Here we present the current state of an exchange between Prof. Andrew Whiten and Dr Damien Neadle, following Neadle et al.’s talk on ape novel action copying at the Cultural Evolution Society’s (https://culturalevolutionsociety.org/) online 2021 conference – hosted by Dr Masanori Takezawa (Hokkaido University). Prof Whiten’s points/questions are presented in italics and Dr Neadle’s are below in each case. Note that this is an ongoing discussion – with plans to formalise it into manuscript formats. For the time being, the following is merely to make the current state of the debate accessible beyond the CES website’s lifetime.
Hi again Damien – I have many questions there was not time for in the live Q&A so I’ll add some now!. First, I think that Zanna’s study and yours are highlighting child-chimpanzee differences in what they are prepared to copy – this may be either the ‘overimitation’ difference Vicky Horner and I first reported back in 2005, or a difference in preparedness to copy a ‘gesture’ like Mike T and others reported back in 1997 – or both – so my first question is – do you think you might be able to disentangle these two?
Hi Andy, yes a shame we were not able to talk about these things face to face, alas I will do what I can to answer these here. I’ll respond to each paragraph in turn to aid in clarity and so that I don’t forget parts. I agree with you here, we are clearly highlighting a species difference (though note that this isn’t just a child/chimpanzee one – in fact Zanna & Claudio’s is child/bonobo and ours is essentially child/non-human ape). However, where we might disagree is with regards to your phrasing: ‘prepared to copy’; this is assuming that apes can copy a novel action under this kind of ‘action only’ paradigm (although note that answers to later questions will allow me to unpack this more).
Whilst the literature on overimitation to date (Hoehl et al, 2019) is clear, there is a difference between humans and non-humans in their propensity for overimitation – this appears to extend beyond just primates (I am referring to work here by Daphne Buchsbaum and colleagues that is as yet unpublished but I have been fortunate enough to read). Whilst I do not agree that you and Victoria identified ‘imitation’ in 2005, as the task was not sufficiently novel to assume this (based on your criteria laid out in 1998 in the Journal of Comparative Psychology), the study is relevant here and that you found ‘over-enhancement’ maybe?. Also, the fact that there were environmental results present means that we cannot rule out other social learning mechanisms like enhancement or emulation (detailed critiques along these lines can be found in Clay & Tennie, 2017). Regardless, the study I presented in my talk does follow an overimitation paradigm – just as you suggest. The task we used was to copy actions that are not ”obviously” required – and this is also the reason why Clay & Tennie had the term overimitation in their title.
You ask about the begging gesture study (Tomasello et al., 1997) and its relation here, I briefly (due to the time constraints enforced by the conference) mentioned that I consider this study to be one of the three remaining pieces of relevant data to this ‘pure’ novel action copying debate (the others being Clay & Tennie, 2017 and Tennie et al., 2012). You will note that Tomasello et al 1997 considered theoretically that there might be ape social learning of their gestures, however based on the substantial variation in the performance of these gestures, they largely dismiss imitation as an explanation (presumably because of a lack of copying fidelity, even if these gestures are copied at all). However, when they later attempted to test specifically for action copying in a begging gesture context (study 2 in the 1997 paper) they came up with no evidence in the tested chimpanzees. So, this evidence is consistent with our findings and interpretation. While Tomasello et al. 1997 did indeed use a gesture copying paradigm, all other studies (including the one I presented) went beyond mere gestures. All three other studies used actions on tasks – i.e., actions that were designed to accompany solutions – that appeared necessary for solutions (and yet, novel and arbitrary – the first to test for human like copying; more on this below), the second to test for the copying of actions, rather than for copying of task structures etc. (achievable in principle via emulation).
In addition, we used two actions, to test additionally for anchoring effects. The ‘rub’ action is in contact with the box (probably more like overimitation tasks in general) and the rotate action is not (perhaps a bit more like gestural tasks) – see Clay and Tennie, 2017 for further discussion of this. So, we are testing for the ability to copy novel actions in technical tasks, and at the very least we achieve this with the anchored action. We are not attempting to fully disentangle these things here for our second action, and neither do we need to (as we have the other action, too).
Other puzzles and questions arise from your dismissal of all ‘two-action’ experimental designs for documenting action copying. This raises many questions, because I think virtually all of these methods’ many appearances in the animal social learning literature is typically offered as specifically used to distinguish copying (of one action form rather than the alternative action form) from simpler processes, notably stimulus or local enhancement; moreover in diffusion versions incorporating the two-action idea, the further interesting question is whether the different forms spread differentially in the experimental groups to become incipient traditions. So – second question – if what is of interest is a species’ capacity to copy in a way that sustains cultural transmission across multiple individuals – aren’t two-action experiments, and specifically two-action diffusion experiments, exactly what we need to test for that spread? (After all, your prediction seems to be that non-human animals, unlike humans, cannot copy different forms and so will not be able to reliably transmit different forms of action along cultural diffusion chains?)
Indeed, if we were just looking for social learning mechanisms that might or might not sustain behaviours for a while in a laboratory transmission chain then the ‘two action’ tasks (as you call them) are suitable. However, in this study we aren’t looking at this, we are instead looking at the capacity to copy novel actions in a technical task (under majority demonstrations). However you will note that when Dawson and Foss initially introduced the two action paradigm there was an error in their application, namely the mixing of actions (body movement/parts) with object movement (sadly, this was not disentangled). This error carries through (as also highlighted by Heyes before) in the literature. To this day, this means that it is possible to ‘succeed’ in the “two-action task”, without needing to actually copy actions at all. This is different from what we are testing for here and in other related studies. Which is why the name of ‘two-action’ task has been questioned previously, preferring to call them two-target tasks (Tennie et al., 2010).
In the study I presented it is important that we go beyond the baseline of the species concerned, for this a ‘two action’ task (at least how they have been implemented) were not suitable. This is because if we stay within the baseline we cannot even exclude individual learning playing a very large role (let alone other social learning mechanisms). This means that when the targets to copy are within the baseline capacity (as with all ape iterations of the two-target paradigm; Tennie et al. 2020 did a review of these) we can’t be sure that actions are copied and we know that something has been copied that did not go beyond baseline performance (begging the question whether this should even be called copying in a human sense) Therefore, these types of tasks aren’t useful to answer the question we pose. That is not to say that are not useful (e.g., to study social learning biases), just not here.
Your assumption about our prediction “that non-human animals, unlike humans, cannot copy different forms and so will not be able to reliably transmit different forms of action along cultural diffusion chains” is somewhat off. We are interested in the first part – we had negative data as to whether non-humans can copy novel actions (in suitable tasks). We wanted to try again (this time with a majority bias) – the transmission and diffusion would have been the next step (it would have been had we identified action copying). There is no need – no possibility – for us to test the latter until we can confirm the former. As I see it, we don’t have this data yet so cannot assume that this is the case (and indeed, the negative data pattern seems consistent by this point).
The types of cultures identified in the lab via the two target method remain real – but they remain within the realm of baseline performances. But again, we never denied that apes can and do sustain cultures made up of such latent solutions.
Third, related question – for you and your colleagues, what counts as an ‘action’? You seem to want to limit that to only what goes on ‘within the skin’, like limb movements. But what about actions that employ tools, such as ‘poke stick into hole X’ versus ‘lift blockage with stick’ in the example of the pan-pipes you illustrated (Whiten et al, Nature 2005)? Aren’t they alternative ‘actions’? And involve differential know-how? It is not implausible that most of the culture we humans acquire concerning what we do with our hands involves forms of tool actions and manipulations of other objects? If so, don’t we need to include these in our tests of action copying? I hypothesise that chimpanzee action copying is relatively ‘pragmatic’ in that it is in play in tool use and other actions on objects, but does not extend to over-imitation nor copying gestures.
This is a continually interesting point that you raise. For us actions are the behavioural forms themselves not the results they produce. So, yes, tool use like this does contain actions. However, and as I said in response to your previous comment, to answer the questions we specifically set out to answer (do apes copy beyond their ZLS; in the current study specifically, do they copy actions beyond their ZLS) these actions must go beyond the individual capacity. This brings me to the pan-pipes. It has been claimed that these experiments might be the exception to the rule of ‘two-action’ studies not going beyond baseline. As you mention in the 2005 study the lift behaviour does not occur at baseline (the poke behaviour does also, in 1 subject in that study and others in other studies, if I recall correctly). However, what one would not be acknowledging here is that the ‘alternative action’ (i.e. alternative solution) was found in another group; i.e., lift was innovated in the poke seeded group (as pointed out in Tennie et al., 2020 citing Hopper et al., 2007 and Whiten et al., 2005). Therefore, it is not possible to assume that the apes have copied (actions or other types of info) here anything novel – anything beyond their ZLS, if you will – as both solutions clearly are within the baseline level of ape performance (thus individual learning capacity) of the subjects concerned. Indeed, both of these behaviours have appeared in the baseline conditions of other studies using the pan-pipes (Tennie et al. 2020), showing once more that they are within the individual learning capacities of these species (their ZLS).
So, in sum, yes. We could include these kinds of behaviours, providing they
- Go beyond the individual/baseline capacity of the subjects
- Do not allow for the task to be solved by other social learning mechanisms that ultimately have to rely on individual learning (e.g., enhancement)
- Allow at least equal testing time for baselines and demonstration/live conditions (and tested with properly powered baselines (Bandini et al. 2020))
Your point about human culture is an interesting one, but our study does not set out to explain human culture in this way, we are interested in ape culture and the mechanisms underpinning it – in particular if they copy beyond their ZLS. If it were shown that modern human culture is not often or never based on action copying (I think we can all agree this is unlikely) then this raises other issues.
Then turning to your three criticisms of the power of two-action tests. From your slide: 1. “No baseline”. But in the example you illustrate, of the pan-pipes, in addition to the two alternative seeded actions groups, there was a third group that first experienced the pan-pipes with no model for an hour individually, then 4.5 hours as a group, without achieving any success, although they tried to apply the stick tool to the panpipes. Just to take diffusion two-action studies alone, as reviewed by Mesoudi and me just up to 2008, we tabulated 8 studies with this three groups design with birds and mammals, plus another 3 with baseline no-model conditions. (That review updated in Whiten, Caldwell, Mesoudi 2016). So question 4 is really – what do you mean by ‘no baseline’ in these studies?
See my above comment about the appearance of the target behaviours in baseline conditions of some iterations of the ‘two-action’ tasks. Clearly, my comments do not target any one study. I used the pan-pipes as a well-known, and well cited, example – and given that you yourself mentioned it above). I refer to general problems, with the application of the two target methodology as an inappropriate technique to identify action copying generally. So, some are not using novel actions, allowing for social learning mechanisms other than action copying (including in some cases, even local enhancement) and yes, some lack a baseline. The 2005 pan-pipes study you refer to has a baseline but, as the behaviours can (and did) occur in the absence of a demonstration (see above) we cannot possibly assume that copying of novel solutions (solutions beyond the ZLS of apes) is at the heart of it. Furthermore, other iterations of studies like the pan-pipes one (using two-target methodologies) often have very underpowered and/or significantly shorter baseline testing time compared to the social learning conditions (even less than half the subjects/time). But, yes, perhaps ‘no baseline’ was the wrong terminology in my talk, I now would correct this to include target behaviours that can occur without the need for demonstrations (either in baseline or by occurring in an alternatively seeded group; see Tennie et al., 2020).
As for other species, we do not deny that other species than apes may copy beyond their ZLS (again, the thing to look out for would be copying beyond baseline performance), our study here was on apes – our main study focus in general. And so, we would like to stick with the ape data.
Then you had 2. “Possibility of result-based (emulative) copying”. But the result for both pan-pipe actions, for example, is the same – a grape comes out of the lower pipe. But chimpanzees see either of two forms of action to achieve this, and they tend to copy the form they have seen and thus acquire the corresponding know-how. Perhaps more importantly, it is exactly this kind of match between what is done with a tool or performed on an object and what the observer then does that is the basis not only for what counts as copying, or more often the term is ‘imitation’, in the child literature generally, but in nearly all of the corpus of even that labelled ‘overimitation’. So question 5 – are you arguing that all these researchers are misguided in treating this as action copying, or as imitation? Remember also that in Horner and Whiten 2005, when the box was opaque, after watching the model do so the chimpanzees did tend to probe the stick in the top hole like children did, its just that they selectively did not when the transparent box revealed it was unnecessary, whereas the children persevered, later called overimitation by Lyons et al. And also remember that Buttelman et al,. with Tomasello as a co-author, described what chimpanzees matched a human doing with their head, or hands, or bum, on an object, as ‘rational imitation’.
As you have been very frank in your questions/critiques of our study here, I will do so too. In short, yes; we are arguing here that labelling the findings of these studies as ‘imitation’/action copying is incorrect. That is not to say imitation (action copying) did not occur, especially in the case of children (for which we have good independent evidence for action copying). However, we cannot exclude other social learning mechanisms in these tasks generally (we know this since Heyes critiqued them) and therefore to assume one mechanism over the others is not supported by data – this it is too soon to draw this conclusion.
Do we dismiss the utility of the ‘two-action’ task? No (see above) – but yes,we do dismiss it as a ‘gold-standard’ for identifying action copying because it lacks the precision required to separate the mechanisms at play. Again, this issue is not novel and we are not the first to suggest it. Cecilia Heyes has often said that these tasks contain confounds that cannot be overlooked (when used in this context) and Lydia Hopper’s had a study that identified copying in ‘ghost conditions’ in apes – where no actions were at all demonstrated. If imitation really were the only possible social learning mechanism at play here ‘ghost conditions’ would have to fail – clearly this is not the case. And again, in some such tasks used, even enhancement effects were not excluded (e.g. the two target locations at the pan-pipes were spatially separated).
The issue with the Horner and Whiten (2005) study is that the tested chimpanzees we not naïve to stick tool use, they were well aware that sticks were functional tools and could be used to solve tasks like this. This links to the ‘not always novel’ point below (it is not helpful to consider these as granularly as you have split them here, but nonetheless I have responded in the order you have raised the issues). As we know the chimpanzees in the 2005 study were capable stick users (as nearly all chimpanzees are) the only thing we can say (with certainty) that is ‘copied’ is the place where they inserted the stick (which is just local enhancement). That task, whilst perhaps still great for identifying social learning biases, falls short when attempting to identify imitation as you have defined it previously (1998 & 2004). Note that this critique has already been levelled in detail at this task by Clay and Tennie (2017), where it was explained how we cannot differentiate between social learning mechanisms.
Regarding the Buttelman study, this study used enculturated apes and therefore is not relevant to this debate. In the same way as I am dismissing Josep Call’s 2001 of imitation with an enculturated orangutan from this debate. Note that Josep was my examiner for my PhD and he had no issue with my dismissal of this data as not relevant to the arguments here.
P.s. My commitment is to the data, you will note that I was perfectly happy publishing my data from 2020 that went to some degree against the ZLS account’s predictions (with caveats). Had the present study – the one I presented in my talk – identified action copying I would have published it and happily so. I was Claudio’s student, he is my co-author and collaborator, but if the data did not support his theory I would have no reservation in still publishing (and indeed, Claudio always encourages everybody to publish all data, and regardless of whether that data supports things he said before or not). Thus, just because Tomasello is on that paper it doesn’t make it any more or less relevant than any of the other studies that identified ‘imitation’ in enculturated apes.
Then you had 3. “Not always novel”. Well in the case of the pan-pipes, for example, the chimpanzees had never seen a pan-pipes before, and of course they had not seen either the poke or lift technique on it before their model did it, and the no-model individuals did not do these actions. So fifth question – what more criteria for ‘not novel’ can reasonably be expected?
This fact is exactly why the pan-pipes were – initially – considered by Tennie et al. (2020) to be the possible exception to the ‘two-action’ tasks being dismissed as good evidence of anything approaching action copying. However, the issue comes back to that one I outlined above – attempting to breakdown my critique of the ‘two-action’ tasks in this way can lead to a misrepresentation of the argument. The pan-pipes failed to meet the standard, not because the task was not novel but because it proved possible for subjects to solve them without the need for social learning at all (lift in the baseline and poke in a lift-seeded group). And when tested with material tasks whose solutions lie outside the ZLS of the species, apes fail to copy (Tennie et al. 2009). So, my critique stands, some applications of the ‘two-action’ paradigm include tasks that are not novel to the subjects – pan-pipes are an example of this not being the case.
In any case, let me suggest that if the research question is ‘is copying necessary to acquire action X?’ then a baseline or parallel no-model condition is indeed needed: but if the research agenda is instead ‘do these animals copy a technique they witness well enough for it to pass faithfully along a chain, or across a group, of cultural learners? then the key is a two-action design, and a no-model control group is in this case redundant – if the alternative actions diffuse within and even between groups like in our 2007 study, that is the interesting finding in relation to cultural transmission. Don’t you agree? Note, by the way, that when Emma Flynn and I reported a parallel pan-pipes study with children in 2010, we found the seeded alternatives also began to spread differentially in the groups, but soon with less fidelity than in the earlier chimpanzee study.
I agree with your first point. If we are to ask the question whether or not action copying (or indeed any social learning) is required or as I phrase it in my 2017 paper “necessary” for behaviour X then we need a baseline. This is the crux of the Latent Solutions testing methodology that Claudio has been championing for many years now. The issue that we have here is that often (not always, see Neadle et al., 2020 and Bandini & Tennie, 2018) primate behaviours do not require social learning to occur. They occur at baseline, therefore are not beholden to culture in the way that has been suggested. So yes, if we are asking this question that is correct. But, that is not the question here. Ours is can apes copy beyond baseline levels? We had planned baseline testing and a whole bunch of testing to figure out what exactly is going on (had they done it; see our preregistration of this study on the OSF for details: https://doi.org/10.17605/OSF.IO/GPV54).
Your point about fidelity has been addressed by other researchers (e.g., Claidière & Sperber, 2010), the “high” fidelity of the mechanism alone is not sufficient to explain culture in the way that is often suggested. It doesn’t matter to that argument whether or not imitation is identified in non-humans as Claidière & Sperber argue that it is not sufficient (they even go as far to suggest it is not in any case, we would not go that far). But, again, this is not the question here. Based on the data, we know that ape cultures (many aspects at least) do not seem to require action copying (or imitation however one terms it; or know-how/form copying), see the many Latent Solutions experiments that have been outlined by our group and beyond. There are still candidates that require additional attention (here I refer you to Claudio’s work with Alba Motes-Rodrigo and her related talks this year).
Regardless, the ape transmission studies that you suggest, following a ‘two-action’ paradigm, tested within the individual learning capacities (or ZLS) of the ape species tested for the reasons I have outlined extensively here. Therefore, this is not relevant to the question we are asking. That is not to say this data is uninteresting, just that this isn’t the topic at hand. As for the note about Emma Flynn and your study in 2010, biases in culture are common and this has been well championed by Sperber and his many students. The countless cultural traits and ‘know how’ that we support are maintained in spite of this (Motes-Rodrigo & Tennie, 2021).
So, if one is interested in explaining the patterns of wild ape culture that you (and many others besides) have identified, the questions you ask are relevant, and so are the two-target tasks so far performed on apes. However, we do not question whether apes have culture. You might recall that we had this very discussion at CES, in Arizona, last time. I have termed these types of culture ‘minimal culture’, that is where social learning plays a role in the facilitation of a trait that is however in the baseline possibility (the ZLS) of the species. I explained in 2017 how gorilla food cleaning falls into this category. This is where I think a sort of straw man has been erected over the years between your research group and the critics, including Claudio (but also Galef etc.). Moreover, Tennie, Call, Tomasello or any of their students would not ever deny the existence of some sorts of culture in apes. Quite the opposite, we acknowledge it and contrast it with our own (markedly different) culture to infer the answers to questions about the early evolution of our specific human culture (which consists mostly of ZPD) and the behaviours and cultures of early hominins.
So, to answer your suppositions of our research questions, I shall state them clearly (now here I speak only for Claudio, and sort of myself – though I confess in response to the politics in this research area I have begun to move towards other aspects of cultural evolution). We ask whether or not apes can spontaneously utilise now-how copying (including novel action copying) to ever go beyond their ZLS, therefore expressing culture dependent know-how and going beyond the ‘minimal’ culture that we all agree they definitely have. If this copying occurs, what social learning mechanisms are included in it and to what degree are they applied? Perhaps now it is clearer, how the methodologies you suggest do not match the questions we are asking.
I hope that my responses here answer at least some of the questions you have. It really is a shame we weren’t able to talk outside of the Q&A session at the conference. If you’d like to chat we can always Zoom, I’d be very glad to hear what you think about all this and how we can all work together to advance the cultural evolution literature even further. These kinds of debates are useful (in my humble opinion) to consider where we all stand and to ensure that we understand one another’s’ point of view. I think this misunderstanding is often the cause of the some more ‘heated’ debates in our field.
Bandini, E., Motes-Rodrigo, A., Steele, M. P., Rutz, C., & Tennie, C. (2020). Examining the mechanisms underlying the acquisition of animal tool behaviour: Sources of animal tool behavior. Biology Letters, 16(6). https://doi.org/10.1098/rsbl.2020.0122rsbl20200122
Claidière, N., & Sperber, D. (2010). Imitation explains the propagation, not the stability of animal culture. Proceedings of the Royal Society B: Biological Sciences, 277(1681), 651–659. https://doi.org/10.1098/rspb.2009.1615
Clay, Z., & Tennie, C. (2017). Is overimitation a uniquely human phenomenon? Insights from human children as compared to bonobos. Child Development, 89(5), 1535–1544. https://doi.org/10.1111/cdev.12857
Hoehl, S., Keupp, S., Schleihauf, H., Mcguigan, N., Buttelmann, D., & Whiten, A. (2019). ‘Over-imitation’: a review and appraisal of a decade of research. Developmental Review, 51(November 2018), 90–108. https://doi.org/10.1016/j.dr.2018.12.002
Hopper, L. M., Spiteri, A., Lambeth, S. P., Schapiro, S. J., Horner, V., & Whiten, A. (2007). Experimental studies of traditions and underlying transmission processes in chimpanzees. Animal Behaviour, 73(6), 1021–1032. https://doi.org/10.1016/j.anbehav.2006.07.016
Horner, V., & Whiten, A. (2005). Causal knowledge and imitation/emulation switching in chimpanzees (Pan troglodytes) and children (Homo sapiens). Animal Cognition, 8(3), 164–181. https://doi.org/10.1007/s10071-004-0239-6
Motes-Rodrigo, A., & Tennie, C. (2021). The Method of Local Restriction: in search of potential great ape culture-dependent forms. Biological Reviews, 9. https://doi.org/10.1111/brv.12710
Neadle, D., Allritz, M., & Tennie, C. (2017). Food cleaning in gorillas: social learning is a possibility but not a necessity. PLoS ONE, 12(12), e0188866. https://doi.org/10.1371/journal.pone.0188866
Neadle, D., Bandini, E., & Tennie, C. (2020). Testing the individual and social learning abilities of task-naïve captive chimpanzees (Pan troglodytes sp.) in a nut-cracking task. PeerJ, 2020(3), e8734. https://doi.org/10.7717/peerj.8734
Neadle, D., Chappell, J., Clay, Z., & Tennie, C. (2018). Testing the effects of conformity bias on great apes’ imitative abilities. https://doi.org/10.17605/OSF.IO/GPV54
Tennie, C., Call, J., & Tomasello, M. (2009). Ratcheting up the ratchet: on the evolution of cumulative culture. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 364(1528), 2405–2415. https://doi.org/10.1098/rstb.2009.0052
Tennie, C., Call, J., & Tomasello, M. (2012). Untrained chimpanzees (Pan troglodytes schweinfurthii) fail to imitate novel actions. PLoS ONE, 7(8). https://doi.org/10.1371/journal.pone.0041548
Tennie, C., Greve, K., Gretscher, H., & Call, J. (2010). Two-year-old children copy more reliably and more often than nonhuman great apes in multiple observational learning tasks. Primates, 51(4), 337–351. https://doi.org/10.1007/s10329-010-0208-4
Tennie, C., Hopper, L. M., & van Schaik, C. P. (2020). Consideration of the Role of Copying in Culture- Dependent Traits and a Reappraisal of the Zone of Latent Solutions Hypothesis. In S. Ross & L. M. Hopper (Eds.), Chimpanzees in Context: A Comparative Perspective on Chimpanzee Behavior, Cognition, Conservation, and Welfare. University of Chicago Press.
Tomasello, M., Call, J., Warren, J., Frost, T., Carpenter, M., & Nagell, K. (1997). The ontogeny of chimpanzee gestural signals: a comparison across groups and generations. In S. Wilcox (Ed.), Evolution of Communication (pp. 224–259). John Benjamins.
Whiten, A. (1998). Imitation of the Sequential Structure of Actions by Chimpanzees (Pan troglodytes). Journal of Comparative Psychology, 112(3), 270–281. https://doi.org/10.1037/0735-7036.112.3.270
Whiten, A., Horner, V., Litchfield, C. a, & Marshall-Pescini, S. (2004). How do apes ape? Learning & Behavior, 32(1), 36–52. https://doi.org/10.3758/BF03196005
Whiten, A., Horner, V., & Marshall-Pescini, S. (2005). Selective Imitation in Child and Chimpanzee: A Window on the Construal of Others’ Actions. In S. Hurley & N. Chater (Eds.), Perspectives on Imitation: From Neuroscience to Social Science: Mechanisms of Imitation and Imitation in Animals (Vol. 1, pp. 263–283). MIT Press.