The evolution of human technology has allowed us to spread over the Earth and beyond. A fascinating question is to understand the cognitive origins of this phenomenon, which is known as cumulative technological culture. Some have argued that humans are unique social learners, who can reproduce and even optimize a technology without us understanding how it works. In April 2019, Maxime Derex, Jean-François Bonnefon, Robert Boyd, and Alex Mesoudi published a Nature Human Behaviour article, which provided evidence for this view.

In this article, Derex and colleagues presented an elegant micro-society experiment, in which participants had to minimize the time it took a wheel to cover 1 m on an inclined track. The wheel had four radial spokes. On each spoke, a weight could be moved closer or further from the hub of the wheel. The participants performed the task as members of chains of 5 participants, each participant being considered as a “generation”. All the participants performed five trials in which they tried to increase the speed of the wheel by moving the weights along the spokes. There were two conditions. In the Configurations condition, the participants were provided with the last two weight configurations and the associated speeds of the previous participant in the chain. In the Configurations+theory condition, the participants could also generate a written theory about the wheel system and transmit it to the next participant in the chain. After the five trials, the participants had to complete a test measuring their understanding of how the system worked. The results indicated that the wheel speed became progressively optimized, while the participants’ understanding of the system did not improve over generations. This finding confirmed that technologies can be optimized without us understanding how they work, but simply through the retention of small improvements over generations.

In April 2019, my colleagues and I were conducting a series of fMRI sessions at the CERMEP, which is the neuroimaging department of Lyon, France. I remember reading the title of the paper while waiting for a participant to end her session. Emanuelle Reynaud was in the room with me. I quickly asked her to have a look at the paper. Indeed, we were writing a Behavioral and Brain Sciences article on the topic, in which we had gathered evidence that cumulative technological culture could be strongly driven by the ability to understand our physical world. Our first impression was that Derex and colleagues had a smart idea of designing a micro-society experiment in which participants had several trials to improve the technology and not only a single one as in most of the studies published on the topic. Our second impression was that this finding was contradictory with our view. This was not very encouraging for our Behavioral and Brain Science article – which is now fortunately published (see here).

A couple of weeks later, we discussed again about the paper and noticed that Derex and colleagues had made a specific choice concerning their understanding test. For each of the 10 items of this test, the participants had to choose which of two wheel configurations was faster. There were three options: ‘Wheel 1’, ‘Wheel 2’, and ‘No difference’. The ‘no difference’ option was always incorrect. We were not sure, but we considered that their participants could have generated an equiprobable representation of the answer distribution. Thus, this methodological choice could have led them to have about 33% of incorrect responses in choosing the ‘no difference’ option. If confirmed, this could have partly biased the results obtained by Derex and colleagues. We contacted a colleague, who was expert in both psychology and mathematics, and we submitted the problem to him. He confirmed our interpretation. So, we decided to conduct a replication/extension of Derex and colleagues’ study to explore the impact of this potential methodological bias.

Derex and colleagues had provided a comprehensive and documented description of their physical system. All of us also knew that Joël Brogniart possessed remarkable technical-reasoning skills, as confirmed by his almost perfect score on the psycho-technical test we frequently use to assess technical-reasoning skills. Thus, in November 2019, Joël reproduced the physical system with the exception of the axis of wheel, which was a steel pole and not a wooden pole. Several brainstorming sessions also led us to device two understanding test of 24 items and 4 options (without ‘no difference’ option). We also decided to include a control group, who would not experience the system. Data collection started in February 2020 and lasted one month. Contrary to Derex and colleagues, we found that the improvement of the wheel system over generations was accompanied by an increased understanding of it. We wrote the article and submitted it in April 2020.

The rest of the story is told in the published version of our Nature Human Behavior article entitled “Technical reasoning is important for cumulative technological culture” as well as in the 177 pages (!) of the Peer Reviewer Information file attached to this article.

This article is the opportunity to rehabilitate technical-reasoning skills, which are often underestimated not only in education but also in the socio-economic world. Yet, they might be of great importance for making social learning a salient source of technical inspiration, thereby allowing us to build our technological modern society.