Search

Homo sapiens’ capacity for the inter-generational accumulation of complex technologies, practices, and beliefs is central to contemporary accounts of human distinctiveness. However, the actual antiquity and evolutionary origins of cumulative culture are not known. Here we propose and exemplify a research program for studying the origins of cumulative culture using archaeological evidence. Our step-wise approach disentangles assessment of the observed fidelity of behavior reproduction from inferences regarding required learning mechanisms (e.g. teaching, imitation) and the explanation of larger-scale patterns of change. It is empirically grounded in technological analysis of artifact assemblages using well-validated experimental models. We demonstrate with a case study using a tool-making replication experiment to assess evidence of behavior copying across three 2.6 million-year-old Oldowan sites from Gona, Ethiopia. Results fail to reveal any effects of raw material size, shape, quality, or reduction intensity that could explain the observed details of inter-site technological variation in terms of individual learning across different local conditions. This supports the view that relatively detailed copying of tool-making methods was already a feature of Oldowan technological reproduction ~2.6 mya. We conclude with a discussion of prospects and implications for further research on the evolution of human cumulative culture.

Investments in stone tool skill learning constitute an oft-neglected cost in the evaluation of alternative adaptive strategies. Studies show that the most salient neural and cognitive demands of stone tool-making should occur during learning rather than expert performance. However, the behavioral complexity and extensive learning requirements that make stone knapping skill acquisition an interesting object of study are the very features that make it so challenging to reconstruct. Here we present results from a new study of Late Acheulean handaxe-making skill acquisition involving twenty-six naïve subjects and up to 90 hours training over several months, accompanied by psychometric testing and regular behavioral assessments. Our study’s main objectives were to derive a robust quantitative skill metric for the experimental handaxes using machine learning algorithms, to reconstruct a handaxe-making learning curve, and to explore sources of variation in handaxe making and learning outcomes. The objective, artifact-based model performed well in predicting handaxe making skill levels, in forecasting future performance and rating out-of-sample expert handaxes. The overall learning curve for subjects follows a characteristic power law showing that with training practice performance generally improves. Extrapolations of our individual learning curves suggest that experimental studies of such refined technologies should plan for at least 200 hrs. of training to achieve archaeological validity. We also find that performance in early stages of training correlates with differences in executive function (as tracked by widely-used psychometric measures) whereas performance during subsequent learning was associated with practice density. Our results highlight the significance of technologically focused teaching for achieving high-level skill in handaxe production, the importance of motivation, persistence, and self-control in knapping skill acquisition, and how these processes can be reliably reconstructed from archaeological evidence.