Storytelling, innovation and teaching in cultural transmission

Ottilie Tilston (Institute of Work and Organisational Psychology, University of Neuchâtel), Adrian Bangerter (Institute of Work and Organisational Psychology, University of Neuchâtel), Kristian Tylén (School of Communication and Culture – Center for Semiotics, Aarhus University)

The role of teaching in cultural transmission is so far unclear, as transmission chain experiments typically do not analyse the content of interactions. Storytelling constitutes a teaching method as it enables the sharing of experience without exposure to the real-life costs of gaining that experience. Specifically, we propose that storytelling is specialised for the discussion of non-routine events, and is thus an important mechanism preventing the intergenerational transfer of mistakes and facilitating innovation (choosing to explore new solutions over exploitation of existing ones). One hundred and sixty-eight participants were divided into 28 chains of six who completed an individual task in succession: building a basket to carry as much rice as possible within five minutes. After building, each interacted with the following participant either with or without their basket to hand. This enables us to investigate whether memory constraints stimulate storytelling and foster innovation. Innovation was measured as the decision to make a qualitative change to a different basket design gestalt vs. exploiting the existing one. Evidence of cumulative improvement was found in the majority of chains, as the last basket (M = 1779g, SD = 648g) carried more rice than the first (M = 2116g, SD = 894g: F(1,1)= -5.18, p = .03), but this did not interact with condition (F(1,1)= .003, p =  .096). More qualitative changes in basket design occurred when the basket was absent during teaching (F(1,26) = 12.14, p = .002). There were no main effects for storytelling by generation (F(1,4)= .99, p = .46) or condition (F(1,4)= 1.41, p = .24).  However, we plan to investigate whether storytelling is proportionally more used to discuss non-routine events (e.g. mistakes) than all other talk, and whether this enhances the following participant’s performance. If so, storytelling in teaching may facilitate the ‘ratchet effect’ in cumulative cultural evolution.

 

Robust source-independent biases in children’s use of socially and individually acquired information

Mark Atkinson, Bill Thompson, Elizabeth Renner, Gemma Mackintosh, Christine A. Caldwell

University of Stirling

Culture has an influence on human behaviour which is unparalleled in other species. Some theories propose that this is due in part to children possessing domain-specific mechanisms for social learning which function to promote rapid enculturation. We assess this by comparing children’s performance on simple stimulus choice problems when they are provided with information from either a social source or from their own personal experience. Over three experiments, involving both a WEIRD sample of 18-month- to 5-year-old children recruited in Glasgow, Scotland (Study 1; N = 172), a non-WEIRD sample recruited in Beijing, China (Study 2; N = 159), and a variant of Studies 1-2 with a counter-intuitive reward structure (Study 3; N = 184), we find little evidence that children perform differently in response to information acquired from a social source compared to information received via their own individual exploration. Expectations about the predictive value of information thus appear to be independent of source. Furthermore, error rates show evidence of a consistent bias driven by motivation for exploration as well as exploitation, which was apparent across both conditions in all three studies. We conclude that some apparent peculiarities identified in human social information use likely reflect domain-general learning and motivational biases rather than domain-specific enculturation mechanisms.

 

Social learning strategies regulate the wisdom and madness of interactive crowds

Wataru Toyokawa

University of St Andrews

Why groups of individuals sometimes exhibit collective ‘wisdom’ and other times maladaptive ‘herding’ is an enduring conundrum. Here we show that this conflict is regulated by the social learning strategies deployed. We examined the patterns of human social learning through an interactive online experiment with 699 participants, varying both task uncertainty and group size, then used hierarchical Bayesian model-fitting to identify a complex variety of the individual learning strategies exhibited by participants. Challenging tasks elicit greater conformity amongst individuals, with rates of copying increasing with group size, leading to high probabilities of maladaptive herding amongst large groups confronted with uncertainty. Conversely, the reduced social learning of small groups, and the greater probability that social information would be accurate for less-challenging tasks, generated `wisdom of the crowd’ effects in other circumstances. Our model-based approach provides novel evidence that the likelihood of swarm intelligence versus herding can be predicted, resolving a longstanding puzzle in the literature.

 

No Evidence that Omission and Congruity Biases Affect the Perception and Recall of Vaccination-Related Information. 

Ángel V. Jiménez, Alex Mesoudi & Jamshid J. Tehrani

University of Exeter

Despite the spectacular contribution of vaccines to the prevention of infectious diseases, social fears about their safety and anti-vaccination claims are widespread. One explanation for this might be the existence of universal cognitive biases conspiring against vaccination. One of these biases is known as omission bias, which describes the tendency to perceive as worse, and recall better, bad outcomes resulting from commissions (e.g. vaccine side effects) than the same bad outcomes resulting from omissions (e.g. symptoms of vaccine preventable diseases). Another factor influencing the perception and recall of vaccination-related information might be people’s attitudes towards vaccines. A congruity bias would mean that pro-vaccination attitudes are positively related to perceptions of severity and recall of symptoms of vaccine preventable diseases and negatively related to perceptions of severity and the recall of vaccine side effects. To test these hypotheses, 202 female participants aged 18-60 (M=38.15, SD=10.37) completed an online experiment with a between-subjects experimental design. Participants imagined that they had a 1-year old child who suffered from either vaccine side effects (Commission Condition) or symptoms of a vaccine-preventable disease (Omission Condition). Then they rated a list of symptoms/side effects by their perceived severity on a 7-point Likert scale. Later, they completed a surprise recall test. An additional scale was used to measure their attitudes towards vaccines. Contrary to the hypotheses, perceptions of severity and the recall of symptoms/side effects were not associated with either experimental condition or their interaction with attitudes towards vaccines. Therefore, neither the omission nor the congruity biases were supported by the data. This raises the possibility that the spread of anti-vaccination claims might not be explained by these particular universal cognitive biases.

 

Find the treasure: Children’s use of social information in a Stimulus Selection Task

Wilks, C.E., Mackintosh, G., Rafetseder, E., Renner, E., & Caldwell, C.A.

University of Stirling

Humans appear to be unique in their ability to accumulate beneficial modifications, over generations of learners, through social transmission. This has been termed the ratchet effect. It has been demonstrated experimentally in adult human participants, but it is currently unknown at what age children develop the ability to produce such an effect.

We investigated potential for ratcheting (PFR) in 160 UK children (aged 3-6) by exposing them to social information (provided by a parrot puppet) in a treasure-island themed Stimulus Selection Task. Over three rounds, social information contained different numbers of correct and incorrect selections, enabling us to examine PFR from the data of individual participants.

We were interested in children’s ability to remember and use this information, to inform their own search. We therefore analysed performance following correct/incorrect demonstrations when visual working memory was, or was not, taxed. When accurate performance was memory-dependent, error rates for correct trials varied between 53% (age 3) and 17% (age 6). However, upon virtually eliminating this memory load (transparent condition), even the youngest children used both correct and incorrect social information almost at ceiling level.

Additionally, we used children’s scores on the three rounds to assess their PFR according to defined criteria. We again found evidence that performance was linked to task memory load: over 85% of children displayed PFR from age 3 in our transparent condition, but the point at which most children displayed PFR shifted to age 5 when memory was taxed. We conclude that constraints on visual working memory may limit the contexts in which children exhibit PFR. Social information is often available for a limited time, so demands on memory may mean nonhumans (and young children) fail to benefit from potentially useful information. Memory may have been overlooked as a key requirement for many examples of cumulative cultural evolution.

 

Limited Spread of Introduced Arbitrary Object-related Conventions in Captive Monkeys

Claire FI Watson, Tetsuro Matsuzawa

Kyoto University

Communicative and other social traditions represent a large proportion of potential cultural variants observed in nonhuman primates, yet remain understudied experimentally relative to food-oriented traditions. Social behaviors can be relatively arbitrary in form with meaning conferred by shared convention among group members, and thus arguably more complex to learn than more directly functional, food-related behaviours. Controlled experimental manipulation of captive groups allows unambiguous demonstration that behavior patterns can be transmitted socially. The ability to learn arbitrary object-related action-sequences from others has been demonstrated in captive chimpanzees. Using a somewhat similar paradigm, we investigated whether monkeys, Japanese macaques (Macaca fuscata), could copy an entirely novel, arbitrary, object-related, action-sequence: picking up and placing a token into one of two differently shaped receptacles, in an open diffusion experiment. Following ten baseline sessions, conspecific demonstrators, one from each of two groups (n = 30) housed in large outdoor enclosures, were trained alternate target behaviors. Demonstrators then performed the behavior in view of their respective group-mates, across ninety 45-min-per-day sessions. Observers had adequate opportunity to perform the behavior. One juvenile within each group placed tokens in a receptacle, but no adults. Our results suggest the capacity exists in Japanese macaques, but that the spread of such novel arbitrary action-sequences is limited. These findings aid understanding of cultural transmission of social traditions in wild monkeys and phylogenetic origins of human social culture.

 

Complex social learning strategies are harder than the sum of their parts

Juliet Dunstone, Mark Atkinson and Christine Caldwell

University of Stirling

The use of explicitly metacognitive learning strategies has been proposed as an explanation for uniquely human capacities for cumulative culture. Such strategies are proposed to rely on the use of executive functions (EF) and explicit, system-2 cognitive processes, and to enable advantageous selective copying. To investigate the plausibility of this theory, participants’ ability to make flexible learning decisions under EF-resource load was investigated.

166 participants (predominately white British) were recruited at the University of Stirling. Participants completed a simple win-stay lose-shift (WSLS) paradigm task, intended to model a social learning situation where vicarious information can be used to inform response choice, by copying rewarded responses and avoiding those that are unrewarded. This was completed alongside a concurrent EF distractor task. Participants were split into 3 strategy groups: those that should use a flexible WSLS strategy, those that should always copy (information trial always revealed the target), and those that should always do the opposite (information trial never revealed the target).

There were significant effects of EF load and strategy group; using a flexible strategy was more challenging than using one rule consistently, and copying was less challenging than avoiding stimuli selected in the information trial. However, each information condition was equally affected by competing executive function demands. A significant training effect was also evident after a small number of trials.

These results suggest that learning decisions are underpinned by the use of executive functions even at a very basic level, and that more complex strategies are more challenging than a basic combination of their component strategies. We found no evidence that flexible learning strategies relied on executive functions any more than did non-flexible strategies. However, the observed learning effects suggest that ceiling effects could be masking differences between conditions which might be apparent in other contexts.

 

Food cleaning is not a culture-dependent trait in gorillas (and introducing the “minimal culture” concept).

Damien Neadle (University of Birmingham), Matthias Allritz (University of St Andrews) & Claudio Tennie (University of Tübingen)

A recent report, employing the method of exclusion in wild western lowland and mountain gorillas, suggested that food cleaning may be a candidate for gorilla culture. The term culture however, is broad; it may cover bumblebee laboratory behaviour to human cumulative culture (and so, spacecraft to the moon). In order to determine where on the “spectrum of culture” gorilla food cleaning lies, we re-analysed data where a group of captive western lowland gorillas were provided with sand-covered and clean apples, and recorded the behaviours that occurred. Each subject showed evidence of food cleaning, with four out of our five subjects reinnovating the very behavioural form reported in their wild conspecifics. Our subjects were culturally unconnected to these wild populations, therefore showing that individual learning can be the driving force behind this behavioural form. However, we do not conclude that social learning plays no role; instead we suggest that social learning most likely acts as a catalyst, harmonising this behaviour within communities (increasing its frequency via socially mediated serial reinnovation). The fact that it can clearly be reinnovated from scratch excludes the possibility of it being a culture-dependent trait. In addition, we constructed a new, minimal, definition of culture, consistent with the idea that cultural behaviours can be individually learned, but may also increase in frequency by social learning. This new definition of culture holds much resemblance with other lean definitions in the literature, but, to our best knowledge, all of these definitions contained additional factors. We suggest to start the “cultural spectrum” in this minimal way, ranging from “minimal culture” to “human cumulative culture”.

 

Capuchin monkeys can learn and generalise a win-stay, lose-shift strategy under social information and individual exploration conditions

Donna Kean, Elizabeth Renner, Mark Atkinson, Christine A Caldwell

University of Stirling

The ability to learn discriminatively from the successes and failures of others may be integral to the capacity for cumulative culture. This ability has not been sufficiently tested in non-human primates, but such investigations could provide insights into the apparent lack of cumulative cultural evolution in their natural behaviour.

To investigate this, capuchin monkeys were tested in a visual discrimination task for which success required copying of rewarded, and avoidance of unrewarded, behaviors. Thirteen monkeys housed at Edinburgh Zoo were trained to take part in a touchscreen stimulus selection task beginning with two-stimulus, and progressing to three-stimulus, discriminations. Subjects were either in a social condition where the experimenter performed an information trial (IT) by choosing between the available stimuli, or an individual condition where the subject performed the response during the IT.

Eight monkeys reached our pre-determined performance criterion on the two-stimulus task (four in the individual condition, and four in the social condition). The subjects that reached criterion were significantly more successful following rewarded ITs, compared with unrewarded, suggesting that it was easier to learn the win-stay rule, compared with the lose-shift. Their ability to generalise the win-stay, lose-shift strategy was then evaluated by transferring these monkeys to the three-stimulus version. Selections that were rewarded during the IT were repeated with high accuracy (>80% success) even following transfer to the three-stimulus discrimination problems, suggesting that these monkeys recognised the predictive relationship between the information trial and the test trial, and could use this information with “high fidelity”.

Results suggest capuchin monkeys can learn to use social information in ways that could in principle support cumulative culture, since the application of a win-stay, lose-shift strategy allows an individual to outperform a demonstrator. Future research will investigate the extent to which this could support improvements in performance over multiple generations.

 

The “cumulative culture puzzlebox” is not a culture-dependent task for children

Reindl, E. (University of St Andrews), Gwilliams, A. L. (University of Birmingham & Aston), Dean, L. (University of St Andrews), Kendal, R. L. (Durham University), & Tennie, C. (University of Birmingham & University of Tübingen)

In 2012, Dean and colleagues published a seminal paper investigating the role of a series of socio-cognitive processes for cumulative cultural learning in children. They presented groups of 3- and 4-year-olds with a three-stage puzzlebox and found that children who reached the last stage of the box made use of imitation, teaching, and prosociality, while children receiving no such social support did not reach the final stage. The authors concluded that imitation, teaching and prosocial behaviour were directly responsible for children’s cumulative cultural learning. They did not, however, test whether the puzzlebox was beyond the reach of children tested individually, a potentially stronger test of whether cumulative cultural learning across individuals is required.
We provide this asocial learning condition, showing that 9 out of 35 children from a metropolitan area in the UK reached the last stage of the puzzlebox without social support. Yet, the Dean et al. groups were more successful overall than the asocial controls and there was no difference in performance between Dean et al.’s ‘asocial controls’ and our true asocial controls. Following the definition of cumulative culture by Dean et al. (2014), the increased performance of the social groups compared to the asocially tested children is sufficient evidence for cumulative cultural learning. However, other definitions of cumulative culture require a task to be beyond the reach of individuals (Richerson & Boyd, 2005) and thus social learning as necessary rather than helpful for its solution (‘culture-dependent’, Reindl et al., 2016).
While this particular puzzlebox is a valid simulation of the complexity of ratcheting tasks in general, it is not strictly ‘culture-dependent’ for children as some may solve it without social learning. Thus the diverse definitions of cumulative culture within the field leave the role of imitation, teaching and prosociality for cumulative cultural learning open to question.

 

Intentional Communication facilitates Cumulative Cultural Evolution.

Gemma Mackintosh

University of Stirling

The cumulative nature of human culture is well documented in experimental research and real world data, showing clear evolution of ideas in language, artefacts and behaviours. However, the capacities that allow for the transmission of ideas in such a cumulative way remain poorly understood. This study explored the idea that one of these capacities could be Theory of Mind, in its role supporting intentional communication (an ability that appears to be – at least in its most complex form -unique to humans).

150 British participants completed a grid search task, after which they were asked to select a subset of their own search results to send to the next participant in a 10-generation transmission chain. The communicated information could then be used by the receivers to guide their own search of the same reward landscape.  This intentional information condition was compared with a random information condition, in which an equivalent amount of (computer-generated) information was transmitted.

In the intentional communication condition, participants were able to maximise the grid score (locate all the rewards with the fewest possible search attempts) much more successfully than in the condition where equivalent random subsets of information were sent between participants.

These results suggests that participants were able to identify information that would be particularly beneficial to their successor. This contributed significantly to cumulative improvements in task performance, relative to a condition intended to capture the effects of transmission via inadvertent social information cues, as opposed to intentional signalling. Human propensities for intentional communication may therefore facilitate many cases of cumulative cultural evolution. In contrast, if most animal social learning depends on inadvertent cues obtained as public information, this may place constraints on the potential for accumulation over learner generations.

 

Individual learning suffices for the re-emergence of a potential cultural behavior in captive task-naïve chimpanzees: tool excavation for USOs

Alba Motes Rodrigo, Parandis Majlesi, Matthias Laska, Elisa Bandini, R. Adriana, Hernandez-Aguilar & Claudio Tennie

Department of Early Prehistory and Quaternary Ecology, University of Tübingen, Tübingen, Germany

Social learning mechanisms are responsible for the behavioral differences found in neighboring chimpanzees groups that share genetic and environmental backgrounds. However, it remains unclear if behaviors are transmitted via social learning directly or if instead social learning catalyzes the manifestation of a behavior in other individuals, thereby increasing the frequency of that behavioral form in the group. One such behavior, so far only described using indirect evidence in two chimpanzees populations (Ugalla, Tanzania and Bandafassi, Senegal), is tool excavation to obtain edible underground storage organs of plants (USOs). This pattern of absence vs. presence despite common ecological opportunism and genetic background (a.k.a. method of exclusion), is used to classify animal behaviour as cultural. Furthermore, tool excavation for USOs is considered to have played an important role in human evolution, as it has been hypothesized that USOs were important fallback foods for early hominins. The aim of this study was to determine if tool excavation develops in naïve chimpanzees, and if so, how it occurs, using direct evidence for the first time. In this study, we tested two captive groups of chimpanzees (total n=10). Subjects in both groups spontaneously used wooden tools to excavate food items that we had previously buried in natural soil in the chimpanzees’ outdoor exhibit. Excavation was composed of six different tool use actions, four of which appeared in both groups: probing, perforating, digging and pounding. Our study demonstrates that tool excavating (a cultural behavior according to the method of exclusion) can be individually reinvented and it is not dependent on copying social learning to emerge. Finally, this study provides new behavioral data for modelling USOs extractive foraging by early hominins.

 

The effect of others’ goals on children’s strategic use of social information

Kirsten H. Blakey, Eva Rafetseder & Christine A. Caldwell

University of Stirling

Experimental social learning paradigms are almost exclusively designed so that a demonstrator and participant are motivated to reach the same goal. In reality it may be relatively rare to opportunistically encounter social information in the form of the performance of another individual who shares our immediate goal. Rather, adult humans actively seek, and strategically use, relevant social information about others’ successes and failures to help them fulfil their own particular goals. This may require sophisticated cognitive mechanisms that may be unavailable to other species, which might help to explain why cumulative culture appears to be unique to humans.

As a means to evaluate the cognitive challenges involved, we are examining 3-to-7-year-old British children’s ability to strategically use social information provided by a demonstrator (puppet) who had either the same goal as or a different goal to themselves. Children are required to use the demonstrator’s reaction to the contents of a capsule (acceptance or rejection), selected from one of two possible reward locations, to decide whether to select their own capsule from the same or an alternative location.

We expect that strategic social information use, and thus success, will increase with age, regardless of the demonstrator’s goals. Children are also expected to be more successful when they share the same goal as the demonstrator compared to when their goals are different, even though our task is designed such that both conditions are potentially equally informative.

 

The effect of array size and information type on capuchin monkeys’ (Sapajus apella) use of information in a stimulus choice task

Elizabeth Renner, Donna Kean, Mark Atkinson, and Christine Caldwell

University of Stirling

We used a touchscreen task to compare the use of different types of information by capuchin monkeys (Sapajus apella, N = 15). In the task, an array of stimuli (2 or 3) was presented on the screen, and individuals were given multiple attempts to find the single rewarded item. Information about one of the stimuli (i.e., whether it was the rewarded item) was provided on the first trial, and we compared three within-subjects conditions. In the social condition, the experimenter selected a stimulus on the first trial; in the virtual condition, an animated hand selected a stimulus; and in the individual condition, the monkey selected a stimulus. On half of problems, the stimulus selected on the first trial was the rewarded stimulus, and on the other half, the selected stimulus was unrewarded. On the second trial, participants were given a raisin for selecting the rewarded item. If a participant selected the same item (“stayed”) after seeing/selecting the rewarded item or selected a different item (“shifted”) after seeing/selecting the unrewarded item, the trial was given a score of 1; otherwise, a trial was scored 0. Surprisingly, there was no main effect of array size (2 vs 3 items) on performance (p = 0.51). There was a main effect of first trial information type (whether a rewarded or unrewarded stimulus was selected), such that all monkeys were more likely to stay after seeing selection of the rewarded item than to shift after selection of the unrewarded item (p < 0.001). There was also an array size by first trial information type interaction, such that monkeys were less likely to stay after seeing a rewarded selection when the array size was larger (3 items). Therefore, task complexity affected performance via its interaction with the type of information presented.

 

Defining linguistic complexity: an introduction

Mark Atkinson

University of Stirling

Cultural complexity is often referred to in the literature, yet in many cases it’s not adequately clear what the term means. The complexity of languages, however, has been more carefully considered. This introduction to linguistic complexity will hopefully be useful when considering other definitions of cultural complexity: if we seek a broad definition of cultural complexity which can be applied across cultural domains, then it would need to be applicable to language, and so the issues with quantifying linguistic complexity should be appreciated; but even where complexity is to be defined for a specific (non-linguistic) cultural domain, the approaches to defining the complexity of language may still be informative.

I’ll introduce the key approaches to defining and quantifying linguistic complexity, including (i) acquisition and usage difficulty, (ii) feature-based structural approaches, and (iii) information-theoretic structural approaches. I’ll also highlight the advantages and problems with each, alongside the difficulties of defining appropriate units for analysis, ethnocentric and usage-assumption pitfalls, and the proposed distinction between overt and inferential complexity.