Week 1 – Social Learning as Action Coordination

This early draft was authored by James Strachan, Arianna Curioni, and Luke McEllin.

In our chapter we propose that, to explain the flexibility of technical traditions, it is important to situate social learning within an interaction context. To this end, we show how current discourse on observational learning through imitative copying typically presupposes a unidirectional type of interaction which imposes rigidity on the learned behaviour. We go on to differentiate between knowledge flow and information flow, and show that, while the former is necessarily unidirectional in social learning episodes, the latter rarely is. We go on to show that, by expanding the range of social learning interaction contexts to allow for bidirectional information flow, it is possible to draw clear parallels with the joint action literature on action coordination. Finally, we show that considering social learning as a type of action coordination can help to explain both the flexibility and rigidity of technical traditions in a way that is coherent with the anthropological record on complex skill learning.

The Unidirectionality Problem

Social learning has been defined as “learning that is influenced by observation of, or interaction with, another animal… or its products” (Heyes, 1994), or, more precisely, “learning that is facilitated by observation of, or interaction with, another individual or its products (Hoppitt & Laland, 2013, emphasis added). It is interesting to note that while under either definition observation is only one way by which individuals can socially learn from conspecifics, many discussions of social learning treat copying through observation as the basic phenomenon of interest (Henrich, 2016; Henrich & McElreath, 2003; Mesoudi, 2011).

To illustrate, take the well-documented example of nut-cracking techniques in chimpanzees (Goodall, 1964; Whiten et al., 1999). Chimpanzees will often use rocks as tools to crack open nuts when foraging, and there is evidence that this behaviour is socially learned and is, some argue, cultural in nature (Biro et al., 2006). As the typical example goes, a young chimpanzee will watch an individual who already knows the technique use a rock to crack a nut and, by copying the behaviour, will learn to perform the technique itself.

Such examples of observational learning — and examples like it describing observational learning in humans — are often used as minimal working examples of social learning. Minimal working examples are important for guiding theoretical and empirical work, particularly on such varied phenomena as social learning, where interactions are diverse and complex and individuals can learn using a range of methods such as observation, local enhancement, or even direct teaching (Kline, 2015). Analysing all possible instances of social learning is clearly intractable and so empirical and theoretical work on social learning must abstract minimal working examples, on the assumption that these minimal examples contain within them distillations of the cognitive mechanisms at play in more complex behaviours, in order to make useful scientific claims about the phenomenon.

Observational learning is a very useful minimal working example in many ways. The juvenile chimp who does not yet know the technique watches the model perform it and, through some mechanisms allowing it to infer the causal relationship between the tool use and positive outcome, and to extrapolate the knowledge of this relationship to form and execute an action plan itself, is able to learn it. To say that this example is minimal is not to say that the question of how an individual learns through observation is trivial — far from it — but that explaining this episode of social learning does not require knowing what is going on in the mind of the model, nor anything of the prior knowledge or experience of either agent, nor the local ecology or social environment in which the interaction is situated. One needs only identify the learning processes going on in a single individual’s mind. Another point in favour of using observational learning as a minimal working example is that this kind of learning is very important early in human development: babies cannot act independently and so learning through watching and imitating adults is very important to the acquisition of early motor skills, together with the ability to discern who and what to imitate (Gergely et al., 2002; Gergely & Csibra, 2006; Meltzoff, 1988; Wood et al., 2013). Given that observational learning is the first kind of social learning that humans are capable of, and is also observed in our closest phylogenetic relatives, it is tempting to think of this also as a conceptual foundation for other kinds of social learning that humans are capable of later in life.

To our minds, this treatment is problematic because it carries with it an assumption of unidirectionality. The problem is that the example of a young chimpanzee watching an adult crack a nut involves no mutual exchange of information between the model and the learner. If a rock fell accidentally and cracked a nut as it landed, the chimpanzee could extrapolate the same information as from watching a conspecific perform it. While the quality of the learned information may be better by watching a conspecific, as social cues from the model may enhance or affect what and how the student learns, the structure of the interaction in both cases does not allow a learner to actually contribute as anything more than a spectator. In this case, information goes only one way, from the model to the learner, and although feedback from the learner (which would allow the model to adapt to and accommodate the learner’s needs) may be possible it is considered something extra to the social learning rather than an inherent part of it. In essence, the only thing that is social about observational social learning is the fact that the observation corresponds to another individual. In cognitive science, this problem of unidirectionality is also known as the ‘fourth wall’ problem and has been well documented with regards to eye gaze processing (Gobel et al., 2015; Risko et al., 2016). The crux of the problem is that within laboratory experiments where participants look at scenes or faces on computer screens, their patterns of gaze fixations are very different from live recordings of participants physically interacting with those scenes or people in real life. Truly one-way observation is rare, and in the real world observers are more than just spectators. An explanation of the mechanisms of social learning should embrace this complexity rather than trying to control for it.

One possible way to embrace the complexity of social learning interactions is to examine the distinction between knowledge and information. In the next section we clarify this distinction and discuss the implications of this for broadening the scope of research on social learning.

Knowledge vs. Information

A social learning episode involves at least two people. One, the model, possesses some technical knowledge that the other, the learner, will acquire as a result of the episode. In this sense, it is a necessary precondition that a social learning interaction has an asymmetry in knowledge and a consequent unidirectional flow of knowledge from the model to the learner. However knowledge is not the only thing communicated in a learning interaction; there is other information that need not flow unidirectionally from the model to the learner: students can ask questions of their teachers, or learners can make statements about what they are learning or try out ways of using the technique that the model can then react and adapt to. Considering this, it is important to distinguish knowledge flow from information flow as dimensions along which interaction contexts can vary. In brief, we use knowledge to refer to the context-invariant representation of an action or sequence of actions that are necessary to produce a certain outcome, while information is the context-specific feedback from the social interaction partner that allows an actor to make online adjustments that can satisfy either instrumental or communicative goals.

Imagine a situation in which two people—Nico and Robert—are involved in preparing a meal according to a family recipe of Nico’s. Nico knows the recipe but Robert does not. This is an example of a knowledge asymmetry, and is a necessary precondition for social learning—the knowledge (the family recipe) is held by the model (Nico) and not by the learner (Robert) who has some motivation or desire to learn it. This knowledge (the family recipe) extends beyond this particular interaction—once learned, Robert could go on to cook this meal without Nico in the future.

However, while a social learning interaction must have some knowledge asymmetry necessitating a unidirectional knowledge flow through the interaction, it can have many different types of information flow. This information flow can be unidirectional — Robert could be learning to cook the recipe by reading a cookbook of family recipes that Nico has published — but it need not be. If Robert and Nico are in the kitchen together then Nico can see what Robert is doing as well and adjust his actions accordingly, perhaps by changing his position at the stove so that Robert can look over his shoulder. The information flow is also affected by the dynamics of the interaction itself: Robert could sit in the kitchen and watch Nico prepare the meal, but Robert could also help Nico to prepare the meal. Any of these situations is an example of social learning, and all can result in the successful transfer of knowledge from Nico to Robert, but for the two interactants they are very different scenarios that require very different cognitive mechanisms to act.

These different interaction structures, which in turn create different information flows, highlight a key aspect of social learning interactions that is overlooked in many treatments of social learning, which is the many varied opportunities for coordination between models and learners. Drawing on research in cognitive science in joint action and coordination, we propose that research on social learning can lean into this complexity and variety in interactions, and that the cognitive mechanisms that are most interesting in social learning episodes may not be specialised for learning per se, but rather for action coordination. This approach centres the social learner not as a parasitic spectator or information scrounger, but an actively engaged participant in the behaviour.

Learning by Active Engagement

Considering the social learner as an active participant in an interaction that is defined by particular interactional constraints opens up opportunities for non-vertical patterns of transmission such as what Tomasello et al. (1993) describe as collaborative learning—children can play with their peers at grown-up jobs before they are allowed to participate with adults. For example, in Papua New Guinea, Asabano male children who are prohibited from joining adult hunters on the search for dangerous game like feral pigs or cassowaries will often band together and structure their playtime around hunting smaller targets like lizards (Little & Lancy, 2016). This gives young children important basic experience running around and throwing sticks at lizards, and gives older children the valuable opportunity to plan and manage a team of individuals of various ages and skill levels, which can give them much better functional understanding of the mechanics at play in the behaviour that will serve them well when they eventually join the adult hunting parties. Even in cases where coordination is not strictly necessary to achieve an outcome—as in the case of stone knapping to make tools, for example—established experts will intentionally perform individual actions in social configurations, discussing their tasks with each other in a line or a circle where each can have perceptual access to what others are doing, and where novices can watch multiple experts at work in parallel (Stout, 2002, 2005). Beyond ethnographic evidence, there is evidence from the archaeological record of southern African stone tools throughout the Pleistocene that technological innovations can and do occur through bottom-up or learner-driven contributions to the behaviour, as opposed to through purely top-down mechanisms of transmission (Wilkins, 2020).

We believe that it is this kind of learning through coordination that plays a key role in dictating how traditions become both stable and flexible. That is, traditions are not more or less flexible in themselves in that there is no ontological quality to a tradition that makes it more likely to become stable through transmission, but the stability and flexibility of a tradition are affected by how the task constraints, motivation, and interaction structure during transmission affect local coordination demands and the expression of various coordination mechanisms. In the following section, we explore how different coordination constraints and mechanisms can lead to both flexible and rigid patterns of behaviour, and propose some avenues for future empirical work afforded by this new framework.

Coordination, Flexibility, and Rigidity

Having outlined our position — that social learning of techniques can be understood in terms of action coordination and its relevant mechanisms — we now illustrate this further by giving a brief overview (which will be expanded further in the full draft) of some literature in joint action and action coordination that describes some of these mechanisms and demonstrates how they can inform social learning and research into cultural evolution and transmission. Specifically, we describe how a drive to make oneself predictable during coordination can result in rigid and stable patterns of behaviour, while having to represent another person’s task during a joint action when that actor has different constraints on their actions can lead to dynamic and flexible action modulations to compensate. Finally, we describe the few existing studies that examine social learning and teaching from a joint action perspective.

Coordinating actions between two or more individuals in order to realise some shared or joint outcome is a demanding task subserved by a host of mechanisms (Vesper et al., 2010). A key  driving principle behind how these mechanisms are expressed is whether and how they help to facilitate a partner’s anticipation of the relevant features of one’s actions. Several studies have shown that, when coordinating, actors become less variable from trial to trial than when they act alone as a way of making themselves predictable (Vesper et al., 2011, 2016). While this kind of variability modulation is a very basic way of facilitating coordination, which has even been observed in macaque monkeys (Visco-Comandini et al., 2015), humans are also able to adapt their behaviour in more systematic ways.

For example, in a study of joint improvisation, Hart et al. (2014) found that expert improvisers who were instructed to synchronise their actions modulated the velocity profiles of their actions such that both partners deviated significantly from how either of them would move individually. Furthermore, rather than simply converging on a pattern of behaviour somewhere between the two actors’ styles of movements (i.e. averaging out each other’s idiosyncrasies), the resulting coordinated actions reflected some universal characteristics across dyads. This suggests that both agents adjusted their behaviour in systematic, general ways that they considered easy to predict for any potential interactant. In a similar way, we believe that rigidity in cultural traditions can itself serve as a coordination facilitator. The background knowledge and common ground afforded by a cultural tradition can help to offset the substantial cognitive costs of coordinating — particularly with unfamiliar interaction partners — while still allowing for efficient yet fine-grained and sophisticated action planning, prediction, and monitoring.

On the other hand, considering social learning episodes as instances of coordination allows for a great degree of flexibility in the trial-to-trial expression of behaviours and techniques. The mechanisms supporting coordination in joint action are versatile and sensitive to the situational context and local interaction demands. Actors are sensitive to their partners’ ecological constraints and embody these in their own movements (Schmitz et al., 2017), and individuals with an easy task will make more adjustments to their behaviour to coordinate with a partner who has a more difficult task (Vesper et al., 2013). In cases of unidirectional information flow, coordinating partners will use leader-follower role assignments to dictate the distribution of online adjustments (Curioni et al., 2019). Even specific coordination mechanisms such as adjusting the speed of one’s movement are dependent on the local task demands: in cases where partners are trying to synchronise the timing of oscillatory movements such as tapping or swaying, people speed up in an attempt to minimise variability (Vesper et al., 2011; Wolf et al., 2019), while in cases where partners are trying to synchronise the spatial endpoints of actions they tend to slow down and adjust the ascent-to-descent velocity ratio of their movements to highlight the upcoming target (McEllin et al., 2018). Actors will even tailor their actions on the basis of their partner’s ability, in order to ensure that they do not communicate redundantly (Candidi et al., 2015).

Until now there has been comparatively little work examining how coordination mechanisms are expressed and exploited in instances of social learning, and what role these mechanisms might play downstream in the transmission, propagation, and stabilisation of cultural traditions. However, early work does show that pedagogical demonstrations share some kinematic characteristics with coordinated joint actions. In a study where participants learned to play a series of notes on a modified xylophone, McEllin et al. (2018) examined the kinematics of participants who knew the sequence under three conditions: a turn-taking demonstration, where the participant played the piece through for someone who had to learn to play it themselves, an unequal coordination condition, where the knowledgeable participant and a naive participant had to play together at the same time, and an equal coordination condition, where both participants knew the piece and had to play it together at the same time. In all three conditions, participants (demonstrators, leaders, and co-actors) exaggerated the peak height of their movements relative to an individual baseline. When participants had to play together at the same time (unequal and equal coordination conditions) they slowed down the descent phase of their movements in order to facilitate temporal coordination. However, when participants were coordinating with a naive participant, they also slowed down the ascent phase of their movements in order to facilitate their partner’s spatial predictions about the upcoming end of the movement. These kinematic signatures, which map clearly to different coordination demands and constraints, are clear evidence of individuals flexibly adapting their behaviour to facilitate coordination and communicate information.

 Okazaki et al. (2019) monitored the kinematics of participants during a turn-taking imitation learning task where a teacher demonstrated to a student how to complete the Tower of Hanoi task in order to quantify the interactional dynamics of teaching, which they describe as a type of reciprocal interpersonal coordination. Importantly, when they calculated causality and noise covariance between the model and the learner they found not just feed-forward information flow from the teacher to the learner, but also feedback information flow from the learner to the teacher. Furthermore, this changed over time as teachers interactively scaffolded the learners behaviour by providing more pedagogical information when learners struggled, and as learners improved the transfer of knowledge from teacher to learner became a bidirectional exchange of common ground information.

In a later study, Strachan et al. (2020) examined how learners interpret these pedagogical cues in observed demonstrations (playing a piece of music, this time on a modified set of drums), and specifically whether they incorporate these action modifications into their own reproductions. If participants were observational learning through copying the model’s demonstration, then they would reproduce this behaviour as these exaggerations are embedded features of the actions. However, as we have shown, under a coordinative framework it is not always necessary for both actors to adjust their behaviour if this is not relevant to the end goal, so there would be no need to copy the pedagogical modifications that learners observe. This study did indeed find evidence in support for a pragmatic reconstructive learning process whereby learners did not incorporate the modifications they observed into their own productions.

To finish, we propose one avenue for future work as a way of illustrating how reframing the question of social learning as an action coordination problem can lead to novel insights for both areas of study. The topic of role assignment has been studied in the action coordination experimental literature typically by manipulating the information flow of the interaction—restricting the perceptual access of one contributor means that their partner has no choice but to follow. But leaders and followers also emerge through social dynamics, particularly in the case of social status or dominance. In cultural evolution, a phenomenon called prestige bias describes how the behaviour of high status individuals at one generation have a disproportionate influence on the expression of that behaviour at subsequent generations relative to lower-status individuals (Henrich & Gil-White, 2001). This has been explained as higher-status (or prestigious) individuals being preferentially imitated. However, another way in which status may affect the transmission of techniques is to consider status or prestige as a feature of the interaction structure that affects how individuals coordinate. It may be that dominant or high-status individuals, like the leaders in laboratory experiments, may be less adaptable or more rigid in their idiosyncratic behaviour, meaning that they exert a greater influence on their coordination partners than they are influenced by them (Chang et al., 2017). In a social learning interaction with a prestigious or high status model, social learners would have to make more adjustments to bring their behaviour in line with these individuals to allow for successful coordination. By considering this question through the lens of joint action coordination, we can ask whether it is truly prestige that affects the likelihood of idiosyncratic behaviours being imitated, or whether intransigence may also play a role.

References

Biro, D., Sousa, C., & Matsuzawa, T. (2006). Ontogeny and Cultural Propagation of Tool Use by Wild Chimpanzees at Bossou, Guinea: Case Studies in Nut Cracking and Leaf Folding. In T. Matsuzawa, M. Tomonaga, & M. Tanaka (Eds.), Cognitive Development in Chimpanzees (pp. 476–508). Springer-Verlag. https://doi.org/10.1007/4-431-30248-4_28

Candidi, M., Curioni, A., Donnarumma, F., Sacheli, L. M., & Pezzulo, G. (2015). Interactional leader–follower sensorimotor communication strategies during repetitive joint actions. Journal of The Royal Society Interface, 12(110), 20150644. https://doi.org/10.1098/rsif.2015.0644

Chang, A., Livingstone, S. R., Bosnyak, D. J., & Trainor, L. J. (2017). Body sway reflects leadership in joint music performance. Proceedings of the National Academy of Sciences, 114(21), E4134–E4141. https://doi.org/10.1073/pnas.1617657114

Curioni, A., Vesper, C., Knoblich, G., & Sebanz, N. (2019). Reciprocal information flow and role distribution support joint action coordination. Cognition, 187, 21–31. https://doi.org/10.1016/j.cognition.2019.02.006

Gergely, G., Bekkering, H., & Király, I. (2002). Rational imitation in preverbal infants. Nature, 415(6873), 755–755. https://doi.org/10.1038/415755a

Gergely, G., & Csibra, G. (2006). Sylvia’s recipe: The role of imitation and pedagogy in the transmission of cultural knowledge. In Roots of human sociality: Culture, cognition, and human interaction (pp. 229–255).

Gobel, M. S., Kim, H. S., & Richardson, D. C. (2015). The dual function of social gaze. Cognition, 136, 359–364. https://doi.org/10.1016/j.cognition.2014.11.040

Goodall, J. (1964). Tool-using and aimed throwing in a community of free-living chimpanzees. Nature, 201(4926), 1264–1266.

Hart, Y., Noy, L., Feniger-Schaal, R., Mayo, A. E., & Alon, U. (2014). Individuality and Togetherness in Joint Improvised Motion. PLoS ONE, 9(2), e87213. https://doi.org/10.1371/journal.pone.0087213

Henrich, J. (2016). The Secret of Our Success: How culture is driving human evolution, domesticating our species, and making us smarter. Princeton University Press. https://press.princeton.edu/books/paperback/9780691178431/the-secret-of-our-success

Henrich, J., & Gil-White, F. J. (2001). The evolution of prestige: Freely conferred deference as a mechanism for enhancing the benefits of cultural transmission. Evolution and Human Behavior, 22(3), 165–196. https://doi.org/10.1016/S1090-5138(00)00071-4

Henrich, J., & McElreath, R. (2003). The evolution of cultural evolution. Evolutionary Anthropology: Issues, News, and Reviews, 12(3), 123–135. https://doi.org/10.1002/evan.10110

Heyes, C. M. (1994). Social learning in animals: Categories and mechanisms. Biological Reviews, 69(2), 207–231.

Hoppitt, W., & Laland, K. N. (2013). Social Learning: An Introduction to Mechanisms, Methods, and Models. Princeton University Press.

Kline, M. A. (2015). How to learn about teaching: An evolutionary framework for the study of teaching behavior in humans and other animals. Behavioral and Brain Sciences, 38. https://doi.org/10.1017/S0140525X14000090

Little, C. A., & Lancy, D. F. (2016). How do children become workers? Making sense of conflicting accounts of cultural transmission in anthropology and psychology. Ethos, 44(3), 269–288.

McEllin, L., Knoblich, G., & Sebanz, N. (2018). Distinct kinematic markers of demonstration and joint action coordination? Evidence from virtual xylophone playing. Journal of Experimental Psychology. Human Perception and Performance, 44(6), 885–897.

Meltzoff, A. N. (1988). Infant imitation after a 1-week delay: Long-term memory for novel acts and multiple stimuli. Developmental Psychology, 24(4), 470–476. https://doi.org/10.1037/0012-1649.24.4.470

Mesoudi, A. (2011). Cultural Evolution. University of Chicago Press. https://www.press.uchicago.edu/ucp/books/book/chicago/C/bo8787504.html

Mesoudi, A., & Whiten, A. (2008). The multiple roles of cultural transmission experiments in understanding human cultural evolution. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1509), 3489–3501.

Miton, H., & Charbonneau, M. (2018). Cumulative culture in the laboratory: Methodological and theoretical challenges. Proceedings of the Royal Society B: Biological Sciences, 285(1879), 20180677. https://doi.org/10.1098/rspb.2018.0677

Okazaki, S., Muraoka, Y., & Osu, R. (2019). Teacher-learner interaction quantifies scaffolding behaviour in imitation learning. Scientific Reports, 9(1), 7543. https://doi.org/10.1038/s41598-019-44049-x

Risko, E. F., Richardson, D. C., & Kingstone, A. (2016). Breaking the Fourth Wall of Cognitive Science: Real-World Social Attention and the Dual Function of Gaze. Current Directions in Psychological Science, 25(1), 70–74. https://doi.org/10.1177/0963721415617806

Schmitz, L., Vesper, C., Sebanz, N., & Knoblich, G. (2017). Co-representation of others’ task constraints in joint action. Journal of Experimental Psychology: Human Perception and Performance, 43(8), 1480–1493. https://doi.org/10.1037/xhp0000403

Stout, D. (2002). Skill and Cognition in Stone Tool Production: An Ethnographic Case Study from Irian Jaya. Current Anthropology, 43(5), 693–722. https://doi.org/10.1086/342638

Stout, D. (2005). The Social and Cultural Context of Stone-knapping Skill Acquisition. In V. Roux & B. Bril (Eds.), Stone knapping: The necessary conditions for a uniquely hominin behaviour (p. 10). McDonald Institute for Archaeological Research.

Strachan, J. W. A., Curioni, A., Constable, M., Knoblich, G., & Charbonneau, M. (2020). A methodology for distinguishing copying and reconstruction in cultural transmission episodes. Proceedings of the Cognitive Science Society, 7.

Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. Behavioral and Brain Sciences, 16(3), 495–511. https://doi.org/10.1017/S0140525X0003123X

Vesper, C., Butterfill, S., Knoblich, G., & Sebanz, N. (2010). A minimal architecture for joint action. Neural Networks, 23(8–9), 998–1003.

Vesper, C., Schmitz, L., Safra, L., Sebanz, N., & Knoblich, G. (2016). The role of shared visual information for joint action coordination. Cognition, 153, 118–123. https://doi.org/10.1016/J.COGNITION.2016.05.002

Vesper, C., van der Wel, R. P. R. D., Knoblich, G., & Sebanz, N. (2011). Making oneself predictable: Reduced temporal variability facilitates joint action coordination. Experimental Brain Research, 211(3–4), 517–530. https://doi.org/10.1007/s00221-011-2706-z

Vesper, C., Van Der Wel, R. P. R. D., Knoblich, G., & Sebanz, N. (2013). Are You Ready to Jump? Predictive Mechanisms in Interpersonal Coordination. Journal of Experimental Psychology: Human Perception and Performance, 39(1), 48–61. https://doi.org/10.1037/a0028066

Visco-Comandini, F., Ferrari-Toniolo, S., Satta, E., Papazachariadis, O., Gupta, R., Nalbant, L. E., & Battaglia-Mayer, A. (2015). Do non-human primates cooperate? Evidences of motor coordination during a joint action task in macaque monkeys. Cortex, 70, 115–127. https://doi.org/10.1016/j.cortex.2015.02.006

Whiten, A., Goodall, J., McGrew, W. C., Nishida, T., Reynolds, V., Sugiyama, Y., Tutin, C. E. G., Wrangham, R. W., & Boesch, C. (1999). Cultures in chimpanzees. Nature, 399(6737), 682–685. https://doi.org/10.1038/21415

Wilkins, J. (2020). Learner-driven innovation in the stone tool technology of early Homo sapiens. Evolutionary Human Sciences, 2. https://doi.org/10.1017/ehs.2020.40

Wolf, T., Vesper, C., Sebanz, N., Keller, P. E., & Knoblich, G. (2019). Combining Phase Advancement and Period Correction Explains Rushing during Joint Rhythmic Activities. Scientific Reports, 9(1), 9350. https://doi.org/10.1038/s41598-019-45601-5

Wood, L. A., Kendal, R. L., & Flynn, E. G. (2013). Whom do children copy? Model-based biases in social learning. Developmental Review, 33(4), 341–356. https://doi.org/10.1016/j.dr.2013.08.002

19 Comments

  • comment-avatar
    Mathieu Charbonneau 7 September 2020 (16:05)

    Individual techniques/Joint techniques
    Many thanks to the three of you for this stimulating early draft.

    The examples you choose and the empirical work from joint action that you discuss in your draft focus on cases where the technique being transmitted/learned is one that, once learned, could be later used by a single individual on their own. The key “joint” aspect here is specifically that of social learning. What about techniques that, themselves, can be understood as joint technical actions or activities? For instance, group hunting techniques and sailing techniques—although some can be used alone—often are joint activities, involving the larger coordination of multiple individuals even beyond the learning episode. Often, in these techniques, it is not just a matter of learning specific sequences of actions, but also to learn one’s role in the larger coordinated system and, I would guess, how the larger system is structured. How would you articulate the transmission and use of these ‘joint techniques’ in the framework you propose?

  • comment-avatar
    Dan Sperber 8 September 2020 (01:32)

    Here we are all simultaneously learners and teachers
    Thank you, James, Ariana, and Luke for opening our webinar in this challenging way. I am convinced that the very fine-grained experimental investigation of joint action that the members of the Social Mind and Body group (SOMBY lab) and you in particular have been pursuing is potentially highly relevant to the study of the use and transmission of actual techniques. Your distinction between knowledge flow and information flow and the way it can inform experimental studies is particularly insightful.
    There are specific points you raise that I might want to discuss later in the week, but first, I would like to draw attention to a quite general issue about our interactions. This is addressed not just to the authors of this contribution, but to all the participants in the webinar.
    Most of our participants study prehistorical, historical, or ethnographic cases of actual techniques. Some of them have themselves developed relevant experimental methods to help them in their work. The general questions (about rigidity and flexibility in use and in transmission) addressed by this whole project are about the kind of real-life, concrete, historical, local, cultural practices they study.
    Is the kind of experimental and theoretical work illustrated in this first contribution relevant to this empirical pursuit? I believe it is. But how is it relevant, and how can this relevance be demonstrated? This is far from being self-evident. It cannot, anyhow, be so without experimentalists making the effort of attend to concrete issues raised by historical and anthropological studies and raising specific questions and making prudent suggestions. And not without historians and anthropologists applying to the work of the experimentalists their capacity to see things from the point of view of another culture – here the scientific culture of the experimentalists – and to appreciate the coherence and merits of this point of view well enough, to at least envisage that there might be areas of common interest and to engage a discussion of, methodological, theoretical, and concrete issues.
    In other words, here we are all simultaneously learners and teachers and both the flow of knowledge and that of information should be bidirectional.

  • comment-avatar
    Luke McEllin 8 September 2020 (15:17)

    Individual techniques/Joint techniques: Learning your role in a joint practice
    Dear Mathieu,

    Thank you for the question, and for bringing us all together!

    What you say is completely true, there are many kinds of joint practices or technical activities in which precise spatial and/or temporal coordination is required in order for that activity to be successful. Particularly in practices in which spatiotemporal contingency is crucial for the technique to be performed effectively (e.g. rhythmically pushing and pulling when cutting a log using a two-person saw), understanding how one’s actions relate to the actions of one’s co-actor is of utmost importance. Indeed, it has been demonstrated that when acting in a dyad, individuals are more likely to mimic other dyads performing a joint action compared to individual actors performing the same action (Tsai Sebanz & Knoblich, 2011; Ramenzoni, Sebanz & Knoblich, 2014), which has been interpreted as evidence that joint actions are underpinned by joint internal models which specify not only an individual’s part of a joint action, but also the spatiotemporal relations between their actions and those of their co-actors (McEllin, Knoblich & Sebanz, 2018; Sacheli, Arcangeli & Paulesu, 2018). Considering this, it is likely that the same cognitive mechanisms that allow one to use input from an expert to construct an internal model specifying the spatiotemporal configuration of an individual action, also allows one to use input from an expert (or group of experts) in order to construct an internal model which specifies an individual action (as a component of a joint action) whilst also specifying it’s contingencies with the actions of a co-actor (much in the same way that one’s actions may relate to each other bimanually). Indeed it is likely that having to account for spatiotemporal relations between one’s own and a co-actor’s actions – in particular how precisely these relations need to be specified – is likely to influence how flexibly or rigidly a technique is transmitted.

    The above applies mostly to dyadic (or at a push triadic) actions – whether this would be the case for a larger group activity like cooperative hunting is an open (and hopefully empirical) question. Perhaps it is necessary to understand one’s role in the system globally (i.e. what every member of the hunting party is doing), or perhaps it is simply enough to understand one’s role locally (adapting to those hunting party members who are close to you), allowing for an internal model that only accounts for the actions of others that are in my immediate vicinity. It is likely that whether one is required to learn about their role locally or globally has implications for whether or not transmission of a technique needs to be rigid or can afford flexibility.

    The examples you use of skilled joint actions also raise an interesting point. As you point out, many real life coordination tasks and techniques require people to coordinate complementary actions.. Take sailing as an example — a two-sail sailing dinghy requires two people to operate; one (often a novice) to operate the jib sail at the front, and the other (typically more experienced sailor) to control the mainsail and rudder. This sailing example is an instance where two people are coordinating complementary actions to achieve a joint goal, and one where a social learner is learning through active participation rather than mere observation, which raises yet more questions. What are the cognitive mechanisms for coordinating complementary actions in this way? The ability to resist automatically imitating a co-actor has been demonstrated to be crucial for coordinating complementary actions, for example (Sacheli, Tidoni, Pavone & Aglioti. 2013), and this skill has been described as a feature of some cultural traditions such as the Afro-Brazilian Congado (Lucas, Clayton & Leande, 2011) and traditional Greek group dancing (Sofianidis, Elliott, Wing, & Hatzitaki, 2014). If we assume that the goal is to learn the model’s task (e.g. to captain the boat and control the mainsail), what challenges does performing complementary actions pose for the learning of the model’s task?

    As experimental cognitive scientists, this raises many open questions that we could potentially test empirically. For example, how do the coordination and learning differ as a result of this being a joint skill? For highly interdependent practices such as sailing, the ability to coordinate is an inherent part of the to-be-acquired skill. In addition to being able to control the mainsail and rudder, the experienced sailor also needs to learn how to monitor and anticipate the novice, how to exchange signals effectively in rainy and windy conditions, and even when to be flexible and when to be rigid with regards to their plan. How do novices come to learn how to coordinate in this context? Moreover, how do experts transmit such skills?

  • comment-avatar
    Luke McEllin 8 September 2020 (15:20)

    Here we are all simultaneously learners and teachers: A unique opportunity!
    Dear Dan,

    Thank you for your comment. We wholeheartedly agree with what you say. As you can see from our chapter brief, the supporting evidence for our position is largely inferred indirectly from basic research into joint action coordination. Targeted empirical work into the joint action mechanisms subserving social learning is still in its early stages, which makes this a uniquely opportune time to reflect on our theoretical models and to generate hypotheses and ideas that are grounded and guided by existing historical and anthropological research, and that may prove interesting and useful to researchers in these disciplines. Our sincere thanks to you and Mathieu for organising this book and webinar to stimulate this kind of dialogue.

  • comment-avatar
    Dietrich Stout 10 September 2020 (21:20)

    Broader context
    Hi! Thanks for the interesting paper. I have two inter-related questions/comments:
    1) Would it be beneficial to situate “social learning as action coordination” within a broader frame or taxonomy of teaching? For example: Kline, M. A. (2015). How to learn about teaching: An evolutionary framework for the study of teaching behavior in humans and other animals. Behavioral and Brain Sciences, 38, e31.
    2) This might also let us consider indirect effects, for example the role of action coordination or interactive alignment in establishing affective and social bonds that might facilitate other forms of teaching as well.

  • comment-avatar
    Ildikó Király 11 September 2020 (09:30)

    Is there unidirectionality in a natural pedagogy perspective?

    Dear Ariana, Luke and James,

    I enjoyed your paper very much. I found especially thought provoking your main claim on social learning seen as a coordination process, and found challenging your starting point that “ … current discourse on observational learning through imitative copying typically presupposes a unidirectional type of interaction which imposes rigidity on the learned behavior”.

    I wonder, as the comment of Dietrich on Broader context also raise, how you relate your approach to the broader teaching context, especially to the natural pedagogy model.

    Natural pedagogy as a model highlights and emphasizes that most ’social learning situations’ are instances of ostensive communication, in which both parties are involved and play special roles. Not only the teacher has intentions to share knowledge and guide the novice’s attention to relevant information, but the students are also sensitive to and apprehend the communicative intent of the demonstrator.

    It is true, however, that previous empirical works, especially in the field of early social learning (and I focus on these here) have preferentially used implementations that have endorsed unidirectionality – e.g. choices measured after observing a sampling situation (Gweon, 2011; Ma and Xu, 2011), re-enactment measured after a one shot demonstration, even in the absence of the demonstrator (Király, 2009; Király and Gergely, 2013). Despite these settings, nonetheless, still there is evidence that children handle the demonstration as bidirectional. Infants attribute preference after observation of taking the rare objects out from a container full of other objects (Ma and Xu, 2011). Moreover, infants only show more frequent re-enactment of the demonstrated subefficient action in the presence of the model (in contrast to the condition where the model was absent) when the demonstration is about a complex tool use, not a simple bodily action, suggesting that modulations and/or feedback is requested from the partner.
    As Gergely and Jacob (2012) argue, children take even strong sampling, namely observing an agent’s selective sampling dependant on the targets’ relevant properties ‘as part of communicative, not instrumental, agency’.
    This natural pedagogy perspective supports in general the claims of the paper discussed, yet raises the specific issue on the specificity of such coordinated episodes for technical know-how transmission. In my view, the pedagogical approach could provide framework also for the way how the range of applicability for the learnt information is acquired. (Ostensive communicative context offers a segmentation of the interaction context itself from the to-be learnt, relevant information – highlighting the features of the context in which the target behaviour is adequate, despite its potential opacity). I wonder whether the distinction between knowledge flow and information flow would allow to form some predictions as well?

  • comment-avatar
    Miriam Haidle 12 September 2020 (08:20)

    Ways of transmission, transmitted entities, and environmental context
    Dear James, Ariana and Luke,
    thank you very much for the inspiring thoughts. They open the discussion about transmission of techniques in two directions. One is thinking about ways of transmission with different levels of social involvement (unidirectional, bidirectional) of the learner and the expert in different learning settings (observational, action coordination, others and mixed). The other is reflecting on the transmitted entities. They can comprise individual, joint, or complementary activities, which can range from very simple and small to very complex and large. Successful transmissions of the different entities require different social engagements of learner and expert.
    Yet, another parameter comes to my mind: the variability of the environmental context, in or to which the learned entities are applied. While in the technique “driving a car” the car itself and the situations can vary very much, the setting of a catholic service is largely fixed and accordingly the technique of the liturgy. Depending on the potential variability of the context, the transmitted entities have to be rather abstract or very concrete, and their use and transmission correspondingly adapted, flexible or rigid.

  • comment-avatar
    Rita Astuti 12 September 2020 (17:10)

    The messiness of the real world
    Dear James, Ariana and Luke,
    Thank you for your paper, which was posted immediately after I submitted mine (I was late!)
    My paper, as you will see in a few weeks’ time, is about sailing (in Madagascar) and is highly ethnographic. As an ethnographer, I observe what people do (for example, the fact that they change their sailing technique), I ask them questions (for example, why did you change the way you rig the canoe? How did you learn the new technique?) and I listen to their discussions (for example, about the reckless nature of young sailors). This gives me access to folk theories of learning and of innovation processes, and some very imprecise data on how either learning or innovation actually take place.
    The body of experimental work you discuss in your paper sits at the opposite end of the methodological spectrum. Although I read you to say that one of the selling points of your approach – which acknowledges that “the social learner is not a parasitic spectator or information scrounger, but an actively engaged participant in the behaviour” – is that it embraces (rather than trying to control for) the complexity of social learning, from my stand-point you have barely scratched the surface of that complexity. This is not a criticism, it is a genuine call for collaboration (echoing what Dan says in his post). Had it not been for the coronavirus pandemic, I would have spent two months in Madagascar this summer and my intention was to try to tackle some of the questions raised by this workshop in a more precise way than I have been able to do with my ethnographic approach. To be honest, I am not sure that I would have known what to do, but your paper suggests potential new ways of asking questions about how children and adults learn and innovate their sailing techniques. As you can imagine, in the wild – on a small outrigger canoe whose hull “breathes”, as the local say, as it expands and contracts under the pressure of the waves – it is impossible to control for anything! You have (typically) two sailors who coordinate their actions in response to the wind, the sea, the aim of their sailing expedition (fishing, visiting relatives, transporting a coffin or a pig or some tourists), and much more besides. At times, one is more experienced but the other is stronger; at other times, one is in charge of the expedition (and is prepared to take more risks with the weather) and the other owns the canoe (and is therefore more conservative); or one is convinced that a bad spell has been cast on the canoe and this affects his sailing performance, etc. The hard methodological question, then, is how we can bridge the precision of your methods with the messiness of the world I’m studying.
    I look forward to discussing this further and perhaps start some kind of lab-to-wild collaboration!

  • comment-avatar
    Giulio Ongaro 12 September 2020 (21:15)

    Perspective sharing / embodiment
    My observations of social learning among the Akha – a group of non-literate swidden farmers who live in the remote highlands of Laos – resemble those of Little and Lancy (whom you cite) on the Asabano of Papua New Guinea, insofar as children are often prohibited from joining the activities of the elders. During communal activities as varied as hunting, house building, pig killing, meat cutting, rice threshing, coffin making, fishing, fish trap making, ritual sacrificing etc. children and young teenagers are not supposed to engage with, or ask questions to, their elders (this is a highly gerontocratic society and doing so is improper behaviour). They congregate a few meters from the scene and watch attentively without interacting. Elders, on their part, disregard the presence of children. Obviously, there are numerous occasions in which learning through bidirectional action-coordination takes place (mostly between children and their own parents or older siblings), but the unidirectional type of learning that you suggest is ‘rare’ because ‘in the real world observers are more than just spectators’ isn’t rare at all in an Akha community: in many instances the mechanisms of transmission are purely top-down and children are really just spectators.
    They are, however, a particular type of spectators. As they attend the scene, they do not merely watch, but actively engage with each other. When I looked at this closely clustered assembly of young attendees, I could see a lot of intra-group communication going on. There is undertone chatter about the scene. There is finger pointing. There is whispering on the ear of a close by peer. There is the occasional exclamation. There is observation of another peer’s observing. There is, in short, a lot of what goes under the name of ‘perspective sharing’. What this does to individual spectators is an enhancement of their vicarious learning and perceptual access to the scene. Collectively, they grasp aspects and nuances of the practice that they would not notice or apprehend singly. These contexts offer an collective-induced ‘education of attention’ – the term that ecological psychologist James J. Gibson used to characterise perceptual learning.
    If I remember right, Luigi Cavalli-Sforza once distinguished social learning into vertical (from parents to their own children), oblique (from elders to non-related young people) and horizontal (among peers of the same age group). In the scenario I have just described people seem to exploit all of these three routes simultaneously. I wonder whether this ethnographically informed and socially expanded perspective on learning could be incorporated into your account.

    Another comment I have relates to your use of the concepts of ‘knowledge’ and ‘information’. These are theoretically heavy terms. I’m not sure whether you have the enough space to address the theoretical debates revolving around these concepts – feel free to disregard my comment – but I was wondering whether you could say something more about their ontological nature. I am mentioning this because there are schools of philosophy of mind and cognitive science that would reject outright the idea of knowledge as ‘context-invariant representation’. I am thinking mostly of the enactivism/embodiment strands in philosophy of mind. Making references to these approaches, anthropologist Tim Ingold offers an account of culinary recipe learning that differs substantially from yours, as he emphasises the embodied nature of social learning. I’m copying it here:

    “How do novices actually learn to cook (rather than to reproduce recipes)? They do so, of course, by working alongside already skilled practitioners in the kitchen. Though I have never had the proverbial opportunity to teach my mother to suck eggs, I did have the opportunity many years ago to teach my (then) small daughter how to break them, in the course of learning how to make an omelette. This operation requires no small degree of skill. Keeping a firm hold of the egg, you have to strike it against the edge of a cup or mixing bowl so as to achieve a clean crack of sufficient extent to enable you subsequently to split the shell easily into two halves, releasing the contents into the bowl. If the force of the strike is too light the shell will not crack, or the crack will be so short that when you try to split open the shell you have to apply so much pressure that the whole shell is crushed into pieces, leaving shards of the shell in the bowl and fingers covered in egg. If the force is too great the entire egg splits on impact, and most of the contents end up all over the work surface rather than in the bowl. What makes the task especially difficult is that the force required is not constant. It varies from egg to egg, depending on the thickness of the shell. The problem for the novice is this. How do you know how thick the shell is when you cannot see until the egg is broken? There is a trick to this that you will not find in any recipe book, but which experienced cooks use so routinely that they are scarcely aware of it. First tap the egg lightly against the edge of the bowl. Listen for the sound. This will tell you how hard to strike next time so as to achieve a clean crack. Thin shells and thick shells sound differently when they are tapped. So this is how I teach my daughter to break eggs. She holds the egg in her down-turned hand. I hold her hand in mine, also down-turned. Together we strike the egg against the bowl, first lightly and tentatively, then firmly and with controlled force. In that way, my daughter gets the feel of it for herself, and the sound of it too. She is also looking what she is doing. Through repeated trials she becomes progressively better able to adjust her striking force to a multisensory monitoring of the task as it unfolds, seamlessly combining vision, hearing and touch.”

    The whole paper (Ingold, ‘Making, growing, learning’ 2013) is a critique of the idea of ‘knowledge’ conceived as ‘mental representation’ and of ‘transmission’ accordingly. Personally, I don’t find his critique of cognitivism particularly effective, mostly because he does not engage deeply with the philosophical literature and with the empirical studies on the topic. But I was wondering whether you are aware of this particular criticism, what you make of it, and how you situate yourself in these theoretical discussions in the philosophy of cognitive science (again, feel free to ignore all this if you think it’s beyond the remit of the chapter).

    Thank you for your interesting paper.

  • comment-avatar
    György Gergely 13 September 2020 (15:24)

    Assumption of Unidirectionality – In the Lab vs in „Real Life”
    Week 1: Gyuri’s comments on ’Social Learning and Action Coordination’ by James, Ari, and Luke

    Let me start by saying that I found the first contribution to our webinar by James, Ari, and Luke (JAL) not only thought-provoking and insightful, but also a ’brave’ and ambitious. And I mean ’brave’ here truly as a compliment: as I am aware of how hard it must be to start such a broadly interdisciplinary dialogue and to do it in such a challenging manner as they did. Also, as Dan’s was quick to focus our joint attention in his comments to the fact that the participants of this webinar represent a broad range of different expertise and come from a variety of ’scientific cultures’, which makes it an even harder task to conduct a substantive dialogue requiring „experimentalists to make the effort to attend to concrete issues raised by historical and anthropological studies… and historians and anthropologists applying to the work of the experimentalists their capacity to see things from the point of view of another culture – here the scientific culture of the experimentalists”.
    So in this Sperberian spirit, let me raise some questions about about JAL’s arguments concerning the critical points about the pervasive problem of the Assumption of Unidirectionality (AoU) in previous accounts of Social Learning (SL) (and ameliorate the critical edge of my comments by some self-critical points I’ll make as a fellow experimentalist).

    Social Learning as Action Coordination: The problem of unidirectionality
    JAL advance a new (and radical) proposal that the key cognitive mechanisms that are central and constitutive of ’true’ or ’real’ Social Learning (SL) „may not be specialised for learning per se, but rather for action coordination”. They prepare the ground for this argument by criticizing a variety of ’standard’ and traditional approaches to SL pointing out that they all share the basic „assumption of unidirectionality” (AoU): that „information goes only one way, from the model to the learner, and … feedback from the learner (which would allow the model to adapt to and accommodate the learner’s needs)… is considered something extra to the SL rather than an inherent part of it.” JAL argue that in these standard views „the only thing that is social about observational SL is the fact that the observation corresponds to another individual.” Nevertheless, they claim AoU is commonly shared by a broad spectrum of approaches to SL including Heyes, 1994; Hoppitt & Laland, 2013; Henrich, 2016; Henrich & McElreath, 2003; Mesoudi, 2011; Goodall, 1964; Whiten et al., 1999; Biro et al., 2006; and…, Grice forbid, Gergely, Bekkering, & Király, 2002; and Gergely & Csibra’s (2006) natural pedagogy theory of relevance-guided teaching by ostensive communicative demonstrations in humans – (a slightly painful point of misunderstanding between experimentalists and experimentalists on which I’ll not dwell on, as this issue has been taken up as the focal point of Ildikó Király’s cogent comments about whether (and why not) natural pedagogy shares the AoU).
    According to JAL the AoU is basically mistaken as „truly one-way observation is rare, and in the real world observers are more than just spectators.” „Drawing on research in cognitive science in joint action and coordination” they argue that „a key aspect of SL… that is overlooked… is the many varied opportunities for coordination between models and learners.” It is from this perspective that they derive their central proposal that „the cognitive mechanisms that are most interesting in SL… may not be specialised for learning per se, but rather for action coordination”.
    My first question concerns what JAL’s reference to „the real world” really denotes in this context: „the real world” where, as they say „observers are more than just spectators” and where the social learner is not „a parasitic spectator or information scrounger, but an actively engaged participant in the behaviour”?. It certainly maps well onto the real world of the SOMBY lab’s ’experimental culture’ created and expertly studied by cognitive scientists like JAL, which consists of the behavioral performance of participants induced by the kind of experimental joint action and coordination tasks that involve shared and jointly represented cooperative goals which are explicitly prespecified by the (high-status) Experimenter’s instructions to the paqrticipants. I can attest that these are superb experiments, but I wonder to what degree and scope do they represent „the kind of real-life, concrete, historical, local, cultural practices” that Dan called our attention to as the (prescribed) shared goal that our interdisciplinary webinar intends to focus on and (jointly and interactively) study?

    In particular, does JAL’s reference to the „real world” cover that part of the natural „real world” which also includes social transmission of technologies and traditions in non-human animal species, like apes or song birds, that most of the models of SL (criticised by JAL for their mistaken adherence to the AoU) have been designed to capture (with the exception, I should add, of natural pedagogy theory)? Well, I am not convinced that it does, as the „minimal working examples of SL” such as learning nut-cracking techniques in chimpanzees (e.g., Biro et al., 2006) clearly illustrate that in this non-human domain of social transmission of technological skills there seems to be no significant role of „feedback from the learner” for „the model to adapt to and accommodate the learner’s needs”. Apes do not modify their nut cracking behaviour to facilitate the naive juvenile’s learning task – and this is a point of rare agreement among comparative researchers (with the single exception of two unique observations by Boesch, 1991, never again observed by others either in the wild or in captivity). So in this domain at least, the AoU seems largely correct and accounts for the evidence of this kind of „real life” social transmition which, in fact, seems very unlike the „real world” studied in the SOMBY lab where information flaws bi-directionally to optimize joint efficiency of cooperating participants performing joint action tasks to realize shared goals.

    This may sound as an annoying and unfair criticism, I know, to which the plausible answer is that the domain of „real life” that JAL refers to concerns the cognitive mechanisms that characterize the social transmission of technologies in human cultures. So let me raise the same question again: how representative are the cognitive processes and bi-directional flow of information demonstrated in the experimental world of explititly prescribed joint action and coordination tasks of the domain of „real life” human social transmission of cultural technologies and traditions? Here the answer is likely to be more complex and less trivial: nevertheless, I want to call attention to a significant mismatch between the ’enlightened’ and optimal flow of bi-directional information induced by many of the explicitly specified and co-represented joint goal tasks in the „real world” of the CEU’s ’experimental cognitive science culture’ of social cooperation and coordination, on the one hand, and the ’brute social world’ of historical cultural transmission of technologies and traditions studied by anthropologists, archeologists, historians, etnographers, and cross-cultural researchers. Let me mention two examples of traditional and institutionalized forms of human cultural practices that are specialized, ont he one hand, for the social transmission of technological skills and knowledge, and, on the one hand, for the transmission of the kind of variously opaque practices and know-how of social traditions.
    The first is the historically widely documented social institutions of apprenticeship (exemplified in guilds of various trades, I think – but remember, I am no historian or anthropologist, just another – and, being politically correct, I am not gonna say ’weird’- experimenter from the CEU’s infancy lab). In institutionalized societal forms of apprentiship knowledgeable and high-status expert masters and dedicated (or forced, but certainly underpaid) juvenile novices live together in rigid (and often brutal and impersonal) and hierarchically organized social settings for long-long years (should I compare this – probably non-optimal length of time – to the many years it takes for a juvenile chimp to learn nut-cracking by observational SL?) sharing the non-interactively joint societal goal to acquire – and transmit to the next generation – the expert master’s knowledge and the technological know-how of the trade. During the initial years of apprenticeship, I hear, the novice is relegated to mostly non-skill-relevant low manual labour of cleaning and carrying heavy stuff around, being punished and starving, etc., which, however, allows him to get (indirect) glimpses of the Master at work – but without much direct interaction allowed and no learner feedback being tolerated or go unpunished. In a similar vein, ’education’ in schools in the 19th century centered on rote learning, obediance, and punishment were not characterized either – if I remember correctly – by co-monitored and free bi-directional flow of information between learner and teacher to maximize joint task efficiency of task performance and optimize joint cost/benefit ratio of knowledge transmission. Furthermore, apprenticeship in trade – and to some degree – education in Victorian schools were social institutions specialized to serve inter-generational transmission of (at least some) causally constrained technological skills and know-how, I mean ’objective’ cultural knowledge, where causal efficiency, successful performance, visible and evaluable output were all available to constrain, motivate, and optimalize bi-directional information exchange.
    The other example is the realm of institutionalized social settings specialized for cultural transmition of opaque knowledge of social traditions and cultural practices of opaque skills, such as rituals, taboos, tranditional healing techniques, and quasi-religious activities where historically „real life” social transmission has been taking place in institutionalized and hierarchically structured rigid and traditional forms of expert-novice activities specialized to serve the ’training’ of specialists as in shamanism, medicin-men, (and women?), traditional healing, etc. (a great source is Pascal Boyer’s recent book, 2018). Here again the transmission of such opaque ’skills’ by expert practitioners to novices was often performed in secrecy, and institutionalized in the form of initiation practices which typically involved mostly uni-directional information flow associated with rather brutal and painful pay-offs for the – often non-voluntary – novices (rather unlike the case of payed novice participants of joint action and coordination tasks in the SOMBY lab.) (Clearly, both of these institutionalized societal forms that were specialized for social transmission of various (and variously opaque) cultural technologies and traditions could have learned a great deal from the SOMBY lab’s discoveries to help them optimize their cost/benefit ratio of efficiency of knowledge transmission had they have the chance (and had they chosen to live with it) to internalize the cognitive mechanisms supporting the bi-directional information flow that characterize performance in the joint cooperation and coordination tasks developed by the CEU’s cognitive scientists).

    The points made above simply express my view (resonating with Dan’s advise) that we should be super-cautious as experimentalists when generalizing our results and insights from our ’experimental cultures’ to the ’brute reality’ of actual cultures both pre-historic and current-day. And this caution is aimed not specifically at JAL or the SOMBY lab, I think natural pedagogists studying early cognitive development, like Csibra, Király, and Gergely, are demonstrably just as fallible experimentalists who are partly to be blamed for inducing the kind of interdisciplinary misunderstandings of their theoretical contributions that plague the dialogue between different scientific cultures – and do so even within different groups of experimentalist cultures studying cognitive science. As an example, this kind of misunderstanding can happen even after five years of fruitful and intensive cooperation under the joint goal of our SOMICS Synergy Project between the experimentalist of the joint action lab (SOMBY) and the experimentalists of the cognitive development lab (COMICS) where some of the natural pedagogists reside. So here is the reason (in my self-critical reconstruction) of how our experimental work and the theory of natural pedagogy (NP) based on it could have (mistakenly) ended up in the list of observational SL theories that our friends and colleagues, JAL have criticized for embracing the AoU. And admittedly they did so in the most polite and friendly way:

    „Another point in favour of using observational learning as a minimal working example is that this kind of learning is very important early in human development: babies cannot act independently and so learning through watching and imitating adults is very important to the acquisition of early motor skills, together with the ability to discern who and what to imitate (Gergely, Bekkering & Király, 2002; Gergely & Csibra, 2006;…). Given that observational learning is the first kind of social learning that humans are capable of, and is also observed in our closest phylogenetic relatives, it is tempting to think of this also as a conceptual foundation for other kinds of social learning that humans are capable of later in life.”

    Now Ildikó Király in her comments to JAL has spelled out a number of reasons and identified relevant behavioural signatures that indicate why our experimental demonstrations of teaching opaque and sub-efficient cultural skills to preverbal infants through ostensive communicative manifestations (such as lighting up a touch-sensitive lamp by contacting it with one’s head – rather than by hand), in fact, imply a bi-directional communicative flow of information between teacher and learner that makes NP an evolved form of communicative learning mechanism of mutual design specialized for the social transmission of opaque, but relevant cultural knowledge in humans. These signature features (see Ildikó’s comments) show how ostensive demonstrations induce selective and context-sensitive, relevance-guided inferential learning in human infants which should clearly differentiate NP from SL by observational imitative behaviour copying. I don’t want to argue this here in more detail as Ildikó has already done so. Rather, the point I want to make is that in „real life” situations where communicative transmission of cultural knowledge is taking place the us of NP is typically not a one-shot affair of rigid knowledge demonstration inducing trust-based imitative behaviour copying. However, I realize that this interpretation has been strongly (and unfortunately) suggested and propagated by our experimental paradigms that have relied mostly on one-shot ostensive demonstrations. NP in „real life” typically involves repeated interactions of joint back-and-forth communicative information exchange often involving turn-taking bouts of ostensive manifestations of communicative intent by the teacher coupled with informative and referential manifestations serving relevant information provision followed by (or induced by) ostensive communicative information requests by the recipient naive learner, where such communicative interactions involve mind-reading of communicative and informative intentions as well as co-monitoring and modulation of demonstrations that serve relevance-guidence and corrective feedback. Importantly, rigidity and lack of modulation of the model’s behaviour are not constitutive features of NP: undoubtably, however, , because this specialized system serves the manifestation and transmission of causally and/or teleologically opaque but relevant cultural and social knowledge (both technological and traditional) by knowledgeable communicative agents to naive communicative recipients, our use of one-shot experimental demonstration paradigms to demonstrate NP appears to have induced the unintended interpretation of the necessary rigidity of modelling opaque knowledge skills.
    Finally, to shed further experimental light on these claims, let me direct our interdisciplinary attention to the highly informative large-scale multiple transmission chain study by Morgan et al. (2014) entitled ’Experimental evidence for the co-evolution of hominin tool-making teaching and language’. Nature Communications, 6:6029 . This remarkeable study tested the capability of five different social learning mechanisms ((i) reverse engineering, (ii) imitation/ emulation, (iii) basic teaching, gestural teaching and (v) verbal teaching) to transmit Oldowan stone knapping techniques across multiple transmission events involving (current-day) knowledgeable experts and naive learners. Across six measures the study found that only (iv) gestural communicative teaching interactions (i.e., natural pedagogy) and (v) verbal teaching interactions could result in successful transmission of this most ancient, but in many respects causally and teleologcally opaque technological skill, while there was no evidence that observational SL involving imitative behaviour copying or emulation could enhance or achieve successful transmission and stabilization of the earliest hominin tool-making skill of stone knapping techniques that were around and passed on across generations since at least 2.5 mya.

    György Gergely
    CEU Professor
    Cognitive Developmental Center
    Department of Cognitive Science
    Central European University
    7, Oktober 6th Street,
    1051 Budapest, Hungary
    e-mail:gergelygy@ceu.edu

  • comment-avatar
    Dan Sperber 13 September 2020 (16:45)

    Youtube technical tutorials involve non-interactive natural pedagogy
    Yes, as pointed out by Ildiko and by Gyuri, natural pedagogy typically occurs in a direct interaction between teacher and learner, but it can also occur in a non-interactive manner. The clearest example may be provided by online technical turorial, which for many young people (and older ones too) often and more and more replace interaction with teachers or reading manuals.See for instance https://www.youtube.com/watch?v=ZJy1ajvMU1k&t=54s&ab_channel=GordonRamsay and ask yourself, as a psychologist and/or anthropologist, how you might investigate this very modern (and very efficient) form of technical skill transmission

  • comment-avatar
    James Strachan 14 September 2020 (12:06)

    Broader context: Coordination mechanisms and teaching
    Thank you all for your comments, which we have found extremely insightful and thought-provoking. We will be responding to all comments, but will take our time to ensure that we give each response the consideration it deserves.

    Below is our response to Dietrich Stout’s comment.

    ———————————

    Thank you for your comment, Dietrich.

    We agree that Kline’s taxonomy of teaching is highly relevant to our paper — indeed, this paper inspired many of our early discussions of social learning as action coordination, given that what Kline calls direct active teaching is the clearest example of bidirectional, mutually adaptive coordination to achieve a learning goal. In particular, Kline mentions both haptic guidance and turn-taking in her description of direct active teaching, both of which can be important features of coordination.

    We believe that our framework is wholly consistent with Kline’s, but takes what is essentially a purely functionalist approach and proposes a set of psychological mechanisms that may play a role in producing the behaviours that Kline lays out. Importantly however, these mechanisms do not only apply to teaching scenarios, as they are specialised for coordinating joint actions rather than just for social learning.

    As an additional note, Kline also emphasises the implications of her framework for comparative psychologists who are interested in identifying teaching behaviours in non-human animals. As she notes, while there is scarce evidence for direct active teaching there is evidence that non-human animals are capable of teaching by other means (such as meerkats engaging in opportunity provision by bringing their pups disabled scorpions). A mechanistic understanding of teaching and social learning by action coordination could provide insight into this interspecific difference by investigating coordination and nonverbal communication mechanisms in non-human animals. For example, research along this line has already shown that chimpanzees can use communicative gestures to support and sustain coordination in the event of a break-down (Voinov et al., 2020), and are capable of flexibly adjusting their pointing gestures according to the local context (Tauzin et al., 2020).

    Below, we describe how coordination mechanisms can play a role in the other definitions of teaching in Kline’s framework (teaching by: social tolerance, opportunity provision, stimulus or local enhancement, and evaluative feedback) and then go on to describe some of the indirect consequences of coordination and how these may affect social teaching interactions.

    ———————————
    Teaching by social tolerance
    ———————————
    Teaching by social tolerance refers to where teachers allow intrusive observation by learners. While we are not aware of much empirical work directly relevant to tolerating observation, there is work showing that people are willing to tolerate inefficiencies and inconvenience in joint action that they would not tolerate in individual actions. For example, there is work showing that individuals are willing to incur additional motor costs when coordinating with a partner in order to maximise the overall efficiency of the joint action (Török et al, 2019; 2020; Strachan & Török, 2020). Furthermore, coordination carries an inherent objective cost — coordinated actions are more error prone than individual actions, require more cognitive resources to predict and monitor, and often involve additional motor costs in the form of exaggerated movements — but human adults will still prefer to coordinate with a partner than to act alone (Curioni et al., 2020).

    This latter study also raises an important point (although on a slight tangent to the issue of social tolerance) regarding the evolutionary origins of teaching behaviour. As Kline and others point out, teaching is difficult to explain in evolutionary terms as it appears to be a form of altruism (the teacher gains no immediate benefit from the social learning interaction, and may incur substantial cognitive, physical, and opportunity costs). However, as Curioni et al. (2020) show, people also incur costs when coordinating and yet still choose to do it. Under a rational framework, these objectively measurable costs can only be explained if actors derive some reward function from acting together. Measuring the utility calculations of teachers during interactions with learners may shed some insight into what motivations drive this behaviour in humans.

    ———————————
    Teaching by opportunity provision
    ———————————
    Teaching by opportunity provision is where a teacher creates an opportunity for the learner to practice or observe a behaviour that would otherwise not be possible. Assigning chores is an example that Kline provides, as these can provide learning opportunities for a large variety of tasks. Given that these chores are often jobs that the model would otherwise have to complete themselves (e.g. washing the dishes, sweeping the floor, etc.) this raises questions about how models distribute tasks to the learner. Co-actors frequently divide labour in joint actions (for a review see Wahn, Kingstone, & König, 2018), take into account their co-actors’ constraints (Vesper et al., 2013; Schmitz et al., 2017) and appear to be motivated by rational principles at the level of the dyad (see description of Török et al. papers above). As far as we know, it remains an open question how models distribute tasks in social learning contexts to account for the learner’s skill and access requirements.

    ———————————
    Teaching by stimulus or local enhancement
    ———————————
    Teaching by stimulus or local enhancement is where a teacher stimulates the learner’s interest in a stimulus or location, allowing the student to discover or develop the skill through increased attention to relevant information. There is an extensive literature into the mechanisms of such triadic joint attention and gaze following. Gaze following (the apparently automatic redirection of attention in the direction that another face looks) has been shown to result in increased liking for objects that are the subject of joint attention (Bayliss et al., 2006; Bayliss et al, 2007).

    However, the role of joint attention is more than simply directing a partner towards a feature of the environment: results with eye-tracking show that, when listening to a speaker describe a scene, the listeners with the best comprehension of the narration were those whose eye-gaze fixations were most similar to the narrator’s had been (Richardson & Dale, 2005). That is, attending to the same things with the same relative timing was important for understanding what was being said. A similar effect has been described in real-world skills — recreational tennis players who were trained to use the visual search strategies that expert players used were better able to anticipate the direction of their opponents’ tennis strokes than participants trained on a placebo strategy (Williams et al., 2002).

    ———————————
    Teaching by evaluative feedback
    ———————————
    Teaching by evaluative feedback refers to where a model will encourage or discourage a student from producing a particular behaviour. Providing evaluative feedback requires recognising errors in another individual’s production, and there is evidence that these error detection mechanisms are specialised in joint action. Vicarious reinforcement learning signals are seen in models when they are instructing others (Apps et al, 2015). Similarly, during joint actions, EEG signals of error-related processing indicate that people treat a co-actor’s musical errors as if they were their own when these affect the joint outcome (Loehr et al., 2013), and participants show similar levels of post-error slowing (a compensatory mechanism indexed by reaction times characterised by slower responses on the trials immediately following an error) for a partner’s mistakes as for their own (De Bruijn, Mars, Bekkering, & Coles, 2012; Schuch & Tipper, 2007). Finally, computational modelling work on how people use rewards and punishments to teach indicates that, rather than simply using evaluative feedback as reinforcement, such feedback is instead used to communicate information about the desired behavioural outcome (Ho et al., 2019).

    ———————————
    Indirect consequences of coordination
    ———————————
    As you say in your second point, there are also indirect consequences of coordination to consider. Coordinating with a partner has long-term affective consequences towards the partner (Cross et al., 2020), resulting in both greater helping and increased cooperation, and — potentially important for interactions between an expert and a novice learner — this does not appear to rely on the quality of the coordination outcome (Cross, Wilson, & Golonka, 2019). There are also prosocial consequences for specific coordination mechanisms such as joint attention — when participants are instructed to initiate an episode of joint attention (ie. a face follows their gaze to an object rather than they follow the face’s gaze), this leads to increased affiliation for the faces and objects involved. These prosocial consequences of coordination may play an important role in moderating models’ or teachers’ attitudes towards learners, such as making teachers more likely to tolerate observation by students with whom they have previously coordinated.

    The other edge to this sword, of course, is that failure to coordinate or violations of these mechanisms may result in negative consequences — for example, individuals who consistently provide misleading gaze cues (that is, shifting their attention away from the upcoming location of a target) are judged as more untrustworthy than those that provide valid cues to the target location (Bayliss & Tipper, 2006; Bayliss et al., 2009; Strachan et al., 2016), these judgements are implicit and long-lasting (Strachan & Tipper, 2017; Strachan et al., 2020), and participants are less likely to invest real money in these individuals in one-shot economic games (Rogers et al., 2014).

    ———————————
    References
    ———————————
    Voinov, P. V., Call, J., Knoblich, G., Oshkina, M., & Allritz, M. (2020). chimpanzee coordination and potential communication in a two-touchscreen turn-taking Game. Scientific reports, 10(1), 1-13.
    Tauzin, T., Bohn, M., Gergely, G., & Call, J. (2020). Context-sensitive adjustment of pointing in great apes. Scientific reports, 10(1), 1-10.
    Török, G., Pomiechowska, B., Csibra, G., & Sebanz, N. (2019). Rationality in joint action: Maximizing coefficiency in coordination. Psychological science, 30(6), 930-941.
    Török, G., Stanciu, O., Sebanz, N., & Csibra, G. (2020). Joint action planning: co-actors minimize the aggregate individual costs of actions. In Proceedings of the 42nd Annual Conference of the Cognitive Science Society (pp. 295-301).
    Strachan, J. W., & Török, G. (2020). Efficiency is prioritised over fairness when distributing joint actions. Acta Psychologica, 210, 103158.
    Curioni, A., Voinov, P., Allritz, M., Call, J., & Knoblich, G. K. Crazy for you! Understanding Utility in Joint Actions. In Proceedings of the 42nd Annual Conference of the Cognitive Science Society (pp. 3268-3274)
    Wahn, B., Kingstone, A., & König, P. (2018). Group benefits in joint perceptual tasks—A review. Annals of the New York Academy of Sciences, 1426(1), 166-178.
    Vesper, C., van der Wel, R. P., Knoblich, G., & Sebanz, N. (2013). Are you ready to jump? Predictive mechanisms in interpersonal coordination. Journal of Experimental Psychology: Human Perception and Performance, 39(1), 48.
    Schmitz, L., Vesper, C., Sebanz, N., & Knoblich, G. (2017). Co-representation of others’ task constraints in joint action. Journal of Experimental Psychology: Human Perception and Performance, 43(8), 1480.
    Bayliss, A. P., Paul, M. A., Cannon, P. R., & Tipper, S. P. (2006). Gaze cuing and affective judgments of objects: I like what you look at. Psychonomic bulletin & review, 13(6), 1061-1066.
    Bayliss, A. P., Frischen, A., Fenske, M. J., & Tipper, S. P. (2007). Affective evaluations of objects are influenced by observed gaze direction and emotional expression. Cognition, 104(3), 644-653.
    Richardson, D. C., & Dale, R. (2005). Looking to understand: The coupling between speakers’ and listeners’ eye movements and its relationship to discourse comprehension. Cognitive science, 29(6), 1045-1060.
    Williams, A. M., Ward, P., Knowles, J. M., & Smeeton, N. J. (2002). Anticipation skill in a real-world task: measurement, training, and transfer in tennis. Journal of Experimental Psychology: Applied, 8(4), 259.
    Loehr, J. D., Kourtis, D., Vesper, C., Sebanz, N., & Knoblich, G. (2013). Monitoring individual and joint action outcomes in duet music performance. Journal of cognitive neuroscience, 25(7), 1049-1061.
    De Bruijn, E. R., Mars, R. B., Bekkering, H., & Coles, M. G. (2012). Your mistake is my mistake… or is it? Behavioural adjustments following own and observed actions in cooperative and competitive contexts. Quarterly Journal of Experimental Psychology, 65(2), 317-325.
    Schuch, S., & Tipper, S. P. (2007). On observing another person’s actions: Influences of observed inhibition and errors. Perception & Psychophysics, 69(5), 828-837.
    Apps, M. A., Lesage, E., & Ramnani, N. (2015). Vicarious reinforcement learning signals when instructing others. Journal of Neuroscience, 35(7), 2904-2913.
    Ho, M. K., Cushman, F., Littman, M. L., & Austerweil, J. L. (2019). People teach with rewards and punishments as communication, not reinforcements. Journal of Experimental Psychology: General, 148(3), 520.
    Cross, L., Michael, J., Wilsdon, L., Henson, A., & Atherton, G. (2020). Still want to help? Interpersonal coordination’s effects on helping behaviour after a 24 hour delay. Acta Psychologica, 206, 103062.
    Cross, L., Wilson, A. D., & Golonka, S. (2016). How moving together brings us together: When coordinated rhythmic movement affects cooperation. Frontiers in psychology, 7, 1983.
    Bayliss, A. P. Murphy, E., Naughtin, C. K., Kritikos, A., Schilbach, L. & Becker, S. I. (2013).‘Gaze leading’: Initiating simulated joint attention influences eye movements and choice behaviour. Journal of Experimental Psychology: General, 142(1), 76-92.
    Edwards, S. G., Stephenson, L. J., Dalmaso, M., & Bayliss, A. P. (2015). Social orienting in gaze leading: A mechanism for shared attention. Proceedings of the Royal Society: B. 282 (1812), 20151141.
    Bayliss, A. P., & Tipper, S. P. (2006). Predictive gaze cues and personality judgments: Should eye trust you?. Psychological Science, 17(6), 514-520.
    Bayliss, A. P., Griffiths, D., & Tipper, S. P. (2009). Predictive gaze cues affect face evaluations: The effect of facial emotion. European Journal of Cognitive Psychology, 21(7), 1072-1084.
    Strachan, J. W., Kirkham, A. J., Manssuer, L. R., & Tipper, S. P. (2016). Incidental learning of trust: Examining the role of emotion and visuomotor fluency. Journal of Experimental Psychology: Learning, Memory, and Cognition, 42(11), 1759.
    Strachan, J. W., & Tipper, S. P. (2017). Examining the durability of incidentally learned trust from gaze cues. The Quarterly Journal of Experimental Psychology, 70(10), 2060-2075.
    Strachan, J. W., Smith, A. K., Gaskell, M. G., Tipper, S. P., & Cairney, S. A. (2020). Investigating the formation and consolidation of incidentally learned trust. Journal of Experimental Psychology: Learning, Memory, and Cognition, 46(4), 684.
    Rogers, R. D., Bayliss, A. P., Szepietowska, A., Dale, L., Reeder, L., Pizzamiglio, G., … & Tipper, S. P. (2014). I want to help you, but I am not sure why: gaze-cuing induces altruistic giving. Journal of Experimental Psychology: General, 143(2), 763.

  • comment-avatar
    Arianna Curioni 14 September 2020 (15:56)

    Is there mutual adaptivity in a natural pedagogy perspective? A reply to Ildikó
    Dear Ildikó,

    Thank you very much for your interesting comment that helped us realise there are certain points that we will still need to develop. One of the clearest to us is that we need to explain how our framework of considering coordination mechanisms in social learning fits in with and complements other theories such as Natural Pedagogy. This is a valuable note for us to keep in mind when redrafting the manuscript.

    We entirely subscribe to the claim that the Natural Pedagogy model presupposes both the teacher and the learner in active and well defined roles – on the one hand the teacher having the intention to share knowledge and provide ostensive signals, and on the other hand the learner being sensitive and selective with regards to the received signals and information. Importantly, as shown by Tauzin et al. (2019) not only are learners (infants) sensitive to the kind of ostensive signals they receive, they will also modify the kinds of inferences and generalizations they formulate about a signal’s content given the (spatio-) temporal structure of the communicative interaction. The important findings of Tauzin et al. indicate that the very structure of the turn-taking exchange between two observed agents determined infants’ detection of those agents as intentional and communicative entities, therefore crucially modifying the interpretation of the signals.

    We would like to point out that such sensitivity to fine-grained spatio-temporal structures and contingent modulation of behavior must be present not only when observing third-party interactions, but also when being active participants of a communication episode, with similar consequences on the way the transmitted signals are processed and interpreted. Much research in joint action points at the fact that interaction partners’ ability to mutually adapt to and predict each other’s actions is crucial in achieving joint action coordination (e.g., Knoblich and Jordan, 2003; Keller et al., 2007; Konvalinka et al., 2010; Kourtis et al., 2013; Keller, Novembre, & Hove, 2014; Curioni et al.,2019). This indicates that information asymmetry, role distribution, and adaptiveness to the partner’s behavior are in fact crucial ingredients of the interactive episode, precisely because they influence the way that signals are conveyed and received.

    Importantly, as you also point out in your comment, our proposed criticism about “a lack of bidirectionality” in observational studies is not meant to deny that interactive partners (teachers and children) handle the communication episode as bidirectional. Indeed, this is a crucial point — the important thing in social learning is not the interpretative structure that we as experimentalists impose but how the participants within the interaction interpret that interaction from within. Given that participants seem to treat interactions as bidirectional, we are interested in investigating how expectations about mutual (bi-directional) adaptation during online interactions influence the processing and the retention of the transmitted information. Furthermore, as you point out, exactly because ostensive signals provide specific segmentations of the interaction context, they also afford the potential emergence of turn taking dynamics in which individuals can develop refined strategies of signal modulation and communication as a result of their interaction history.

    In line with both your and Dietrich’s comments, one avenue for work that exploits the distinction between knowledge and information flow could be to examine teaching interactions in unidirectional and bidirectional interactions. As Dan points out in his second comment, pedagogical demonstrations can be unidirectional — instructional YouTube videos are an example, and there are entire television channels devoted to delivering cooking instructions. This is a different kind of unidirectionality than, say, the case of the Akha that Giulio describes above where the inadaptability of the model to the learner is the result of a lack of intention to teach — in demonstration videos, there is intention but no reciprocity. How does this affect the kinds of pedagogical cues expressed — previous work using demonstration videos (e.g. Strachan et al., 2020) cannot answer this as there has been no bidirectional condition with which to compare it. Even if the same cues are present in both cases (exaggerations, eye-contact with the learner/camera, etc.) how they are used to broadcast or tailor a pedagogical experience remains an open question.

    References
    Tauzin, T., & Gergely, G. (2019). Variability of signal sequences in turn-taking exchanges induces agency attribution in 10.5-mo-olds. Proceedings of the National Academy of Sciences, 116(31), 15441-15446.
    Knoblich, G., & Jordan, J. S. (2003). Action coordination in groups and individuals: Learning anticipatory control. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29(5), 1006.
    Keller, P. E., Knoblich, G., & Repp, B. H. (2007). Pianists duet better when they play with themselves: on the possible role of action simulation in synchronization. Consciousness and cognition, 16(1), 102-111.
    Konvalinka, I., Vuust, P., Roepstorff, A., & Frith, C. D. (2010). Follow you, follow me: continuous mutual prediction and adaptation in joint tapping. Quarterly journal of experimental psychology, 63(11), 2220-2230.
    Kourtis, D., Sebanz, N., & Knoblich, G. (2013). Predictive representation of other people’s actions in joint action planning: An EEG study. Social neuroscience, 8(1), 31-42.
    Keller, P. E., Novembre, G., & Hove, M. J. (2014). Rhythm in joint action: psychological and neurophysiological mechanisms for real-time interpersonal coordination. Philosophical Transactions of the Royal Society B: Biological Sciences, 369(1658), 20130394.
    Curioni, A., Vesper, C., Knoblich, G., & Sebanz, N. (2019). Reciprocal information flow and role distribution support joint action coordination. Cognition, 187, 21-31.
    Strachan, J. W. A., Curioni, A., Constable, M., Knoblich, G., & Charbonneau, M. (2020). A methodology for distinguishing copying and reconstruction in cultural transmission episodes. Proceedings of the Cognitive Science Society, 3433-3439.

  • comment-avatar
    James Strachan 16 September 2020 (17:40)

    Ways of transmission, transmitted entities, and environmental context: The question of variability
    Thank you for your comment, Miriam. You raise a very important point about contextual variability in learning and performing techniques. Indeed, the role of variability in acquiring motor skills is the subject of growing interest in the motor learning literature, and these findings have some interesting implications for the social transmission of techniques.

    One source of variability is the endogenous variability of one’s own volitional actions. No two actions are exactly alike, even with the same actor performing the same movement on the same object with the same intention and goal. Classic models of motor cognition have typically conceived of this variability as the undesirable yet unavoidable byproduct of a noisy motor system. Indeed, a key feature of skilled expert performance is a reduction of this trial-to-trial variability. However, more recent work has cast doubt on this and suggests that this variability is very important for early motor learning when coupled with reinforcement as it allows people to explore the possible solution space (see Dhawale, Smith, & Ölveczky, 2017; and Sternad, 2018, for reviews). That is, individuals with high initial variability can explore different motor solutions and, having found an optimal solution, switch to a low-variability exploitation strategy (Wu et al., 2014).

    Variability in learning and variability in practice are also closely linked. The best illustration of this is the phenomenon known in the motor learning literature as contextual interference (Shea & Morgan, 1979; Magill & Hall, 1990). High variability during learning, such as by practicing more than one skill at a time, results in slower learning but better long-term retention of skills. Not only that, but participants who learn in high-interference (or high-variability) conditions are also better at transferring the new knowledge to different task contexts.

    Contextual interference has been studied in some social learning scenarios as well. Rohbanfard and Proteau (2011) showed in an observational motor learning paradigm that, relative to just watching an expert model perform a motor task, participants learned better from watching a mixture of expert and novice performances as the variability of the novice’s actions exposed them to a wider range of outcomes than the expert’s highly consistent actions. In particular, they found that participants who learned from this mixed observation condition were also more adaptable themselves, and could transfer this knowledge to a new set of task constraints better than those who learned from watching an expert only. Furthermore, participants are also sensitive to the practice schedule of their partners during a dyad practice task (Karlinsky & Hodges, 2018) — participants who were allowed to choose their own practice schedule were influenced in their choices by the task that their partner was practicing, and if the partner was practicing in a high-interference condition the self-choosing participant also showed better retention of the behaviour in immediate and delayed retention tests.

    There are many open questions relating to variability and social learning, in particular how these findings from the motor learning literature might fit within the framework of action coordination that we lay out here. We have intuitions about this, but very little empirical data to draw upon — I am currently working on an empirical project that may hopefully shed some light on this question, but it is unfortunately too soon to tell.

    ———————————
    References
    ———————————
    Dhawale, A. K., Smith, M. A., & Ölveczky, B. P. (2017). The role of variability in motor learning. Annual review of neuroscience, 40, 479-498. https://doi.org/10.1146/annurev-neuro-072116-031548
    Sternad, D. (2018). It’s not (only) the mean that matters: variability, noise and exploration in skill learning. Current opinion in behavioral sciences, 20, 183-195. https://doi.org/10.1016/j.cobeha.2018.01.004
    Wu, H. G., Miyamoto, Y. R., Castro, L. N. G., Ölveczky, B. P., & Smith, M. A. (2014). Temporal structure of motor variability is dynamically regulated and predicts motor learning ability. Nature neuroscience, 17(2), 312-321.
    Shea, J. B., & Morgan, R. L. (1979). Contextual interference effects on the acquisition, retention, and transfer of a motor skill. Journal of Experimental Psychology: Human Learning and Memory, 5(2), 179–187. https://doi.org/10.1037/0278-7393.5.2.179
    Magill, R. A., & Hall, K. G. (1990). A review of the contextual interference effect in motor skill acquisition. Human movement science, 9(3-5), 241-289. https://doi.org/10.1016/0167-9457(90)90005-X
    Rohbanfard, H., & Proteau, L. (2011). Learning through observation: a combination of expert and novice models favors learning. Experimental brain research, 215(3-4), 183-197. http://dx.doi.org/10.1007/s00221-011-2882-x
    Karlinsky, A., & Hodges, N. J. (2018). Dyad practice impacts self-directed practice behaviors and motor learning outcomes in a contextual interference paradigm. Journal of motor behavior, 50(5), 579-589. https://doi.org/10.1080/00222895.2017.1378996

  • comment-avatar
    Luke McEllin 21 September 2020 (12:04)

    The messiness of the real world: A reply to Rita
    Dear Rita,

    We thank you for taking the time to comment on our chapter, and also for providing us with a beautiful example of how a real social practice, even though only involving two people, can consist of so many factors. Indeed, as soon as one steps out of the lab, an interaction as simple as a handshake is such a complex affair (especially during these times) that it makes us want to go back in and hide behind our desktop computers. Thus, we are under no illusion that our approach barely gets below sea level with regards to the complexity of real-life social practices. Be as it may, as Cognitive Scientists, it is our job to try and somehow control and quantify as much of this complexity as we can, in order to understand at least some of universal mechanisms which may underpin these practices. This approach has proven very useful in allowing us to understand how a particular cognitive system may process information in the face of complexity – this is particularly the case for perceptual processes (e.g. low level visual and motor processes), even to the extent that we can quite successfully replicate these processes in machines.

    That being said, social systems are infinitely more complex than low-level perceptual systems, meaning that one must be mindful of the fact that much (probably even most) of the richness of social life gets lost in this reductionist approach. The innovation offered by our lab (CEU SOMBY lab) has been to try and move beyond an isolationist approach of studying how the cognitive system interprets and processes social information that is presented to it (e.g. by having one person responding to social stimuli such as faces or moving bodies on a computer) to investigating how the cognitive system may interpret and process information when situated in a real social interaction. As well as informing us how social information is processed, this approach also allows us to understand how the dynamics of a social interaction affect not only how this information is processed, but also how an actor behaves in a social setting.

    Although this approach often inspires protocols which simply employ two people responding to stimuli on a computer instead of one person, there are many studies that attempt to find a compromise between experimental control and ecological validity by using constrained versions of real-life social practices in order to quantify the behaviors in these practices. For example, studies tracking the motion of professional string quartets has allowed for the quantification of information flow between these musicians. By inviting expert musicians into the lab and simply allowing them to do their thing whilst connected to a motion tracking system, researchers were able to understand with temporal precision the causal influences that each of the musician’s movements and gestures had on each other’s performance (Badino et al. 2014). In a similar vein, studies tracking the motion of expert dance improvisers performing together have shed light on what sets expert performances apart from that of novices, both with regards to the quantitative characteristics of how these experts movements relate to each other, and the kinds of ways that experts modulate their movements in order to ensure the performance runs smoothly (Noy et al. 2011; Hart et al. 2014). This kind of approach has even allowed for an understanding of how these movement characteristics may directly influence a spectator’s aesthetic experience (McEllin et al. 2020). We acknowledge that these kinds of studies may also not really scratch the surface of the complexity of a social interaction, but perhaps this kind of approach can allow for a more holistic understanding of the social practices that humans engage in.

    In answer to your question, of how to bridge the gap between the controlled but narrow-focussed approach of experimental cognitive science and the messy but richly informative approach of ethnographic research, we don’t have an easy answer. The studies we describe above are examples of our discipline’s attempt to move out into the real world, but to our knowledge few, if any, of these actually involve engagement with anthropologists and ethnographers. A synergy of these approaches requires dialogue beyond the confines of this webinar — it needs collaboration. While there are researchers who do excellent work at the boundaries between disciplines, it is also important for the more entrenched of us to engage and collaborate outside of our comfort zones. One type of collaboration could be to find some phenomenon that interests both disciplines (we mention music making as an example above, but there are many others) and get both anthropologists and cognitive scientists to analyse and interpret instances of the behaviour using their own tools. A direct comparison of our approaches, our methods, and, yes, our biases and blind spots, would give us an excellent starting point for this synergy and help to highlight the complementarities (and conflicts) in our approaches. Hopefully this kind of useful cross-disciplinary work has already been done (in which case we expect that at least one person in this webinar will be able to signpost us to the studies).

    The benefits of such collaborations could well prove useful for researchers in both disciplines. Developing ethnographic tools that are tailored and informed by psychological mechanisms (e.g. an ethogram-based approach for coordination similar to Kline’s TEACH method, 2017) would give valuable feedback to experimental cognitive scientists on how the mechanisms we study in the lab relate to real-world cultural phenomena. Furthermore, it may also be useful for anthropologists for whom such a fine grain of analysis is interesting as it would help to identify small but significant behavioural adaptations related to coordination and communication that can give insight into otherwise hidden mental states that drive behaviour. The converse benefit is that experimental cognitive scientists can turn to ethnographic observations to inspire testable and working hypotheses for the lab. For us this would be invaluable in allowing us to add more richness and ecological validity to the kind of models that we are trying to construct of social learning (or indeed social behaviour in general).

    Undoubtedly, science itself is subject to the same kinds of processes that lead to the evolution and development of other cultural practices. Thus, we hope that our project, and any collaborations that come out of it will contribute to the advancement of both of our fields. Like you say, some kind of either lab-to-wild or even a wild-to-lab approach combining our expertise could be very fruitful indeed – just let us know if lifejackets will be provided!

    ——————————————–
    References
    ——————————————–

    Badino, L., D’Ausilio, A., Glowinski, D., Camurri, A., & Fadiga, L. (2014). Sensorimotor communication in professional quartets. Neuropsychologia, 55, 98-104.
    Noy, L., Dekel, E., & Alon, U. (2011). The mirror game as a paradigm for studying the dynamics of two people improvising motion together. Proceedings of the National Academy of Sciences, 108(52), 20947-20952.
    Hart, Y., Noy, L., Feniger-Schaal, R., Mayo, A. E., & Alon, U. (2014). Individuality and togetherness in joint improvised motion. PloS one, 9(2), e87213.
    McEllin, L., Knoblich, G. & Sebanz, N. (2020). Synchronicities that shape the perception of Joint Action, Scientific Reports
    Kline, M. A. (2017). TEACH: An ethogram-based method to observe and record teaching behavior. Field Methods, 29(3), 205-220.

  • comment-avatar
    James Strachan 21 September 2020 (15:14)

    Perspective sharing / embodiment: A reply to Giulio
    Thank you for your comment, Giulio, which raises several very interesting points with which we would like to engage. We have split our response into two sections to address what we see as your two distinct points: the first pertaining to unidirectional learning among the Akha, and the second pertaining to our definition of knowledge. We had drafted this response before your chapter went up today, but after reading through it we feel that our comments are not rendered moot by your post (also, very nice post and we may comment on it later in the week!)

    ———————
    Unidirectional learning among the Akha
    ———————
    As you say, the unidirectional type of learning that we claim is rare does not appear to be rare in an Akha community, where children really are just spectators. We have no argument against this, nor would we wish to make one. The existence of unidirectionality in interactions is not incompatible with the framework that we lay out, and when we claim that unidirectionality is rare this is not to say that no community anywhere in the world at any point in time has ever approached social learning in this way. Our aim is to expand tacit models of social learning that we see in areas of the cognitive science literature — particularly in the field of cultural evolution — to encompass non-unidirectional interactions.

    Moreover, we do not want to appear dismissive of such cases. The scenario you describe sounds fascinating and raises many thought-provoking questions. In particular, the phenomenon you describe of collaborative social learning. Learning alongside other learners can have very mixed effects on learning outcomes (Rajaram & Pereira-Pasarin, 2010; Nokes-Malach, Richey, & Gadgil, 2015). For example, many participants in laboratory tasks show collaborative inhibition — when presented with word lists and tested in a group, collaborating participants recall fewer words than the same number of individuals recalling items individually, which is thought to be driven in part by the coordination challenges of collaborative recall. However, in more naturalistic settings collaborative learning has been shown to lead to better retention and transfer of information (Craig, Chi, & VanLehn, 2009). Much of this may relate to the intra-learning-group interactions and the coordination mechanisms that learners employ amongst themselves. For example, one mechanism that has been proposed to result in collaborative facilitation is the joint management of attention (Nokes-Malach et al., 2015), which echoes other lab research showing that collaborating partners distribute their attention efficiently during visual search tasks (Wahn et al., 2020), and individual observation styles in gaze fixation patterns can influence and interact with each other (Brennan et al., 2008). In a scenario where the pooled group of observational learners consists of different age groups, younger learners may gain a particular benefit in action understanding by dynamically following the gaze of their more experienced older peers to relevant features of the actions being performed (Richardson & Dale, 2005; Williams et al., 2002). One question that this scenario raises is what might the consequences of collaborative learning be for the flexibility and rigidity of cultural information — people are more conservative and resistant to outside influence for decisions they have made together (a ‘myopic underweighting of external viewpoints’; Minson & Mueller, 2012): might this also apply to techniques that are learned together, making these traditions more resistant to change and contributing in part to the rigidity of traditions you describe in your chapter?
    Importantly, although we emphasise other kinds of learning, learning through observation (or joint, collaborative learning) is wholly consistent with our framework. Indeed, one of the principles that should be emphasised is that people will coordinate to the point that they need to — if the prevailing understanding among a community is that children will learn through simply observing adults (a wholly justified position given that all adults holding it were presumably once children who learned their skills through observing adults) then it would be unnecessary to engage in other, more costly types of skill transmission. Just let the children watch so long as they keep out of the way.

    It would be very interesting to hear more about these skilled techniques and these learning interactions. For instance, are there any features of expert techniques that cannot be learned or are difficult to learn through observation because they involve adaptation to rare environmental circumstances (e.g. dealing with challenging materials or having to construct something to uniquely atypical specifications)? Presumably, as transmission is purely unidirectional, the experts who perform these actions make no concessions to allow child observers to exploit such learning opportunities by adjusting or exaggerating their actions, so then where do the adults learn it? In other words, what do young adults who are finally allowed to engage in the tasks do when encountering a non-standard problem that they may have never seen or encountered before? Do they solve it individually? Do they ask their peers or elders for advice?

    ———————
    Definition of knowledge
    ———————
    We were mindful of the fact that knowledge and information would be theoretically laden terms, particularly given the association that some readers may have between knowledge and the ontological nature of mental representations, which is not something to which we want to commit ourselves either way. Our intention was to distinguish between these two types of signal available to a participant in a social learning interaction: between what is learned (knowledge) and what is used to interact (information). Perhaps we should make it clearer in future drafts, but we do not use knowledge to mean something reliant on mental representations — you will note that although we refer to knowledge as a representation, at no point do we call it a mental representation. Representation in the case of technical transmission could refer to an action plan, or a specific motor or proprioceptive representation, without requiring any ‘mental’ content (e.g. Wolpert, Doya & Kawato, 2003).

    We do agree, however, that our use of the term context-invariant is unclear. In light of how you have interpreted it, it seems like ‘context-general’ may be a more appropriate term that would be consistent with Ingold’s description of cracking eggs. Indeed, how we define knowledge is as something that is not restricted to the local context of the immediate interaction, but something that can be applied across many different cases. As such, knowledge as we use it refers not to the specific muscle contractions required to crack this particular egg, but the action plan (which includes the shell-evaluating tap) that will allow you to crack any egg you encounter against any bowl.

    ———————
    References
    ———————
    Rajaram, S., & Pereira-Pasarin, L. P. (2010). Collaborative memory: Cognitive research and theory. Perspectives on psychological science, 5(6), 649-663.
    Nokes-Malach, T. J., Richey, J. E., & Gadgil, S. (2015). When is it better to learn together? Insights from research on collaborative learning. Educational Psychology Review, 27(4), 645-656.
    Craig, S. D., Chi, M. T., & VanLehn, K. (2009). Improving classroom learning by collaboratively observing human tutoring videos while problem solving. Journal of educational psychology, 101(4), 779.
    Brennan, S. E., Chen, X., Dickinson, C. A., Neider, M. B., & Zelinsky, G. J. (2008). Coordinating cognition: The costs and benefits of shared gaze during collaborative search. Cognition, 106(3), 1465-1477.
    Wahn, B., Czeszumski, A., Labusch, M., Kingstone, A., & König, P. (2020). Dyadic and triadic search: Benefits, costs, and predictors of group performance. Attention, Perception, & Psychophysics, 1-19. https://doi.org/10.3758/s13414-019-01915-0
    Richardson, D. C., & Dale, R. (2005). Looking to understand: The coupling between speakers’ and listeners’ eye movements and its relationship to discourse comprehension. Cognitive science, 29(6), 1045-1060.
    Williams, A. M., Ward, P., Knowles, J. M., & Smeeton, N. J. (2002). Anticipation skill in a real-world task: measurement, training, and transfer in tennis. Journal of Experimental Psychology: Applied, 8(4), 259.
    Wolpert, D. M., Doya, K., & Kawato, M. (2003). A unifying computational framework for motor control and social interaction. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 358(1431), 593-602.

  • comment-avatar
    James Strachan 22 September 2020 (14:15)

    Assumption of Unidirectionality – In the Lab vs in „Real Life”: A reply to Gyuri
    Dear Gyuri. Thank you for your extensive comment. In our response we focus on what we feel are the main points to address. Our responses to these will hopefully also indirectly address some of the many other points in your comment, but we are happy to continue the conversation with you. See our response below.

    ———————
    We mischaracterise or misunderstand the principles of natural pedagogy and claim that it suffers from the assumption of unidirectionality
    ———————
    Please see our response to Ildikó above, posted by Arianna. As we say, we do not currently discuss natural pedagogy or the many ways in which our framework fits with it, as our focus in this first draft has been to engage with theoretical and empirical work on cultural evolution, and we have not situated our position within the broader cognitive science literature. We do, as you point out, cite some work that is relevant to natural pedagogy, but this is cited to acknowledge that even cultural evolutionists who argue in favour of an evolved mechanism for high-fidelity imitation would agree that imitation is more than ‘blind copying’.

    “Given that observational learning is the first kind of social learning that humans are capable of, and is also observed in our closest phylogenetic relatives, it is tempting to think of this also as a conceptual foundation for other kinds of social learning that humans are capable of later in life.” Let us clarify the meaning of this sentence. When we say that babies cannot act independently and so must learn through observation and imitation rather than active participation, this is not to argue that such interactions are unidirectional (although they may superficially appear so as babies cannot engage in what we might think of as paradigmatic action coordination). These points are not criticisms of natural pedagogy, but of cultural evolutionary theories that model social learning on high-fidelity imitation as the key component of human cultural transmission. Natural pedagogy, as you say, does indeed posit a bidirectionality of information flow as, for example, infants’ responsiveness to ostensive signals can help to perpetuate interactions by engaging adult models through eye contact, facial expressions, etc. We simply argue that as this bidirectionality persists beyond infancy and childhood it becomes more coordinative as the learners’ coordination skills develop. This is particularly important when learning those complex technical skills that are typically learned later in life (i.e. as young adults).

    As we say in our response to Ildikó, our revised chapter will explain these points in more detail.

    ———————
    The ‘real world’ of the lab is far from the ‘real world’ of cultural practices, where practices such as apprenticeships or formal (ie. WEIRD-style classroom) teaching often limit the amount of interaction possible
    ———————
    We do not pretend that laboratory experiments are accurate proxies of the real world, nor do we use the real world to refer to the “real world of the SOMBY lab’s experimental culture” — although see our response to Rita above for some of our thoughts on the ecological validity of basic research into cognitive mechanisms. Rather, our intention is to describe a set of plausible psychological mechanisms (informed by empirical data in the lab) and consider how these relate to real cultural phenomena in practice.
    The instances you describe of unidirectional social learning with little opportunity for interaction (guarded apprenticeships or learning by rote instruction in a classroom) relate somewhat to Giulio’s comment above. As we say in our response there, the existence of unidirectional interaction structures does not undermine the importance of investigating social learning interactions that encompass more than unidirectionality. We do not pretend that these cases do not exist — we just propose to treat a different kind of social learning interaction as the starting point for understanding the mechanisms involved.

    In fact, the cases you describe are interesting under our framework, as these are particular cases where the cultural context in which social learning is embedded leads to particular social affordances and constraints on the social learning interaction that elicit different (perhaps even competing) motivations from the model and the learner. These may have consequences for the learned behaviour. See also Dan’s second comment and our response to Ildikó for some discussion of how unidirectionality in pedagogical demonstrations may offer avenues for empirical research.

    ———————
    Many models of social learning have been designed to explain social learning in non-human animals, and do not seem to suffer in this endeavour from assuming unidirectionality
    ———————
    From the perspective of non-comparative psychologists, these models do indeed seem to explain instances of social learning in non-human animals well. As you say, apes do not appear to modify their nut cracking behaviour to facilitate learning in a juvenile, suggesting they do not treat the interaction as bidirectional. If such models can explain social learning in non-human animals then that is valuable.

    The problem of assuming unidirectionality instead comes into play when building psychologically plausible models of human social learning in adults. As we say in the chapter, when trying to figure out what makes human social learning so special, it can be tempting to look to the developmental and comparative literature and find something that human babies can do but that our closest great ape relatives cannot (or that they do in a qualitatively different way). High-fidelity imitation is one such skill — babies frequently overimitate, while chimpanzees typically emulate actions. The problem arises in drawing conclusions from such comparisons to explain complex, cumulative cultural practices and techniques that are invariably innovated, transmitted, and sustained by adult humans without actually studying how those adult humans behave. And this is where mechanistic theories that have either a tacit or explicit assumption of unidirectionality will fail to capture the complexity of social learning in human adults (and, indeed, children and infants) because this interactivity matters. See our response to Dietrich, above, for some additional discussion of how thinking of social learning in terms of coordination mechanisms may also be an approach with some relevance to comparative research.

    ———————
    We should be cautious when generalising our experimental results to actual cultures
    ———————
    We absolutely agree, as I’m sure would any experimentalist. While it is of course our hope that this work is engaging for other disciplines, we propose this framework as a way of expanding the current empirical literature on social learning in adults to encompass a broader range of phenomena under a plausible mechanistic framework — where we draw from other literatures in our chapter it is to situate the mechanisms we describe rather than to prescribe our framework onto these literatures.

    Furthermore, we do not make a normative argument that bidirectional coordinative learning is the ‘best’ way to learn, nor that understanding social learning in terms of coordination mechanisms is always the appropriate grain of analysis. Rather, our goal is just to describe the psychological mechanisms that subserve social learning in humans — opening the black box of cultural transmission (Heyes, 2016) — in a way that is at least coherent with documented observations of cultural phenomena.

    ———————
    Morgan et al. (2014) show that transmission improves with teaching and language, but not with imitation or emulation
    ———————
    Yes, that paper and a couple of other studies (Caldwell & Millen, 2009; Saldana et al., 2019) are excellent illustrations of how social learning is more complex than the imitation strategies that cultural evolutionists typically describe. Indeed, while Morgan et al. show that teaching and language are selected for by human reliance on stone tool-making, the latter two show that features of cumulative cultural evolution — previously thought to rely on high-fidelity imitation — are possible under laboratory conditions without imitation, highlighting the need for alternative explanatory models.

    ——————–
    References
    ——————–
    Heyes, C. (2016). Blackboxing: social learning strategies and cultural evolution. Philosophical Transactions of the Royal Society B: Biological Sciences, 371(1693), 20150369.
    Caldwell, C. A., & Millen, A. E. (2009). Social learning mechanisms and cumulative cultural evolution: is imitation necessary?. Psychological Science, 20(12), 1478-1483.
    Saldana, C., Fagot, J., Kirby, S., Smith, K., & Claidière, N. (2019). High-fidelity copying is not necessarily the key to cumulative cultural evolution: a study in monkeys and children. Proceedings of the Royal Society B, 286(1904), 20190729.

  • comment-avatar
    Josep Call 30 September 2020 (15:20)

    big bodies and low pitches
    Dear James,

    Thanks for your last post. I read it with great interest. The “big objects and low pitches” association reminded me of the argument made in evolutionary biology regarding the relation between body size, honest signalling, and sexual selection. In general, mammals with larger body sizes tend to produce vocalisations with lower pitches.

    This rule holds firm for between-species comparisons, at least in mammals, (e.g., elephants produce calls of lower frequencies than mice) and also within-species comparisons although with some qualifications. For within-species comparisons, the rule still holds for species with high intra-sexual competition based on fighting contests (e.g., red deer). Those are precisely the species that use the quality of male vocalisations, especially those that are notoriously difficult to fake, as a reliable indicator of body size, which in turn correlates with fighting ability and subsequent female preference. In contrast, the relation between body size and vocal frequency is weak or non-existent for those species with low intra-sexual competition based on fighting contests (e.g., rhesus macaques). Humans have been suggested to belong to the first group, something that would explain why deep male voices are found attractive and intimidating by females and males, respectively.

    So although I agree with you that culture may shape, and even reverse, some aspects of our primitive dispositions (you cited several examples), others may still be anchored to our evolutionary history. Do you or anybody in the group know if some cultures associate big objects with high pitches?

  • comment-avatar
    James Strachan 2 October 2020 (15:45)

    big bodies and low pitches: Reply to Josep
    Hi Josep. I don’t think there are any established instances of culture where the mapping of this type of correspondence is reversed, and I don’t expect this would occur without substantial intervention. These cross-modal associations are learned through repeated exposure to co-occurrence of particular stimulus features within the environment, so while many of these can be cultural or social in nature (such as the SNARC effect being subject to shaping by cultural writing practices, or how people that conceive space using absolute (east/west) rather than relative terms (left/right) do the same with time, Boroditsky & Gaby, 2010, Psych Science), others are strictly related to biological or physical principles. The relationship between size and sound, for example, has a physical foundation: larger, heavier objects tend to have lower resonant frequency than smaller, lighter objects because heavier objects simply do not vibrate as fast.

    So yes, the cross-modal correspondence between size and pitch is likely driven by larger animals producing lower vocalisations, and this may well be tuned in humans because of its relevance for sexual selection. What is interesting here is that, while these associations work as general rules (larger individuals tend to have deeper voices), they can also interfere with accurate perception in specific instances. The relationship between voice pitch and body size is not strictly linear, and there is substantial variability in adult human males along both dimensions orthogonally, which leads to systematic biases in estimating men’s heights that appear to be driven by their vocal pitch (Armstrong, Lee, & Feinberg, 2019, Animal Behaviour).

    Importantly, to our knowledge such correspondences have typically been studied in terms of their effect on perception. What we found so interesting about Giulio’s post was how it highlighted how these may also play a role in action, particularly relating to technical traditions.