00:01
Hi, my name is May Hanzhi Zhang, and today I'd like to talk to you a little bit about how we study extinction and diversification on cultural phylogenies.

00:14
I want to quickly introduce you to the message of cultural phylogenetics. Then I'll go through three case studies applying this method to study the diversification and extinction process of both the cultural populations and the cultural traits.

00:33
A phylogeny is a tree-like shape, which represent the hierarchical descent of organisms from your common ancestors to the extant species, which we observe today. So, on this tree diagram, we can see the different species that we observe today near the tip of the tree—and these nodes along the branches represent a common ancestor that give rise to today's diverse... diversity, which we observe today.

01:06
Because genetic materials are more or less accurately passed down from the parents to the offspring generation—with rare modifications or mutations, which also gets passed down—we can sequence the genetic material of contemporary species and use this information to reconstruct the evolutionary history and the descent of the species. And as we see on this chart, the more closely related organisms will show similar patterns in their genetic material. So, that is phylogenetics as a method which was originally developed in evolutionary biology.

01:56
So, next, we will look at how we can apply this method to study cultural diversity. So human societies exhibit an extraordinary extent of diversity in terms of languages, customs and norms. There are these systematic differences between cultural groups that are relatively stable and consistent within the group. To understand the evolutionary history of this cultural diversity, we can model cultural history with various proxies. But in terms of applying phylogenetic methods, languages really stand out as the best proxy we can use. And this is because language and genes share a lot of similarity in terms of how they're transmitted, and how they can be broken down into discrete chunks of information, which is passed down from parents to offering with relative accuracy. So similar to the genetic material.

03:10
In languages, there is a set of core vocabulary, which tends to be fully universal (which means "prevalent in almost all" human languages). And they tend to be quite stable, which means that they're unlikely to be borrowed across languages or be replaced. So these core vocabulary are similar to the genomic material, which gets passed down accurately from parents to offspring. Similarly to how nucleotides which make up the genes that are used to reconstruct the evolutionary history, in languages in these core vocabulary, linguists can identify the cognates—which means the words that share a common ancestry—and use this information to infer and reconstruct the evolutionary history of languages.

04:09
And where possible phylogenetic inferences can incorporate fossils. These are any information we know about the timing ..., the timing of extinction events or the ancestral relationship between some languages or it could be the timing of the speed of branching events—any information we have from archaeological sources or written history or even genetic studies, we can incorporate that information into the phylogenetic models to get a more realistic estimate of the the transition rates and overall the tree topology.

04:59
So, just to give you an example of using fossils to calibrate phylogeny, this phylogeny is a biological tree of different species and where we have fossils of extinct organisms. If we know the timing, and if we can trace the ancestral relationship—or at least extinct historical organism to their contemporary offsprings—then we can use that information to calibrate the time of the phylogeny.

05:35
So now I'll go through three case studies, that show how we can apply cultural phylogenetics to study the diversification and extinction process of cultural populations and cultural traits.

05:49
And the first case study is how we use reconstructed cultural phylogeny: In this case, a language phylogeny to understand the diversification of cultural populations. Now, when we study diversification, in macroevolution, it often comes hand in hand with extinction. When there's limited resources and populations are in competition with each other, the diversification of one population often means extinction of another. If they take over their territory, or if they conquered another population, in order to expand and diversify. So it's really the two processes are really two sides on the same coin. But we're going to focus on diversification for this part. And later I'll talk about another study where we focus on the extinction process.

06:52
So our recent study looked at the evolutionary history of Sino-Tibetan languages. Sino- Tibetan language family is the second largest language family in the world, with 1.4 billion speakers and with territories that stretch from the eastern Pacific all the way to the Himalayas and Southeast Asia. So there are three main hypothesis regarding the origin of Sino-Tibetan languages. The first one are used to... the earliest Sino-Tibetan language arise from agricultural habitats near the Yellow... Yellow River Basin somewhere around 4000 to 6000 years ago. And there has been an archaeology evidence and some genetic evidence of a major farming dispersal. So out from the Yellow River Basin around that time period.

07:53
And the other two hypotheses claims the homeland to be somewhere either in northeast India around 9000 years ago, or Southwest Sichuan more than 10 thousand years ago. These two hypothesis are basically saying the earliest Sino-Tibetans are not an agricultural group at all. They are foragers in... near the Himalaya area, and the contemporary Sino-Tibetan languages are really the result of these forger groups, kind of gradually dispersing into the eastern territory of the Sino-Tibetan area.

08:34
So we use the cognate database from a website called Tower of Babel, by a Russian linguist Starostin. We use the Sino-Tibetan cognate data to reconstruct the phylogeny of these languages. And we did find some evidence of the Yellow River basin homeland hypothesis as we found a Sinitic language to be out-group of the entire tree. (So that out-group here at the bottom is more distantly related to any other subgroups on the tree, and likely to be the first one branched off from the common ancestor). So we felt this bifurcation between the Sinitic languages and the Tibeto-Burma languages. The root age of our inferences are somewhere around 8000 years ago. Now this is thousands of years before the first major... major farming dispersal out from the Yellow River Basin. So we think this really calls for a more nuanced understanding of how we interpret the root age of language phylogenies. Initial... linguistic divergence may not be triggered by physical distance that can be caused by increased social distance. And this.. this can be caused by increased social stratification and many other reasons.

10:18
So now we've looked at a case study of the diversification of cultural populations. Here I want to demonstrate how we can use phylogenetic methods to study the diversification/extinction of cultural traits. A big caveat of this application—this method—is it is only suitable for cultural traits or system or whatever you call it—social institutions—that are mainly inherited from a parent cultural group to an offspring cultural group. For cultural traits that are more likely to be copied and borrowed between groups, perhaps the phylogenetic method is not the most appropriate method to use. And there should be some reason for you to believe if you're using the phylogenetic method; there should be some reason for you to believe this trait is relatively stable within the unit of a cultural population or cultural group is passed on over generations rather than borrowed across groups.

11:35
So there are many analyses you can do with phylogenetic comparative method. You might be interested to know what's an earlier form of a cultural feature. That is, the ancestral state of a cultural trait. You might be interested to know what kind of which direction did the cultural trait change from—and that is transformation—and the mode and tempo of these changes. You might be interested in understanding whether two or more traits had been evolving in a correlated manner, whether there's any co-evolution.

12:22
So, I want to give an example to just contextualise what we have said before. So I'll give an example of a discrete cultural trait, which is the kinship system of human societies. Now, this is a very reductionist summary of human kinship systems. We... in this case, we're interested in two kinds of kinship system: the descent and the residence pattern. And descent—again a very reductionist summary—is we... can categorise it either by descent via the female kin which is a matrilineal system, or via the male kin which is the patrilineal, or both types: bilateral. Of course, there are many other more nuanced forms of descent. And similarly with the residence. In this case, it was the postmarital residence. We can categorise it into matrilocal, patrilocal, ambilocal—various categories that are commonly observed in human societies.

13:37
So, this is one of the earlier papers reconstructing the ancestral state of descent system of Austronesian population. In this case, the kinship system shows a very strong phylogenetic signal, which means it tends to be passed on from parental language group to offspring language group. So in in this case, it is appropriate for phylogenetic comparative inference. And this study found the ancestral form of descent among Austronesian groups are most likely to be matrilineal, which is the red colour here.

14:25
Because the phylogeny we reconstruct is inherently uncertain And this is why we get a posterior sample of something like a thousand trees rather than a single tree. So the numbers (which you can't really see here), actually represent the level of certainty we have, when we say this ancestral society are likely to be matrilineal or patrilineal.

14:50
Another kind of comparative inference you can do is to look for correlated evolution of multiple cultural traits. So, as you can see in this diagram, what we mean by correlated evolution is not the correlation of contemporary traits, we mean... is the two traits in their evolutionary history show some coinciding or closely occurring changes in their transition. So, if one... if the changes of one trait tend to lead to the changes of another trait, that is what we mean by correlated evolution. And this is only possible to know once we reconstruct that evolutionary history on the phylogeny.

15:42
So, an example of a co-evolution analysis is this 2003 paper looking at whether matriliny or patriliny co-evolved with cattle pastoralism among Bantu populations. So, this map shows the distribution of the two traits where there is clearly a belt—or what they call the matrimony belt—where you find a lot of matrilineal groups and most of them do not practice cattle pastoralism. So, on a surface level it looks like there is a correlation between matrimony and cattle pastoralism, but does this correlation still hold after we account for the phylogenetic history?

16:32
So, the way you test correlated evolution is you set up two models: one is the independent model which assumes the two traits evolve independently; and the other dependent model assumes the transition of one trait or fact the transition of another. And you... and you then putting the observed data and phylogeny into the model and compare which of the two model are more likely.

17:05
In this study, they found evidence of the dependent model being more likely than the independent model and they actually found this directional trajectory of the transition between matriliny and cattle. So, the patriliny and cattle pastoralism tend to be the evolutionary stable state or an evolutionary equilibrium. While the matriliny with cattle or patriliny without cattle is an unstable state. And the direction of phase transitions tend to lead to the equilibrium of either matriliny without cattle or patriliny with cattle.

17:56
So in our recent studies, we looked at the kinship evolution of Sino-Tibetan cultures using the method discussed above. We sampled around 110 Sino-Tibetan groups and their kinship systems. Most of the groups are patrilocal and patrilineal. But we also included information on the historical and extinct Sino-Tibetan groups, where possible, to inform the inferences of our model.

18:29
So this is the ancestral-state reconstruction of post-marital residence. The inferences suggests the ancestral state of Sino-Tibetan kinship were most likely female dispersal which means either patrilocal or neolocal, in this case. We did not find evidence for the hypothesis—which is commonly believed in some of the Asian countries—that Sino-Tibetan society evolved from a matrilineal and matrilocal form into the patrilineal and patrilineal form.

19:12
There are in some clades on the Sino-Tibetan tree; we do observe a concentrated matrilineal or ambilineal groups. But our inferences suggest these group most likely adopted female-centred kinship systems more recently rather than having inherited from the earliest ancestor ancestral groups of Sino-Tibetans.

19:43
We also looked at the co-evolution of the kinship system with their subsistence—that is, agriculture or cattle pastoralism—among Sino-Tibetans. And unlike the results in the Bantu inferences we do not find evidence that descent or residency co-evolved with domestic cattle, cattle pastoralism among Sino-Tibetans. We did find some evidence that residency, female dispersal co-evolved with agriculture. And we think this is because in Sino-Tibetans, the practice of a mixed agriculture pastoralist economy has a very long history and many of the ecologies in the Sino-Tibetan area are actually suitable for this mixed economy. Whereas in the Sino-Tibetan case, you have a very distinct boundary of where pastoralism is possible—which is outside the matriliny belt—and within the matriliny belt because of the tsetse fly where cattle pastoralism is quite impossible. So there's a more sharp divide between the forms of the fitness.... subsistence that groups can adopt.

21:09
So how do we use phylogenetic comparative method to study the extinction process of cultural populations? This is actually a very difficult question to answer using empirical data. Because unlike the biological evolution process, in cultural macroevolution, we have very little material fossils of the historical and extinct languages. And without these empirical evidence in place the inferences—using some of the very complicated model developed in in biology—tend to have really high uncertainty and really cannot tell us a precise answer that we're looking for.

22:00
So in evolutionary genetics, there have been a lot of papers published, finding various phenotypes or genotypes related with accelerated extinction, or promoted diversification of species, using these speciation extinction models. Such as the BiSSE, which is the binary state SSE model. There are also there are other extension of the models applied mainly in biological studies. And it's actually.... Well, it seems pretty straightforward. If we want to study the extinction of cultural populations, we should just borrow the method from biology. But actually, in practice, it is quite difficult because this model is not robust. Using... if the sample size of the comparative data set is small. And in most cases of comparative cultural dataset, this sample size is quite limited.

23:04
And another paper—Louca and Pennell—published last year, in 2020 actually, used simulations... simulation studies to prove that when we do not have any empirical calibrations on the tree, the... a given phylogeny can be reconstructed with many possible combinations of diversification and extinction rates. And this is why we cannot estimate extinction rates alone using... cultural phylogeny this is exacerbated by the fact also we do not have the company sampling of all the taxa in the financial means that mean no extinct taxa. So these are the kind of the challenges going ahead. If you want to study cultural extinction on the phylogeny we have to overcome—even with perhaps new methods—or more suitable for cultural datasets, or some other method that will be appropriate.

24:19
So the case study, I'll give for... for looking at cultural extinction on a phylogeny is really a very unique case where we do have a lot of empirical data on the extinct groups. And we do have a fairly, fairly complete sampling of most of the taxa, including extinct taxa on this phylogeny. So my collaborator, Kiran Basava, reconstructed the phylogeny of historical Islamic sects, in this case, these are sects that were formed from 600AD to 1900 and does not include any contemporary Islamic groups. Because they don't necessarily always have a coherent ideology according to the way we categorise the evolution of these religious groups, on our phylogeny. And it is by no means the only way to represent the evolution of Islam, just one of the many possible way and suitable for looking at the traits that we study in this paper.

25:33
So we coded for afterlife belief, and various forms of religious violence in these religious groups. And I looked at whether the presence or absence of these beliefs or violence will predict the duration or the survival of these religious groups. So in this case, our phylogeny, we do have the data from historical records, we do know many of the extinct groups, the exact timing of extinction, exact timing, when they branch out from their ancestral group. So this allows us to, to investigate this hypothesis.

26:24
With fairly complete empirical data, we use the survival analysis, the Cox regression model, to study the duration whi... which on the tree is represented as the length of the tree branches. Any of... so in this case, if it's leading to an extinct group, that will be an extinction event. Well, if it bifurcates, or it leads to a contemporary tip, where our sampling ended, that will be counted as a censoring event.

27:01
So, we found there is significant evidence that apocalyptic beliefs, religious violence, and revolutionary violence is correlated with the duration of the religious group. But there is an issue of the confounding effect of the phylogenetic history. So, if you recall two slides before, sometimes we have a clade of religious groups that all have apocalyptic beliefs that or went extinct fairly quickly. So, it might not be the afterlife beliefs itself causing extinction, but something else that was passed on inherited by offspring groups.

27:51
So, after controlling for the phylogenetic associations, we run the test again. In this case, after the ancestral relationship is controlled, only apocalyptic beliefs still show some correlation with the duration of the group, but this does not necessarily imply the presence of apocalyptic beliefs will be causal to a shorter duration or faster extinction of these groups. It could also be a circumstance that were inherited by this group, which facilitated the adoption of apocalyptic belief, that also facilitate faster extinction. That is a question that cannot be answered explicitly with the survival analysis.

28:50
So, I gave some very brief summaries of some of the case studies where we can look at diversification and extinction of cultural macroevolution. Some take home messages will be: (i) cultural phylogeny is useful a useful tool to examine the mechanism of cultural changes (diversification/extinction) given the observed diversity at present, (ii) and a unit of analysis will be populations or the cultural traits that are inherited and passed on from a parent to offspring, (iii) a cultural population and empirical data are absolutely essential for a robust estimate from these influences. As in most cases we do not have the complete stamping of the taxa. And we we do not have a full knowledge of the historical and extent cultural form and therefore, we should caution fitting, very complicated evolutionary models with restricted, smaller, comparative cultural datasets.

30:15
So, that will be all if you have any questions, feel free to reach me on Twitter or my email. I would like to thank my mentors and collaborators, Professor Ruth Mace at UCL Professor Mark Pagel and his colleagues and the University of Reading, Dr. Ting Ji at The Chinese Indian Institute of Science, and Kiran Basava, at the University of Oxford, and thank you for listening.