British Journal of Psycbolqgy (1993), 84, 145-1 69 Printed in Great Britain 145

0 1993 The British Psychological Society

The mathematical modelling of human

culture and its implications for psychology
and the human sciences

Kevin N. Laland*
The Galton Laboratory, Department of Genetics and Biometry, University College London

Recent years have seen the growth of a new and exciting field of theoretical research
concerned with the mathematical modelling of human culture, and of its interaction
with genetics. Drawing on analogies between genetic and cultural processes,
mathematically sophisticated biologists have used population genetics models as the
basis for the development of analogous models of cultural transmission, cultural
evolution and gene-culture co-evolution. These models are designed to describe and
analyse the diffusion of cultural traits through populations, under the influence of
various cultural and evolutionary forces. They have already been applied to address
many problems of interest to psychologists. Here I present an introduction to these
models, explaining the mathematics in simple terms, and giving examples of the
work of leading theorists. I go on to discuss the findings of most relevance to
psychology, critically analysing the most important conclusions. Finally, I suggest
some areas of psychology in which these models might usefully be applied. I argue
that these models constitute a major theoretical innovation, and that there is
considerable potential for their application in psychology.

In 1981 Lumsden & Wilson, two mathematically minded biologists, published a

book which ‘proved’ mathematically that ‘tabula rasa is an unlikely state’ of mind.
Four centuries of heated debate in psychological and philosophical circles could all
have been prevented if only John Locke had known some maths! O r could it? Over
the last 20 years several biologists have employed mathematical modelling as a tool
to explore human beings, our nature, our thinking, our culture and our societies.
Having spotted analogies between genetic and cultural processes, these biologists
have used population genetics models as the basis for analogous models of cultural
transmission, cultural evolution and gene-culture co-evolution. The problems they
address are of fundamental interest to psychology and the human sciences. Do our
genes restrict and delineate the nature of our culture? Is there a genetic basis to cross-
cultural or gender differences in beliefs and values? Are our societies destined to
maintain a particular structure? How did culture evolve, and how has it affected our
subsequent evolution? And, How does our culture change? For us in the human
sciences this is our territory and the problems are ones with which we are familiar.
This paper is written in the belief that the mathematical modelling of human culture
*Requests for reprints should be sent to Kevin Laland at the Department of Integrative Biology, University of
California, Berkeley, CA 94720, USA.
6 P S Y 84
146 Kevin N. Laland
is an exciting and expanding area of research which has a valid and important
contribution to make to psychology and the human sciences. Yet for many these
mathematical models are difficult and inaccessible, and would involve a large
investment of time to get to grips with. Here, I present a ‘guided tour’ to these
models, describing them in simple terms, explaining the aims and assumptions of the
modellers, and critically analysing the nature of their conclusions. I begin by giving
examples of the types of models employed by three leading practitioners in the field.
I then go on to discuss the findings of most relevance to psychology.. Finally, I
discuss how these models might usefully be applied to many areas in psychology.

The models
Introduction
Over the last 15 years in particular, mathematical modelling of cultural evolution,
and of the interaction between genetic and cultural transmission has received
considerable attention (Aoki, 1991; Aoki & Feldman, 1987, 1989, 1991; Boyd
& Richerson, 1976, 1982, 1983, 1985, 1987, 1988a, b, 1989a, b ; Cavalli-Sforza &
Feldman, 1973a, b, 1981, 1983a, b ; Chen, Cavalli-Sforza & Feldman, 1982; Eshel &
Cavalli-Sforza, 1982; Fagan, 1981; Feldman & Cavalli-Sforza, 1975, 1976, 1977,
1979,1984,1989; Feldman, Cavalli-Sforza & Peck, 1985; Findlay & Lumsden, 1988;
Lumsden, 1988, 1989, 1991; Lumsden & Wilson, 1981, 1985; Pulliam, 1982, 1983;
Pulliam & Dunford, 1980; Richerson & Boyd, 1978, 1989; see Durham, 1990 for a
review). Many biologists have argued that culture is itself an evolutionary process
(Boyd & Richerson, 1985; Cavalli-Sforza & Feldman, 1981 ; Dawkins, 1976;
Durham, 1982, 1990; Hull, 1982; Plotkin & Odling-Smee 1981), cultural variants
being generated, selected and socially transmitted. Mathematically sophisticated
evolutionary biologists have taken advantage of this analogy to apply ‘population
genetics models to describe and analyse culture. As an introduction to the nature of
this modelling, a selective review of the principal contributions of three sets of
theorists is given.
The first team is that of Lumsden & Wilson, who developed a model of gene-
culture co-evolution in their text Genes, Mind and Culture (1981). The second is
Cavalli-Sforza & Feldman, who have published a string of papers and an influential
book Cultural Transmission and Evolution (1981), applying evolutionary models to
culture. Finally, a review of Boyd & Richerson’s (1985) Culture and the Evoiutionary
Process is given, in which they develop a ‘dual-inheritance model’ of gene-culture
interaction.

Lumsden & Wilson’s ‘Gene-Culture Co-evolution’

Lumsden & Wilson (1981) set out to investigate ‘gene-culture co-evolution ’, which
they describe as:
...a complicated, fascinating interaction in which culture is generated and shaped by biological
imperatives while biological traits are simultaneously altered by genetic evolution in response to
cultural innovation (p. 1).

Like most other modellers, Lumsden & Wilson (henceforth LW) adopt an atomistic
 Mathematical modelling of human culture 147
description of culture for mathematical convenience, defining it as a pool of
‘culturgens’, units of culture. For LW a culturgen is ‘a relatively homogeneous set
of artifacts, behaviors, or mentifacts’ (p. 27). ‘Culturgens’ can be considered
synonymous to Dawkins’ (1976) ‘memes’. LW argue that an individual’s choice of
culturgen is directed by genetically determined ‘epigenetic rules ’ :

I
The rules comprise the restraints that the genes place on development ..., and they affect the
probability of using one culturgen as opposed to another’ (p. 7 ) .

LW focus on gene-culture translation, where for each genotype more than one
culturgen is possible and at least two culturgens differ in the likelihood of adoption
because of innate epigenetic rules. They argue that the epigenetic rules will tend to
channel development towards the adoption of certain culturgens as opposed to
others. They describe this as the ‘leash principle ’ : ‘ ...natural selection operates in
such a way as to keep culture on a leash’ (p. 13). Many reviewers have questioned
the appropriateness of this metaphor (Boyd & Richerson, 1985; Kitcher, 1985).
LW develop two main models, called ‘gene-culture translation ’ and the ‘co-
evolutionary circuit’. One of their principal goals is to predict the shape of the
‘ethnographic curve ’, which is the probability distribution that plots the frequency of
alternative culturgen use across many societies. This allows them to predict the
spread of the culturgens in societies in general, rather than in any given society.
In gene-culture translation, individuals choose between two culturgens, c l and c2,
periodically reassessing their choice, influenced by their epigenetic rules and the
pattern of culturgen usage in the population. The rates at which individuals adopt
alternative culturgens is determined by the product of a ‘ raw ’ (innate) bias function
and an ‘update assimilation’ (copy others) bias function. LW place most weight on
an ‘exponential trend watcher’ model, where as more people adopt one culturgen the
probability of the others switching to it increases exponentially. This means that the
ethnographic curve will have a bimodal, polarized distribution, as populations will
tend to be mainly c l users, or c2 users (see Fig. 1a). If individuals have an innate bias
for one of the culturgens, then the bimodal distribution is likely to be asymmetrical
(Fig. 1b). LW claim that ‘ ... a barely detectable amount of selectivity in an epigenetic
rule operating during the behavior of individuals can strongly alter social patterns ’
(p. 144). Or, as Kitcher (1985) puts it, ‘if people are strongly inclined to follow what
others around them do, small initial differences in switching propensity can be
dramatically magnified at the level of the group’ (p. 357).
This conclusion, which LW call the ‘amplification law’, only follows if there is
very strong cultural transmission (i.e. exponential, positive frequency-dependence).
Clearly other types of social transmission are possible. For instance, Cavalli-Sforza &
Feldman (1981), Maynard-Smith & Warren (1982) and Boyd & Richerson (1985)
have all developed models in which cultural transmission by itself (unbiased by
genetic predispositions) does not alter the frequency or distribution of cultural
variants.
This model is then extended to the full co-evolutionary circuit by allocating
fitnesses to genotypes, and calculating changes in allele frequency (see Fig. 2). A
genotypes’ fitness is a function of the rewards (food, resources) it gathers, offset by
6-2
148 Kevin N.Laland

nl/N

nl/N

Figure 1. Lumsden & Wilson’s ethnographic curves. (a) An exponential trend watcher model of social
transmission polarizes the distribution. (b) Small innate predispositions strongly bias cultural patterns.

Pattern of
cdturgenuse \
\
\
Rewards of
I
culturgen use

Probability of
switching culturgen
for each genotype

Number of gametes
produced by
each genotype

Epigenetic rules

Figure 2. Lumsden & Wilson’s co-evolutionary circuit.

 Mathematical modelling of human culture 149
the costs of building and running the appropriate brain structures. A fertility
function translates this into gametes - the more resources gathered, the greater the
fertility. Finally, LW generate a recursive expression for allele frequencies, in terms
of the fertility functions of each genotype. (For a more detailed account of their
model see Pulliam [1983] and Kitcher [1985].)
Following simulation, LW reach five major conclusions: (1) pure tabula rasa is an
unlikely state; (2) sensitivity to usage patterns increases the rate of genetic
assimilation; (3) culture slows down the rate of genetic evolution; (4) changes in
gene frequency during the co-evolutionary process can nevertheless be rapid ;and (5)
gene-culture evolution can promote genetic diversity. These conclusions are
discussed in the next section.
LW’s findings have received severe criticism (Kitcher, 1985; Lewontin, 1983;
Maynard-Smith & Warren, 1982). However, published as it was in the midst of the
sociobiology debate, LW’s theory has not always been judged completely objectively.
Lumsden, a physicist recruited by Wilson for his mathematical skills, used
mathematical tools familiar to physicists, but not biologists and social scientists. For
almost all readers, this theory was completely opaque, and as a consequence their
assessment of it was probably heavily influenced by the hostile views of two
mathematically competent sets of reviewers (Kitcher, 1985; Maynard-Smith &
Warren, 1982). Many social scientists were suspicious, and there was an uneasy
feeling amongst many of them that the mathematics was more sophisticated than it
need be (Kitcher, 1985). Kitcher even goes so far as to accuse them of abusing
mathematics - ‘Complex mathematics is employed to cover up very simple.. .ideas ’
(p. 393) - principally because, in some cases, he could develop much simpler models
which would reach the same conclusions. However, there is a difference between
those mathematical models which attempt to describe, or at least to capture, the
important dynamics of the processes being modelled, and those which are simply
employed as tools to explore the processes. While the former are likely to be more
complex, their component parameters are more likely to be meaningful, and they can
form a theoretical framework for subsequent research in a way that the latter cannot.
Kitcher’s criticism is unfair because it underestimates Lumsden & Wilson’s aims :
their models reflect their wish to create a theoretical framework for the analysis of
much of the social sciences. As Pulliam (1983) puts it: ‘Lumsden and Wilson make
an important contribution to social theory by specifying the numerous steps involved
in going from genes to culture and back to genes again’ (p. 428). Genes, Mind and
Cuhure was a pioneering piece of work, and its controversial findings have provided
an impetus for much of the more recent modelling (e.g. Boyd & Richerson, 1985).
Unfortunately, for most social scientists the theory is probably too intimidating to be
useful.

Cavalli-SforTa & Feldman’s models of human culture

The field of the mathematical modelling of cultural evolution was begun by Cavalli-
Sforza & Feldman (CF) in 1973, with their introduction of a dynamic model of
cultural transmission into the nature-nurture debate (Cavalli-Sforza & Feldman,
1973b). Their approach is less ambitious than LW, and their models more
150 Kevin N. Laland
conventional. These authors argue that the forces of mutation, selection, migration
and genetic drift in biological evolution have analogues that describe cultural change,
and consequently develop analogues of genetic models for culture. They are well
qualified to do so : Cavalli-Sforza is an internationally renowned authority on human
genetics and Feldman is one of the world’s leading theoretical population geneticists.
CF’s Cultt/ral Transmission and Evolrrtion was published at about the same time as LW’s
Genes, Mind and Ctrlttlre. Unlike LW, however, in their text CF focus solely on the
cultural level: no assumptions are made about the genetic basis of traits or trait
(cultural) fitness, and no attempt is made to look at any feedback effects that cultural
evolution may have on gene frequencies. Their objectives are to determine the
equilibrium frequency states, and time-dependent behaviour of the traits under
cultural forces. Ultimately, they have two goals : Firstly, they endeavour to develop
a theoretical framework for the analysis of cultural change, a framework that can be
used for explaining phenomena as diverse as linguistics, epidemics, social values and
customs, and the diffusion of innovations (p. v). Secondly, they wish to address
theoretical questions concerned with the nature of cultural change, for example, ‘Can
non-adaptive cultural traits evolve? ’.
CF distinguish between natural and cultural selection. Some traits have a direct
affect on survival (e.g. smoking), thus natural selection may change their frequency.
Other traits, such as the spread of drinking Coco-Cola, or playing frisbee, do not
greatly affect Darwinian fitness. In both cases, however, ‘cultural selection ’ may
occur, memes being generated, replicated, selected and socially transmitted. CF also
distinguish three types of transmission: (1) vertical - parent to offspring; (2) oblique
- parental generation to offsprings (excluding parents) ; and (3) horizontal - within-
generation transmission. They develop models of the spread of cultural variants
under all three types of transmission, for both discrete and continuous traits. By way
of illustration, a description of CF’s vertical transmission model (1981, pp. 78-107) is
given (see Table 1).

Table 1. Cavalli-Sforza & Feldman’s (1981) vertical transmission model (Table

2.2.1)

Mating type Probability of Frequency of mating

Mother Father H child h child General Random

H H b3 1-b3 p 3 (t)
H h 62 1-62 p 2 (t) ntvt
h H 61 1-bl p l (t) ntvt
h h 60 1-60 PO (t) V;

Table 1 may look intimidating to those unfamiliar with population genetics, but
is is actually quite simple. H and h represent two states that a trait can take (say
drinking and not drinking alcohol), and bO, b l , b2, b3 the probabilities that an H child
(drinker) results from the parental matings h x h (neither parent drinks), h x H
 Mathematical modelling of human culture 151
(mother drinks), H x h (father drinks), H x H respectively (both parents drink). If the
parents mate at random (unaffected by whether they drink), the frequency of H
(proportion of individuals that drink) in the offspring generation (t+1 ) is given by
the product of the probability of any given mating and the probability that their
offspring will be H (drinkers), summed for all possible matings. Thus
ut+1 = C,,(t) 4
= u:B+tltC+bO, (2.2.3 p. 79)
where B = b3+60-b1-62
and c = b2+61-260.
If individuals with trait H have a different probability of survival to those with h (say,
if drinkers are more likely to suffer from heart disease), then natural selection will
alter the frequency of the trait. T o investigate the impact of selection CF allocate
fitness 1 + s to H and 1 to h, where s represents the difference in the probability of
survival of individuals with each trait. Thus if s were -0.2 then only four drinkers
would survive for every five non-drinkers. The relative proportion of H individuals
after selection (u:) is then

(2.6.1 p. 102)

If selection is followed by random mating between adults, and vertical transmission

according to the rules in Table 1, then the frequency of juveniles in the next
+
generation (tit 1 ) is
ut+1 = 63(u~)2+(61+b2)tr,*(l-ut/:)+60(1-u~)2. (2.6.2 p. 103)
This, together with the expression for change in frequency due to selection, allows
the frequency of the trait in any generation to be expressed in terms of its frequency
in the previous generation (a one-step recursion). CF analyse the properties of this
model and describe the conditions under which transmission laws produce fixation
of the trait (when everybody drinks alcohol) or loss of H (when nobody drinks
alcohol), or a memetic polymorphism (only some people drink alcohol).
Many diverse traits are culturally transmitted from parent to offspring, including
political and religious beliefs, attitudes, values, fears, dietary tastes and recreational
habits (documented in Boyd & Richerson, 1985, pp. 50-51, and Cavalli-Sforza &
Feldman, 1981, p. 75 fl). CF demonstrate that, using empirical data on trait usage in
consecutive generations, it is possible to estimate the transmission coefficients (bO, b l ,
etc.) for five vertically transmitted cultural traits (p. 83): salt usage, frequency of
praying to God, frequency of swimming, belief in ability versus luck, and political
interest. This illustrates that, provided such data are available, it is possible that the
dynamics of vertically transmitted traits can be described in a formal way, as well as
being predicted. Clearly, to be able formally to model and predict the frequency of
cultural traits at any time could be of great utility to psychologists. CF have
developed similar models for oblique and horizontally transmitted discrete traits, for
continuous (quantitative) traits, and for investigation of the effects of migration,
population structure, innovation and other factors. In most cases, the models are
152 Kevin N . Laland
relatively straightforward, and could be applied directly to the analysis of cultural
change.
In most of CF’s studies in which evolution is studied at both the genotypic and
phenotypic levels (Cavalli-Sforza & Feldman, 1983 ;Feldman & Cavalli-Sforza, 1976,
1984, 1986) the (Darwinian) fitness differences have been primarily between
phenotypes (e.g. between drinkers and non-drinkers) with the genotypic effects on
the transmission efficiency. Although there are no direct fitness effects of the
genotype, the resulting genotypic selection is ‘induced’ by that on the phenotypes
(e.g. if there was a gene making individuals less likely to drink, it might spread).
More recently (Feldman & Cavalli-Sforza, 1989) CF have developed a class of models
in which the selection on a dichotomous phenotype depends on the genotype,
fitnesses being allocated separately to each pheno-genotype (a genotype with a
particular cultural trait). (For example, instead of allocating fitnesses to drinkers or
non-drinkers, they would be allocated to drinkers with the gene for abstinence,
drinkers without the gene, non-drinkers with the gene, and non-drinkers without the
gene.) More sophisticated diploid models would take account of the fact that humans
carry two copies of the gene. CF are concerned with modelling the spread of lactose
absorption genes, suggesting that both Darwinian selection based on nutritional
properties of milk and the cultural transmission of milk use have been influenced by
the lactose-absorbing genotype. They construct a series of models in which vertical
and oblique cultural transmission are followed by selection on the pheno-genotypes.
This allows them to model the spread of initially rare lactose absorption genes and
milk use. These models are an important step towards understanding the dynamics
of the interaction between genes and culture, because they are truly dual-level
models. They are consequently a valuable extension of CF’s approach. Aoki &
Feldman have gone on to use these models to explore the evolution of communication
(Aoki & Feldman, 1987) and the co-evolution of sign language and hereditary
deafness (Aoki & Feldman, 1991).

B y d & Richerson’s ‘dHal-inheritance models ’

Boyd & Richerson’s (BR) book CnltHre and the Evolthonar_y Process is the most recent
major text concerned with cultural evolution. This work has been highly acclaimed,
and BR w-ere awarded the J. I. Staley Prize in recognition of the book’s contribution
to anthropology. Although BR build on the work of CF, they explicitly link genetic
and cultural evolution into a unified ‘dual-inheritance model’. In the process, they
address many of the issues raised by the sociobiology debate. They investigate how
cultural change is affected by individual and social learning, and how natural
selection can act directly and indirectly on culturally transmitted information. BR are
interested in two types of problems: (1) what patterns of behaviour will become
common if a particular social transmission process operates? and (2) when should
natural selection favour the evolution of various cultural transmission processes ?
Their goals are to develop a unified theoretical framework (their dual-inheritance
models) that will allow them to analyse how biology and culture interact. BR also go
to some lengths to sample the psychological and social science literatures in a serious
attempt to assess whether their models fit the data. By culture BR mean:
 Mathematical modelling of human culture 153
... the transmission from one generation to the next, via teaching and imitation, of knowledge,
values, and other factors that influence behavior (p. 2).
BR restrict their definition of cultural transmission to ‘imitation and teaching’,
hypothesizing that only this kind of social learning is likely to have ‘dynamic
properties analogous to genetic inheritance’ (p. 35). Unlike the other (more
primitive) types of social learning found amongst animals as well as humans,
imitation gives individuals cheaply acquired information. They argue that because
this information is cheap to store and replicate, it is easily transmitted : ‘ ... culture is
cheaply acquired information, encoded in memory, that is capable of producing
major phenotypic effects’ (p. 35). Other, more primitive forms of social learning are
not considered to be cultural transmission because the nai’ve individual must acquire
her/his own information by a process of reinforcement that is almost as costly as
ordinary individual learning. The validity of this assumption and BR’s conclusions
about the evolution of culture are considered in the next section.
BR are principally concerned with three ‘cultural forces’ (or agents of cultural
change) which interact with natural selection. The first is ‘guided variation’ -
individually learned behaviours are culturally transmitted (in an unbiased manner)
resulting in a force that increases the frequency of those behaviours. The second is
‘biased transmission’, where social transmission is biased in favour of some variants
as opposed to others. This bias can be ‘direct’ (i.e. genetic), ‘frequency-dependent ’
(if the tendency to adopt the trait depends on its popularity), or ‘indirect’ (if the
behaviour is associated with other attractive variants). The third force is ‘the natural
selection of cultural variants ’, in which individuals are selected according to their
variant (as opposed to ‘cultural selection of cultural variants’, in which traits are
selected according to their ‘cultural fitness’ [Cavalli-Sforza & Feldman, 19811).
Adopting a similar approach to CF, BR begin by developing a linear model of
cultural transmission which they incorporate into subsequent analyses (see Appendix).
Individuals acquire cultural variants c or d after interacting with N cultural models
(e.g. parents, teachers, peers). They conclude (p. 80) that, if the formation of sets of
models is random, cultural transmission does not by itself change the frequency of
variants. This is equivalent to the Hardy-Weinberg rule in population genetics,
where genetic transmission alone does not change allele frequencies. It is important
to note, however, that there is no a priori reason to expect cultural transmission to
be linear. For instance, LW’s ‘exponential trend-watcher’ model is an example of a
non-linear, frequency-dependent model that drastically changes variant frequencies.
Some cultural traits, for example, voting patterns or pop record sales, are highly
frequency-dependent, and their dynamics may be better described by non-linear
transmission models.
If traits are culturally transmitted in an unbiased manner, but modified by
individual learning, then the resultant is a force that changes the frequency of traits.
This is the logic behind BR’s model of guided variation which combines the unbiased
transmission model with a simple learning rule

Y=
vex+L (r(w+ s) = a X + ( 1 -a) (T(H)+.s), (4.9 p. 95)
L+ v,
where L measures the extent to which an individual relies on individual learning,
154 Kevin N. Laland
T(H) is the goal of the learning rule in habitat H , and e is a normally distributed
random variable that represents the effect of errors made during the learning process,
with mean of zero and variance V,.This is formally equivalent to the simple linear
learning models of stochastic learning theory (Atkinson, Bower & Crothers, 1965).
For instance, Bush & Mosteller’s (1955) linear-operator model can be represented by
the equation
+
P,,, = aP, (1 -a) v,,
where a is the learning parameter, and v, is the reinforcement value. For example,
+
the probability of a rat pressing a lever on trial n 1 is a linear function of the
probability of it pressing the lever on trial n and the reinforcement v, experienced
following trial n. The reinforcement values might be vn = 0 if the rat is shocked, and
vn = 1 if it receives a food reward. The larger the value of a the smaller the impact
of the last trial on the rat’s behaviour, and the more it is guided by its earlier
experiences. In BR’s model, learning trials are effectively separated by a generation,
with unbiased transmission of the trait value (or probability) occurring between
trials. The parameter a = V,/(L+V,)(equivalent to a in the Bush & Mosteller
model) gives the relative importance of cultural transmission (equivalent to earlier
trials) and individual learning (which is equivalent to the latest trial) in determining
the mature phenotype. Thus for BR ‘social learning and individual learning are
alternative ways of acquiring a particular behavioral variant’ (p. 97), alternative in the
sense that individuals can acquire a phenotypic value ( y)directly through cultural
inheritance or through individual learning modifying an alternative culturally
inherited state. Nalve individuals first socially learn their parents’ behaviour, and
then modify it through learning in an adaptive way. Each generation learning moves
the mature phenotype (Y)a fraction [L/(L+V,] towards the equilibrium state or
habitat’s adaptive peak, T(H). The learning model is combined with a transmission
model for a quantitative character, to give recursive expressions for the mean and
variance of the trait after learning and transmission. If the environment does not
change, then the equilibrium value of the trait is H.
BR’s models of direct bias are analogues of CF’s cultural transmission models.
Linear transmission is modified by introducing a bias parameter ( B ) which increases
the probability of offspring adopting one of the cultural variants. If two cultural
models have weights (relative importance) a, and a2 then the frequency of the
favoured variant in the next generation is

p’ =p + P (1 -p) (1-B y a ,
4Ba,a2 ).
- a2)2
(Table 5.6 p. 140)

Thus the magnitude of the force can be seen to depend on the strength of bias ( B ) ,
and the variance in trait frequency p ( l -p).
Like CF’s models, BR’s models can potentially be exploited to analyse aspects of
cultural change of interest to psychologists. This is pursued in Section 4.

Findings of interest to psychologists

Apart from developing mathematical tools to describe and analyse cultural
transmission, this body of theoretical work has addressed a number of questions
 Mathematical modelling of human culture 155
related to the relationship between genes and culture. Much of this analysis is of
direct relevance to psychology and the human sciences. Some of the more pertinent
findings are discussed below.

Does biology constrain culture?

Lumsden & Wilson (1981) reached the conclusion that culture unbiased by genetic
constraints is unlikely, whilst Boyd & Richerson (1985) found that only a ‘general
purpose’ bias can evolve, and then only in a heterogeneous environment. Cavalli-
Sforza & Feldman (1981) concluded that non-adaptive traits can spread in a constant
environment. Under closer scrutiny these apparently conflicting conclusions unveil
a consistent picture which demarcates the conditions under which biases can evolve.
LW’s finding demonstrates that, in a constant environment in which a particular
phenotype is advantageous, a genetically determined learning rule that makes
individuals more likely to adopt the fitter cultural trait is superior in fitness to
choosing traits at random. Using a similar, but much simplified model, Pulliam
(1983) reached the same conclusion, but found that in a temporally variable
environment genes promoting more variable (i.e. culture-dependent) behavioural
responses will replace genes promoting more fixed responses. There are two
important considerations related to LW’s finding. Firstly, their conclusion is
restricted to a constant environment. Secondly, genetic biases affect the probability
of adopting alternative cultural traits, so an unbiased (tabula rasa) genotype adopts
each trait with equal probability, and consequently has an average fitness. Any
genotype with a bias which increases the probability of adoption of the advantageous
cultural trait will have a fitness advantage over the unbiased genotype. LW claim that
by this mechanism genetic differences between cultures in culturgen preference will
be selected after roughly 50 generations, their ‘ 1000-year rule’. This rule only follows
from their assumptions of very strong selection and weak culture (Kitcher, 1985;
Maynard-Smith & Warren, 1982). Obviously, different parameter values will give
different results. BR reach what appear to be conflicting conclusions to LW
principally because their assumptions are different. For BR there is a fitness cost to
the bias because ‘bias may entail costly experiments or additional neurophysiological
machinery’ (p. 149). LW clearly do not consider such a cost significant. Moreover,
BR’s unbiased genotype does not adopt each cultural trait at random, but in
proportion to the frequency of the trait in the parental generation, for they assume
that cultural parents and nai’ve offspring associate randomly with regard to the
offspring’s genotype. This significantly decreases the fitness disadvantage of the
unbiased genotype. Whilst for LW there is always variation in cultural traits, for BR
this is eventually eliminated by selection. When this has occurred the cost of bias
leads to selection again the biased allele. Even when the cost of bias is zero, if there
is no cultural variation there can be no fitness differences between genotypes, and the
unbiased allele is not lost. BR conclude that genetically biased transmission is unlikely
to evolve in homogeneous environments, and while in a heterogeneous environment
a general purpose bias may evolve, a habitat-specific bias is unlikely (p. 171).
(However BR do find that a frequency-dependent bias will be favoured in a spatially
heterogeneous environment [p. 2201.) Thus, while in a constant environment
156 Kevin N. Laland
genetically biased transmission has a fitness advantage over LW’s conception of
unbiased cultural transmission this is not the case with BR’s conception. Two
important empirical questions are : (1) Which conception of unbiased transmission is
the more valid? and (2) Is there a fitness cost to bias? An important theoretical
question is to investigate the effects of disequilibrium between genotypes and cultural
traits by relaxing BR’s assumption that cultural parents and na‘ive offspring associate
randomly with regard to the offspring’s genotype. CF have begun this analysis
(Feldman & Cavalli-Sforza, 1984, 1986).
CF’s conclusion (p. 107) that, depending on the value of the transmission
parameters, even a trait that reduces genetic fitness can spread through a population
in a constant environment, is intuitively easy to understand. If a trait has high enough
cultural fitness, then it can spread rapidly even if its carriers are more likely to die.
BR (p. 150) reach the same conclusion.

How did cultare evolve?

BR (1985, 1988a 1989a) and Rogers (1988) investigated the relative efficiency of
social and individual learning in tracking environmental variability, asking under
what ecological conditions natural selection should favour the evolution of social
learning. These models focus on ‘guided variation’, a process in which individuals
acquire behavioural traits culturally from their parents and then modify them on the
basis of their personal experience. In BR’s (1988~)model individuals effectively
have to guess which of the two habitats they are in on the basis of their individual
experience, and adopt the appropriate behaviour. When individuals are uncertain
about their habitat they rely on social learning and imitate others. Thus individuals
with quantitative character trait ( x ) greater than a threshold value d adopt behaviour
1, and those with a value less than - d adopt behaviour 2, the individuals being
confident that they are in habitat 1 or 2 respectively. If - d < x < d then the
individual imitates the behaviour of a single individual chosen at random from the
population. The value of d thus determines the relative importance of social and
individual learning. BR investigate the evolution of social learning by calculating the
stable equilibrium value of d. They concluded that a significant dependence upon
social learning is most adaptive when individual learning is inaccurate (i.e. results in
a lot of errors) and the chance that an individual’s social models experienced the same
environment the individual experiences is reasonably high. Thus they argue that
culture evolved in a constant environment. By the same logic, any reduction in the
accuracy of social learning, perhaps as a consequence of noisy transmission systems,
is likely to favour increased reliance on individual learning.
For BR, the evolution of cultural transmission occurred in a hominoid lineage. If,
as BR suggest, imitation is necessary for social transmission to have dynamic
population level properties then social transmission would not be expected to be
common amongst animals. Laboratory studies suggest that social learning by
imitation is rare amongst animals, relative to other social learning mechanisms
(Galef, 1988). However, recent laboratory and aviary studies have found that some
mammal and bird species are capable of social transmission (Curio, Ernst & Vieth,
1978; Giraldeau & Lefebvre, 1986, 1987; Laland & Plotkin, 1990, 1992; Lefebvre,
1986). Field and laboratory findings imply that animal social learning may function
 Mathematical modelling of human culture 157
to enhance foraging efficiency, allowing individuals rapidly to ‘home in ’ on
appropriate behaviours, with the transmitted information usually of only transient
value (Lefebvre & Palameta, 1988). While population and field studies have found
much evidence for rapid, horizontally transmitted behaviours amongst animals
(Fisher & Hinde, 1949; Itani & Nishimura, 1973; Lefebvre, 1986; Lefebvre &
Palameta, 1988; Takasaki, 1983), aside from birdsong, only a few traits appear to be
vertically transmitted (Itani & Nishimura, 1973; Kawai, 1965), and the reliability of
these is open to question (Galef, in press). Thus it would appear that most animal
social transmission functions principally as an adjunct to individual learning, and
may enhance foraging efficiency in rapidly changing, unpredictable environments.
These empirical findings suggest: (1) that animal social learning can have dynamic
population level properties; (2) that the assumption that culture evolved in a
hominoid lineage is open to doubt; (3) that the most primitive forms of culture were
probably horizontally transmitted ; and (4) that such horizontally transmitted traits
are advantageous in rapidly changing, spatially heterogeneous environments.
The conflict between the empirical and theoretical findings arises principally
because the models have been developed with human populations in mind, and are
not designed to address the specific features of animal social transmission. The
theoretical work has established that stable, vertically transmitted cultural traditions,
like guided variation, must have evolved in a relatively constant environment.
However, there is little evidence for ‘guided variation’ operating in animal
populations. The spread of cultural traits horizontally through animal populations
may be better conceptualized as the interaction of biased cultural transmission and
individual learning (Fisher & Hinde, 1949 ; Laland & Plotkin, 1990, 1991 ; Lefebvre
& Palameta, 1988; Sherry & Galef, 1984). BR (1989a) have introduced biased
transmission into their models by exploring the situation in which individuals sample
a number of models, and subsequently use this information to adopt a behavioural
variant. This model could be usefully extended to take account of empirical findings
in the animal social learning literature. The fact that horizontal social transmission
has been found amongst social foragers living in unpredictable and rapidly changing
environments (Lefebvre & Palameta, 1988) suggests the possibility that selection
may favour rapid horizontal social transmission in an unpredictable environment if
it increases the fitness of social foragers over and above that of individual foragers.
In such an environment individual learning is likely to be costly, and social learning
would be advantageous provided that traits that reduce fitness do not spread.
Laboratory experiments on rats suggest that social learning mechanisms exist that
allow rats to avoid the social transmission of preferences for toxic foods, or traits that
might reduce fitness (Bond, 1982; Galef, 1985; Lavin, Freise & Coombs, 1980).
Mechanisms, such as scent marking in mammals, may allow individuals to sample
the behaviour of the population rather than specific models (Laland & Plotkin,
1991). Finally, theoretical analyses have found that social learning is less effective
than individual learning in tracking environmental variability, partly because
transmission acts as a conservative force that lags behind environmental variability.
‘Cultural traditions should not change instantly in response to changing en-
vironmental conditions’ (BR, 1985, p. 56). In reality, individuals may be forced
temporarily to ignore social cues (for example, those indicating which foods to eat)
when they conflict with the state of the environment (i.e. if the food supply is
158 Kevin N. Laland
exhausted). Models that examine the consequences of (1) restricting population size,
(2) considering more than two behavioural variants, and (3) making environmental
variability a function of population behaviour, may uncover circumstances in which
natural selection will favour reliance upon rapid horizontal social transmission in
unpredictable environments.

Evolution of cooperation
Campbell (1975) argued that the fact that most moral beliefs are altruistic suggests
that they have been shaped by group selection. Group selection is the idea that
natural selection can operate between groups of individuals, and not just between
individuals. This idea became popular in the 1950s and 1960s, principally because
many biologists found it difficult to see how natural selection acting at the level of
the individual could lead to the evolution of traits which do not benefit the
individual, but rather its group. Evolutionary biologists have since developed
alternative mechanisms to account for altruistic behaviour (kin selection, reciprocal
altruism) and have largely dismissed group selective arguments. However, kin
selection can only explain altruism between related individuals. Reciprocal altruism,
the trading of altruistic acts in which benefit is larger than cost, is expected to evolve
when two individuals associate long enough to exchange roles as potential altruist
and recipient (Trivers, 1985). However, while analysis of the repeated interaction of
pairs of individuals has found that cooperation via reciprocal altruism can be stable
(Axelrod & Hamilton, 1981), the conditions that allow the evolution of reciprocal
cooperation become extremely restrictive as group size increases (Boyd & Richerson,
1988b, 1989b). BR (1982, 1985) argue that certain types of cultural transmission can
make group selection workable, and lead to the evolution of cooperation amongst
large groups of unrelated individuals.
The conditions necessary for the traditional form of group selection envisaged by
Wynne-Edwards (1962) are generally thought to be too stringent to be realistic
(Maynard-Smith, 1976). If individuals are either ‘selfish’ (S) or ‘altruistic’ (A), and
if the appearance of a single S immigrant in an A patch will result in its rapid
conversion to an S patch, then the conditions for the survival of A patches are very
low migration between patches, very short survival of S patches, and new patches
must be colonized by one or a few individuals (Maynard-Smith, 1989). If ‘altruistic’
groups cannot survive then it is difficult to envisage traits ‘for the good of the group’
resulting in group selection. In human populations, however, and possibly amongst
hominoids, cultural mechanisms may have existed which considerably relax the
conditions necessary for group selection. Building on earlier work by CF, BR
develop a model of ‘conformist’ cultural transmission in which individuals adopt the
most common cultural variant (positive frequency-dependent bias). They find that
conformist frequency-dependent bias improves the chance of acquiring the locally
favoured variant, and increases the amount of cultural variation. BR show that one
of the consequences of this is an increase in the strength of the group selection of
cultural variation : ‘ since selection between groups may favor beliefs and attitudes
which benefit the group at the expense of the individual, this provides an explanation
for human cooperation’ (1985, p. 227).
 Mathematical modelling of human culture 159

How does culture affect our evolution?

LW conclude that ‘sensitivity to usage patterns can increase the rate of genetic
assimilation’ (p. 290). This claim refers to the fact that a genotype with a strong bias
for the favoured cultural trait will replace an unbiased genotype more quickly than
a genotype with a weak bias. However, LW can only reach the conclusion that this
results in an acceleration in the rate of evolution by assuming extremely high fitness
costs for the development of a nervous system, an assumption that both Maynard-
Smith & Warren (1982) and Kitcher (1985) believe to be unwarranted. In fact, at first
sight culture seems much more likely to slow down evolutionary rates : if organisms
respond at the cultural level to changed selection pressures a genetic response is
unnecessary, and genetic variability that would have been selected out is preserved
(Laland 1990, 1992a; Maynard-Smith & Warren, 1982). However a capacity for
cultural transmission is likely to have complex effects on the rate of genetic evolution
which have only begun to be understood (Bateson, 1988; Laland, 19926; Odling-
Smee, 1988; Plotkin, 1988; Wilson, 1985). Culture may not only lead to genetic
assimilation, but also increase the likelihood of group selection (Boyd & Richerson,
1985; Simon, 1990) and sexual selection (Boyd & Richerson, 1985; Richerson &
Boyd, 1987), generate novel selection pressures (Laland, 1992a; Wilson, 1985),
preserve genetic variability (LW, 1981 ; Laland, in press, a), build up disequilibrium
(non-random associations) between genes and learned behaviours (Laland, 1992 b ;
Lumsden, 1988), and affect population structure (Laland, 19926), all of which may
affect evolutionary rates. Gene-culture co-evolutionary models provide the tools to
investigate these phenomena further.

Why model culture?

Wh_ymodel a t all?
Population biologists make accurate predictions about field data despite assumptions
such as random mixing, discrete generations and no inbreeding. Similarly, physicists
have made enormous progress by assuming that bodies can be treated as particles,
with no friction, no air resistance, and so on. Much theoretical work that employs
mathematical modelling has been conducted in the belief that reducing the analysis
of the processes under consideration to a few relevant parameters, and formalizing
the relationship between them to capture the essence of the processes, can greatly aid
our understanding of the phenomenon. Substituting a complex model for a complex
world does not aid our understanding. Formalization forces models to be simple,
otherwise we should never be able to interpret the results of analysis and simulation.
Using mathematics forces the modellers to be explicit about their assumptions, and
often clarifies ideas or relationships between variables.
Some processes are not accessible to experimental analysis. We cannot, for
example, conduct breeding experiments using human beings in order to test
hypotheses about our evolution. We can, however, develop mathematical models of
such processes, subject them to simulation, and use the data generated to test the
feasibility of the hypotheses. For example, CF investigated whether there are
circumstances in which cultural traits might spread despite the fact that they lower
genetic fitness. Without mathematical modelling as a tool such general questions
160 Kevin N.Laland
would be difficult to address. Further, sometimes it is possible to apply these models
to specific problems. For example, CF ask what kinds of cultural transmission and
migration best explain the divergence of languages in Micronesian island populations.
Their analysis rules out some explanations and confirms that others are feasible.
Similarly, LW apply their modelling to gain insight into incest avoidance, the
dynamics of women’s fashions and the splitting up of villages in Venezuela. CF and
BR have provided us with the tools to address other theoretical questions of interest
to psychology.

In which areas of psychology could these models be usejdb applied?

It is difficult to conceive of an area of psychology in which the social learning and
transmission of information, beliefs, attitudes or behaviours is irrelevant. Whatever
the specific focus, the dynamics of transmission is typically important for a complete
understanding of the phenomena. For those fields in psychology in which diffusion
dynamics are central, for example, social psychology and psychopathology, the utility
of the transmission models for describing, and predicting the behaviour of traits
(beliefs, attitudes, psychopathological disorders) is clear. If this were the sole
potential application of the theory of relevance to psychology then it could still have
a major impact. However, this theory has considerably more to offer psychology than
the description and formal analysis of psychological and behavioural traits. This
section endeavours to illustrate the kind of projects that could be undertaken by
psychologists utilizing the methodology or findings of the body of theory described
in the preceding sections.
The most obvious application of cultural evolution and dual-inheritance models is
to describe, model and ultimately predict the patterns of behaviour change found in
populations of human beings. Clearly there are many behavioural traits which are
culturally transmitted, and few people would dispute the value of being able to model
and predict the ways in which patterns of trait usage change over time. All of the
modellers whose work is discussed above have given examples of cultural traits for
which there is empirical evidence that they are vertically, obliquely or horizontally
transmitted. BR have gone to some lengths to document these cases (pp. 50-55).
Vertically transmitted traits include personality traits, cognitive development,
attitudes, attainments, educational and occupational status, upward/downward
mobility, patterns of socialization, occupation, sex-role conceptions, sexual activity,
attitudes towards feminism, political beliefs, political activity, attitudes to war,
religious beliefs, drug usage, alcohol usage, dietary habits, phobias, self-esteem,
language and linguistic usage. Horizontally and obliquely transmitted traits include
attitudes, career and social mobility aspirations, sex-role and sexual behaviour,
adolescent behaviour, aggressive behaviour, altruistic behaviour, morals and social
values, conformity, language and dialect, technological innovations, clothing
fashions, consumer behaviour, children’s games, rituals, stories and rhymes. The
dynamics and diffusion of all of these traits, and many more, can be modelled,
analysed and predicted using the cultural evolution models that CF and BR have
developed. Some of this work has already been undertaken. For instance, CF have
developed models applicable to the study of changes in language structure, word
usage and pronuciation change (Cavalli-Sforza & Feldman, 1973a, 1983).
 Mathematical modelling of human culture 161
There are also many topics of interest to psychologists that could benefit from the
application of gene-culture co-evolutionary models. This is particularly true of
psychopathology. For example, the dynamics of many psychiatric disorders (such as
schizophrenia, obsessive-compulsive, psychosexual, unipolar and bipolar disorders),
personality traits, aggression, criminal behaviour, alcoholism, drug usage, eating
disorders, may all be modelled in terms of an interaction between genetically and
culturally transmitted factors. As dual-inheritance models formally incorporate both
genetic and cultural processes, and by virtue of the fact that genetic and cultural
forces can be modelled as conflicting pressures, these models can capture many of the
essential elements of current medical, sociocultural, social learning and psycho-
analytic models of abnormality. In fact BR go so far as to develop a ‘Freudian’ model
in which the co-evolution of genes and culture becomes an arms race, cultural and
genetic traits evolving in opposite directions, with behaviour retaining an
intermediate value (1985, p. 194). Dual-inheritance models could also be usefully
employed to explore the dynamics of other traits such as handedness and cerebral
dominance, intelligence, and possibly religious and political beliefs. The infra-
structure for this kind of analysis has been developed and is available for our
exploitation.
The potential application of this cultural evolution theory in psychology is
considerably broader than the mere description and analysis of the ways in which
patterns of beliefs, attitudes, behaviours and disorders change over time. The models
also furnish psychologists with a new means to test between alternative
developmental hypotheses. In the same way that medical researchers and behaviour
geneticists can use family pedigrees to test the goodness-of-fit of different genetic
models of disease and psychiatric disorder, psychologists can use both genetic and
culturally transmitted pedigree data to test between different gene-culture co-
evolutionary models. This approach could lend insight into the nature of many
behaviours of concern to psychologists, from alcoholism, to political beliefs, to
sexual behaviour. For instance, it might allow psychopathologists to investigate
whether a gene-culture co-evolutionary model of the transmission of schizophrenia,
which took into account child-rearing practices, or the nature of interpersonal
relationships in the home, might account for more of the variation in incidence than
purely genetic models. As gene-culture co-evolutionary theory has established that
cultural processes can allow maladaptive traits to spread (CS, 1981), and can screen
fitness reducing genes from selection (Laland, 1992a), such a model might also
be able to account for schizophrenia’s puzzlingly high incidence. Secondly, in cases
where a particular model is established, this approach could be employed to estimate
the transmission coefficients. For example, social psychologists might be able to
establish the differential weighting individuals place on their cultural models
(mother, father, teacher, peers) for a particular culturally transmitted behaviour (say,
smoking behaviour) by investigating which estimates of the transmission coefficients
give the best fit to family histories. Thirdly, as the theoretical models all make
different assumptions about the nature of social learning and transmission, they
provide an impetus for the empirical investigation of these processes. D o cultural
traits spread through ‘ cultural selection ’, natural selection, ‘guided learning ’, or
‘trend watching ’? Which parameters determine the transmission dynamics ? How is
socially transmitted information integrated with individual learning and innate
162 Kevin N.Laland
biases ? Do different social learning mechanisms have different transmission
dynamics? The answers to such questions remain elusive, and our understanding of
the transmission process is primitive. BR have emphasized the potential of multi-
‘generation ’ social learning experiments, in which traits are transmitted along a chain
of subjects, with varying degrees of subject overlap per generation (e.g. Insko e t al.,
1980, 1982, 1983; Jacobs & Campbell, 1961; Zucker, 1977).
This experimental design could be modified so that the effects of all the forces could be
investigated. Selection could be modeled by removing variant individuals or groups from the
experimental population, migration by exchanging individuals between groups, and random
errors by measuring the divergence between replicated groups. The experimenter can also
manipulate the environment by giving individuals or groups tasks to perform, the payoffs of
which are manipulated. Subjects from different cultures or subcultures could be used to
investigate the effects of natural cultural variation on the various forces (BR, 1985, p. 297).

Given the fundamental nature of the cultural transmission process to the research
interests of a large fraction of the psychology community, such basic research is
surely warranted.
These considerations also apply strongly to cross-cultural psychology. Hewlett &
Cavalli-Sforza’s (1986) study of cultural transmission among Aka pygmies is a good
example of basic empirical research which was motivated at least partly by CF’s
modelling of cultural transmission. The theory suggested that quantitative data on
mechanisms of transmission of cultural traits (such as hunting and food-gathering
skills, mating behaviour, child care, singing and dancing) could be used to predict
within-group variability, stability of cultural traits over time and space, and patterns
of cultural change. The predictions were largely confirmed: for instance, the
hypothesis that vertical (parent-offspring) cultural transmission would allow
considerable variation amongst individuals in the population, and exert a
conservative force on rates of cultural change, was given clear empirical support.
Different patterns of transmission generated different levels of variability and
exhibited different rates of change, again as predicted. The cultural evolution theory
provides the impetus and conceptual rigour to stimulate further research into
questions of interest to cross-cultural psychologists and anthropologists.
Moreover, in the same way that sociobiology and behaviour ecology generated a
number of plausible hypotheses concerned with human behaviour (e.g. parent-
offspring conflict, adultery, xenophobia) primarily through the use of game
theoretical approaches, the cultural evolution theory can also be employed to address
these kinds of problems and generate testable hypotheses. The advantages of the
formal apparatus of gene-culture co-evolutionary models over that of game theory
are many. Firstly, they facilitate analyses of the ways and rates at which dynamical
systems change, rather than simply finding the equilibrium states. Secondly, they
allow a consideration of the effects of culturally transmitted traits on the system
dynamics. Thirdly, some of the assumptions of game theory, for instance that
selection will favour strategies that maximize mean fitness, can be relaxed. These
advantages are likely to enhance the accuracy and specificity of any hypotheses
generated by the use of this theory over those of sociobiology.
An example of how this theory can be employed to generate testable hypotheses
is its application to understanding human cooperation. The view that human beings
are by nature cooperative is not easily explained from an evolutionary perspective,
 Mathematical modelling of human culture 163
whilst the facts of human cooperation are difficult to reconcile with the view that
human nature is fundamentally self-serving. Reciprocal altruism, the game theory
explanation, can only account for a subset of cases of human cooperation. BR’s
models of cultural group selection provide an additional explanation for cooperative
behaviour which may have a broader jurisdiction. As BR (1982, p. 349) point out,
the models provide qualitative predictions about the kind of transmission rules that
might explain human cooperative behaviour, i.e. ‘a cultural transmission rule that
increases the frequency of the more common variant can cause group selection to be
a strong force in determining the kinds of behaviours that characterize different
human societies ’.
BR find that group selection on cultural variation is probably more likely when the
variants are also subject to indirect bias. Indirect bias refers to cases where
individuals use particular traits (‘indicator traits ’) or (charismatic/high status)
individuals to guide their choice of behaviour. A ‘ runaway ’ process can result which
selects for clearer and more overt indicators of status, which eventually become
divorced from adaptive superiority. ‘Much as peacock tails and bowerbird houses are
thought to result from runaway sexual selection, the indirect bias runaway process
will generate traits with an exaggerated, interrelated, aesthetically pleasing but
afunctional form’ (BR, 1985, p. 278). BR argue that this ‘may explain why the
altruistic, group functional behavior of humans seems to be so commonly embedded
in systems of supernatural sanctions and costly rituals ’ (p. 276). This theory promises
insight into the evolution of sacred traditions and creation myths, and the
significance of ritual and symbolism. The co-evolution of cultural beliefs and
practices could readily be modelled using CF’s or BR’s theory, and, for example,
could be used to investigate the ways in which religious beliefs change over time. For
instance, belief in a shepherd god of the Judaeo-Christian model probably co-
evolved with a pastoral way of life (Lenski & Lenski, 1970). Theoretical analyses
suggest different properties of culture should be stable amongst populations in
different kinds of environment, with different social structures, levels of migration,
and so on. Careful application of the cultural evolution theory tuned to the specific
properties of the population under consideration is likely to reveal a number of
predictions which can be put to empirical test. Which religions will dominate the
globe in the next decade? What properties do successful religions have, and why? It
is something of an irony that fundamentalist religious beliefs, which often challenge
the authority of evolutionary theory, are themselves spreading through Darwinian
natural selection. Measures of religious affiliation are strongly correlated with
fecundity and mortality, and, horizontal and oblique cultural transmission aside, a
significant component of the spread is likely to be due to natural selection (BR, 1985,
p. 176). Moreover, conservative religions, by impeding any cultural response to
other cultural changes (for instance, resulting from advances in technology) may set
up a genetic response to the novel selection pressures which more rapidly changing
religions will buffer out (Odling-Smee, in press).
Comparative psychology is another area in which this theory might find useful
application. Theoretical investigations of the evolution of communication, and the
evolution of cultural transmission are far from complete. Aoki & Feldman’s (1987)
finding, from analysis of a diploid two-locus model, that a mutant allele that allows
the learning of communication signals will not spread from low frequency is
164 Kevin N. Laland
puzzling. Similarly BR’s (1985, 1988a) conclusion that culture should evolve in a
constant environment does not square easily with empirical observations of social
transmission amongst mammalian and avian foragers living in unpredictable
habitats. In both cases further theoretical investigations are required to elucidate with
greater specificity the conditions under which communication and learning processes
will be favoured by selection. There are numerous other questions of interest to
comparative psychologists which could potentially be addressed by dual-inheritance
models. In particular, these models provide a framework for an investigation of the
co-evolution of brain and nervous system structure and behaviour : for example, the
evolution of intelligence, the evolution of psychopathological conditions, the co-
evolution of laterality and behavioural asymmetry, or the evolution of ‘modules of
mind’ (Fodor, 1983). Although the terms ‘dual-inheritance ’, or ‘gene-culture ’,
imply that the models apply exclusively to cultural organisms, similar models can be
developed for learners incapable of social transmission. Such a theory could, for
example, address the conditions under which constraints on learning might evolve.
The models also give psychologists a new opportunity to frame their work in an
historical or evolutionary context. For example, those psychologists interested in
language will be interested in the evolution of communication and symbolic
reasoning, and in the co-evolution of communication systems and brain structure
(Aoki & Feldman, 1987,1989; Boyd & Richerson, 1985; Cavalli-Sforza & Feldman,
1981, 1983). Cavalli-Sforza, Piazza, Menozzi & Mountain (1988) have found a
remarkable correlation between genetic and linguistic boundaries, suggesting that
recent human phylogeny and linguistic evolution are inextricably tied. Aoki &
Feldman’s (1991) investigation of the co-evolution of recessive hereditary deafness
and sign language is a particularly interesting example of the kind of problem that
gene-culture co-evolution models can address. Most hereditary deafness is believed
to be caused by recessive genes, which means that an affected child may be born to
normal heterozygous parents. This means that the transmission of a particular sign
language across generations would be interrupted if it was only learned by the deaf.
By assuming that deaf and hearing individuals learn to sign with different
probabilities, Aoki & Feldman develop theoretical models which investigate the
conditions for the persistence of sign language use in the population. They find that
these conditions are difficult to satisfy, and interpret this as suggesting a fairly recent
origin for extant sign languages. The persistence of sign language is facilitated by
increased levels of oblique and horizontal transmission, and by assortative mating
according to whether or not an individual is deaf or not. Interestingly, assortative
mating according to whether an individual can sign or not decreases the likelihood
of the persistence of signing, since it reduces the number of families with deaf
children that have one or more signing parent. As assortative mating, and
oblique/horizontal transmission are density-dependent phenomena, the authors
suggest that sparsely distributed populations would seem unlikely candidates for the
establishment of sign language.
This theory also encourages psychologists to focus on how contemporary
psychological and biological processes can interact to determine the way in which
human beings are evolving now. One of the classical sociobiological arguments is
that ‘genes may keep culture on a leash’. Models of gene-culture co-evolution
 Mathematical modelling of human culture 165
suggest that the opposite scenario is also plausible, i.e. that the selection of genes may
be a function of the frequency of particular cultural traits. A well-known example is
the spread of dairy farming leading to selection of genes for lactose absorption.
Another is the finding that culturally transmitted factors affecting choice of mate may
generate sexual selection in human populations (Richerson & Boyd, 1989). This will
occur if the mating preferences influence the intensity of selection on genetically
transmitted physical and behavioural traits. This is another example of how high
status or charismatic individuals in a population can exert a strong affect on the
dynamics of the population’s behaviour, in this case generating sexual selection. If
individuals select mates that resemble idolized cultural heros of the opposite sex, then
a single Marilyn Monroe could lead to strong selection in the population for blonde-
haired, blue-eyed females. This theory illustrates disadvantages of divorcing research
into evolutionary biology from the study of the mechanisms involved in the
development of human behaviour.
Cultural evolution and dual-inheritance models have a proven record of useful and
valid application in the human sciences. As reported in this and the preceding
sections, the theory has been successfully employed to describe, analyse and predict
the diffusion of cultural traits through populations ; to test specific models against
data; to generate hypotheses with considerable predictive power; and to place the
work of psychologists in a meaningful evolutionary context. The theory can be
considered a major methodological innovation, and there are many possibilities for
its fruitful application to psychological problems.

Summary
Recent years have seen the growth of a new and exciting field of theoretical research
concerned with the mathematical modelling of culture, and of its interaction with
genetics. These models have already been applied to address many problems of
interest to psychologists. The models constitute a major theoretical innovation, and
there is considerable potential for their application in psychology. Suggestions are
made as to the areas of psychology in which these models might have some utility.

Acknowledgements
This work was partially carried out under a science and engineering postgraduate studentship. I am
grateful to the Tregaskis Bequest of London University and the Experimental Psychology Society for
supporting a study visit to the University of California, Davis and Los Angeles, which facilitated very
useful discussion with Prof. P. J . Richerson and Prof. R. Boyd. I would also like to thank Prof. M. W.
Feldman, D r C. M. Heyes, Dr C. Lumsden, D r F. J. Odling-Smee, D r H. C. Plotkin, Prof. P. J.
Richerson, Prof. E. 0. Wilson, and an anonymous referee for their comments on this manuscript.

References
Aoki, K. (1991). Time required for gene frequency change in a deterministic model of gene-culture
coevolution, with special reference to the lactose absorption problem. Tbeoreticul Population Biology,
40, 354-368.
Aoki, K. & Feldman, M. W. (1987). Toward a theory for the evolution of cultural communication:
Coevolution of signal transmission and reception. Proceedings oftbe National Academy ofsciences, U S A
84, 7164-7168.
166 Kevin N.Laland
Aoki, K. & Feldman, M. W. (1989). Pleiotropy and preadaptation in the evolution of human language
capacity. Theoretical Population Biology, 35, 181-1 94.
Aoki, K. & Feldman, M. W. (1991). Recessive hereditary deafness, assortative mating, and persistence
of a sign language. Theoretical Population Biology, 39 (3), 358-372.
Atkinson, R. C., Bower, G. H. & Crothers, E. J. (1965). A n Introduction to Mathematical Learning Theory.
New York: Wiley.
Axelrod, R. & Hamilton, W. D. (1981). The evolution of cooperation. Science, 211, 1390-1396.
Bateson, P. (1988). The active role of behavior in evolution. In M-W. Ho & S. W. Fox (Eds),
Evolutionary Processes and Metaphors. Chichester : Wiley.
Bond, N. W. (1982). Transferred odor aversions in adult rats. Behavioraland Neural Biology, 35,417422.
Boyd, R. & Richerson, P. J. (1976). A simple dual inheritance model of the conflict between social and
biological evolution. Zygon, 11 (3), 254-262.
Boyd, R. & Richerson, P. J. (1982). Cultural transmission and the evolution of cooperative behavior.
Human Ecology, 10, 325-351.
Boyd, R. & Richerson, P. J. (1983). The cultural transmission of acquired variation: Effect on genetic
fitness. Journal of Theoretical Biology, 100, 567-596.
Boyd, R. & Richerson, P. J. (1985). Culture and the Evolutionary Process. Chicago: University of Chicago
Press.
Boyd, R. & Richerson, P. J. (1987). The evolution of ethnic markers. Cultural Anthropology, 2, 65-79.
Boyd, R. & Richerson, P. J. (1988~).An evolutionary model of social learning: The effects of spatial
and temporal variation. In T. Zentall & B. G. Galef J r (Eds), Social Learning. Hillsdale, NJ : Erlbaum.
Boyd, R. & Richerson, P. J. (19886). The evolution of reciprocity in sizable groups. Journalof Theoretical
Biology, 132, 337-356.
Boyd, R. & Richerson, P. J. (1989a). Social learning as an adaptation. Lectures on Mathematics in the Lz$e
Sciences, 20, 1-26.
Boyd, R. & Richerson, P. J. (1989b). The evolution of indirect reciprocity. Social Networks, 11,213-236.
Bush, R. R. & Mosteller, F. (1955). Stochastic Modelsfor Learning. New York: Wiley.
Campbell, D. T. (1975). On the conflicts between biological and social evolution and between
psychology and moral tradition. American Pychology 30, 1103-1 126.
Cavalli-Sforza, L. L. & Feldman, M. W. (1973~).Models for Cultural Inheritance 1. Group mean and
within group variation. Theoretical Population Biology, 4, 42-55.
Cavalli-Sforza, L. L. & Feldman, M. W. (19736). American Journal of Human Genetics, 25, 618-637.
Cavalli-Sforza, L. L. & Feldman, M. W. (1981). Cultural Transmission and Evolution: A Quantitative
Approach. Princeton, N J : Princeton University Press.
Cavalli-Sforza, L. L. & Feldman, M. W. (1983~).Paradox of the evolution of communication and of
social interactivity. Proceedings of the National Academy of Sciences U S A , 80, 2017-2021.
Cavalli-Sforza, L. L. & Feldman, M. W. (1983b). Cultural versus genetic adaptation. Proceedings of the
National Academy of Sciences U S A , 79, 1331-1335.
Cavalli-Sforza, L. L., Piazza, A., Menozzi, P. & Mountain, J. (1988). Reconstruction of human
evolution : Bringing together genetic, archaeological, and linguistic data. Proceedings of the National
Academy of Sciences U S A , 85, 6002-6006.
Chen, K. H., Cavalli-Sforza, L. L. & Feldman, M. W. (1982). A study of cultural transmission in
Taiwan. Human Ecology, 10 (3), 365-382.
Curio, E., Emst, U. & Vieth, W. (1978). Cultural transmission of enemy recognition: One function of
mobbing. Science, 202, 899-901.
Dawkins, R. (1976). The Selfish Gene. Oxford: Oxford University Press.
Durham, W. H. (1982). Interactions of genetic and cultural evolution. Human Ecology, 10, 28S-323.
Durham, W. H. (1990). Advances in evolutionary culture theory. Annual Review of Anthropology, 19,
187-210.
Eshel, I. & Cavalli-Sforza, L. L. (1982). Assortment of encounters and evolution of cooperativeness.
Proceedings of the National Academy of Sciences U S A , 79, 1331-1335.
Fagan, R. (1981). Animal Play Behaviour. New York: Oxford.
Feldman, M. W. & Cavalli-Sforza, L. L. (1975). Annals of Human Biology, 2, 215-226.
Feldman, M. W. & Cavalli-Sforza, L. L. (1976). Cultural and biological evolutionary processes,
selection for a trait under complex transmission. Theoretical Population Biology, 9 (2), 238-259.
 Mathematical modelling of human culture 167
Feldman, M. W. & Cavalli-Sforza, L. L. (1977). The evolution of continuous variation: 11, complex
transmission and assortative mating. Theoretical Population Biology, 11 (2), 161-181.
Feldman, M. W. & Cavalli-Sforza, L L. (1979). Aspects of variance and covariance analysis with
cultural inheritance. Theoretical Population Biology, 15 (3), 276-307.
Feldman, M. W. & Cavalli-Sforza, L. L. (1984). Cultural and biological evolutionary processes: Gene-
culture disequilibrium. Proceedings of the National Academy of Sciences U S A , 81, 16061608.
Feldman, M. W. & Cavalli-Sforza, L. L. (1986). Towards a theory for the evolution of learning. In S.
Karlin & E. Nevo (Eds), Evolutionary Processes and Theory. New York: Academic Press.
Feldman, M. W. & Cavalli-Sforza, L. L. (1989). On the theory of evolution under genetic and cultural
transmission with application to the lactose absorption problem. In M. W. Feldman (Ed.),
Mathematical Evolutionary Theory. Princeton, N J : Princeton University Press.
Feldman, M. W., Cavalli-Sforza, L. L. & Peck, J . R. (1985). Gene-culture coevolution: Models for the
evolution of altruism with cultural transmission. Proceedings of the National Academy of Sciences U S A ,
82, 58165818.
Findlay, C. S. & Lumsden, C. J. (1988). The creative mind: Toward an evolutionary theory of
discovery and innovation. Journal of Social Biology and Structure, 11, 3-55.
Fisher, J. & Hinde, R. A. (1949). The opening of milk bottles by birds. British Birds, 42, 347-357.
Fodor, J. A. (1983). Modularity of Mind. Cambridge: MIT Press.
Galef, B. G. J r (1985). Direct and indirect behavioral pathways to the social transmission of food
avoidance. In P. Bronstein & N. S. Braveman (Eds), Experimental assessments and clinical
applications of conditioned food aversions. Annals ofthe New York Academy of Sciences, 443, 203-215.
Galef, B. G. J r (1988). Imitation in animals: History, definitions and interpretation of the data from the
psychological laboratory. In T. Zentall & B. G. Galef J r (Eds), Social Learning. Hillsdale, N J :
Erlbaum.
Galef, B. G. J r (in press). Tradition in animals: Field observations and laboratory analyses. In M. Bekoff
& D. Jamieson (Eds), Interpretation and Explanation in the Stub of Behavior. Boulder, CO: Westview
Press.
Giraldeau, L. A. & Lefebvre, L. (1986). Exchangeable producer and scrounger roles in a captive flock
of feral pigeons: A case of the skill pool effect. Animal Behaviour, 34, 797-803.
Giraldeau, L. A. & Lefebvre, L. (1987). Scrounging prevents cultural transmission of food-finding
behavior in pigeons. Animal Behaviour, 35, 387-394.
Hewlett, B. S. & Cavalli-Sforza, L. L. (1986). Cultural transmission among Aka pygmies. American
Anthropologist, 88, 922-934.
Hull, D. L. (1982). The naked meme. In H. C. Plotkin (Ed.), Learning, Development and Culture: Essays
in Evolutionary Epistemology. Chichester: Wiley.
Insko, C. A., Gilmore, R., Drenan, S., Lipsitz, A., Moehle, D., & Thibaut, J. (1983). Trade versus
expropriation in open groups: A comparison of two types of social power. Journalof Personaland Social
Psychology, 44, 977-999.
Insko, C. A., Gilmore, R., Moehle, D., Lipsitz, A., Drenan, S. & Thibaut, J . (1982). Seniority in the
generational transition of laboratory groups : The effects of social familiarity and task experience.
Journal of Experimental Social Psycbology, 18, 577-580.
Insko, C. A., Thibaut, J., Moehle, D., Wilson, M., Diamond, W. D., Gilmore, R., Soloman, M. R. &
Lipsitz, A. (1980). Social evolution and the emergence of leadership. Journal of Personal and Social
Psychology, 39 (3), 431-448.
Itani, J . & Nishimura, A. (1973). The study of infrahuman culture in Japan. A review. In E. W. Menzel
(Ed.), Precultural Primate Behavior. Basel : Karger.
Jacobs, R. C. & Campbell, D. T. (1961). The perpetuation of an arbitrary tradition through several
generations of laboratory microculture. Journal of Abnormal Social Psychology, 62, 649-658.
Kawai, M. (1965). Newly acquired pre-cultural behavior of the natural troop of Japanese monkeys on
Koshima Inlet. Primates, 2, 1-30.
Kitcher, P. (1985). Vaulting Ambition: Sociobiology and theQuestfor Human Nature. Cambridge, MA: MIT
Press.
Laland, K. N. (1990). Social transmission in Norway rats and its implications for evolutionary theory.
Unpublished PhD thesis, University of London.
168 Kevin N.Laland
Laland, K. N. & Plotkin, H. C. (1990). Social learning and social transmission of digging for buried
food in Norway rats. Animal Learning & Behavior, 18 (3), 246-251.
Laland, K. N. & Plotkin, H. C. (1991). Excretory deposits surrounding food sites facilitate social
learning of food preferences in Norway rats. Animal Behaviour, 41, 997-1005.
Laland, K. N. & Plotkin, H. C. (1992). Further experimental analysis of the social learning and
transmission of foraging information amongst Norway rats. Behavioural Processes, 27, 53-64.
Laland, K. N. (1992~).A theoretical investigation of the role of social transmission in evolution.
Ethology & Sociobiology, 13, 87-1 13.
Laland, K. N. (1992b). Learning and evolutionary rates: A critical review. Cognition & Evolution, 2,
63-78.
Lavin, M. J., Freise, B. & Coombes, S. (1980). Transferred flavor aversions in adult rats. Behavioral &
Neural Biology, 28,25-33.
Lefebvre, L. (1986). Cultural diffusion of a novel food finding behavior in urban pigeons: An
experimental field test. Ethology, 71,295-304.
Lefebvre, L. & Palameta, B. (1988). Mechanisms, ecology and population diffusion of socially learned
food finding behavior in feral pigeons. In T. Zentall & B. G. Galef J r (Eds), Social Learning.
Hillsdale, NJ : Erlbaum.
Lenski, G. & Lenski, J. (1970). Human societies: A macro-level introduction to sociology. New York:
McGraw Hill.
Lewontin, R. C. (1983). Review of ‘Genes, Mind and Culture’. Proceedings of the N e w York Academy of
Sciences.
Lumsden, C. J. (1988). Gene-culture coevolution: Culture and biology in Darwinian perspective. In D.
de Kerckhove & C. J. Lumsden (Eds), The Alphabet and the Brain. Berlin: Springer-Verlag.
Lumsden, C. J. (1989). Does culture need genes? Ethology & Sociobiology, 10, 11-28.
Lumsden, C. J. (1991). Culture as a semantic fractal: Sociobiology and thick description. Journalof Ideas,
2 (l), 11-18.
Lumsden, C. & Wilson, E. 0. (1981). Genes, Mind and Culture. Cambridge, MA: Harvard University
Press.
Lumsden, C. & Wilson, E. 0. (1985). The relationship between biological and cultural evolution.
Journal of Social Biological Structure, 8, 343-359.
Maynard-Smith, J. (1976). Group selection. Quarter4 Review of Biology, 51, 277-283.
Maynard-Smith, J. (1989). Evolutionary Genetics. Oxford : Oxford University Press.
Maynard-Smith, J. & Warren, N. (1982). Review of Lumsden, C. & Wilson, E. 0. (1981). ‘Genes,
Mind & Culture’. Evolution, 36,620-627.
Odling-Smee, F. J. (1988). Niche constructing phenotypes. In H. C. Plotkin (Ed.), The Role of Behavior
in Evolution. Cambridge, MA: MIT Press.
Odling-Smee, F. J. (in press). Biological evolution and cultural change. In V. Reynolds & E. Jones
(Eds), Religion and Survival: Biological Evolution and Cultural Change. Oxford : Oxford University Press.
Plotkin, H. C. (1988). Learning and evolution. In H. C. Plotkin (Ed.), The Role of Behavior in Evolution.
Cambridge, MA: MIT Press.
Plotkin, H. C. & Odling-Smee, F. J. (1981). A multiple-level model of evolution and its implications
for sociobiology. Behavioural Brain Sciences, 4, 225-268.
Pulliam, H. R. (1982). A social learning model of conflict and cooperation in human societies. Human
Ecology, 10, 353-363.
Pulliam, H. R. (1983). On the theory of gene-culture co-evolution in a variable environment. In R. L.
Mellgren (Eds), Animal Cognition and Behavior. Amsterdam : North-Holland.
Pulliam, H. R. & Dunford, C. (1980). Programmed to Learn: A n E s s q on the Evolution of Culture. New
York: Columbia University Press.
Richerson, P. J. & Boyd, R. (1978). A dual inheritance model of the human evolutionary process: I,
basic postulates and a simple model. Journal of Social Biology Structure, 1 (2), 127-154.
Richerson, P. J. & Boyd, R., (1989). The role of evolved predispositions in cultural evolution: Or,
Human sociobiology meets Pascals Wager. Ethology & Sociobiology, 10, 195-219.
Rogers, A. R. (1988). Does biology constrain culture? American Anthropologist, 90 (4), 819-831.
Sherry, D. & Galef, B. G. J r (1984). Cultural transmission without imitation: Milk bottle openings by
birds. Animal Behaviour, 32, 937-938.
 Mathematical modelling of human culture 169
Simon, H. A. (1990). A mechanism for social selection and successful altruism. Science, 250, 1665-1668.
Takasaki, H. (1983). Mahale chimpanzees taste mangoes - Toward acquisition of a new food item.
Primates, 24, 273-275.
Trivers, R. (1985). Social Evolution. Menlo Park, CA : Benjamin/Cummings.
Wilson, A. C. (1985). The molecular basis of evolution. Scientific American, 253 (3), 148-157.
Wynne-Edwards, V. C. (1962). Animal Dispersal in Relation to Social Behaviour. New York: Hafner.
Zucker, L. G. (1977). The role of institutionalization in cultural persistence. American Sociological Review,
42, 726-743.

Received 15 August 1991; revised version received 6 May 1992

Appendix : BR : Culture and the Evolutionary Process

Chapter 3 : The Cultural Inheritance System
Individuals acquire variant c or d from n models. BR assign the value 1 to c and 0 to d. Let XI be the
numerical value of the cultural variant of the ith model, in a particular set of cultural parents. Let A,
be the importance of the ith parent in transmission. Then in the case of n models, the probability that
a nai've individual acquires c given that n cultural parents have traits that take on the particular values
X,, ... , Xn is
n
Prob (cIX,,... , X,) = 'c A, Xi, (3.13 p. 66)
I-1

where A, = 1.
To predict the frequency of c after transmission, p', BR weight the probability that the nai've individual
acquires variant c given that he or she is exposed to a given set of models, by the probability that the
set of models is formed, summed over all sets of models. This gives
1 1
p' = C ... C Prob (+,, ... , xn)Prob (X, = xl, ... xn= xn). (3.14 p. 66)
z,-0 x,-0

If the formation of sets of models is random then p' = p .

This model is easily modified for horizontal transmission. BR consider a population at time t. During
an interval d t each individual contacts n- 1 individuals (models). At t+ 6t each individual either retains
his/her pre-existing cultural variant with probability A,, or adopts the variant of the ith individual
encountered with probability A,. Then
n
Prob(cJXl, ..., X,) = E AIXI. (3.19 p. 69)
I-1

This has the same form as 3-13. The only difference is conceptual: to model horizontal transmission
BR allow an individual to be one of its own cultural parents.