Thread by @fusaroli

TLDR: we applied cognitive science & cultural evolution to investigate some of the earliest human engravings (100k year old), finding that they were likely used to express implicit style and human intent. Background, pics and nerdy methodological observations in the thread. 1/n

https://twitter.com/PNASNews/status/1229858205183901696

After a long journey "The evolution of human symbolic behavior in Homo sapiens" is finally out in PNAS. The international interdisciplinary group included
@kristian_tylen, @johannsen_niels, @ARCHAEOfelix, @katheimann, @Nicolas_Fay, @SergioGdlR and Marlize Lombard. 2/n

Here a picture of (most of) the team getting familiar with an ancient (200k) ochre production site. 3/n

The archeologists brought to the group's attention the existence of two South African sites where engravings had been found, the oldest ones being 100k year old and spanning more than 30k years. In one site the engravings were on ochre, in the other on ostrich shells. 4/n

These unusual and exciting longitudinal corpora had been reported in a series of really high impact papers, soundly focused on the dating and description of the engravings, but often reporting brief (very speculative) statements as to their meaning. 5/n

The engravings are variously argued to be serving aesthetic functions (decorations?), or to express group-specific specific styles and cultural tradition; or again to be denotational systems and even precursors of language. 6/n

As semioticians and cognitive scientists we were intrigued. The puzzle sounded part Dan Brown, part Indiana Jones. We started wondering how to test these hypotheses and we relied on cultural evolution ideas, given the longitudinal nature of the corpora. 7/n

If the engravings are used for one or more functions, over time they will be selected and improved to fulfil those functions best, just like more physical tools are 8/n

Further, the presence of two independent sites allowed us to check whether we would see similar trends across sites (which would then yield a tad more generalisability to our findings). 9/n

H1) If the engravings have aesthetic purposes (e.g. decorations), later engravings will be more perceptually salient and recognisable as intentionally made than earlier ones.
10/n

H2) If the engravings have a social purpose, as implicit or explicit markers of group identity (this is how we do things here), over time they will become easier to remember, reproduce and recognize as belonging to one of the two sites. 11/n

H3) If the engravings have a denotational purpose (indicating specific referents), they should become easier to discriminate within site (assuming they refer to different referents). 12/n

We used both the original engravings, and more controlled stimuli composed each by 6 identical lines inspired by the original ones to reduce visual confounds due to the materials on which the engravings were made, the number of lines, etc.

13/n

We found full support for H1 (aesthetic function) and support for H2 (the engravings being implicit markers of group identity, albeit not explicit ones). No support for H3 (denotational function), but this might also be due to the sparsity of the findings). 14/n

Read the paper for the experimental details (which were in large part masterfully designed by @kristian_tylen).
15/n

This is exciting work merging complementary archeological, semiotic, cognitive and cultural evolution competences in an innovative fashion, and we hope it will pave the way for more work on these lines. 16/n

The study has also a number of shortcomings. We used WEIRD undergrads to start making inferences about ancient minds (but we needed to start somewhere). 17/n

We isolated 3 functions hypothesised in the literature and operationalised them in 5 experiments, but other functions and other operationalizations could be made (and we'd love to see them made). 18/n

So we would love more people jumping into the field and we made all our materials, experimental paradigms and analysis scripts available on OSF: osf.io/rbtk4/
19/n

From a statistical perspective it was a fun paper. The first where I got to implement a full Bayesian workflow (using @mcmc_stan and #brms), designing bespoken models, testing their predictions (prior predictive checks) and obsessively evaluating and improving their quality. 20/n

As any paper, now there are a few things I would want to do differently. We modelled the engravings' age in term of within-site "early", "intermediate" and "late" ones. This sidestepped the many controversies on the dating. 21/n

However, it would have been fun to try to model the actual datings (and uncertainties surrounding them). It would also have been fun to develop and include more explicit causal inference reflections and DAGs. 22/n

Identifying plausible observed and unobserved causal parameters would have enabled us to set up simulations and assess whether our model could actually retrieve the right estimates for simulated data. 23/n

All in all one of the best scientific challenges I've participated in. 10/10 would/will do it again.

This would not have happened w/o the support and inspiration of @interact_minds, and @aroepstorff. Also special thanks to @mcxfrank, @judyefan, @hawkrobe, Bruno Galantucci and the many others who discussed w us early versions of the study. 25/n

And keep an eye on this space, more will happen on these themes.