« Wait until we're wise.... | Gene Expression Front Page | Rational thinking on immigration »
May 04, 2003


This continues the discussion in my note of 24 April. It’s a long one, so take a deep breath (as the rock star said to the groupie).

First, a couple of preliminaries.

(a) I have had several useful comments on my previous note, both on the message board and by email. I will try to take these into account, but if I don’t respond to each comment in detail, it doesn’t mean I don’t value them.

(b) For the purposes of a blog, one must omit all the nuances and elaborations that would be necessary in a published book or article. Also, part of the aim is to provoke debate. However, I will not deliberately say anything that I don’t think is broadly correct.

I finished my last note by saying ‘I do not believe that cultural traits have been produced by any process closely resembling natural selection’. The present note expands on this.

First, there are two distinct ways in which cultural traits might, in principle, evolve by a process resembling natural selection. One way is roughly as follows: social groups (populations, societies, etc.) possess cultural traits; these social groups reproduce and undergo selection; and their cultural traits evolve as a by-product of this process. The second possible way is that the cultural traits themselves ‘reproduce’ and undergo selection directly. The implications of these two processes are radically different, and they must not be confused. Unfortunately they often are.

The present note is only concerned with the first process, which I call ‘cultural evolution by group selection’. The second process (an example of which is Richard Dawkins’s ‘memes’) is more plausible, though not without difficulties of its own. I will return to this in another note.

To evaluate the first process, it would be helpful to have a clear and concise account, by a believer in it, of how it is supposed to work. Alas, I haven’t found any. However, the following, from Hayek’s ‘Law, Legislation and Liberty’ (vol. 1, 1973, pp. 17-19), gives the gist of the doctrine: ‘Mind...is the result of man having developed in society and having acquired those habits and practices that increased the chances of persistence of the group in which he lived...The cultural heritage into which man is born consists of a complex of practices or rules of conduct which have prevailed because they made a group of men successful but which were not adopted because it was known they would bring about desired effects...such rules come to be observed because in fact they give the group in which they are practised superior strength, and not because this effect is known to those who are guided by them.’ In support of these assertions Hayek cites the (now-discredited) work of Wynne-Edwards on group selection among animals, and the classic work of A. M. Carr-Saunders on human demographic evolution: ‘Men and groups of men are naturally selected on account of the customs they practise just as they are selected on account of their mental and physical characters. Those groups practising the most advantageous customs will have an advantage in the constant struggle between adjacent groups over those that practise less advantageous customs’ (Carr-Saunders, ‘The Population Problem: A Study in Human Evolution’ (1922), p.223.)

Unfortunately the advocates of this kind of group selection never seem to bite the bullet and follow through the details of the process. So I will do it for them! In order for the process to allow evolution by natural selection, it must have the following elements:

1. Groups must have ‘reproduction’, i.e. groups must produce other groups.

2. Groups must have ‘heredity’ of cultural traits, i.e. descendent groups must have traits resembling those of their ‘parents’.

3. There must also be some variation in inherited traits. Crudely, the offspring must resemble their parents, but not be identical to them.

4. If the process is to be closely analogous to natural selection, the production of variant traits must be a blind, undirected process.

5. There must be variance in reproductive success. Some groups must have more ‘offspring’ than others, or some 'offspring' must survive better than others, or both. Evolution could in theory proceed entirely by differential reproduction, but obviously if there were ‘reproduction’ without ‘extinction’ the number of groups would increase indefinitely. Since this does not appear to be the case in human history, there must be group ‘deaths’ as well as ‘births’.

6. Variance in reproductive success must be based in part on heritable traits. There will of course also be a random component in reproductive success, but this should not significantly influence the direction of evolution (except maybe in Sewall Wright’s ‘shifting balance’ theory, which incidentally has recently been given a thorough trashing: see ‘Evolution’, vol. 52, 1997).

7. If evolution is to be cumulative, the offspring must themselves have offspring, and so on. As Dawkins has emphasised in his ‘Blind Watchmaker’, the important thing about natural selection is its ability to produce complex adaptations, and this requires iterative, cumulative selection over many generations.

The most obvious objection to this scenario as an explanation of cultural evolution is that societies simply don’t reproduce or die in the way required for the process to work. Or at any rate, if they do, the event is so rare in relation to the amount of cultural change, and the number of cultural traits to be accounted for, that it cannot be a major factor in their evolution. I think this objection is broadly correct, but it is necessary to examine some processes that might be considered sufficiently close to group ‘reproduction’ to invalidate the

Colonisation: by this I mean not just imperial conquest, but the settlement of populations to form new societies, displacing or overwhelming any existing inhabitants, and bringing with them most of the cultural traits of their society of origin. Examples include the British and French settlements in North America, the British settlement of Australia and New Zealand, and the Polynesian settlements in the Pacific islands. Colonisation in this sense is a comparatively rare process, because it requires either that the areas colonised are uninhabited, or that the colonisers have an overwhelming technological superiority. As a basis for cultural evolution, its main limitation is that it is usually a ‘one-off’ event: once a colonial society has been formed, it does not usually create further colonies of its own. It therefore does not lead to cumulative adaptation.

Growth and expansion: some populations grow and expand their territory at the expense of others: for example, the boundaries between Germans and Slavs in Eastern Europe have shifted backwards and forwards over the last 1500 years. Some advocates of group selection treat expansion as analogous to ‘reproduction’; for example, in the book ‘Unto Others’ Sober and Wilson cite the expansion of the Nuer at the expense of the Dinka in Sudan as their prize example of group selection in action. But as with colonisation, growth and expansion by itself does not lead to cumulative adaptation.

Spreading and splitting: if, however, populations expand and then split into distinct groups, some of which themselves expand and split, and so on, we do get an iterative process. This possibility had occurred to me, but an email comment has persuaded me that I should take it more seriously. The ‘spread-and-split’ scenario is similar enough to the reproduction of individual organisms to be plausible as a basis for group selection of cultural traits. There is also little doubt that the process has sometimes occurred in human history; for example, the spread of the Indo-Europeans in Europe, and the Bantu in Africa, have probably involved repeated expansion and splitting. My main reservation is about the frequency and importance of the process. I will come back to this shortly.

The frequency of ‘reproduction’ is not the only problem. For evolution of culture to occur by group selection, it is necessary that variance in reproductive success should be based in part on heritable cultural traits. But it is not clear that this is usually the case. For example, the formation of colonies by Britain was due to a combination of circumstances, most of which are not inherited by its ‘offspring’. The Australians and New Zealanders show no tendency to form colonies of their own, unless you count Hollywood. In the case of the Indo-Europeans and the Bantu, they probably had the accidental advantage of technology (such as iron weapons) which was not possessed by the populations they displaced. And even when the ‘reproduction’ of a society is based on some cultural practice or belief, there is no guarantee that this will persist long enough to be passed on to a third ‘generation’. One of the commonest phenomena of history is that tough nomadic peoples conquer their sedentary
neighbours, then settle down themselves.

Returning to the problem of frequency, in principle one should set up an explicit population genetics model to explore this. I won’t do this because:
(a) I don’t have the mathematical skill
(b) most of the relevant parameters are unknown, and
(c) we don’t even have a clear idea of the structure of the model, e.g. what counts as a separate unit of selection, or what counts as a genetic ‘locus’.

My gut feeling is that group selection will be too slow and too weak to be a major factor. However, I can’t expect people to share my gut feelings, so I will at least make a gesture towards quantifying the relevant parameters:

1. I assume that the number of distinct cultural units (‘societies’) in the world at any one time is around 2,000. This is rather more than the number recognised in ethnographic databases, but these are not complete.

2. For simplicity, I assume that the number of ‘societies’ has been roughly constant since soon after the emergence of Homo Sapiens Sapiens about 100,000 years ago. Obviously this isn’t strictly true - for example, the number of societies will have increased when HSS entered the Americas - but I don’t think variations in the number materially affect the outcome, they just complicate the model.

3. I assume that to count as a distinct unit a society must exist for at least 10 human generations (about 300 years). The average ‘lifetime’ of a society as a distinct unit is more difficult to estimate, but from historical and archaeological data I suggest that an average duration of 1,000 years is reasonable. Admittedly, primitive societies may have a fluid sub-structure, splitting and recombining sub-groups repeatedly over a shorter time-scale than this, but the offset is that these sub-groups are culturally very similar to each other.

4. From the assumption that the total number of societies is roughly constant over time, it follows that there must be one ‘birth’ for each ‘death’: each society must on average produce one ‘offspring’ before it ‘dies’. To maintain the steady state, around 2,000 societies will ‘die’ and 2,000 be ‘born’ around the world in any 1,000 year period.

5. There will be some variance in ‘reproductive success’. Although the average number of offspring is 1, some societies will have none, and others will have more. I suggest a range of about 0-5 ‘offspring’, in approximately a Poisson distribution, with the mode and median both at 0, the mean at 1, and the variance also approximately 1.

6. This modest variance in reproductive success gives some scope for selection, but only a part of the variance, probably less than half, will be based on heritable cultural traits.

7. The fitness effect of any particular cultural trait will usually be relatively small. I assume that there are at least 100 cultural traits (analogous to genetic loci) to be accounted for (this is probably a gross underestimate), and that each of these exists in at least two variant forms (analogous to alleles). The most favourable variant (allele) would not usually have a fitness advantage of more than 10 percent.

If these assumptions are in the right ballpark, group selection of cultural traits must be a weak force. Suppose first that a new cultural variant appears as a ‘mutation’ in one society out of the total ‘population’ of 2,000. The mutation may well go extinct by chance fluctuation before it gets a chance to spread. However, suppose that by a combination of good luck and selective advantage it manages to reach a frequency of 1 percent of the population (i.e. 20 societies). It will then be reasonably safe against chance extinction, and will gradually increase in frequency. But with an advantage of only 10 percent, this will initially be a very slow process: from 1 percent of the population to 1.1 percent, from 1.1 percent to 1.21 percent and so on. The process will gradually accelerate by the ‘compound interest’ effect, but when the ‘allele’ has become relatively common, the rate of increase will be damped down again, because the allele will be competing against itself as well as the less favoured variants. I haven’t done a precise calculation, but I reckon that it would take about 50 ‘generations’ for a new mutation to spread to half the ‘population’. But with a ‘generation’ length of 1,000 years, that is about half the entire lifespan of Homo Sapiens Sapiens since its origin! More important, any new mutation that has appeared in the last 20,000 years - the part of human history we are mainly interested in - would still be at a very low frequency.

It may of course be said that some cultural traits might have a greater selective advantage than 10 percent, and would therefore spread more rapidly. But given the limited reproductive variance, only a small proportion of the 100 or more traits could be in this position, and all the other traits that happened to be linked with these highly advantageous alleles would be carried along as ‘hitch-hikers’, regardless of their fitness effects. In the absence of anything resembling sexual reproduction there would be no easy way of getting rid of undesirable variants by counter-selection (see Maynard Smith, ‘Evolution of Sex’).

I have not discussed the ‘mutation rate’ for cultural traits, but I would guess that it would be orders of magnitude higher than for genetic mutations. It might well be high enough for mutation pressure to swamp the effects of selection. (The point to note here is that the probability of mutation to a particular allele applies to all the members of the population who do not already possess it, whereas a selective advantage applies only to those members of the population who do possess it, which is initially a small proportion. For example a mutation rate of 1 percent applying to 99 percent of the whole population would have more effect than a selective advantage of 20 percent applying to the remaining 1 percent.)

I also haven’t discussed ‘sideways’ transmission of cultural traits from one society to another, e.g. by imitation, but on plausible assumptions this would also tend to swamp the effects of group selection.

Finally, even if group selection of cultural traits does occur, there is no reason to suppose that all or even many of the traits of a ‘successful’ group have contributed to its success, let alone that they are beneficial to its individual members. The most we can say is that some of its traits give it a 'reproductive' advantage relative to other competing groups. To return to the example of the Nuer and the Dinka, one of the traits of the Nuer which emerges most clearly from Evans-Pritchard’s classic study is that they are extremely proud and warlike. This in turn is related to practices like the blood-feud and initiation rites. According to E-P, ‘initiation rites, more than anything save language, distinguish Nuer culture and give Nuer that sense of superiority which is so conspicuous a trait of their character’ (‘The Nuer’, p. 260). Before rushing to imitate them, nota bene the following description (ibid., p. 249): ‘All male Nuer are initiated from boyhood to manhood by a very severe operation. Their
brows are cut to the bone with a small knife, in six long cuts from ear to ear. The scars remain for life, and it is said that marks can be detected on the skulls of dead men’. No wonder they are such hard bastards!


Posted by David B at 05:38 AM

Don't have time to be thorough, but new religious groups are something to look at. The Mormons are tremendously successful and are growing. They're thought of as just another Protestant sect, but actually they're heretical to all other Christians because they claim a new revelation and a new prophet. They are one of the few survivors of a large number of very odd cults, mostly quasi-Christian, which appeared in the US XIX c. (Seventh Day Adventists, Jehovah Witnesses, and Christian Science are others).

Incidentally, guys, if you convert to Mormonism, you can marry a nice blond 16-year-old more or less right away. They are probably the sexiest Christian sect and use that to lure converts. Not kidding. Bigoted maybe, but girls invited my son to dances in their church basement when he was in ninth grade.

Posted by: zizka at May 4, 2003 12:56 PM

>> They are probably the sexiest Christian sect and use that to lure converts.

I wonder if there are sex differences in converts. Do males, or females tend to switch religions more frequently? I don't know any studies on the matter. My guess is that different religions appeal to different genders. For example, I'll venture a guess that New Age religions attract more females and Islam attracts more males.

Posted by: Dienekes at May 5, 2003 12:49 AM

David -
if I were to make an argument for group selection I would look at a longer time scale and a smaller group size than you are. The examples you cite - historical cases where one society displaced another - are the final result of group selection that took place over the previous fifty thousand+ years, not really illustrative of the mechanism in general.
In my opinion one of the strongest pieces of evidence for group selection is manifestation of group loyalty and patriotism. Not terribly useful to the individual, indeed rather more dangerous than cynical self-interest, but necessary for group survival in any environment where there is violent inter-group competition (which is to say, almost the entire world).
Of course it could be argued that that's culturally transmitted (a meme), and in its specifics it usually is - but the tendency and desire to feel oneself as part of a greater unit seems to be inbred.

Posted by: bbartlog at May 5, 2003 06:37 AM

Just a quickie - why do you (bbartlog) attribute group loyalty to group selection rather than to kin selection? Or to selfish (but not necessarily conscious) motives such as desire of status and approval within the group? Etc, etc, etc...

I just don't understand this constant desire to find group selection under every stone!

In any event, my note was not concerned with the evolution of altruism but with the evolution of culture. Altruism is a different issue, which I hope to deal with some other time.

Posted by: David B at May 5, 2003 07:52 AM

And another thing - I don't know why you assume that distinct cultural units (societies) were shorter-lived in prehistoric times than modern societies are. One of my reasons for assuming an average 'lifetime' of societies of 1,000 years was that archaeology shows a persistence of distinctive material cultures over this sort of time scale, e.g. 4 or 5 distinct cultures in the 5,000 years of Neolithic and Bronze Age Britain. Also (though I didn't think of this at the time) Cavalli-Sforza's genetic and linguistic cladograms show only about 10 branching points since HSS came 'out of Africa'. This probably understates the number of successive cultural 'splits', but language and culture are very closely associated.

Posted by: David B at May 5, 2003 08:01 AM

Maybe we are simply talking about different things. I assume that we are talking about actual genetic selection, and the expression of those genes through group behavior (very indirectly, in other words, given the memetic influences on group behavior). So for me there is no clear boundary between kin selection and group selection. Moving from a family (10 people) to an extended family / tribe (100 people) to a larger tribe of a thousand, there is a difference in the degree of relatedness and hence the amount of altruism that kin selection would predict, but no clear point where we would expect it to stop functioning entirely.
But this may simply mean that I don't understand what is meant by group selection. You are looking at the perpetuation or extinction of whole cultures, whereas I am thinking more of certain types of behavior. I should probably read some more about group selection so that I can tell how it differs from extended kin selection...

Posted by: bbartlog at May 5, 2003 10:32 AM

i went to a mormon dominated HS, i can agree with zizka on on mormon chix. but they won't do ANYTHING but missionary and only once they're married.

mormons went and did a lot of missionary work in northern europe during the 19th century and some of these migrated directly to utah-explaining the nordic look of AMERICAN mormons. but over half of mormons today are non-white (latin america, the pacific, korea, etc.).

Posted by: razib at May 5, 2003 01:15 PM

Bbartlog: thanks for your further comment, which I agree explains some of the difference between us.

My note was specifically concerned with cultural rather than genetic traits.

For cultural evolution to occur by 'group selection', the relevant 'groups' are populations which share a common culture. This would normally be a 'people' or 'nation' - e.g. the Navaho or the Nuer - with a common language, intermarriage, and similar if not identical customs and beliefs. Of course, as my note recognised, in primitive societies a 'people' may often contain many sub-groups, which split and recombine over a relatively short timescale, but as these groups share a common culture this doesn't matter.

Of course, I might be wrong about my '1,000 year average' - it is just a rough guess - but at the level of the 'people' there does seem to be persistence over at least several centuries. E.g., the peoples encountered by early European explorers in the 16th-17th Centuries were usually still recognisable in the C19.

I am posting a separate note on the whole subject of altruism and group selection.

Posted by: David B at May 6, 2003 02:24 AM