This short paper details two experiments conducted by Joseph Henrich and his colleagues which ‘purportedly’ explores the importance of sociality for successful cultural transmission. The aim of these lab based studies is to provide another piece of convergent evidence for work in theoretical biology using modelling techniques as well as ethnographic data sets and palaeoanthropological findings. The wider claim here is that the size of a social group and the degree of interconnectedness between members of the group is the crucial factor for accumulative cultural evolution – a feature that has been designated as unique hallmark of our species by Michael Tomasello (1999 [This claim is moderated that by the fact that there is some evidence of accumulative culture in other species – for instance, Andrew Whiten and colleagues have recently claimed that there is some accumulative culture in chimpanzees – but even then it does not occur to such an extreme scale in humans for which it is abundant. Also see Kim Sterelny 2003, ch8]). And Henrich has claimed in a recent book, Secrets of our Success (2016), that it is our extreme pro-sociality that is to account for the fact that our species occupies almost every terrestrial habitat (also see Sterelny 2003, ch7). In week 10 we discussed his work with Richerson and Boyd (2011) that outlined this in greater detail – and how they contrast with the views of evolutionary psychologists such as Steven Pinker who place a greater emphasis on human intelligence. Boyd and colleagues (2011; also see Henrich 2016) argue against this view by pointing to a series of natural experiments: e.g. lost European explorers; fortuitous inventions through recombination of cultural knowledge rather than de novo and solo creativity; and ecological disasters which eliminate cultural knowledge (see week 10 for more details).
I will first describe the two experiments and then discuss some general limitations that were raised by our group discussion.
In both experiments one hundred WEIRD participants were put in to one of two groups (A and B) and then placed in to one of ten ‘generations’ comprised of five people. In group A members of the next ‘generation’ had access to only one ‘cultural parent’ whereas in group B agents had access to every member of the proceeding generation. This experimental set up is nicely shown in the diagram below (taken from Muthukirshna et al 2014, p. 2):
Participants were tasked with copying a ‘cultural skill’ (either using a piece of graphic design software or tying a knot from rock-climbing) as accurately as possible and then generating instructions (either written or a short instructional video) which were passed on to the next generation. Participants only had a limited amount of time to perform the skill and then create the package that was then passed on to their descendants. Participants were rewarded for both their own performance and those of the following generation. However, it was raised in the group discussion that the motivational component here was a complicating factor. And that it would have been better if participants were rewarded for the pay-off of the cultural skill they were copying and transmitting. Interestingly, another lab based experiment on the importance of sociality for cultural evolution dissolves these tensions. In a computer game trial Derex and colleagues (2013) motivated their participants by getting them to collect in game resources in order to maintain their health. Participants were able to do this using either a simple or complex cultural package to achieve this (and they played in groups ranging from 2 to 16). Their results were similar to those discussed here – that larger groups outperform smaller groups.
Experiment 1: in the first experiment the aim was to create an image using a piece of software. Participants, who were all novices, were given written instructions by the previous generation as well as the image the previous generation created. However, as one can see in the diagram below (adapted from Muthukrishna et al 2014, p. 4), they presumably also had access to the initial target image as some sort of Platonic ideal – because how else are we to explain how there could be cultural transmission in group A when the previous generation failed badly or utterly? I think this is a serious issue for the experiment because if one is trying to test genuine cultural transmission then these events should simply count as an extinction event (i.e. that the skill is lost). Indeed, the very existence of these sorts of events would support the conclusion that group size is vital to the preservation cultural skills.
One can see that in group A there are multiple instances where there is a complete failure to make the image (e.g. generations 3, 5, 6 and 10) and many other instances in which performance is so poor that it is extremely odd that performance increases so radically in the next generation if all they had access to was the previous generation’s attempt . For instance, generations 2, 4, 5, 6, 7 and 9 all contain very poor images that are then followed by marked improvements. If the transmission in these channels was genuinely only from one model to the next then these moments would have led to extinction events without some form of Platonic ideal (as is often the case when cultural skills are genuinely lost ‘in the wild’). However, despite this criticism the above diagram does also show that performance in group B is far superior and more robust – so much so that by the tenth generation Muthukrishna and colleagues comment that every single member of group B is superior to group A (and this matches the findings of Derex et al 2013).
Experiment 2: although this experiment had much the same results as the previous experiment there were some important differences in the experimental design:
 The ‘cultural skill’ to be transmitted was a special knot from rock climbing.
 The first generation of cultural parents were composed of expert rock climbers and all other generations were composed of novices. As such, there was an instant decrease in the proficiency of how well the skill was performed from generation two onwards. However, this decrease had stabilised by the tenth generation – albeit that it was at a much lower level of proficiency in group A compared to group B. And there was also a continuing steady but slow loss of enskillment in group A.
 Transmission of the skill was by the making of a video rather than written instructions. This is particularly interesting because the time limits imposed by the experimenters prevented members of group B from actually being able to view all videos made by the previous generation. As such, even though they ‘theoretically’ had access to five models in principle, there was not actually not sufficient time for them to access this information in practice. The joke was made in our group discussion that this was equivalent to the USA’s policy on universal healthcare where in principle all citizens have access to healthcare but in practice only a much smaller portion can actually afford it. Even more troubling was how agents went about choosing the one video they had time to view. Rather than making a choice solely by themselves – which would then be a measurement of how individuals utilise prestige bias (where humans preferentially copy those who are successful at a task – and indeed this copying can carry over into other unrelated domains, as is seen in advertising) – they actually just relied on test scores that the experimenters had given to the previous generation. As such, this just measures the extent to which humans are prepared to rely on expertise of others. Whilst one might argue that this still measures prestige to some extent, I find it troubling that there is this additional factor introduced since it complicates the results.
General comments: Our group discussion also raised a number of other issues and queries. But first I would like to note that it was recognised that many of these criticisms were made in light of the observation that this is a lab based study. Lab based studies need to be designed to control as many variables so as to either eliminate them or reduce them sufficiently so that one can manipulate a target variable and get a genuine measurement that is not complicated by too many other factors. That being said, it is slightly worrying that this experiment was designed to explore sociality but did not include any actual proper interactions between the agents. This raised the question of what is meant here by sociality: in the introduction to the paper Muthukrishna and colleagues seem to define sociality as involving group size and group interconnectedness. This then reduces sociality down to how many other agents one can communicate with. But is this what we want to mean by sociality? – especially given that this factor is being heralded for our tremendous success as a species and the basis for accumulative cultural evolution which then separates us from the other primates. I think that the mediated interactions in experiments 1 and 2 – through written instructions and video – sever what is actually really interesting about human sociality. Another experiment by Dean and colleagues (2012) that I have discussed previously shows this nicely. They compared the performance of groups of macaque monkeys, chimpanzees, and five-year-old children on a three stage puzzle box (a standard experimental set up in primate studies). Despite the groups of children having exceedingly far less time to solve the puzzle box they generally outperformed the other primate groups. But what is most interesting is that Dean and colleagues also counted numerous acts of cooperation amongst the children compared to none in the other primate groups. And this also included spontaneous teaching – where one child, having completed the challenge themselves, would demonstrate how to do it to another child. I think this shows another element that is severely missing in this experiment – the importance of peer-to-peer interactions in cultural learning (see Sterelny 2012).
Another related issue raised in the group discussion was the general lack of a discussion of the developmental niche – the scaffolded environment which human novices inhabit and which improves and constrains trial-and error learning so that it becomes more efficient. One could counter-argue that the written instructions and videos could be counted as scaffolds, but I think this would be to give a severely limited notion to just how much the developmental trajectories are shaped and sculpted by the previous generation (see Sterelny 2003, 2012). However, one could again argue in defence of Muthukrishna and colleagues that they are precisely attempting to gain insights into the earliest stages of the cultural niche when cultural transmission was not perhaps reliable nor high fidelity. Instead, their aim is to show that there can be accumulative cultural evolution even in conditions of low fidelity transmission as long as their is sufficient number of learners (size) and that they are sufficiently interconnected. Henrich captures this sentiment with a just-so-story (2016, ch12). Imagine two extreme populations, one composed of geniuses who are antisocial, and another that is composed of dolts who are very social. Henrich labels these populations ‘geniuses’ and ‘butterflies’ and argues that no matter how brilliant the geniuses are, they will fail to retain any of the amazing inventions and innovations that they create. In contrast, although the butterflies will make good innovations very rarely these will always be retained. And just through the shear number of social interactions that they engage in novel innovations will occur through recombinations and errors (as well as through occasional insight).
In discussing accumulative cultural evolution Sterelny (2003, 2012) not only discusses the importance of the cultural niche but also the internal cognitive adaptations that must have necessarily co-evolved with the cultural niche and which prime agents for cultural learning. Several members of the group felt that this paper overlooked these factors, and the experimental design also occluded them in the manner in which it cleaved generations from one another. (Also see week 6 for more on this topic)
A last point of consideration was about whether one could take this experimental design to look at cognitive skills – e.g. memory, social cognition, etc. The issue here is that the focus of the experiment is on artefacts, but how would we examine more nebulous or intractable topics such as beliefs, norms, and techniques, etc.? One suggestion was the teaching of an algorithmic procedure in mathematics. Much of our previous weeks’ discussions have been taken up with the general question of how to make the concept of culture empirically tractable (e.g. see week 2). So this is perhaps asking for quite a lot, but it is also a pressing issue. Another suggestion is that the field of cognitive archaeology is already engaged in trying to draw out conclusions of the cognitive states of our ancestors from material traces (e.g. Malafouris 2008). Furthermore, the more recent work of cognitive psychologist Don Norman (1988) and others (e.g. Zhang & Norman 1994) is precisely designed to make inferences about the human mind from how we use various technologies and artefacts – and to then redesign these so as to improve their functionality, given our primate neuro-cognitive profile, so as to enhance our performance – even enabling us to achieve previously impossible or infeasible cognitive goals (Menary & Gillett 2017).
Lastly, despite these criticisms, I do think that this is a good lab based experiment. It nicely supplements a range of other convergent pieces of evidence to support the notion that sociality (particularly the size of the social group and how interconnected they are) is crucial for an accumulative cultural niche; and that this is a definitive feature of our success as a species.
Boyd et al (2011) The cultural niche: Why social learning is essential for human adaptation. PNAS
Derex et al (2013) Experimental evidence for the influence of group size
on cultural complexity. Nature 503, 389–391.
Henrich (2016) The Secrets of Our Success. Princeton University Press.
Malafouris (2008) Beads for a Plastic Mind: the ‘Blind Man’s Stick’ (BMS) Hypothesis and the Active Nature of Material Culture. Cambridge Archaeological Journal 18 (3), 401-414.
Menary & Gillett 2017 Embodying Culture: Integrated Cognitive Systems and Cultural Evolution. (pp. 72-87) in Kiverstein (ed.) The Routledge Handbook of Philosophy of the Social Mind. New York: Routledge.
Muthukrishna et al (2014) Sociality influences cultural complexity. Proceedings of the Royal Society: B 281 (1774), 2511.
Norman (1988) The Psychology of Everyday Things. [Reprinted as The Design of Everyday Things (2002)]. New York: Double Day Books.
Sterelny (2003) Thought in a Hostile World: The Evolution of Human Cognition. Oxford: Blackwell Publishing.
Sterelny (2012) The Evolved Apprentice: How evolution made humans unique. Cambridge, MA: MIT Press.
Tomasello (1999) The Cultural Origins of Human Cognition. London: Harvard University Press.
Zhang & Norman (1994) Representations in Distributed Cognitive Tasks. Cognitive Science, 18, 87-122.
There was an extended discussion about the loss of cultural achievements. In regards to this, here is a recent Australian documentary charting one Indigenous lady’s attempt to revive and preserve specialist canoe making in the Northern Territory.