Skip to main content

The Cog in the Ratchet: Illuminating the Cognitive Mechanisms Generating Human Cumulative Culture

Periodic Reporting for period 3 - RATCHETCOG (The Cog in the Ratchet: Illuminating the Cognitive Mechanisms Generating Human Cumulative Culture)

Reporting period: 2018-09-01 to 2020-02-29

In human populations, skills and knowledge accumulate over generations, giving rise to behaviours and technologies far more complex than any single individual could achieve alone. This ratchet-like property of human culture appears absent in nonhuman species, as socially transmitted behaviours in animal populations are generally no more complex than those that can be acquired by trial and error. Scientists from a wide range of disciplines have offered high-profile speculative theories about the underlying differences that might be responsible for this striking evolutionary discontinuity, but adequate empirical evidence is still lacking. In the RATCHETCOG project, we aim to implement a comprehensive systematic investigation into cumulative cultural evolution, using experimental methods that offer sufficient flexibility to generate valid comparisons across three critical research domains: species differences across the primate family tree; age differences over human development; and learning condition differences in groups of adult human participants. The methods that we have devised within the project have allowed us to both measure and manipulate the complexity of transmitted behaviours, and as such they provide a tool for analysing the extent of ratcheting under different conditions and across different populations. Each of the three research strands provides a vital source of evidence. Studies of nonhuman primates will reveal the limits on learning in these species, and studies with children will provide key opportunities to determine which cognitive abilities predict the development of capacities for cumulative culture. Finally, comparing different learning conditions in groups of adults is critical, as these experiments will allow clear causal conclusions regarding prerequisites and constraints, in relation to task complexity. The project will therefore fully expose the cognitive machinery responsible for the uniqueness of human culture.
The current report covers the first 30 month period of a five-year project, and therefore reflects our interim progress towards the overall goals of the project. Two of the research team members were appointed in the first month of the project, and the remaining four joined in October 2016 (month 14), so much of the planned data collection began in earnest midway through the current reporting period, following on from development and pilot work that was completed in the first year. During this period we have made considerable progress towards realising the objectives outlined in the proposal. The work performed during the period covered by this report has been summarised below, in relation to the stated objectives of the original proposal and grant agreement.


Work Package 1: Task Development

In the research proposal it was stated that, “A task will be developed which can be used to study the potential for ratcheting in the social learning of members of diverse populations, including children of varying ages, and primates of different species, as well as adult humans”.

The project involves developing a novel method for assessing the potential for cumulative culture by testing how effectively learners can make use of information that has been acquired vicariously. Specifically, our methods are intended to evaluate whether individual participants’ performances suggest that repeated social transmission could generate behaviour that is increasingly advantageous in terms of the resulting payoffs. In addition, the methods can assess whether payoff increases actually do occur under real conditions of social transmission (in human adults). We have achieved this objective, and (as per the proposal) we have developed both computer-based, and physical stimuli equivalent, versions of the experimental task. So far we have used these tasks to test hypotheses about the necessary conditions for cumulative culture in young children (see WP2 and WP3 progress below) and in human adults (see WP4 and WP5 progress below). We have also shown that capuchin monkeys can be trained to a learning criterion indicative of basic competence in relation to the demands of the task (see WP6 progress), and work is now under way to investigate whether the performance of task-competent individuals could potentially generate a ratchet effect (also detailed under WP6), and how their performance corresponds to that of human children (see WP7 progress).


Work Package 2: Testing ontogenetic development of capacities for cumulative culture in human children

In the research proposal it was stated that, “Studies with young children of a range of ages (18 months to 6 years old) will establish at what age ratcheting over learner generations becomes possible in young children”.

As noted above, we have already embarked on data collection with human children using our experimental task. Our first study (manuscript currently in preparation) focussed on children’s performance following observation or personal experience of a simple binary choice. This was an important starting point to help us to understand whether there were differences in the extent to which children were able to learn from observation of another’s experience, compared with learning from their own. Broadly speaking, children were able to use the social information as effectively as information gleaned from their own personal choices and experience, and both abilities improved with age. In terms of dissemination activity (please see table of activities) this work was presented at the 2016 Culture Conference in Birmingham (June 2016) by Renner, and at the New Directions in the Evolutionary Social Sciences workshop in Cambridge (December 2016) by Caldwell, and also at the BCCCD meeting in Budapest (January 2017) by Renner. Over and above what was planned in the original proposal, we also formed a new collaborative link with researchers from Peking University in Beijing, China, which allowed us to obtain an equivalent cross-cultural dataset (which will be reported in the same manuscript). This additional dataset has allowed us to determine the extent to which our findings generalise across populations with different cultural backgrounds.

This basic groundwork using binary discrimination problems established that young children readily grasped the demands of the task, making use of vicariously acquired information (or indeed information from their own experience) to guide their choices in test trials involving corresponding sets of stimuli. This has allowed us to further develop the task to address the question of the potential for cumulative culture more directly. As noted above, ultimately our methods are intended to evaluate whether individual participants’ performances suggest that repeated social transmission could generate behaviour that is increasingly advantageous in terms of resulting payoffs. This requires tasks that permit a continuum of potential payoffs that can be earned for any given problem, rather than just a single dichotomous (correct/incorrect) choice. We have now completed data collection on such a version of the task (i.e. involving multiple stimuli, and multiple selections, generating a continuum of performance success), which has allowed us to assess at what age children exhibit the potential for ratcheting in their use of social information, as well as factors influencing the age at which this becomes possible. Linking this to our dissemination activities, this work was presented at the 2017 Culture Conference in Birmingham by Wilks (May 2017), where it was awarded first prize for best poster. A manuscript reporting this study is currently in preparation.

Work Package 3: Testing effects of complexity of social information on children’s capacity for cumulative culture

In the research proposal it was stated that, “The performances of human children will also be compared across age groups in which potential for ratcheting can be established (Objective 2), using social demonstrations of varying complexity.”

As should be clear from the above extract, this particular objective reflects work that builds on findings generated as part of WP2, described earlier. Work carried out under WP2 will aim to identify how at what age the potential for ratcheting can be identified in young children’s learning using versions of the task designed to be maximally transparent and involving low cognitive load. Work carried out under WP3 is therefore intended to extend this research to further determine how the developmental timeline is affected by properties of the task itself, and the degree to which optimal use of social information places demands on cognitive resources and/or conceptual understanding (e.g. requiring memory, inference about mental states, or active information-seeking). To date we have completed data collection comparing two alternative versions of the task designed to evaluate potential for ratcheting in children (described under WP2 above), which vary in the extent to which memory is required in order to make effective use of the social information. This manipulation impacted on the age at which the potential for ratcheting became apparent from children’s performances, providing a clear indication of how cumulative culture can be constrained by basic limitations of cognitive resources. These results will be reported in a manuscript currently in preparation (referred to previously in WP2 above). The study will also be presented in June by Wilks at the Swiss Graduate Summer School on “Perspectives on Human Memory: Memory Functioning and Memory Failures”, for which she has been awarded a funded place.

In addition we have begun to embark on studies investigating the extent to which children’s developing understanding of others’ minds influences how they use social information. Data collection for this is currently ongoing.


Work Package 4: Investigating the relationship between transmission mechanisms and behavioural complexity in adult humans

In the research proposal it was stated that, “The project will also establish whether certain types of social information are critical for the transmission and/or ratcheting of certain types of behaviour.”

We have published one study related to this objective (Caldwell, Renner & Atkinson, see publications) which investigated the relationship between behavioural complexity and benefits gained from intentional teaching, compared with information about process of completion in the absence of active instruction from an experienced partner. This study found that, in line with our predictions, interactive teaching particularly facilitated the transmission of hard-to-learn skills. This study was also presented by Caldwell at the “Teachers’ Cognition” workshop in Paris (December 2015), and was included in the work presented by Caldwell in a keynote lecture at the inaugural meeting of the Cultural Evolution Society (September 2017), as well as in an invited seminar presented at the University of Abertay (November 2016).

We have also published a review paper (Caldwell, Cornish & Kandler, 2016) discussing the relevance of different task goals in cultural evolution experiments, and the likely effects of different task goals on the behavioural forms that emerge through transmission. This led to the completion of an empirical study directly comparing the effects of transmission on behavioural forms depending on the goal of the task from the perspective of the participant. The ideas presented in the review article, and the findings of the study, have been discussed in a number of our dissemination activities, including at the 2016 Culture Conference in Birmingham (presented by Atkinson, June 2016), the Cognitive Science Society Conference Philadelphia (presented by Caldwell, August 2016), a workshop on Culture, Cognition and Actions in Milan (presented by Caldwell, March 2017), and a workshop on Ecologies of Creativity in Cultural Evolution in Sandbjerg, Denmark (presented by Caldwell, February 2018). We also have a manuscript currently in preparation which will report the results of the empirical work.

In addition, we have recently completed data collection for a study comparing the effect of intentional communication on ratcheting in transmission chains of adult human participants, using one of the computer-based stimulus-selection tasks developed under WP1. This study tested the hypothesis that when there are bottlenecks in transmission (i.e. receivers of social information do not have access to a complete record of the behaviour of those they learn from), the specific subset of information to which they learners are exposed may strongly influence the potential for the retention and accumulation of beneficial information. We thus compared a full information condition (complete history of a predecessor’s task performance), with two conditions involving transmission of a highly restricted subset of this performance history. In one condition, this subset was selected by the predecessors themselves, with a view to passing on information that would be maximally beneficial to a naïve individual. In the other, the (quantity-matched) subset was randomly selected from the complete record of the predecessors’ responses. Analysis of this dataset is currently under way and we expect to disseminate the results in a conference presentation later this year. We also plan to prepare a manuscript for publication in due course.


Work Package 5: Investigating the influence of adult human capacities for understanding others’ minds on cumulative culture

In the research proposal it was stated that, “Again using microsociety research designs, the project will also investigate the causal influence of theory of mind in generating the ratchet effect in cultural transmission.”

We have completed data collection for three studies related to this objective, all using computer-based stimulus selection tasks developed specifically for the project (under WP1). In one line of research we are investigating dual-processing views on the cognitive requirements of cumulative culture (which propose that distinctively human culture may be explained by human capacities for System 2 metacognition and/or explicit theory of mind). To this end we carried out pilot work to investigate the effect of additional executive function load on reaction times (as a means of assessing effortful System 2 processing) in a very simple binary-choice version of our stimulus selection task. We have also further extended this work to investigate whether an executive function distractor task impacts differently on copying of successful choices, versus avoidance of unsuccessful choices, and also strategic switching between copying and avoidance. Analysis of this dataset is currently under way and we expect to disseminate the results in a conference presentation later this year. We also plan to prepare a manuscript for publication in due course.

In a second line of research, we recently completed data collection for a study investigating how inferences about others’ knowledge might facilitate use of social information, by allowing learners to make valid inferences about others’ past experience of a problem from exposure to a only a small snapshot of their behaviour. We predicted that non-random behaviour is likely to be attributed to an informed strategy on the part of the demonstrator, and that the learner might thus be able to infer that apparent avoidance of certain possibilities may mean that these have already been sampled and rejected. In order to test this we have compared the performance of participants in two different social information conditions. In one condition, the participants received information about their predecessor’s final attempt at the task, structured exactly as it was performed. In the other condition the participants received matched information, but this time shuffled such that it cannot be used to draw valid inferences regarding a systematic or informed search. Analysis of this dataset is currently under way and we expect to disseminate the results in a conference presentation later this year. We also plan to prepare a manuscript for publication in due course.


Work Package 6: Investigating limits on ratcheting in the social learning of nonhuman primates

In the research proposal it was stated that, “Studies with primates (chimpanzees, capuchins and squirrel monkeys) using the experimental task, will establish whether ratcheting over learner generations is possible in nonhuman primates.”

We have now completed data collection for a total of four studies involving nonhuman primates (two with squirrel monkeys and two with capuchin monkeys), all using stimulus selection tasks developed for the project under WP1 (starting with a simple binary choice version, analogous to the method used in our first study with children for WP2). Our first study, with squirrel monkeys, used a computer touchscreen version of the task. Although we were able to collect data on task performance for thirteen individuals, only one reached the pre-determined performance criteria we had set as a means of determining whether they were task-proficient with respect to the reinforcement contingencies. This limited the extent that we could draw conclusions about nonhuman primate performance generally, and therefore also the extent to which we could make valid comparisons with human performance (see WP7, below). We subsequently carried out a follow-up study using a physical object-choice version of the task, in which the monkeys simply selected one of two cups, one of which concealed a food reward. Although we expected this version of the task would be more intuitive for the squirrel monkeys, they still apparently failed to learn the predictive relationship between the information they were exposed to, and
the cor
Understanding differences between nonhuman learning and traditions and human cumulative cultural evolution (CCE) continues to be an extremely hot topic, and recent demonstrations of CCE-like effects in nonhumans have generated great interest (e.g. Claidiere et al., 2014, Proc Roy Soc B; Sasaki & Biro, 2017, Nature Comms). However, to our knowledge we remain the only group developing flexible methods for studying cumulative culture which will allow us to directly manipulate variables such as the complexity and opacity of the transmitted traits, whilst holding other aspects of the task structure constant. This will permit a much fuller understanding of the mechanisms required to transmit cumulative cultural traits, and the extent to which we can observe similar phenomena in nonhumans. We therefore expect that the findings from this project will have substantial impact across multiple academic disciplines (including Anthropology and Zoology, as well as Psychology).