Innovation in chimpanzees

The study of innovation in non‐human animals (henceforth: animals) has recently gained momentum across fields including primatology, animal behaviour and cultural evolution. Examining the rate of innovations, and the cognitive mechanisms driving these innovations across species, can provide insights into the evolution of human culture. Especially relevant to the study of human culture is one of our closest living relatives, the chimpanzee (Pan troglodytes). Both wild and captive chimpanzees demonstrate an impressive ability to innovate solutions to novel problems, but also a striking level of conservatism in some contexts, creating a unique and at times puzzling, picture of animal innovation. Whilst the animal innovation field is rife with potential for expanding our knowledge of human and non‐human cognition and problem‐solving, it is undermined by a lack of consistency across studies. The field is yet to settle on a definition of the term ‘innovation’, leading to studies being incomparable across and even within the same species. Here, we fill two gaps in the literature. First, we discuss some of the most prevalent definitions of ‘innovation’ from different fields, highlighting similarities and differences between them. Secondly, we provide an up‐to‐date review of accounts of innovations in both wild and captive chimpanzees. We hope this review will provide a resource for researchers interested in the study of innovation in chimpanzees and other animals, as well as emphasising the need for consistency in the way in which innovations are reported.


I. INTRODUCTION
Many animal species are able to innovate new behaviours, modify existing behaviours in response to changing environments, and invade new niches by developing novel behavioural repertoires (Reader & Laland, 2002). One of the most celebrated cases of animal innovation is that of Imo, a Japanese macaque (Macaca fuscata), who began removing dirt from provisioned sweet potatoes by washing them in the sea before eating them (Kawai, 1965). Imo's behaviour spurred research into the innovative abilities of animals, as it was identified as a field that could provide insight into various aspects of human and non-human primate (henceforth: primate) behaviour and cognition. Since then, innovation has been considered to be of critical interest across multiple biological fields due to its relationship with brain size (Reader & Laland, 2002;Lefebvre, Reader, & Sol, 2004), impact upon species' fitness and ecology (Giraldeau, Caraco, & Valone, 1994;Reader & Laland, 2002;van Schaik & Pradhan, 2003;Lefebvre et al., 2004), and potential utility as a proxy measure for cognitive ability and behavioural flexibility (Sol, Timmermans, & Lefebvre, 2002;Reader & MacDonald, 2003). An increasing capacity for innovation may drive the evolution of cultural transmission by generating non-genetic information in a population, thereby increasing the benefit of social learning (Enquist et al., 2008;Hoppitt & Laland, 2013), and innovation is also a key process driving (human) cumulative culture (Tomasello, 1999;Dean et al., 2012;Legare & Nielsen, 2015). Due to the shared ancestry of humans and other primates, data from primate studies may provide insight into the evolution of innovation and culture in our own lineage. Thus, for researchers interested in the evolution of human and primate culture, studying the cognitive mechanisms behind innovations can be invaluable (Reader & Laland, 2002, 2003. Perhaps more pressingly, species with more extensive innovative abilities might be able to adapt better when under threat from rapid environmental change or from invasion by other species (Whittaker, 1972;Reader & Laland, 2002;Ramsey, Bastian, & van Schaik, 2007;Sol et al., 2007). This is particularly pertinent to chimpanzees, given the continued impact of human activities upon great ape habitats, with less than 10% of African great ape habitat expected to be undisturbed by human activity by 2030 (Nelleman & Newton, 2002). An understanding of innovation is therefore timely for a wide range of fields of study [see also Kühl et al., 2019;Gruber et al., 2019].
This review will focus primarily on innovations in the technological domain in chimpanzees. As mentioned above, for researchers interested in the evolution of human innovative abilities, examining how and when innovations emerge in our closest living relatives, chimpanzees (Pan troglodytes) and bonobos (Pan paniscus) is critical. Although both species are equally relevant, more extensive records of innovation are available for chimpanzees, likely due to the greater research effort on this species (Gruber & Clay, 2016), but also perhaps due to the fact that chimpanzees (currently) demonstrate a larger repertoire of behaviours in the technological domain than bonobos (Furuichi et al., 2015). Whilst innovations can occur in all realms (e.g. gestures, vocalisations and social behaviours), there is a bias in the literature towards reporting technological and foraging innovations. This bias may be due to the fact that it is more straightforward to study technological and foraging behaviours experimentally in captive chimpanzees, and it may be easier to observe technological innovations than new social behaviours in wild chimpanzees. It may also be easier to distinguish novel technological and foraging behaviours from an existing repertoire than social behaviours. Furthermore, the purpose of new tool-use behaviours can often be construed from observations of these behaviours. On the other hand, identifying and understanding the purpose behind a new chimpanzee gesture, vocalisation, or other social behaviour may require a much deeper analysis after observation. As our review is based on existing reports of innovation in chimpanzees, it is necessarily also limited by this bias towards technological and foraging behaviour in the literature. However, we urge interested researchers to explore potential innovations in other behavioural realms and in other species to ensure a more complete understanding of primate innovation. Despite focusing on innovations in chimpanzees, inferences regarding the innovative abilities of other species, especially closely related primates (such as humans) can be drawn from our review.
(1) Chimpanzee culture The broad repertoire of behaviours observed in chimpanzees, alongside evidence of variation of these repertoires between communities in the wild (e.g. Kühl et al., 2019;Whiten et al., 2001Whiten et al., , 1999, has led researchers to suggest that chimpanzees possess 'cultural traditions' (Boesch, 1995;Whiten et al., 1999;Whiten et al., 2001;de Waal, 2002). Innovation is often central to definitions of culture (Ramsey et al., 2007;Nishida, Matsusaka, & McGrew, 2009;Legare & Nielsen, 2015), and therefore a full understanding of culture in chimpanzees requires examination of the impact of innovation at a group level. Despite being (to a certain extent) an innovative species (Hockings et al., 2015;Reader, Hager, & Laland, 2011; although see also Gruber et al., 2011;Gruber et al., 2009), many novel behaviours in wild chimpanzees are never acquired by the rest of the group, remaining limited to the original innovator ). This may result in some innovations appearing only transiently (as they are dependent on both the survival of their innovator and on-going use by just one individual). Furthermore, if the rest of the group fails to develop the behaviour, the population-level impact of an innovation is limited. A more complete understanding of the mechanisms behind innovations may provide insight into why some behaviours are never acquired by other members of the group, and how current chimpanzee cultures are catalysed and sustained across time. A first step towards this understanding is reviewing existing reports of innovations in chimpanzees. The cognitive mechanisms behind the emergence of novel behaviours in individual great apes and across different populations are still heavily debated. Some have suggested that once an innovation is observed in the first individual, various forms of social learning (most often high-fidelity social-learning mechanisms) are the likely means by which other individuals of the group acquire the same behaviour, and are necessary for it to be sustained in the population over generations (e.g. Whiten et al., 1999;van Schaik & Pradhan, 2003;Whiten & van Schaik, 2007;Gruber et al., 2011;Robbins et al., 2016). Others have argued that individual learning is sufficient for all individuals in a population to acquire the form of behaviours within their behavioural repertoire [the 'Zone of Latent Solutions' (ZLS) hypothesis (Tennie, Call, & Tomasello, 2009;Tennie & Hedwig, 2009;Bandini & Tennie, 2017)], but that increases in frequency, and differences in repertoires observed across populations, are facilitated and sustained by low-fidelity social learning. These authors suggest that non-human behavioural repertoires are the product of 'socially mediated serial reinnovations' (SMSRs; Bandini & Tennie, 2017;Tennie, Call, & Tomasello, 2012).
Whether chimpanzee cultural variation is primarily due to (high or low-fidelity) social or individual learning, or, most likely, a combination of these (Reader & Laland, 2002;Bandini & Tennie, 2017;Whiten & van de Waal, 2018), the focus of this review is not to distinguish between these hypotheses [although see Bandini & Tennie, 2017, Moore, 2013and Whiten, 2011 for further discussion of this topic]. Instead, we focus here on the fact that both approaches centre on individual innovation as the process by which novel behaviours are added to apopulation's or individual's repertoire. The first hurdle in studying such innovations is understanding which behavioural forms should be classified as innovations in the first place. To this end, several authors have offered definitions of innovation. Despite a growing interest in the field, there is, as yet, no universally agreed-upon definition of the term (see also Ramsey et al., 2007). The lack of any agreed multidisciplinary definition has led many authors to adopt their own definition to fit the requirements of their specific aims (see Table 1). This lack of agreement over the definition of innovation has created an absence of comparability between reports of innovations within and across disciplines, further resulting in a lack of integration.
Thus, the objective of this review is twofold. Our first aim is to provide an overview of the definitions of innovation used in the fields of primatology, cultural evolution and psychology, in order to examine the ways in which these definitions may shape our understanding and recognition of innovation in chimpanzees. Secondly, we review reported cases of innovation in both wild and captive chimpanzees, drawing inferences from these reports on the possible mechanisms and contexts that foster or hinder innovation in chimpanzees and other primates, including humans.
(2) Literature review on definitions of innovation We conducted a literature review on definitions of innovation within the fields of animal behaviour, cultural evolution and psychology. Key words for the literature search, conducted using search engines such as Google Scholar and Web of Science, included 'innovation definition', 'innovative behaviour', 'animal innovation' and 'primate innovation'. Parameters for inclusion were that the authors proposed a novel definition rather than citing previous definitions, and that the paper pertained to the fields of animal behaviour, cultural evolution or comparative psychology (this excluded definitions from the fields of business management and computer science, for example, which were not relevant to the subject at hand).
(3) Literature review on chimpanzee innovations We also carried out a literature search on published papers reporting innovations in wild and captive chimpanzees. We used key words such as 'chimpanzee novel', 'chimpanzee innovations', 'chimpanzee never seen before', 'chimpanzee new', and 'chimpanzee inventions' in search engines such as Google Scholar and Web of Science, following the protocol of previous reviews of animal innovation (e.g. Reader & Laland, 2002). To avoid subjective decisions by the authors, all studies that were found using these key words are summarized in Tables 2 and 3 as observational or experimental studies of innovation, respectively, regardless of whether these would be considered innovations following more recent definitions of the term (see Section II). The literature search yielded primarily reports of novel behaviours observed in wild chimpanzees or early studies on experimentally induced innovations in captive chimpanzees. However, we were also interested in studies in which eliciting innovations was not the aim, but occurred nonetheless (for example, studies on social learning in which chimpanzees in asocial control conditions produced novel, unseeded, solutions to the task). To find such studies, we used the key words 'chimpanzee spontaneous innovation' 'chimpanzee individual learning', 'chimpanzee asocial learning' and 'chimpanzee invention'.
Each behaviour has a separate entry in Tables 2 and 3, even if described within the same paper. We include a column indicating whether the behaviour is a probable innovation (whether the behaviour was likely to be genuinely novel to the individual or group in which it was observed), based upon the information provided by the authors. Cases in which the behaviour is indisputably novel (e.g. foraging on novel fruits not previously available) are categorised as 'probable innovations'. Cases in which the behaviour is likely to be novel, but strong evidence for this is still lacking, are categorised as 'possible innovations' (e.g. behaviours such as leaf clipping, which are relatively common in a population, but were only observed for the first time after researchers had been documenting behaviour for a long period). Cases in which the behaviour may not be novel (e.g. behaviours documented for the first time in the target paper, but have since been found to be customary behaviours in multiple communities) are categorised as 'unknown' to reflect the uncertainty regarding their status as innovations. In making these categorisations, we attempted to follow as broad a definition of innovation as possible and thus have focused on the reported novelty of the behaviour, regardless of whether this novelty was reported at an individual or population level. Additionally, we included any information provided regarding the original innovator, and whether the behaviour was suggested to have been acquired by other individuals within the group (as this is one of the requirements of some definitions of innovation: see Table 1; Hopper, 2016;O'Brien & Shennan, 2010). We note that many of these reports are observational, and many did not set out to investigate whether social learning of the reported innovation subsequently occurred. The details given in the column "Suggested to have been acquired socially" in Tables 2 and 3 are based on the information and interpretation provided by the original authors of each report.

II. DEFINING INNOVATION
Although it might seem at first that defining innovation is a relatively straightforward task, perhaps simply by following standard dictionary definitions, such as "the introduction of something new" (Merriam Webster, 2019), in reality providing a useful definition of the term for use in human and animal research has proved to be a challenge. Literature on this subject has been, and will continue to be, informed by the current definitions of innovation, which dictate the situations in which an animal's behaviour is perceived as being innovative. A selection of definitions of innovation is presented in Table 1; this list is not intended to be exhaustive, but to present key definitions that have influenced primatology, cultural evolution, and psychology.
Definitions of innovation differ in a few important ways, especially when one considers the operational implications of these differences. A crucial distinction is whether the behaviour is considered an innovation based upon novelty in the context of an individual's repertoire or in terms of the population's repertoireso, whether the behaviour is novel to the individual (but may already exist within the broader population) or whether it is a behaviour never before seen in the population. Contrast the definition suggested by Reader & Laland (2003;p. 14) "A process… that introduces novel behavioural variants into a population's repertoire", with that of Ramsey et al. (2007;p. 395) "The process that generates in an individual a novel learned behaviour". For Reader & Laland (2003), the primary characteristic of innovation is "novelty, relative to the baseline of the population's existing behaviour" (Arbilly & Laland, 2017, p. 1), whilst Ramsey et al. (2007) instead use the individual's existing repertoire as the baseline for judging the novelty of a behaviour. Both definitions have merits, both operationally and theoretically; identifying novel behaviours on a population level may be operationally simpler than identifying their emergence on an individual level, and does not even require identifying an original innovator. A population-based approach also avoids the possibility of  Carr et al. (2016Carr et al. ( , p. 1515 Human "In the physical realm, a behavioural innovation is a new, useful, and potentially transmitted learned behaviour, arising from asocial learning (innovation by independent invention) or a combination of asocial and social learning (innovation by modification), that is produced so as to successfully solve a novel problem or an existing problem in a novel manner." Hopper (2016, p. 1) see also Brosnan & Hopper et al. (2014) Non-human primates "…innovation is process that can be broken into three component steps. It begins with the initial invention, which is then transmitted to other members of the inventor's group, and is then adopted by other individuals and maintained within the society. These three stepsinvention, transmission, and maintenanceare all required for innovation and yet the factors that influence each step vary…" Mesoudi (2010, p. 175) Human "'Innovation' describes the processes by which a novel trait (an 'invention') emerges and becomes fixed in a population" Ramsey et al. (2007, p. 396) Non-human Primary: "The process that generates in an individual a novel learned behaviour that is not simply a consequence of social learning or environmental induction" Secondary: "Repertoire modification involving the addition of a new behaviour, or the modification of an old one, underdetermined by maturation, the environment, and the behaviour of conspecifics" Reader & Laland (2003, p. 4437) Non-human "A process that results in new or modified learned behaviour and that introduces novel behavioural variants into a population's repertoire" Kummer & Goodall (1985, p. 205) Non-human primates "An innovation can be: a solution to a novel problem, or a novel solution to an old one; a communication signal not observed in other individuals in the group (at least at that time) or an existing signal used for a new purpose; a new ecological discovery such as a food item not previously part of the diet of the group." Tebbich et al. (2016, p. 2) Non-human "the discovery of a new behavioural interaction with the social or physical environment, tapping into an existing opportunity and/or creating a new opportunity"             falsely identifying an individual's acquisition of a novel behaviour as being due to innovation rather than some form of social learning. However, this approach also risks overlooking instances of innovation if multiple individuals in a population independently invent similar or identical behaviours. It may be that different definitions are suitable for different study contexts; in the wild, it may be difficult to gather sufficient data to determine the role of individual and social learning in the acquisition of a behaviour by other members of the group, once the behaviour has appeared. In this scenario, a definition based upon novelty at a population level may be more useful [for example, see Kalan & Boesch, 2018, in which the leaf clipping gesture re-emerged in a wild chimpanzee group after an absence of 2 yearsthis behaviour is not considered an innovation due to its previous presence in the group]. However, in captive experimental situations, e.g. where access to a novel foraging device can be carefully controlled and all interactions monitored, researchers may be able to say confidently that multiple individuals arrived at an innovation independently. Imposing a population-based definition in these scenarios would underestimate a species' propensity to innovate. The choice of whether to use an individual-or population-based definition may therefore depend somewhat on the focus and methodology of a given study. The second key distinction is the inclusion of the necessity for the transmission of the novel behaviour. Several definitions include, either explicitly or implicitly, a requirement for the innovated behaviour to be acquired by other group members, beyond its innovator [Hopper (2016, p. 3): "transmission, and maintenance -are…required for innovation"; Mesoudi (2010, p. 175): "a novel trait… emerges and becomes fixed in a population"; Reader & Laland (2003, p. 14): "A process… that introduces novel behavioural variants into a population's repertoire"; although note that Reader & Laland (2003, p. 14) also specify that: "population repertoire is not meant to imply that all individuals in a population will necessarily acquire the novel behaviour, but rather that at least one individual in the population will behave in a manner not previously seen"]. Tebbich et al. (2016) include social diffusion as the fourth of their suggested six "phases of innovation", arguing that relevant and significant innovations with adaptive value must be acquired by other group members and across generations, a point also made by Reader & Laland (2003). However, Tebbich et al. (2016, p. 7) point out that their six phases should not be thought of as necessarily following a linear order ("Discovery of an opportunity", "Discovery of a favourable interaction", "Repetition and testing", "Social diffusion", "Exploitation of an adjacent opportunity", "Production of an opportunity") but rather, as a cyclic system (such that "Exploitation of an adjacent opportunity" may feed back into "Discovery of a favourable interaction", generating a continuing sequence of innovation). Tebbich et al. (2016) argue that some phases may be skipped in this cyclic process, and so social transmission does not appear to be an essential component in order to classify a behaviour as an innovation (although it may be an indication of its evolutionary relevance). Carr, Kendal & Flynn's (2016, p. 1516 definition does not call for transmission of the novel behaviour, but the authors still include a requirement for transmission in order for an innovation to be classed as 'high level' innovation (defined as "an individually learned innovation that is acquired by others"). Distinguishing in this manner between those innovations which remain at an individual level and those which contribute more extensively to population repertoires may be useful. Indeed, as Tebbich et al. (2016) point out, such social transmission may be necessary for an innovation to have evolutionary relevance. The framework of Carr et al. (2016) still classifies behaviours that are not acquired socially as innovations, although at a lower level, and may be particularly useful in the case of chimpanzees, a species in which innovators are likely to be low ranking (Reader & Laland, 2001). Furthermore, low-ranking individuals may be more exploratory [Hopper et al., 2015a; see also Kalan et al., 2019 for evidence that younger wild chimpanzees are more exploratory], alongside other factors that may also influence whether an individual innovates a new behaviour, such as sex, age, temperament, developmental stage, motivation, and so on (Reader, 2003). Low-ranking innovators may also be unlikely to seed novel behaviours due to a potential rank bias in chimpanzee social learning (Horner et al., 2010;Kendal et al., 2015; but see also Watson et al., 2017). Similarly, other authors, while not putting forward a new definition of innovation, have suggested referring to novel behaviours which are not acquired by others as 'inventions', reserving the term 'innovation' for those behaviours which are then adopted by other group members (McGuigan et al., 2017). Definitions which explicitly require the transmission and maintenance of behaviours in order to classify such behaviours as innovations at all may therefore lead to the exclusion of examples from species in which social learning is biased towards individuals which are unlikely or infrequent innovators, or from species which are largely solitary.
However, incorporating 'learning' [here, following Ramsey et al. (2007, p. 396), taken to mean "modifying the (brain of the) organism so that it behaves differently in the future"] in a definition of innovation that allows innovations to be distinguished from accidents or improvisations [following Ramsey et al. (2007, p. 397), improvisation is "a novel behaviour that fits all of the criteria of being an innovation except that it is not learned by the individual" and an accident is an unintentional behaviour that is not learned by the individual]. The argument in favour of including transmission of a novel behaviour beyond its innovator in the definition of this term should therefore be distinguished from the inclusion of learning in the individual innovator, something the majority of definitions appear to agree upon. Carr et al. (2016) suggest categorising innovations as 'high, medium and low level' depending on the extent of learning involved (i.e. a low level indicates the innovator themselves does not learn or reproduce the behaviour; medium indicates only the innovator adopts the behaviour; and a high level indicates transmission of the behaviour to other group members). This places the emphasis on the outcome rather than the source of innovation. Again, this focuses on the impact Biological Reviews 95 (2020)  of innovation in terms of population-level change. Carr et al. (2016) are not the only authors to suggest categorising innovations. Arbilly & Laland (2017) argue for a conceptualisation of the magnitude of innovation, based upon the deviation of a behaviour from the current population repertoire. Whiten & van Schaik (2007) distinguish 'cognitively simple' and 'cognitively complex' innovations, based upon the necessity for causal reasoning and deliberate action (termed 'complex'), as opposed to innovations that occur accidentally (but which may later be deliberately reproduced). Rendell, Hoppitt, & Kendal (2007) make a similar distinction in their argument for delineating 'passive' and 'active' innovation, with active innovation being more likely to reflect the cognitive abilities of the innovator, rather than relying upon chance events. While distinguishing between cognitively taxing, deliberate innovation, and accidental innovation is of great interest to those studying the cognitive processes underlying innovation, accidental or 'passive' innovations have proved important in the origins of widespread human cultural phenomena (for example, the semi-accidental discovery of penicillin by Alexander Fleming, although here we note Pasteur's famous quote: "Chance favours only prepared minds"famous instances of 'accidental' innovation in humans are often part of, or are followed by, a process of intentional exploration and refinement; van Andel, 1994;de Rond, 2014). Ramsey et al. (2007) suggest a continuum rather than strict categories, with weak innovations being behaviours that, while primarily explained by the process of innovation, draw significantly upon social learning or environmental induction, and inventions being those innovations that are rarer, more novel and involve more cognition. Ramsey et al. (2007) also suggest that the 'passive' innovations of Rendell et al. (2007) might fall within their description of weak innovations. These categorisations broadly fall into two types: those that focus on either (i) the innovation within the context of the population [i.e. transmission to others, or extent of deviation from the norm (e.g. Carr et al., 2016;Arbilly & Laland, 2017)] or (ii) the cognition of the innovator and corresponding complexity involved in generating the innovation (e.g. Ramsey et al., 2007;Rendell et al., 2007 ;Whiten & van Schaik, 2007).
Despite the reference made to cognition in the abovementioned categorisations of innovation (Ramsey et al., 2007;Rendell et al., 2007;Whiten & van Schaik, 2007), the definitions listed in Table 1 are largely focused upon the novelty of the emergent behaviour, rather than the process underlying it. Yamamoto, Humle, & Tanaka (2013) suggests that invention of novel behaviours (particularly tool-use behaviours) can be the result of three types of mechanisms: accident, trial-and-error, and insight. 'Accidental' invention requires only that the individual learns and repeats the behaviour. Trial-and-error invention implies the individual engages in a process whereby potential solutions are attempted and abandoned or refined until a satisfactory outcome is achievedthis mechanism does not require any causal understanding of the problem at the outset. Finally, insightful invention, in which a solution is invented without a trial-and-error process preceding success, implies a sufficient understanding of the causal relationships involved prior to attempting to solve the problem. Each of these mechanisms is put forward by Yamamoto et al. (2013) as a potential route to the invention of novel (tool-use) behaviours. Similarly, Köhler (1925) argued for a distinction between 'chance' and 'insight' in his landmark studies of chimpanzee problem-solving. One could consider Yamamoto's (2013) 'accidental' and 'insightful' invention analogous to Köhler's (1925) 'chance' and 'insight' problemsolving. It must be noted that while more descriptive of potential mechanisms than the product-focused definitions described previously, Yamamoto et al. (2013) still does not define specific cognitive processes.
An additional potential source of novel behaviour in a population is conserved behaviours. For example, Wegdell, Hammerschmidt, & Fischer (2019) found that when exposed to a flying drone stimulus, West African green monkeys (Chlorocebus sabaeus), produce an alarm call distinct from their usual repertoire, but highly similar in structure to the aerial predator alarm call given by closely related East African vervet monkeys (Chlorocebus pygerythrus). It is possible that when exposed to certain environmental stimuli, animals may produce novel behaviours not seen before in their population, but that are "hard-wired" (Wegdell et al., 2019(Wegdell et al., , p. 1040. Such behaviours, relying upon environmental induction, would not be classified as innovations under the definition of Ramsey et al. (2007), but might be considered innovations under that of Reader & Laland (2002), as they are novel to the population and are repeated over time, thus adding to the population's behavioural repertoire. Whilst knowledge of the mechanism underlying the invention of a novel behaviour may not always be critical, depending upon the question one aims to address, it is important to recognise that behaviours reported as 'innovations' may result from very different cognitive processes.
There is also a potential distinction to be made between definitions which discuss innovation as a product (i.e. the resulting behaviour, e.g. "An innovation can be: a solution to a novel problem, or a novel solution to an old one"; Kummer & Goodall, 1985, p. 205) and those that discuss innovation as a process (i.e. the act of innovating). Ramsey et al. (2007) in fact provide two definitions, one for the process (see Table 1) and one for the product ("An innovation (sensu product) is any learned behavioral variant created through the process of innovation"; Ramsey et al., 2007, p. 397). However, those definitions which describe innovation as a process generally focus upon the emergent behaviour as the criterion by which innovation is judged (e.g. "The process that generates in an individual a novel learned behaviour"; Ramsey et al., 2007, p. 396). Therefore, even in those definitions which conceive of innovation as a process, the key factor remains the novelty of the resulting behaviour, rather than any specific cognitive mechanism that could be identified independently of the resulting behaviour. Thus, the novelty of the emergent behaviour, relative either to the existing population or individual repertoire, remains the key focus of definitions of innovation.
Broadly, it seems that each definition may prove useful depending on the question one is asking. If one is interested in the innovative capacity of individuals, perhaps in the Biological Reviews 95 (2020)  context of captive-animal experimental work, definitions which focus upon individual-level novelty and complexity of cognition may be more appropriate. On the other hand, for population-level questions, perhaps related to fitness or longterm cultural dynamics, definitions which incorporate social transmission and focus on population-level repertoires may be useful. Difficulty may arise, however, when gathering evidence of innovation from across wild and captive animals, and observational and experimental work, when authors' criteria for assessing innovation are informed by disparate definitions.

III. CHIMPANZEE INNOVATIONS
(1) Chimpanzee innovations in the wild As described above, we included all behaviours that were classified by the authors as 'innovations', regardless of the definition they adopted. We discuss the implications of their choice of definition below. One of the first systematic reports of innovations in wild chimpanzees described the "new" behaviours of chimpanzees in the Gombe Stream Reserve, Tanzania (Kummer & Goodall, 1985). The loose definition of innovation adopted by the authors (see Table 1), alongside the fact that chimpanzee behaviour had yet to be exhaustively documented by researchers at this early stage, led to several behaviours being categorised as innovations that might not be considered as such following more recent definitions (see Table 1). For example, Kummer & Goodall (1985, p. 208) describe as an innovation a young male's deceptive response to banana-feeding, in which he would get up and walk away with a "purposeful gait… [and] as is often the case when one individual sets off as though with a goal in mind, the [group] followed"; he would then return later to retrieve more bananas without competition (see Table 2). This behaviour was considered an innovation as it was only recently observed by the researchers, but would likely no longer be categorised as such without additional evidence. However, Kummer & Goodall (1985, p. 210) also describe 'ecological and technical' innovations seen in juveniles during play, for example: "Occasionally an infant will taste a novel food object: behaviour which we have never observed in an adult chimpanzee in the wild". Tasting novel foods would be considered an innovation following most of the definitions above. However, again, it is unclear whether these behaviours were truly innovations for the individuals performing them, or whether these reports simply document the first time researchers observed them. Although the authors did not test the mechanisms behind the acquisition of these (potentially) novel behaviours, Kummer & Goodall (1985) concluded that other group members later developed some of these behavioural forms through observational learning.
Following Kummer & Goodall's (1985) reports of novel behaviours in wild chimpanzees, Boesch (1995) documented the innovative behaviours of chimpanzees in Tai National Park, Cote d'Ivoire. Boesch (1995) followed Kummer & Goodall's (1985) definition of innovation (see Table 1) and operationalised this to identify as innovations all behaviours observed in the Tai chimpanzee's repertoire after the first review of behaviours (Boesch & Boesch, 1990). Boesch (1995) additionally included behaviours that had been observed previously, but had since varied in the context of their use, or in frequency, as innovations. Three tool-use behaviours (see Table 2) and three non-tool-use behaviours were counted as potential innovations, as Boesch (1995) also acknowledged that it is possible that these behavioural forms were practiced by the chimpanzees before the second review, and were simply not observed previously. However, Boesch (1995) described one "genuine invention": the insertion of small sticks into holes to remove larvae from their envelopes. This behavioural form was argued to have been invented recently, and alongside other innovations, to have been acquired by the rest of the group through social learning, creating chimpanzee "fashions" (Boesch, 1995). Again, although the cognitive mechanisms behind these behavioural forms were not investigated, Boesch (1995) also suggested that social learning of some form must have driven their acquisition in other group members. Nishida et al. (2009) described novel behaviours of chimpanzees in Mahale, Tanzania. Similarly to Boesch (1995), Nishida et al. (2009, p. 24) operationalised innovation as "a behavioral pattern seen by observers for the first time only after 1981". Innovations, following this operational framework, varied in context, and included new feeding, human-directed behaviours, courtship, hygiene, maternal, play, intimidation and grooming behavioural patterns (see Table 2). The authors also recorded whether the innovations were observed in other members of the community, and found that whilst innovations were not rare (N = 32 new behavioural patterns were observed), more than half of the behaviours (N = 21) were not reported in other members of the group, and six behaviours disappeared completely, even from the original innovators . Thus, contrary to Kummer & Goodall's (1985) and Boesch's (1995) reports, where the potential role of social learning in the propagation of innovations is emphasised, Nishida et al. (2009, p. 34) concluded that most of the behavioural forms practiced by the chimpanzees were "so simple that anyone could learn to do them via trial-and-error learning", and suggested that it is rare for a behaviour to propagate from an innovator to the rest of the community, despite opportunities for social learning. It is therefore possible that many of these behaviours were not adaptive in the wild (otherwise, given their simplicity, one would expect them to spread via either social or asocial learning), and this lack of utility might be why they never 'caught on'. Such behaviours, which fail to spread beyond their innovator, might be better thought of as idiosyncrasies (and indeed would fail to meet the criteria of some definitions of innovation due to their failure to spread to other group members).
a Observational studies of chimpanzee innovations that were acquired by group members Although these early reports present interesting data on the behavioural repertoire of chimpanzees across field sites, recent reports on newly observed behaviours in the wild are Biological Reviews 95 (2020)

Innovation in chimpanzees
substantially more conservative in labelling behaviours as innovations, acknowledging the possibility that the behavioural forms may have already been present in the population before they were observed for the first time by researchers. Thus, it is very difficult to determine with confidence that a behaviour in the wild constitutes a genuine innovation, if the historic behaviour of all the individuals within a group is largely unknownas was often the case in early field studies (e.g. Kummer & Goodall, 1985;Boesch, 1995;Yamakoshi & Sugiyama, 1995;Nishida et al., 2009). Note that this problem can also extend to studies with captive subjects (see Section III.2 for further discussion). This can be illustrated by a study documenting the re-emergence of the leaf clipping gesture in a group of wild chimpanzees in the Taï forest in Côte d'Ivoire (Kalan & Boesch, 2018). After an absence of 2 years, the behaviour re-emerged in the group during an alpha male takeover. Without a long-term record of the group's behaviour, this gesture could have been misidentified as an innovation. With the establishment of increasingly long-term field sites, first instances of novel behaviours can be more confidently identified in wild chimpanzees [and, as in Kalan & Boesch, 2018, the re-emergence of behaviours can be discriminated from innovation]. Hobaiter et al. (2014) reported the emergence of two behavioural variants in Budongo Forest, Uganda: moss-sponging and leafsponge re-use. These two behavioural variants emerged in response to the discovery of a watering hole that had been flooded throughout the previous rainy season. Using network-based diffusion analysis, the authors were able to track the spread of these behaviours throughout the group. From this analysis, Hobaiter et al. (2014) argued that social learning accounted for 85% of observed events of mosssponging, after the alpha male and female independently innovated this behaviour. On the other hand, social learning did not account for the propagation of the second behaviour, leaf-sponge re-use, as several individuals were observed to innovate the same behavioural form independently, without any previous experience. This study is rare in that repeated innovations in the wild were observed and their diffusion throughout the chimpanzee population could be tracked; therefore it is one of only a few which meets requirements, implied in some definitions of innovation, for a behaviour to spread beyond its initial innovator (e.g. Reader & Laland, 2003;Mesoudi, 2010;Hopper, 2016). In a followup study conducted 3 years after the original observations, moss-sponging was still being practiced by most of the group (Lamon et al., 2017). The authors of the new study concluded that the persistence of this innovation was due to a combination of individual and social learning (Lamon et al., 2017). Further research has shown that moss-sponges are both more effective at absorbing water than leaf-sponges, and can be manufactured more quickly, suggesting that this innovation represents a functional improvement upon the existing behaviour of leaf-sponging (Lamon et al., 2018). By some definitions, this behaviour may therefore meet the requirements of cumulative culture, as it is a more efficient behaviour that arguably builds upon a previous behaviour (e.g. Sasaki & Biro, 2017;Schofield et al., 2018). However, many researchers argue that cumulative culture must consist of behaviours which go beyond those that an individual could invent alone Dean et al., 2014), which has not been shown to be the case for moss-sponging.
Another observational report of innovation comes from van Leeuwen et al., 2014 who reported the emergence of a new, seemingly non-adaptive, 'grass in ear behaviour' (GIEB) in semi-free-ranging, sanctuary-housed chimpanzees. This behaviour was originally innovated by an adult female chimpanzee residing at the Chimfunshi Wildlife Orphanage in Zambia, Africa, and consisted of placing a stem of grass in the ear and leaving it in whilst carrying out other activities (van Leeuwen et al., 2014). The behaviour was gradually observed in other individuals within the same group. Only one observation of GIEB was made in a different group to the innovator's, leading the authors to conclude that GIEB was socially learnt by naïve chimpanzees in the original innovator's group (van Leeuwen et al., 2014). The behaviour continued for a brief period after the death of the innovator, but individuals gradually ceased the behaviour, which was continued by only two chimpanzees at the time of van Leeuwen et al.'s (2014) publication. This report provides a rare glimpse into the origins and end of an innovation. Furthermore, it is particularly interesting as the behaviour did not have a clear adaptive value, yet was still observed across one chimpanzee population and once in a second population.

b Attempts to induce innovations experimentally in wild chimpanzees
As tracking the first instance of an innovation in the wild is so rare, past studies on innovation in chimpanzees have attempted to recreate this process experimentally by providing novel foraging opportunities to wild groups. For example, Biro et al. (2003) carried out field experiments on the acquisition of nut-cracking in chimpanzee populations in Bossou, Guinea. These chimpanzee communities are particularly interesting for the study of nut-cracking as they seem to differ in their selection of which species of nuts to crack, differences which cannot be attributed to ecological features (Biro et al., 2003). The authors presented two new species of nuts (Coula nuts Coula edulis and panda nuts Panda oleosa) to one of the wild groups, to examine whether individuals would spontaneously crack the new types of nuts despite having no prior experience with them. One female in the group, Yo, started cracking both types of new nuts almost immediately. The rest of the group varied in their response, with all juveniles adopting coula-nut-cracking soon after Yo (at least 60% of juveniles continued to crack coula nuts over the years after the experiment), but no adults attempted to crack the coula nuts during the first trial (Biro et al., 2003). In subsequent presentations of the nuts, the number of adults cracking coula nuts gradually rose from 11% in 1993 to 67% in 2002 (Biro et al., 2003). Two adults (including Yo) and two juveniles attempted to crack the panda nuts in the first experiment, but soon after, all individuals abandoned panda-nut-cracking. The difference between the acquisition of coula-nut-cracking and panda-nut-cracking is striking, but may simply have been the product of differences in taste and in the toughness of the shell, as panda nuts are one of the hardest nuts to crack (Boesch & Boesch, 1983). Although both types of nuts were novel to the group, Biro et al. (2003) argue that these observations do not constitute 'true' innovations, as Yo (the only individual who spontaneously cracked both types of nuts) immigrated into the group as an adolescent and might have had experience of these nuts from her previous group. Although it is possible that Yo was familiar with the nuts beforehand, it is also conceivable that Yo simply adapted her pre-existing knowledge of nut-cracking to the new species of nuts, successfully innovating a new type of nut-cracking. However, if Yo was indeed naïve to the new nuts, an open question would be whether there were any factors that made Yo more likely to innovate compared to the other chimpanzees in the group (see Section IV for further discussion).
A case of a novel foraging technique failing to 'catch-on' was described by Gruber et al. (2009Gruber et al. ( , 2011, who examined honey-dipping behaviour in two communities of East African chimpanzees (the Sonso community of Budongo Forest and Kanyawara community of Kibale National Park). Whilst the chimpanzees in Kanyawara frequently use sticks to retrieve honey from tree holes, the chimpanzees at Sonso use their hands instead or, occasionally, leaves (as a leafsponge). To examine whether the Sonso chimpanzees could be encouraged to use sticks to retrieve honey, both groups of chimpanzees were provided with similar artificial logs with drilled holes baited with honey. Both groups persisted in their community-specific techniques to retrieve the honey. Even after Gruber et al. (2011) scaffolded the alternative method by inserting sticks into the holes for the Sonso individuals, these chimpanzees simply removed the sticks and continued using leaves to sponge the honey, whilst the Kanyawara chimpanzees continued exclusively to use sticks, and never innovated the leaf-sponging technique (Gruber et al., 2011). Simply providing the ecological challenge, along with 'scaffolding' towards a certain technique, proved unsuccessful in encouraging innovation in wild chimpanzees. Gruber et al. (2009Gruber et al. ( , p. 1807 concluded that: "The fact that all the chimpanzees reacted in a community-specific way supports a culturally based rather than an individual acquisition of the behaviour", and in a subsequent publication suggested that these results may be explained by a "'cultural bias', which constrains how chimpanzees of different communities perceive and evaluate their environment" (Gruber et al., 2011, p. 5). Thus, the authors argued that the Sonso chimpanzees did not innovate stick use for honey-dipping because their previous (lack of) experience constrained the way in which they perceived the problem. This raises the interesting possibility that chimpanzees' innovations are shaped, and potentially restricted, by their existing behavioural repertoire. Others have argued that, since the Sonso chimpanzees already had a method to retrieve the honey, they may have not been motivated to switch to a new method to achieve the same goal [Tennie & Hopper, 2011comment on Gruber et al., 2011; although see also Gruber, 2016]. Although no data exist on the efficiency of the two methods, Gruber (2016) stated that the methods are likely equivalent, as the most frequent stick-tool user and leaf-sponge user did not differ in the time taken to empty the honey trap. Chimpanzees have demonstrated a relative inflexibility in switching to new techniques if the difference in efficiency between the existing and the new technique is small [e.g. Davis et al., 2016;Vale et al., 2017; although see Jacobson & Hopper, 2019 for evidence of chimpanzees switching to an only marginally more efficient technique than their known technique]. Therefore, it may be that a combination of lack of motivation to innovate a new, similarly efficient technique, and the absence of sociallearning opportunities may have hindered the Sonso chimpanzees' motivation to innovate. These factors may have also contributed to Nishida et al.'s (2009) finding that most innovations in Mahale never 'catch-on'.

c Social innovations
A variety of innovations in the social or play domains are also reported in wild and captive chimpanzees via observational study; these constitute 24 (44%) of the 54 reports of innovation found in our review of observational studies (see Table 2), although the majority of these come from the two seminal site-specific reviews of Kummer & Goodall (1985) and Nishida et al. (2009). The behaviours described include novel play behaviours, generally observed in infants or juveniles [e.g. novel locomotor play in an infant female (Kummer & Goodall, 1985); water play first observed in a juvenile female ], but also include novel social strategies, such as incorporating paraffin cans in display (Kummer & Goodall, 1985) and novel courtship behaviours ('shrub bending' courtship and 'stem pull through' courtship; Nishida et al., 2009).
Candidate innovative social behaviours have also been observed in captive chimpanzee groups, with de Waal & Seres (1997) reporting the spontaneous emergence of handclasp grooming (HCG) in a captive group. This behaviour was initiated by an adult female, and the authors tracked its propagation over 5 years, reporting that the behaviour increased in frequency, and that while initially the originator or her kin were involved in the majority of instances of HCG, over time other individuals participated and also began to initiate HCG bouts. Of course, HCG requires at least two individuals to perform the behaviour, although one participant need only tolerate the form of grooming initiated by the other. The fact that over time, the originator's kin and some non-kin began not only to participate in, but to initiate HCG bouts suggests that the behaviour may have been socially facilitated by others, a view supported by a later study showing that dyads with strong affiliative ties were more likely to develop HCG, and that in most new dyads, at least one individual had previously handclasp groomed (Bonnie & de Waal, 2006 observed initiating HCG, this behaviour has also been observed in other groups, both in captivity (van Leeuwen et al., 2012) and in the wild (McGrew & Tutin, 1978).
This raises the interesting question of why certain behaviours may arise seemingly independently across multiple populations. Dean et al. (2014, p. 288) point out that similar behaviours may arise independently due to there being a "salient, or easily discoverable, method" for achieving a particular goal. Similarly,  argue that independent reinnovations [see Bandini & Tennie, 2017 for an extensive definition of this term] of similar behavioural forms across unconnected populations are due to species sharing a zone of latent solutions (ZLS).  argue that behavioural forms within a species' ZLS are available to all individuals, and will develop when the individual is in the appropriate developmental, motivational and environmental context, primarily via individual learning. It seems plausible that if separate populations share ecological or social factors, the same behaviours may prove to be rewarding, and thus arise independently, with shared morphological and cognitive factors making some solutions more salient than others. (

2) Chimpanzee innovations in captivity
Although it can be challenging to observe the emergence of innovations in the wild, experimental tests on innovation with captive chimpanzees provide an opportunity for the cognitive mechanisms involved in the emergence of a new behaviour to be identified, whilst also allowing researchers to control testing conditions (an aspect which is often impossible in wild studies). Researchers may also be able to control for individuals' background knowledge and thus more confidently categorise novel behaviours as innovations, although we note that this is not always possible, depending on the individual history of the chimpanzee and the extent to which its prior experience has been documented. Past research with captive chimpanzees has generated a large body of data on targeted innovations (studies which focused on encouraging innovation) and on unexpected innovations (innovations which occurred in studies whose main focus was on another aspect). Table 3 provides an overview of these studies. Köhler (1925) described some of the earliest experimental studies on captive chimpanzee tool use and innovation. Köhler (1925) tested chimpanzees at the research station of the Prussian Academy of Sciences on the island of Tenerife by placing food just out of reach and providing his subjects with tools to help them retrieve it. Köhler did not give any social information on how to retrieve the food to examine what solutions the chimpanzees could innovate independently. Alongside his experiments, Köhler also documented all innovations of tool use he observed. Under these conditions, the chimpanzees demonstrated an ability to innovate a wide variety of tool-use solutions, a finding Köhler (1925) attributed to "insight learning" (what we would now interpret as a form of individual learning; see also Reindl, Bandini, & Tennie, 2018). With these studies, Köhler may have been one of the first to induce chimpanzee innovations experimentally. Birch (1945) carried out a similar study with captive chimpanzees (the subjects were born and raised in a testing facility), and found that two chimpanzees spontaneously used provided sticks to rake in out-of-reach food. One individual, who had previously used sticks to manipulate a light-switch outside her enclosure, was able to innovate the use of a stick as a rake within 12 s of presentation of food outside her enclosure. The second individual accidentally brushed against the provided stick, causing it to pivot and move the food reward. He then used the stick to sweep food towards himself. This report is interesting, as it indicates two potential mechanisms of innovation; the first individual appears to have been able to generalise existing knowledge to a novel problem, perhaps in a manner akin to what Köhler (1925) called "insight", while the second individual appears to have reached a solution via accident or trial-anderror learning. Exposure to sticks increased the likelihood of innovation, as Birch (1945) reports that after a longer period of exposure, the rest of the group (four other individuals) used stick tools as rakes in a similar manner to the two initial innovators. This study therefore provides evidence that chimpanzees can innovate stick-tool use, despite varying amounts of experience with the tools [contrary to the findings of Gruber et al., 2009Gruber et al., , 2011 with wild chimpanzees (see Section III1b)although Birch's chimpanzees knew no existing alternative solution to the task, unlike the wild chimpanzees in Gruber et al., 2009Gruber et al., , 2011. Paquette (1992) focused on the mechanisms behind the innovation of wild chimpanzee behaviours by providing naïve captive chimpanzees with artificial bee hives. All the chimpanzees in the group innovated the same behavioural form of honey-fishing soon after the apparatus was placed in the enclosure. A more recent study (Bandini & Tennie, 2017) followed the same logic by providing two groups of naïve, captive chimpanzees with an ecological problem analogous to algae scooping (the chimpanzees were provided with buckets and floating bread, to simulate algae); a behaviour practiced by wild chimpanzees in Bossou, Guinea (Humle, Yamakoshi, & Matsuzawa, 2011). Prior to testing, it was established (through questionnaires and interviews with keepers) that the chimpanzees did not have previous experience of the behaviour or any similar tasks. In both groups of chimpanzees, individuals spontaneously demonstrated the same behavioural form as their wild counterparts, despite never having seen the behaviour beforehand. Thus, the authors identified another case of innovation of a wild behavioural form by naïve chimpanzees. Similarly, Huffman & Hirata (2004) provided captive chimpanzees with Helianthus tuberosus leaves, similar in texture to leaves habitually swallowed by wild chimpanzees, to examine whether the behavioural form was driven by social learning, or whether naïve individuals could innovate this leaf-swallowing behaviour. Two chimpanzees (one male and one female) displayed leaf-swallowing behaviour in the first trial, without any opportunity to observe others (both innovators were captive born). Collectively, these findings, alongside some of the studies included in Table 3, lend support to the view that at Biological Reviews 95 (2020)  least some captive chimpanzees have extensive innovative abilities that allow them individually to learn some of the same behavioural forms as their wild counterparts (see also Bandini & Tennie, 2017. A further open question remains over whether these innovative abilities are enhanced in captive environments, due perhaps to a 'captivity effect' [i.e. reduced neophobia, increased exploration and more extensive tool-use and problem-solving abilities in captive animals than in their wild counterparts (van Schaik et al., 1983;Benson-Amram, Weldele, & Holekamp, 2013;Forss et al., 2015;Cheng & Byrne, 2018)], or whether innovations are simply easier to observe and elicit in captivity. In addition, various studies have found that exposure to human artefacts may lead to increased technical skill even in wild primates (van de Waal & Bshary, 2010), and that extensive human contact and/or training ('enculturation'; see also van Schaik & Burkart, 2011) has an important effect on the cognition of non-human primates. Enculturated chimpanzees have been found to be more successful than non-enculturated individuals in various experimental paradigms, including in object-choice and seeing-begging tasks (Buttelmann et al., 2007;Leavens, Russell, & Hopkins, 2010).
Furthermore, a recent meta-analysis of innovative problem-solving in both wild and captive animals across multiple taxa also indicates that different factors may influence innovation in wild and captive settings (Amici et al., 2019). The meta-analysis, which was based upon data from experimental foraging tasks, found that the relationship between innovation and exploration tendency, and innovation and neophilia, was stronger in captive than in wild animals, while social facilitation played a stronger role in wild animals (Amici et al., 2019). This raises the possibility that not only may captive animals possess traits that make them more likely to innovate than their wild counterparts (Benson-Amram et al., 2013), but that these factors may influence the probability of innovation to a different extent dependent upon the setting. Therefore, along with the 'captivity effect', the past rearing history of subjects (i.e. their level of enculturation, but also their level of deprivation, as some subjects that have been kept in deprived conditions whilst in captivity may not develop the same cognitive abilities and skills as their wild counterpartsalthough this is hopefully less often the case with extant captive chimpanzees) must to be taken into consideration when assessing studies on innovation in captive chimpanzees. Future studies should focus on examining the effects of captivity (including the lack of predators and more free time afforded by these conditions), and human contact and/or training, on the innovative abilities of chimpanzees and other animals (e.g. Haslam, 2013; Damerius et al., 2017).

a Failures to innovate in captivity
Despite this potential captivity effect, not all behaviours in captivity seem to be as easily innovated, and external influences can hinder chimpanzees from innovating certain behavioural forms (as in wild chimpanzees; Gruber et al., 2009Gruber et al., , 2011. For example, no chimpanzee at Leipzig Zoo was able to innovate the solution to the floating peanut task (in which, in order to retrieve a peanut from the bottom of a long tube, water needs to be inserted to make the peanut float; Hanus et al., 2011). However, five chimpanzees at the Ngamba Island chimpanzee sanctuary, Uganda, successfully innovated the solution (Hanus et al., 2011). The authors suggested that functional fixedness (the inability to develop a new function for a known tool or object; Hanus et al., 2011) might have blocked the Leipzig chimpanzees' ability to innovate, as these chimpanzees were tested using the same tube that they regularly drank water from. Meanwhile, the Ngamba Island chimpanzees were provided with a new tube that did not have a preexisting use for these chimpanzees. Davis et al. (2016) assessed the ability of captive chimpanzees to switch to a new, more productive or efficient technique to solve the same problem. The authors first seeded an inefficient technique for a foraging task in five captive groups. Three of the five groups were then exposed to a conspecific demonstrating a more efficient method, and the authors found that most individuals only switched to a new technique when their existing technique became significantly less efficient in comparison. The two remaining 'control' groups were not given any opportunities to learn the more efficient techniques socially, yet one individual in the control group innovated the efficient method (Davis et al., 2016). This study provides further evidence for the relative inflexibility chimpanzees seem to demonstrate towards acquiring new techniques when they already possess a satisfactory pre-existing method; only when their existing technique was rendered highly inefficient (in comparison to the novel technique) were the chimpanzees in this study likely to acquire the novel behaviour (see also Manrique et al., 2013;Yamamoto et al., 2013). Studies have also shown, however, that even when their known technique is rendered entirely unrewarding, chimpanzees may still be unable or unwilling to acquire novel solutions to a task (Hrubesch, Preuschoft, & van Schaik, 2009;Bonnie et al., 2012;Harrison & Whiten, 2018). This lack of behavioural flexibility, or tendency towards conservatism, may hinder chimpanzees' innovative abilities, perhaps setting a high threshold of dissatisfaction before chimpanzees will explore novel behaviours (Brosnan & Hopper, 2014; although see also Jacobson & Hopper, 2019). However, the study of Davis et al. (2016) also demonstrates that, when motivated, chimpanzees are able to innovate new solutions, even when faced with relatively complex tasks. Indeed, several studies have unintentionally encouraged innovations in captive chimpanzees, despite their focus being on other aspects of animal behaviour.

b Unexpected chimpanzee innovations during captive experiments
Innovations of alternative techniques were observed by Hopper et al. (2007), when the authors seeded two methods of opening a puzzle box (the 'panpipes' apparatus;  in two different groups. The reward inside the apparatus could either be retrieved by poking a stick through a blocking door, or by lifting a T-bar attached to the door. The 'lift' methodology was originally observed only in the Biological Reviews 95 (2020) 1167-1197 © 2020 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. group in which it was socially seeded, but the 'poke' method was unexpectedly innovated by one individual in the lift group, and became the dominant method in both the poke and lift groups (Hopper et al., 2007). Whilst the 'poke' method was innovated, the authors argued that the 'lift' technique requires social information to be acquired [Hopper et al., 2007; see also Hopper et al., 2015b for a follow-up study in which chimpanzees did not invent the 'lift' technique when presented with the 'panpipes' with the 'poke' technique blocked]. In a similar study, Vale et al. (2017) also described the innovation of an unseeded technique by a naïve chimpanzee. These techniques were predicted to be too complex for the chimpanzees to learn individually without some form of social facilitation (Hopper et al., 2007;Davis et al., 2016;Vale et al., 2017). Unexpected innovations by naïve chimpanzees in controlled conditions suggest that whilst some behaviours may require higher levels of social information to be discovered (although all behaviours should be explicitly tested for the role of social learning in their emergence before it can be confidently stated that social demonstrations are required), the innovative abilities of chimpanzees may have been underestimated. However, it could also be argued that the behaviours under investigation in these studies were relatively 'simple' behaviours, and that more complex behaviours (for example, those involving a sequence of actions, or consideration of the causal relationships between multiple objects) could not be innovated without social learning.
Perhaps one of the most complex chimpanzee behaviours is nut-cracking (Boesch et al., 1994), a composite-tool technique in which two tools are used simultaneously, in different modes, to achieve a single goal (Shumaker, Walkup, & Beck, 2011). The behaviour is complex in that it requires consideration of the relationships between multiple objects (hammer, anvil, and target nut; Biro et al., 2003). Several authors, including Biro et al. (2003; see Section III.1b) have examined how this complex behaviour emerges in naïve chimpanzees. Nut-cracking seems to be a difficult behaviour for chimpanzees to innovate without previous experience of the materials or the target actions, and some studies have found that even when the materials were provided, naïve chimpanzees did not innovate this behavioural form (Sumita, Kitahara-Frisch, & Norikoshi, 1985;Hayashi, Mizuno, & Matsuzawa, 2005). Yet, cases of nut-cracking innovations have also been recorded. For example, in a systematic study on nut-cracking with the semi-free-ranging chimpanzees of Ngamba Island, Uganda, Marshall-Pescini & Whiten (2008) provided naïve chimpanzees with the ecological materials for nut-cracking, but no social information. One individual (Mawa) spontaneously started cracking nuts within the first 10 min of the baseline trials, and after Mawa's success, seven more chimpanzees started cracking nuts (Marshall-Pescini & Whiten, 2008). However, the authors hesitated to classify Mawa's nut-cracking as an innovation, but argued instead that "it is likely that Mawa acquired the behaviour during his stay in a human home" (Marshall-Pescini & Whiten, 2008, p. 190). Whilst it is possible that Mawa may have been exposed to nut-cracking before coming to the sanctuary, this remains only one potential explanation for the innovation of this behaviour, and other explanations, including the possibility that Mawa innovated the behaviour through individual learning, should also be considered (see also Biro et al., 2003). This also highlights the challenge of ascertaining, even in captive populations, the pre-existing behavioural repertoire or knowledge of chimpanzee experimental subjects.

IV. QUANTITATIVE FINDINGS
Whilst the aim of this review was not to conduct a metaanalysis on primate innovation (for this, see Reader & Laland, 2001), and therefore formal statistical analysis would not be appropriate, some broad inferences can be made by examining the cases of innovation found in our literature search. We note that the majority of innovations reported in observational studies of wild chimpanzees were technological or foraging based (55%; 30/54 cases), although many social or play innovations were also reported (44%; 24/54 cases). This appears to be broadly in line with the findings of Reader & Laland (2001) that almost 50% of records of primate innovation in their meta-analysis were foraging related. In captive chimpanzees, we found few observational accounts of innovation (8%; 5/59 cases described in Table 2). Of these, three were technological or foraging innovations, and two were social innovations. All reports of innovation in an experimental context were technological or foraging based (see Table 3), reflecting a strong bias towards the use of artificial foraging tasks in experimental designs.
Fourteen of 33 (42%) reported innovations in the technological or foraging domain in observational studies of both wild and captive chimpanzees were considered (by us, based on information provided in the reports) to be probable or possible innovations, while nine of 26 (35%) reported innovations in the social or play domain were considered likely to be probable or possible innovations. It may be more feasible to categorise a behaviour decisively as an innovation when it is in the foraging or technological domain (for example, consumption of a novel food not previously available) than when it is in the social domain, where it may be more challenging to distinguish convincingly the novel behaviour from the existing social repertoire.
Regarding social transmission of the potential innovations recorded in observational studies, 11 of 33 (33%) technological or foraging behaviours are categorised (by us, based upon information provided in the reports) as possibly being socially transmitted to group members, while 13 of 26 (50%) social or play behaviours are categorised as potentially being socially transmitted. This may partially reflect the fact that some social behaviours require multiple individuals to perform the behaviour (e.g. handclasp grooming; de Waal & Seres, 1997). It is also possible that social behaviours (such as incorporating paraffin cans into display; Kummer & Goodall, 1985) are easier for other individuals to observe than foraging or tool-use behaviours may be, facilitating social learning. Of course, this interpretation should be approached with caution, given that not all of these reports are considered by us to constitute probable innovations (thus making problematic any inference that an 'innovation' spread, if the behaviour was never a true innovation to begin with). The extent of evidential support for behaviours to be categorised as 'possibly' or 'probably' spreading socially also differed in the reports, due to varying amounts of information provided and differences in the research effort towards identifying social learning. The figures given above therefore include studies in which social learning was formally statistically detected (Hobaiter et al., 2014), and those in which the behaviour was observed in individuals beyond the innovator, and suggested to have been socially acquired by the original authors, but no formal investigation was made. It is therefore quite possible that these categorisations overestimate the number of cases in which the behaviour spread socially. While there are of course limitations in assessing whether a behaviour may have spread socially based only on the information provided in a report (in which social learning may not have been of interest to the original researchers), it appears that social behaviours may be more likely to spread socially (or at least, more likely to be interpreted as having spread socially). In considering which innovations are likely to spread, a potential domain difference should therefore be investigated.
This overview of the innovations reported in the literature may highlight some areas in need of further research effort. While all but one of the 20 experimental studies reporting innovation were conducted with captive chimpanzees, we found relatively few instances of observational studies reporting innovation in captive chimpanzees. The lack of reports of innovation in captive chimpanzees is surprising, considering the seemingly important role that the 'captivity effect' plays in increasing the innovative capacity of captive animals in experimental studies (see Section III.2). It is possible that the lack of observational reports is due to a publication bias, in which observations of innovations in captive populations are not deemed 'interesting' enough for publication. It is also possible that captive groups are not subject to the same longterm observational data-collection methods as habituated wild groups, making identification of the first instance of a behaviour challenging. Animals in captivity may experience a less variable, or more predictable, environment than those in the wild (Morgan & Tromborg, 2007;Clark, 2017), and may therefore encounter fewer novel situations (outside of behavioural experiments) that elicit innovation. Enrichment activities, which are increasingly provided in zoos, are not always (or even often) scientifically documented (Clark, 2017), and therefore any novel behaviours that such activities might elicit may not be documented or published. Therefore, we strongly encourage any researchers who observe innovations in captive subjects to publish these findings, in order to fill this gap in the literature.
Regarding the identity and characteristics of innovators, while we were limited by the information provided in the reports gathered, we found that of the 40 cases for which such details were provided, the most frequent age and sex class reported as innovators were adult males (17 of 40; 43% cases). This is in line with the findings of Reader & Laland (2001), whose meta-analysis found that male chimpanzees were more likely to innovate than females, and with Amici et al. (2019), whose meta-analysis found that members of the larger sex were more likely to innovate in novel foraging tasks. In terms of specific personality traits, studies across multiple taxa have shown that there are traits that vary among individuals that may increase an individual's likelihood of innovating, including neophilia (Day et al., 2003, exploratory tendency (Overington et al., 2011, and motivation or persistence (Morand-Ferron et al., 2011). These traits may themselves vary across an individual's lifetime [Biondi, Bó, & Vassallo, 2010; see also Kalan et al., 2019 for evidence that younger wild chimpanzees may be more exploratory], and may therefore contribute, or interact with, effects of age and rank upon the likelihood of innovation (Reader & Laland, 2001). In addition, these interindividual differences may affect the likelihood of innovation differently depending upon the context, with a recent meta-analysis finding that neophilia and exploration had a stronger effect upon the likelihood of innovation in captive than in wild animals (Amici et al., 2019). In corvids, exploration tendency has been shown to vary across contexts (Vernouillet & Kelly, 2020), and so an individual that explores in one situation may not always explore in others. The extent to which individual traits such as neophilia and exploratory tendency remain consistent across time and context requires further investigation, as does the question of whether findings regarding these traits in one taxon might apply to others.
External factors, such as the presence of conspecifics, may also facilitate innovative problem-solving (social facilitation: Dindo, Whiten, & de Waal, 2009 argued that innovation may be more likely in groups or species with higher levels of social tolerance, as individuals are likely to have greater opportunities for uninterrupted object manipulation without being displaced by others, suggesting that it is not only the presence of conspecifics, but also their behaviour, which may influence the likelihood of innovation. Environmental factors are also likely to impact the likelihood of innovation, via two potential routes; opportunity and necessity (Koops, Visalberghi, & van Schaik, 2014). These two hypotheses have primarily been discussed in the context of primate tool use (Fox, Sitompul, & van Schaik, 1999;Sanz & Morgan, 2013;Lee & Moura, 2015) but may also apply to innovation more broadly. Under these (not mutually exclusive) hypotheses, innovation is more likely when resources are scarce, increasing the benefit of discovering new foraging techniques to exploit new resources (the necessity hypothesis; Fox et al., 1999), and/or innovation is more likely when individuals frequently encounter foraging situations that have the potential to be exploited using novel behaviour (the opportunity hypothesis; Fox et al., 1999). Koops et al. (2014) argue that the weight of current evidence more strongly supports the opportunity hypothesis as being a driver of innovation, citing evidence that tool-based foraging such as termitefishing does not increase in prevalence during periods of fruit scarcity, and that chimpanzees are more likely to nut-crack when nuts grow at higher densities in their environment (Koops, McGrew, & Matsuzawa, 2013). A recent experimental study of wild chimpanzees found that individuals who had travelled more and fed less prior to encountering a novel foraging opportunity (an artificial tree cavity containing honey that could be accessed using a stick tool) spent more time exploring the device and were more likely to attempt tool use (Grund et al., 2019). The authors argued that necessity and opportunity work in concert to influence innovation, with necessity in the form of resource shortage prompting individuals to direct more time and attention towards alternative opportunities, especially high-quality, difficult-toaccess food resources (Grund et al., 2019). Grund et al. (2019) also argued that opportunity, rather than necessity, influences the maintenance of resulting novel behaviours, which may explain the findings of Koops et al. (2013); once an adaptive novel foraging technique has been discovered, its use may not be restricted to the period of ecological necessity in which it may have been invented, as long as the opportunity to perform it persists.
With the information available, it was not possible to look for trends in any other individual characteristics that may impact an individual's propensity to innovate (e.g. rank, personality factors, motivation), although such factors are highly likely to influence innovation along with sex and age (Reader & Laland, 2001;Brosnan & Hopper, 2014;Hopper et al., 2014;Amici et al., 2019).

V. CONCLUSIONS
(1) Definitions of innovation vary in several aspects, the most prominent being whether innovation is regarded as a process or a product (and some provide different definitions for the product and the process; e.g. Reader & Laland, 2003), whether innovations are considered at the individual or population level, and whether the innovation has to be socially transmitted to other members of the community for it to count as such. Given the evidence suggesting that it may be unusual for novel behaviours to become populationwide patterns in chimpanzees , requiring that the behaviour be observed in other members of the innovator's group may set overly stringent parameters for innovation, particularly when formulating a definition to encompass multiple species, including those in which social learning may not be the most common acquisition strategy for novel behaviours. Such a definition may lead to innovation failing to be recognised in species that are biased towards learning from specific individuals (e.g. higher-ranking individuals; Kendal et al., 2005) or learn socially only in specific contexts (e.g. "copy if better" or "copy if unsure" social-learning biases; Kendal et al., 2005).
Requiring the behaviour to be observed in multiple individuals may lead to cases of innovation being dismissed as idiosyncratic behaviour. Recent research making use of novel statistical methods demonstrates that the social transmission of behaviours can be documented in wild chimpanzee groups (Hobaiter et al., 2014), and so while definitions incorporating the development of the behaviour in other group members should not be considered operationally impossible, the possibility remains, as suggested by Nishida et al. (2009), that the social acquisition of innovations may be relatively unusual, and thus difficult to observe due to its rarity.
(2) We agree with multiple authors (Ramsey et al., 2007;Rendell et al., 2007;Whiten & van Schaik, 2007;Carr et al., 2016;Arbilly & Laland, 2017) who suggest delineating different 'levels' of innovation. This may facilitate comparative work that can draw upon the same basic definition of innovation, whilst recognising qualitative differences between instances of innovation or the innovative capacities of different species. However, agreement must still be reached upon what these categorisations are based upon, and it seems that here, we return to the same tension in the field regarding what aspect of innovation is of interest: the individual cognition involved, or the population-level cultural impact of innovation. While a unified definition of innovation that can be applied across disciplines seems desirable, it may be that this distinction (individual-versus population-level interest) dictates what definition is suitable for any given study.
(3) The differences in terminology used to describe innovations have made comparison of innovations among and even within species a difficult task. As the study of innovation is inherently interdisciplinary, it might be impossible to propose one definition that encapsulates innovation across all these fields. Therefore, perhaps the most practical solution to this issue is not to find one definition that fits all fields and all species (which may indeed be a Herculean task), but instead researchers could continue to use the definition they prefer, so long as they provide enough information on the novel behaviour that it can be interpreted using most (if not all) definitions. We propose that when describing an innovation, researchers should also include information on the type of innovation, how many individuals demonstrated the innovation (before and after it was first observed), the likelihood of the behaviour being socially learnt by other individuals, or whether individual learning can explain any increases in frequency (and whether socially mediated serial reinnovations are likely; Bandini & Tennie, 2017& Tennie, , 2019, alongside other relevant data. By including all this information, the novel behaviour can then be assessed using other definitions as well. This will allow comparisons of behaviour among field sites, and with captive subjects, to be made. (4) Innovations that occur unexpectedly in experimental studies of captive subjects provide valuable insight into the cognition of chimpanzees, the extent of their innovative and problem-solving abilities, and on the difficulty of the task being presented. Such innovations should be highlighted by researchers when reporting the study. In addition, studies on the role of social learning in the acquisition of novel behaviours in captive chimpanzees may provide evidence of multiple instances of innovation from within the same populations (for example, see Davis et al., 2016;Vale et al., 2017;Watson et al., 2017). These studies provide valuable evidence on the characteristics of innovative individuals, as an individual's performance may be tracked across multiple studies. Indeed, it has proved possible to conduct a meta-analysis of individual differences in social learning across multiple studies conducted at the same captive care facility, establishing quite stable variations in whether individuals tend towards learning from others, or instead rely on their own exploration of the opportunities presented (Watson et al., 2018). Social-learning studies are also carried out under carefully controlled conditions, with individuals in asocial control groups often having the opportunity to attempt an artificial foraging task in isolation. This allows for a high degree of certainty when identifying a behaviour as innovative, as it can allow complete exclusion of the possibility of social information being used to solve a task. As in studies of wild chimpanzees (Hobaiter et al., 2014), the use of novel statistical techniques such as network-based diffusion analysis (NBDA) in conjunction with these studies has the potential to provide evidence of innovated task solutions being socially transmitted [Watson et al., 2017 demonstrate social learning of seeded behaviours in captive groups using NBDA]. (5) The studies included in Tables 2 and 3 highlight the different contexts in which innovations appear. This variance has led to a healthy debate on the role of different learning mechanisms in the acquisition of innovations by naïve members of the innovator's group. It is now clear that chimpanzees possess a wide range of low-fidelity social-learning mechanisms that facilitate the acquisition of some of their behavioural forms (e.g. see Whiten et al., 2004). However, recent work on the learning mechanisms behind chimpanzee (and other animal) behaviour has also demonstrated that at least some behavioural forms can be individually learnt by naïve members of the species [e.g. Bandini & Tennie, 2017 and see also the studies included in Tables 2 and 3]. Thus, whilst logically the original innovator must express the new behaviour via individual learning, it is still unclear in many cases how the other members of the group then acquire the behaviour: whether high-fidelity social learning is required, or whether the behavioural form can be innovated by all individuals given the appropriate conditions (as argued by the ZLS hypothesis; . In the latter case, low-fidelity forms of social learning such as local enhancement may facilitate the acquisition of the behaviour by placing individuals in favourable conditions for the behavioural form to emerge, or individuals may use social information as a shortcut to avoid the cost of individual exploration and innovation. (6) This review has highlighted the extent of chimpanzees' innovative abilities across domains and environmental settings. However, it has also demonstrated that chimpanzees can be remarkably conservative in sticking to their pre-existing techniques, often resulting in innovations not 'catching-on' in other members of a community ). This creates a somewhat puzzling picture of innovation in chimpanzees, making it hard to estimate confidently which types of novel behaviours will spread throughout a group, and which ones will be lost with their innovator (and therefore be categorised as a likely idiosyncrasy, rather than a true innovation, following some of the definitions described previously). However, it seems that certain innovations are more likely to persist within a community: namely ones that appear in response to novel ecological stimuli (e.g. moss-sponging, which emerged as a response to a new watering hole; Hobaiter et al., 2014) and ones that are rewarding enough relative to pre-existing behaviours to warrant the cognitive load of (either individually or socially) learning a new behaviour. On the other hand, behaviours that do not represent a considerable improvement in efficiency over pre-existing methods (such as using a stick instead of a leaf sponge to retrieve honey; see Gruber et al., 2009Gruber et al., , 2011 are less likely to be acquired by a chimpanzee group. Chimpanzees' reluctance to acquire novel behaviours in certain contexts may have an adverse effect on their survival as an endangered species, especially now that deforestation in chimpanzee habitats is peaking (Nelleman & Newton, 2002). Chimpanzees have also demonstrated a remarkable ability to respond to novel ecological situations (e.g. Hobaiter et al., 2014;Gruber et al., 2019), which may allow them to adapt to their changing environment. However, while some forms of human impact may provide novel opportunities to chimpanzees (for example, the availability of human-cultivated foods; Takahata et al., 1986), these opportunities may at the same time bring chimpanzees into contact and conflict with human communities (McLennan & Hill, 2012). In addition, human activity may impact the environment in a manner beyond that which innovation of novel behaviours can overcome; for example, it may reduce the carrying capacity of the forest, thus reducing the group size of chimpanzee communities, limiting opportunities for social facilitation of adaptive behaviours  and reducing the behavioural repertoire (Lind & Lindenfors, 2010;Kühl et al., 2019). Innovation in chimpanzees (7) It is very likely that individual characteristics and personality influence the probability of a chimpanzee becoming an innovator. The reports on innovations compiled for this review did not provide enough information on the innovators themselves to allow for a detailed analysis of what these characteristics are. However, it is likely that no single characteristic makes an innovator, but instead a combination of factors such as personality, developmental stage, and environmental context (see also Brosnan & Hopper, 2014). Furthermore, sex and rank seem to affect whether an individual will innovate a new behaviour, as lowranking male chimpanzees are observed most often making these innovations (Reader & Laland, 2001). A more recent meta-analysis incorporating multiple taxa (not only primates), and examining performance in experimental novel foraging tasks in both wild and captive settings, found that innovation is more common in older animals, and in those of the larger sex (Amici et al., 2019), a finding in agreement with that of Reader & Laland (2001) of a sex difference in chimpanzee innovation. Personality factors such as neophilia and exploration tendency were also found to influence propensity to innovate, though the extent of their influence differed in wild and captive subjects (Amici et al., 2019). Further research is required to explore how the findings of multi-species metaanalyses such as Amici et al. (2019) apply to chimpanzee innovation, but they indicate that the individual factors driving innovation may be shared across taxa. Thus, studying which individuals innovate in chimpanzee populations may provide broader insights into the role of individual traits in the emergence of innovation across species. (8) Whilst this review focused only on chimpanzees, implications from these findings can be drawn on the innovative abilities of other primates, especially closely related species such as humans and other great apes. Indeed, chimpanzees are often used as models for early hominins (e.g. Toth & Schick, 2009), and some insight into the evolutionary path of our own innovative abilities can be inferred from the findings of this review. For example, it is very likely that the last common ancestor (LCA) of humans and chimpanzees also used tools; perhaps even in similar ways to how extant chimpanzees use them today (e.g. marrow picking, which is observed in extant chimpanzees and most likely was also practiced by early hominins; Sanz & Morgan, 2007). Furthermore, it is also likely that the repertoire of the LCA consisted of behavioural forms that were so simple that they could be innovated by most individuals, therefore not requiring any particularly complex cognitive abilities (as seems to be the case for many chimpanzee innovations), and thus perhaps our propensity to learn complex behaviours via highfidelity social-learning mechanisms evolved later on in the evolutionary timeframe. Indeed, at some point, our path diverged significantly from chimpanzees' (and other great apes), resulting in the extremely complex, culture-dependent niche that modern humans have created for themselves. Modern human culture has created an environment that promotes and fosters countless innovations to be made daily around the world (these innovations are so frequent, and so valued by society, that we have even developed copyright systems to ensure that innovations are attributed to the rightful innovator). This environment is very different to what we observe in extant chimpanzees, which innovate relatively rarely, only in certain contexts, and often other group members do not value these innovations enough to develop them themselves. Therefore, the stimuli and conditions that created this prolific innovation environment for our species, but seemingly not for other great apes, remain to be examined. Hopper (2016) suggests that one important difference between the two species may be that chimpanzees innovate primarily for themselves, whilst humans tend to innovate behaviours that benefit a group as a whole (see also Hopper & Torrance, 2019). Additionally, (modern) humans greatly benefit from what Muthukrishna & Henrich (2016) call the 'collective brain', which enhances our ability to innovate by building upon the ideas and work of previous innovators. This is facilitated by our ability as modern humans to transmit and acquire information reliably through highfidelity social-learning mechanisms, an ability that may be less extensive, or deployed less frequently, in (non-enculturated) chimpanzees Tennie et al., 2012). Future directions for research should involve a further examination of the similarities and differences between the motivations, goals and results of chimpanzee and other primate innovations. (9) Several issues and open questions remain in the field.
Firstly, to continue studying these abilities in primates, and indeed other animals, more information on the innovations themselves needs to be provided, to allow researchers in other fields to assess the relevance of the behaviour for their work. Furthermore, more data are required from both wild and captive chimpanzees to provide a full understanding of the cognitive mechanisms driving innovation in our closest living relatives. Future studies into the innovative behaviour of chimpanzees can provide further insight into the role innovation plays in modern human culture, and how it may have evolved in the hominin lineage.
to the editor and two anonymous reviewers for their valuable and constructive comments. E.B. is supported by the Institutional Strategy of the University of Tübingen (Deutsche Forschungsgemeinschaft, ZUK 63). R.A.H. is supported by a Swiss National Science Foundation grant (PP00P3_170624) awarded to Erica van de Waal.