Nijmegen Lectures - Day 3: The evolution of culture without miracles

Part of the Nijmegen Lectures 2014, given by Professor Russell Gray. See previous post. 

Bio-cultural analogy

On the third and final day of the Nijmegen Lectures, Russell Gray expanded on the evolution of culture ‘without miracles’. There has been a debate for some time whether cultural evolution can be likened to biological or genetic evolution. Darwin himself had this to say about it:

"The formation of different languages and of distinct species, and the proofs that both have been developed through a gradual process, are curiously parallel." (Darwin, 1871)                                         

 A few of the major arguments against the analogy are:

• Variation in cultural ideas is not as random (as genetic mutations), but rather intentionally designed
• Cultural features have low fidelity transmission, including:
• Information loss
• Vertical AND horizontal transmission
• Blending (copying one ancestor, like genes, but also copying many ancestors, unlike genes)

People from a tribe in Northern Thailand (August 2008)

Transmission fidelity

Russell discussed these points briefly in his lecture. He brought forward some ideas from Dan Sperber's 'Explaining Culture' (book) to show that perhaps we should not see cultural variance as being replicated or copied, but rather as interpreted, or reconstructed. When you reconstruct, you can take a lot of variants you come across, and then make your own out of that set (blending) - this suggests a difference from high fidelity copying. Similarly, when we tell stories (a cultural feature), we may not even always tell the same story ourselves. Still, Russell argues, such cultural attractors create discrete units of culture that are stable despite noisy transmission.

One way such units of culture may stabilize could be on the population level: Henrich and Boyd (2002) have argued that accurate replication need not occur at the level of cultural innovations, but rather that conformist bias can stabilize these units at the population level. Prestige bias is another form of such stabilization, whereby the most successful cultural features are blended, resulting in adaptive evolution. In this way transmission fidelity can be very low at the individual level but high at the population level.

Horizontal transmission / hybridization

Another issue is that cultural trees don't just diverge, but syncretize and form anastomoses too (when two branches of a tree come back together again).  In this sense, the tree of culture is a ramification of assimilations and acculturations (Kroeber, 1948). This led Stephen Jay Gould (1987) to oppose the bio-cult evolution analogy; he argued that biological lineages become separate, and don't amalgamate (except through rare hybridization), while cultural evolution has rampant transmission and anastomosis. Russell argued that this judgement was made too fast: while Gould worked on vertebrates, the evolutionary trees of virus are a lot 'messier' and can include horizontal transmission as well (and may be more similar to cultural evolution). 

But most of all we need methods to test these different ideas! We need to develop data and methods to test whether there is cross-lineage transfer, what the fidelity of cultural features is, etc. 

It seems that some cultural aspects show high fidelity transmission - there are certain words in languages in the Pacific that after 446,000 years of cultural evolution are still very similar. That is striking. It may be that some aspects are transmitted with high fidelity (traditional decorations, for example) and others via horizontal transmission (iPods, for example). How can we look at this? Russell suggested that rather than looking at these aspects by means of evolutionary trees or 'waves', perhaps we should use networks (to capture the combination of horizontal and vertical transmission). One way to do this is via NeighbourNet, developed by Bryant and Moulton (2002). 

Network analyses

This method was developed for biological data, to understand bacterial and viral lineage conflicts. It allows one to visualize whether there is strong historical lineage or more conflicting signal. When there is a good match, the data looks tree-like, but when there are conflicting signals, the figure looks more like a network, or a spider web. In this way, hybrid origins of languages can for example also be visualized, which shows up as double lines grouping a language with two sets of languages (parallel lines represent spliuts, and the line length represents the split weight) - see Gray, Greenhill and Bryant (2010). Such a method may prove very useful for the science of cultural evolution, and even the story of Little Red Riding Hood has been analysed with it so far (Tehrani, 2013 - open access article).

So, to round off the final talk of the Nijmegen Lectures, there is some evidence that cultural evolution may be likened to biological evolution - in some case more like the biological evolution of bacterias and viruses. In either case, we need methods to test such ideas directly, and network analyses may be helpful to this regard.

I hope you have enjoyed reading about the Nijmegen Lectures, I certainly enjoyed being there! /V.

P.S. For those readers who have noticed the tags used for this post - what's the difference between 'evolution of culture' and 'cultural evolution'? Merely in terms of blog tags, I am trying to distinguish between discussing (specific) examples of cultural features and their evolution ('evolution of culture') and the discussion of cultural vs biological evolution ('cultural evolution'). Cheers!

Selected references

- Boyd R., Borgerhoff Mulder M., Durham W. H., Richerson P. J. 1997. Are cultural phylogenies possible? In Human by nature, between biology and the social sciences (eds Weingart P., Richerson P. J., Mitchell S. D., Maasen S., editors. ), pp. 355–386, Mahwah, NJ: Lawrence Erlbaum Associates.
- Bryant, D., & Moulton, V. (2002). NeighborNet: An agglomerative method for the construction of planar phylogenetic networks. In Algorithms in Bioinformatics (pp. 375-391). Springer Berlin Heidelberg.
- Darwin, C. (1871). The Descent of Man. D. Appleton and Company.

- Gray, R.D., Greenhill, S.J., and Bryant, D. (2010). On the shape and fabric of human history. Philosophical Transactions of the Royal Society London, B, 365, 3923-3933.

- Henrich, J. & R. Boyd (2002) On Modeling Cognition and Culture: Why replicators are not necessary for cultural evolution. Journal of Cognition and Culture 2(2): 87-112. 
- Pagel, M., & Mace, R. (2004). The cultural wealth of nations. Nature, 428(6980), 275-278.

- Sperber, Dan. (1996) Explaining Culture: A Naturalistic Approach. Blackwell Publishers.
- Tehrani, J. J. (2013). The Phylogeny of Little Red Riding Hood. PloS One, 8(11), e78871.

Darwinizing culture (video)

On the third day of the Nijmegen Lectures, Russell Gray talked about the evolution of culture (including language). A highly informative video that covers some similar material, can be found through UCLA (Youtube), posted below. If you are interested on how phylogenetic methods can be applied to cultural features, I recommend you watch this video. Enjoy!

For the discussion afterwards, go here.

Nijmegen Lectures - Day 2: The evolution of language without miracles

Part of the Nijmegen Lectures 2014, given by Professor Russell Gray. See previous post. 

On the second day of the Nijmegen Lectures, Russell Gray talked about the evolution of language ‘without miracles’. He started with ubiquitous tension in language research between unity and variety, between similarity and diversity. While languages have striking similarities, the diversity of languages is also impressive: it is typically estimating that there are over 7,000 languages in the world, with some areas of the world having a particularly dense variety (i.e., 500 languages in Vanuatu alone).

image source

Interestingly, in linguistics, people have typically tried to focus on the similarities: while Chomsky postulated a ‘universal grammar’ applying to all langauges, Greenberg tried to find probabilistic constraints. The idea behind these set-ups was that, despite the seemingly large variety at the surface, underlying on these so-called 'language phenotypes' are very similar 'language genotypes', constrained by the innate structure of the mind.

In (evolutionary) biology on the other hand, people have generally emphasized the variety of species rather than similarities between them. Indeed, since Darwin, the variation is considered to be the central reality, not an irrelevant disturbance to take away from a sort of underlying Platonic form (Lewinton, 1983). According to Darwin, if there is any unity of type, this is to be explained by unity of descent.

Alright, but what does this tell us about the evolution of language? Well if there is one principle underlying all languages, then one mutation could perhaps suffice to give language to all humans (as is argued, if somewhat simplified, by Chomsky). If, on the other hand, you suggest that languages evolved in incremental steps, this provides more leeway to arrive at the languages seen today.

However, one of the problems with comparing languages and tracking their evolution is that languages are not independent. This is typically referred to as Galton’s problem: Cultures are not statistically independent, their features are not independent data points. Thus, the history of the languages matters, and sampling from a wide variety of language families doesn’t solve the problem (because even apparently shared dependencies can have different historical pathways and vice versa). 

In order to get around this issue, Dunn, Greenhill, Levinson and Gray have applied phylogenetic methods (from evolutionary biology) to word-order in languages around the world. By fitting trees probabilistically on phylogenies, they were able to test whether instances of language evolution were correlated or uncorrelated. Through this research, they show that functional dependencies between word-order traits are lineage-specific (i.e., to what 'family' the language belongs) rather than universal tendencies, as argued by Greenberg. This means that word-order is determined through cultural evolution rather than being driven by innate constraints of the mind. 

Another example of how cultural evolution may drive language evolution is the diversification of Austroasiatic languages (Gray, Drummond and Greenhill, 2009). Russell and colleagues used Bayesian phylogenetic methods to construct a phylogeny of 400 languages on the basis of basic vocabulary. Basic vocabulary is used because it typically evolves slowly and is less likely to be borrowed. Then, given their priors and the model, they tested whether which of two prior hypotheses (the 'pulse-pause' model or the 'slow-boat' model, see paper) was most likely to be true. They found that Austroasiatic languges change through time, though through a constant diversification, but rather by an initial burst of diversification which then plateaus out. Then, after this 'pause', there is another increase in diversification, etc. This model fits best with the 'pulse-pause' model, which was also correct about the origin (Taiwan) and trajectory (via Philippines, Indonesia, and so on) of this language diversification. This rises the question however, what is driving the increases of diversification? One hypothesis is that it may follow the development of rice cultivation: When rice cultivation becomes more serieous and systematic, people are tied to a place, and the population can increase, which leads to a density increase, and possibly an increase in diversification of language. 

In sum, I believe that Russell and colleagues have convincingly shown that phylogenetic methods can be applied to cultural phenomena as language. Though reported somewhat simplified here, I encourage people to read Gray, Drummond and Greenhill (2009) and Gray & Atkinson (2003) to learn more about these exciting studies (and to learn more about validation techniques and robustness checks they employed).

For future research of the evolution of language then, we need to figure out how language can come about from a continuous system, through incremental steps in the brain. Thus, we need a better linguistic theory so that we know what to look for in the brain. Important is that we start appreciate the wide variety of languages in the world again, rather than reducing all of these to a variation on, randomly, English.

Selected references

- Chomsky, N. (2010). Some simple evo devo theses: how true might they be for language? The evolution of human language (eds. Larson R. K., Deprez V., Yamakido H., editors. Cambridge:Cambridge University Press), 45-62.

- Dunn, M., Greenhill, S.J., Levinson, S.C. and Gray, R.D. (2011). Evolved structure of language shows lineage-specific trends in word-order universals. Nature, 473, 79–82. doi:10.1038/nature09923

- Gray, R. D., Drummond, A. J., & Greenhill, S. J. (2009). Language phylogenies reveal expansion pulses and pauses in Pacific settlement. Science, 323(5913), 479-483.

- Levinson, S.C. and Gray, R.D. (2012). Tools from evolutionary biology shed new light on the diversification of languages. Trends in Cognitive Sciences, 16 (3), 167-173 doi: 10.1016/j.tics.2012.01.007

Further reading

- Gray, R. D., & Atkinson, Q. D. (2003). Language-tree divergence times support the Anatolian theory of Indo-European origin. Nature, 426(6965), 435-439.