge as maximum likelihood methods. Three are known: the Camin-Sokal parsimony method, Farris's Dollo parsimony method, and the Estabrook-Johnson-McMorris compatibility method. A new method, the polymorphism parsimony method, also emerges. It explains parallelism and convergence by persistence of character-state polymorphism after a unique origin of the derived character state, and attempts to find that evolutionary tree which requires the least extent of polymorphism. Details of implementation of the polymorphism parsimony method are given. Some variants of the evolutionary model are discussed, involving unrooted character state trees. The use of the model to resolve a paradox which arises when we attempt to apply the Dollo parsimony method to multiple-state characters is briefly considered.

The Comparative Method in Anthropology [and Comments and Reply]

Figures

Recommended publications

The Collapse into the Inorganic

The Comparative Method in Anthropology

Soziobiologie, Anthropologie und menschliche Kultur

An Adaptive Model of Human Reproductive Rate Where Wealth Is Inherited: An Anthropological Perspecti...

Commentaries and Author's Reply on “EvolutionaryPsychology: A New Perspective in the Behavioral Scie...