This paper suggests (i) that while work on animal innovation has made good progress in understanding some of the proximate mechanisms and selective regimes through which innovation emerges, it has somewhat neglected the role of the social environment of innovation; a neglect manifest in the fact that innovation counts are almost always counts of resource-acquisition innovations; the invention of social tools is rarely considered. The same is true of many experimental projects, as these typically impose food acquisition tasks on their experimental subjects. (ii) That neglect is important, because innovations often pose collective action problems; the hominin species were technically innovative because they were also socially adaptable. (iii) In part for this reason, there remains a disconnect between research on hominin innovation and research on animal innovation. (iv) Finally, the paper suggests that there is something of a disconnect between the theoretical work on innovation in hominin evolution (based on theories of cultural evolution) and the experimental tradition on human innovation. That disconnect is largely due to the theoretical work retreating from strong claims about the proximate mechanisms of human cultural accumulation.
1. Hominin exceptionalism and its implications
Prima facie, the capacity to innovate seems highly adaptive, especially in heterogeneous and/or fluctuating environments, as a means of responding successfully to new risks and opportunities, but also in more stable or uniform environments, as a means of easing the competitive squeeze on resources. The widespread existence of complex, canalized responses to environmental challenges documented in the niche construction and behavioural ecology literature—nest-building, burrow-making, migration in response to seasonal change—reinforces this impression . For these behaviours presumably had their origins as innovations, prior to their coming under genetic control. Moreover, while as we shall see, some kinds of innovations may well be cognitively demanding, a basic capacity to innovate is almost certainly phylogenetically widespread, as it depends only on reinforcement learning: on an agent's capacity to register the environmental state it finds itself in; register the response it has produced in that state; and to evaluate the outcome of that response . If innovating would often be adaptive, and if innovation often depends only on capacities that are ancient and widespread, then what explains the scope and limits of innovation in the animal lineages, and the apparently significant differences in innovation between different lineages? Our lineage is an especially vivid illustration of these differences; our lineage is an extreme outlier. But though humans are an outlier, that may not be a deep feature of our evolutionary history. One notorious and puzzling fact about hominin evolutionary history is that between about 3.4 Ma and 250 kya, the pace of technological change seems to have been very slow . It increased quite sharply between about 250 kya and 50 kya, and has increased further since then (very obviously in recent times) [4,5]. The accelerating pace of hominin cultural evolution seems not to be driven by a matching change in the biology of humans. Indeed, in general, the technical history of hominin innovation, and hominin expansion into new habitats, does not map in any clean way onto speciation events in our lineage, or onto obvious changes in hominin morphology . That is even true of brain size and change in brain size. Though Gamble and co-workers organize their recent monograph on hominin evolution with separate chapters on smaller and larger brained hominins, even among large-brained hominins there are spectacular differences in material culture. There have been very large-brained hominins for over half a million years, but the material culture of (for example) Homo heidelbergensis was very limited compared to more recent hominins [4,7].
These facts about innovation in hominin history are relevant, in my view, to the approach to innovation exemplified by many of the papers in this issue and of the papers they review. The strange history of hominin evolution suggests that the array of innovations we see in a population at a time, and in a lineage through time, is not a simple reflection of the adaptive capacities of the individuals in the target population. This is the central message of Muthukrishna & Henrich . They offer a radical, distributed cognitive theory of innovation as the product of ‘collective brains'. Individual capacities are relevant to innovation and innovation rates, but those capacities depend intimately, and in ways that vary across cultures, on the ways those agents are embedded and linked in their social environment. In developing this perspective, Muthukrishna and Henrich distinguish three sources of innovation and argue that the size and organization of social and informational networks is central to all three sources of information.
As they see it, the sources of information are:
Fortunate accident: The history of science and technology details many inventions which began with fortunate accidents: rubber, antibiotics and X-ray technology are three poster examples, but while these are detailed from the recent historical record, it is very likely this is an ancient source of innovation. While these innovations begin with good fortune, recognizing the significance of these accidents is essential and far from trivial, and typically depends on culturally acquired intellectual resources.
Incremental improvement: Early versions of technology are tinkered with and improved. Tinkering is a skilled and thoughtful activity, and individual cognitive skills are important. But in incrementally improving existing techniques and material culture, horizontal and oblique transmission give agents access to many sources of relevant expertise. Moreover, the attractions even of early versions of a new technology normally guarantees that a number of agents will be working on optimizing projects in parallel, obviously making population-level success more likely.
Cultural recombination: The concept of natural selection is the classic example of innovation through combining information from different local nodes of the cultural resources of the community, for both Darwin and Wallace were familiar not just with the biological literature of their time, but also that of early political economy. They drew on this most famously via Malthus and population pressure, but Adam Smith (in The Wealth of Nations) and Bernard Mandeville (in The Fable of The Bees: or, Private Vices, Public Benefits) pioneered early versions of hidden-hand explanations, showing how organization could emerge without an organizer, as a side-effect of individual decision-making. Of the many Darwin biographies, Desmond & Moore  most emphasizes this social and cultural context of Darwin's work. The lesson from this example and others is that connection matters: in cultures with horizontal and oblique transmission, and with fairly open network organization, the cultural group as a whole will typically include a number of individuals with broad connections across local sources, giving each an opportunity to combine ideas and techniques in the right way.
In all of these routes to innovation, innovation depends on the social flow of information, not on unique and unrepeatable individual insight. This analysis of the social dimension of human innovation depends, of course, on the prior evolution of cumulative cultural evolution, and the establishment of a synthesis between extensive cognitive capital, high fidelity intergenerational transmission and incremental addition to that cognitive capital. Central to this perspective is the intimate connection between innovation and social learning: innovation improves upon the cognitive capital of the community, and hence depends upon reliable access to that capital. The theoretical analysis offered by Muthukrishna and Henrich is plausible, and it is supported by connections to historical case studies. But there is something of a disconnect between the body of theoretical work on hominin innovation and cultural accumulation exemplified by their paper, and experimental work on cultural transmission represented in this theme issue [10–12].
The theoretical work is focused on population-level processes: selection, demographic size and connectedness ; the economics of the emergence of a division of labour [14,15]; on the power of ‘attractor’ models in explaining cultural stability . These theoretical analyses are typically quite well tied to the archaeological record [17,18]. But the experimental work presented and reviewed in this collection is not tightly coupled to these broader theoretical and historical analyses. The experimental papers are focused on the emergence of social learning in development, its accuracy under differing conditions, and the circumstances under which agents choose to learn socially or individually. In this experimental work, social learning is often contrasted with innovation; innovation requires individual learning, and so results only when agents choose or are forced to learn asocially. The results are often interesting in themselves, but they typically do not bear directly on the theoretical frameworks, in part because these frameworks have become less explicit about both the proximate mechanisms of social learning and of innovating on the foundations that social learning provides. Thus, Flynn et al.  show that there is evidence that children begin to assess their own learning capacities reliably remarkably early: by about five, children who prefer to learn asocially are good at learning asocially; they solve the test problems as quickly by themselves as they do from social input. Likewise, Caldwell and co-workers report a series of experiments on skill transmission in microsocieties, showing that the balance between individual and social learning is modulated by individual assessments of their own reliability. As the experimenters increase uncertainty about how successful a design might be (for example, by introducing an unknown test on the structural stability of a tower made of pasta, hence making it difficult for agents to assess the success of their own designs), agents shift to a greater reliance on social learning . The general message seems to be that learning choices are nuanced, strategic, and that these capacities to assess self, task and environment begin to emerge quite early. But while these results are certainly valuable, they connect to the theoretical frameworks only in a broad and general way, for these theoretical frameworks make room for social and individual learning, and hybrids between the two. The main reason for this loose connection between the theoretical and experimental work is that theoretical analyses of hominin innovation and cultural accumulation now rest much less on specific claims about the mechanisms of social learning. At one stage, these analyses linked hominin cumulative culture to the claim that imitation learning was both uniquely human and uniquely powerful , but the link between imitation and cumulative culture is now less emphasized. Instead, one of the main messages of the theoretical and integrative work is that a range of quite different cognitive mechanisms can explain, in the right contexts, both high fidelity cultural transmission, and the creation of variation [20–22]. Theory is now somewhat decoupled from proximate mechanism.
The package of extensive cognitive capital, high fidelity intergenerational transmission and incremental addition to that cognitive capital, seems to be a unique feature of our lineage. So we face a critical question: does an account of hominin innovation—an obvious and extreme outlier—tell us something important about innovation in non-human animals? This paper will suggest that the hominin case does illuminate the animal case, by making salient constraints on innovation, especially constraints that do not derive from intrinsic features of agents' cognitive capacities.
2. Environment and innovation
Hominin innovation history suggests that innovation rates depend on the physical and social environment, not just individual capacity. Perhaps most obviously, the innovation rate in a population is influenced by the ways individuals are embedded in their environment. The heterogeneity of the environment at a time, and its variability over time influences the rate at which animals encounter novel phenomena, with their opportunities and threats. But the environment also structures the costs and benefits of innovation. Experiment is often risky; a point made forcefully by van Schaik et al. . Novel items sometimes carry hidden dangers; a novel species of frog, for example, can turn out to be poisonous, as many Australian animals have found out to their cost. But less obviously, van Schaik and his colleagues point out that the solution of technical challenges (in particular) often requires extended periods of focused attention, and for animals at risk of predation, prolonged focused attention can be very dangerous indeed . Resisting distraction is, equally, resisting peripheral signals of danger. So as they show, the predation threat in natural environments is one factor constraining innovation (see also Morand-Ferron & Quinn  for a similar point, but using great tits as the model species). Captive orang-utan populations are safer, less neophobic and more innovative. Van Schaik et al. suggest that the neophobia of wild populations of orang-utans forecloses many potential pathways of innovation. For example, he and his colleagues doubt whether necessity is the mother of orang-utan innovation, for they respond to stress by limiting movement and falling back on emergency, low preference resources like bark. Likewise, they doubt whether curiosity-driven exploration has been important in orang-utans or other non-human great apes, as they are too neophobic. Instead, they suggest that most innovation in their lives begins with a resource whose value they already know and so they attempt to develop new ways of reaching that resource. Their example is working out how to use a stick to break into a beehive.
Likewise, Lefebvre et al. recognize the role of predation threat in their choice of experimental models, choosing broad-niched island species, on the grounds that reduced predator threat makes these species more likely to innovate . However, while a number of the papers in this volume discuss rates of innovation within different clades, and broad correlations between rates of innovation and morphological features (for example potential correlations between rates of technical innovation and brain size, see Navarrete et al. ), and there is some discussion of whether innovation rates depend on ecological factors (in particular, on harsh environments) there is much less discussion of the relations between rates of innovation and broad geographical variables; for example, comparisons of island versus mainland species, or the relation between innovation rates and narrow versus broad geographical distribution. Again, the hominin comparison is quite instructive: there is a reasonably well-supported latitudinal gradient in hominin technological sophistication, with toolkits becoming larger in populations distant from the equator . Defeyter et al.  are suggestive in this respect, hinting at a role for social context on children's willingness to innovate.
It is equally true that the social environment impacts on the rate of innovation. For one thing, there is a link between innovation rate and the extent and nature of social learning . There is an evolutionary connection: individual and social learning depend on the same, or on overlapping, cognitive mechanisms, so selection for one probably enhances the other . There is an ontogenetic connection too, even setting aside the special case of cumulative culture. For as Tebbich et al. and others note, innovation is more likely if an agent has a large baseline repertoire of skills and capacities [31,32], and social learning is an important means through which agents build their repertoire. Finally, an innovation could simply be a social learning misfire that happens to have beneficial consequence, though, plausibly, in their review paper Caldwell et al.  regard this as an unlikely source of innovation.
Secondly, innovation often requires the solution of collective action problems, not just technical problems, and in such cases innovation is not just a move by an agent in a game against nature. It is also a move in a game within a group. If a group-living animal innovates, and thereby generates extra resources, that will have an impact on its relations within its social world. Most obviously, if the animal is not high ranked, and if the resources are valuable and cannot be instantly consumed, those extra resources are likely to make it the target of aggression. Agents will not be rewarded for investing time, effort or risk in crafting resources (as in making tools) or collecting resources (for example, processing manioc to remove its poisons) if those resources are liable to be seized by more dominant individuals or coalitions of individuals. The more valuable the resources crafted or collected, and the more investment they represent, the greater the temptation for the powerful, and the greater the disincentive to innovate, if the risk of seizure is significant. (It is perhaps surprising that there are not more reports of nest and burrow piracy, though it is also true that many colonially nesting birds—birds repeatedly exposed to the opportunity to pirate—build very minimal nests). Humans have long operated a delayed-return economy, depending on resource-gathering practices which involve long latencies between action and return . Delayed-return economies help explain hominin innovation rates, as many technical innovations pay only through long-time horizons. By and large, animals operate immediate return economies, and those that do not need to be wary of pilfering [34,35]. There is not much evidence of ontogenetically flexible, contextually sensitive, long-time horizon behaviour in animal life, and that is typically supposed to reflect a cognitive constraint: they do not have much in the way of representational resources for future planning [36,37]. These considerations about the interaction between collective action and delayed return suggest that the capacity to resist the pull of the immediate and look to the future would be advantageous only in cooperative, or at least tolerant, social worlds (and, perhaps, for sufficiently solitary individuals). With respect to future-oriented behaviour, many agents will find themselves in risk-dominant versus pay-off-dominant situations, and the pathways to innovation will be barred, even if they have the impulse control, the future-orientation and the technical and motor capacities to innovate.
One of the strengths of Lane's paper is his recognition of the link between innovation and collective action . His poster example of innovation is the invention of printing, but the technical advances were rewarded only given the establishment of reliable sources of raw material, reliable ways of distributing, marketing and returning the profits from book printing. None of these were in place prior to printing, so existing networks, input and output distribution and money flow systems had to be re-jigged to turn the technical innovation of book printing into book publishing, as a new form of material production. This is typical of important innovations, and the paper gives a nuanced discussion of the interplay between organizational adaptation and incremental technical improvements, as cleaner types and layout made the books easier to read, and the reduction of their weight made them more portable.
3. Missing links
If the physical and social environment in which agents find themselves is as important as I have just suggested, two features of much work on animal innovation are somewhat puzzling. First, much of the work on animal innovation is organized through a bottom-up framework: that is, the strategy of explaining innovation differences in populations and lineages by differences in the characteristics of the individual agents from which populations are composed. Adaptable individuals form high-innovation lineages. That strategy is of limited validity if the social organization of those agents plays an important role in rewarding or suppressing innovation; individual adaptability is not the only variable that matters. That general feature of the literature is reflected in this theme issue. Lefebvre and co-workers exemplify this framework as they develop and test the West-Eberhard's ‘flexible stem hypothesis' . On islands where we find the classic examples of adaptive radiation, we also find migrant species which have not radiated. West-Eberhard suggested that those species which form the stems of adaptive radiations do so because of the high degree of adaptive plasticity of the individuals forming the founder populations. Tebbich et al. suggested that Darwin's finches might exemplify this pattern , and Lefebvre and colleagues reasoned that if that were right, the clade which contains Darwin's finches should show high rates of innovation, and members of the clade on other islands should likewise show high levels of individual adaptability. They confirmed both predictions . A macroevolutionary pattern—adaptive radiation and high rates of population-level innovation—is explained by the characteristics of individual agents. We see the same explanatory strategy in those papers that distinguish food-type innovations (merely adding a new kind of food to the menu) from technical innovations: the development of a new technique for harvesting resources. The idea is that technical innovations (especially complex ones like tool use) are more cognitively demanding and that conjecture is then tested by determining whether the rate of technical innovation is correlated with evidence of neural and cognitive complexity . Likewise, Quinn et al.  explore the relationship between personality types, measures of developmental environment quality and innovation, and there were some effects, though no very robust pattern.
The second feature of the animal innovation work is an almost exclusive preoccupation with games against nature, though that was less true of earlier work , and there has also been some attention to the effects of social context on strategy choice in those games [24,43]. In this theme issue, it is also true that the experimental studies on human innovation focus on games against nature, but that is not in general true of experimental work on human innovation, nor of theoretical reflection on it . There are many studies which induce agents to invent on the fly coordination and communication tools, and which track how those tools evolve over the course of the experiment (e.g. [45,46]). In contrast, the work on animal innovation seems largely focused on resource-gathering innovation. Very little work on animal innovation focuses on innovations in social life: on the establishment of communication and coordination tools, or social tools for leveraging an agent's place in a social hierarchy. This is very striking, for animals have developed some social tools, and these are studied and analysed in a different theoretical context (e.g. [47,48] and more recently [49,50]). There are important issues to press here: is the supply of social innovation constrained by the same cognitive, environmental and social factors that constrain the supply of technical innovation? There is, for example, a debate on the extent to which great ape communication traditions can be explained by ritualization; an associative mechanism that shapes an initially functional act into a signal; turning a grab into a begging gesture. There are great ape gestural practices which seem not to have plausible origins in associative learning . Likewise, the motivational contexts and constraints on social innovation seem different from those that constrain ecological innovation. For example, it is hard to see how neophobia could constrain social innovations in the way that it constrains feeding and technical innovations. There is, though, some reason to think that the cognitive constraints on technical and social innovation overlap: Reader et al.  found that rates of tactical deception correlate with rates of innovation.
4. Individual adaptability
While social and environmental factors are relevant to the generation of innovations, so too is individual capacity and differences in individual capacity, and these differences are probably easier to measure and manipulate than social and environmental factors. So despite the complaints of the previous section, individual adaptability is a natural starting point in identifying the source of new behaviours, and especially in the experimental study of new behaviours. Perhaps most obviously, the capacity to innovate depends on the potential motor repertoire. The larger that potential repertoire, and the more that repertoire is under fine-grained, top-down control, the more likely it is that an agent will have the behavioural potential to produce a new and adaptive act/act sequence. Motor repertoire is in part cognitive, depending on perceptual and control capacities, but morphology has a huge imprint on repertoire: compare elephants and horses. Elephants have far greater capacities to manipulate their environment. It is notable that ‘innovation’ is not even in the index of a fairly standard collection on elephants’ behaviour . Obviously, repertoire and adaptive plasticity coevolve with other characteristics. The sheer size and power of elephants presumably lowers their vulnerability to predation, and hence the cost of experiment, while their long lives and large range sizes makes it likely that they encounter changed conditions, increasing the pay-offs for adaptive plasticity. (The role of extended life histories  makes it somewhat surprising that low innovation rates are reported for seabirds, as many of them are very long-lived and wide-ranging). Cetaceans are an important contrast case, as some of them are highly encephalized, with complex social lives, and yet their capacities to manipulate their environment are relatively limited. It would be very difficult to identify and study the rates of (say) orca innovation, but even information on a single such lineage would test the relationships between morphology, motor repertoire, individual adaptability and innovation rates.
There is relatively little on morphological constraints on, or preadaptations for, individual adaptability in this issue (with the partial exception of Tebbich et al. ). Instead, most of the analysis is on the influence of cognitive and motivational factors on individual adaptability. For example, Sol et al.  develop and defend a version of their ‘cognitive buffer hypothesis'. Population-level innovations are occasional manifestations of individual adaptive plasticity, and that plasticity evolves by reducing adult mortality in the face of unpredictable environmental events. Individual adaptive plasticity varies between lineages (and hence so does innovation rates) because the costs of plasticity are worth paying for longer lived ecological generalists that spread their reproductive effort over many episodes. Plasticity is relatively cheap for generalists, as they need good learning skills to learn the recognition, assessment and manipulation skills that support their broader foraging capacities. So innovation propensity, a future-directed life-history strategy, and increased brain size tend to coevolve in generalist species that are frequently exposed to unpredictable conditions.
One ongoing focus of discussion is whether there is an important difference between simple feeding innovations—adding a new food to the resource base—and technical innovations, adding a new technique to the foraging repertoire. Tebbich et al. argue that innovations in technique are more likely to generalize, leading to innovation cascades, for they expand the agent's motor repertoire, making further innovation possible . Learning to open molluscs by dropping them from a height, or under the wheels of a moving vehicle, is a behavioural recipe that can be applied to other protected resources like nuts. However, in some cases, simple feeding innovations can lead to further innovation too. Avital & Jablonka  point out that an innovation like potato washing by Japanese macaques automatically places those monkeys far longer in an edge environment, increasing their probability of encountering new opportunities, and if and as they exploit any of those new opportunities, that process ramifies, by still further entrenching the new pattern of time use, and hence increasing their exposure to new items. The most obvious and important form of these innovation-through-exposure cascades is via the accidental colonization of new habitats, with their suite of new opportunities and challenges. Darwin's finches on the Galapagos are a classic example of a lineage that has found itself in a new habitat, and has diversified both morphologically and behaviourally (with some of those innovations now frozen). Habitat archipelagos with sharp ecological gradients will offer many adaptive pathways of this kind, where resource profiles change abruptly at edges.
So technical innovations may be especially significant because they open potential pathways to further innovation . But do they also require distinctive cognitive capacities? Some technical innovations could be learned associatively. A raven might accidently knock a nut under the wheel of a moving car and discover that it has been opened. A gull might drop an item of prey while taking it to a safer place, and discover that the fall has broken it open. However, some innovations are beyond the reach of associative learning, either because the rewards are too distant in time, or because the innovative act (while it will be immediately rewarded) is very low-probability unless the agent recognizes and responds to causal structure in its environment. Some of the experimental work on New Caledonian crows suggest that they have a capacity to innovate that depends on some recognition of causal structure in their environment; for example, their ability to exploit the properties of water displacement in unobvious ways to bring rewards within reach [55,56]. Tool-using innovations may not require the representation of causal structure in the environment, but even when they do not, tool-using innovations are difficult to explain through associative mechanisms alone because, typically, a long process of experimentation without reward precedes reward. Tebbich et al. suggested that an agent's motivational profile is central to this form of adaptive plasticity: those animals that find play and experiment intrinsically rewarding are much more likely to find these delayed-return innovations.
What is the upshot of this theme issue and of the research it reviews, represents and extends? I shall venture three suggestions. (i) The research community is making reasonable progress in understanding the evolution, distribution and cognitive foundation of individual adaptive plasticity in non-human animals, as that plasticity is manifest in those agents' encounters with their physical and biological environment. A couple of the contributors to the issue [23,32] develop useful taxonomies of innovation types and pathways to innovation, and of the motivational and cognitive demands imposed by those pathways. Tebbich et al. also pay a good deal of attention to the role of the target of exploration, pointing out that while some are tolerant of a range of approaches, other opportunities can be exploited only via precise and error-intolerant sequences. Likewise, the distinction between technical and feeding innovations is rough and ready, but it seems to mark an important cognitive and evolutionary difference . In this work, there is recognition of the importance not just of those agents' cognitive mechanisms, but of their motivation and personality types as well. Individual difference and its importance is increasingly well recognized. In particular, there is significant research focus on whether plasticity is manifest most in animals under pressure, or whether experiment is an expression of bold, exploratory and fit agents . Innovation in populations and lineages is an occasional manifestation of individual adaptive plasticity, but while innovation depends on individual adaptive plasticity, innovation rates do not depend only on that plasticity, and as noted above the role of physical and social environment is less salient in this work, though the work of Quinn et al.  is something of an exception. (ii) In particular, this research community has focused much less on innovations in social life, and on the evolutionary regimes and cognitive mechanisms which make innovative behaviour in social life possible. I have argued that we need to understand the flow of innovations in social life, and constraints on that flow, because without social innovations, the pathway to an array of potentially profitable technical innovations is blocked. Hominins are a cooperation outlier as well as a technical innovation outlier, and those facts are intimately connected. Why are there so few pathways to technically innovative, socially tolerant lifeways? We need to understand the constraints on social innovation to answer this question. (iii) There remains something of a disconnect between the research effort on hominin innovation, and research on non-human innovation. In the work on hominin innovation, the role of social tools is obviously central. That is true of the microworlds studies of the interplay between individual learning and social transmission, and of the developmental studies showing the growing capacities to make strategic choices between individual and social learning. It is most obviously true of the rich body of work, exemplified here by Muthukrishna and Henrich, that takes cumulative culture to be the core adaptation of our lineage. But in part because the animal innovation literature concentrates on resource-acquisition innovations, we do not have a good picture of why our lineage is such an outlier. The hominin innovation avalanche is evolutionarily recent, but critical changes were much deeper, as our ancestors combined stone technologies with cooperative foraging [57,58]. What were the barriers to socio-technical innovation in early hominin evolution, when the first stone technologies were developed, and how were those barriers overcome?
The author declares no competing interests.
The author gratefully acknowledges the Australian Research Council's generous support for his work on the evolution of social behaviour.
One contribution of 15 to a theme issue ‘Innovation in animals and humans: understanding the origins and development of novel and creative behaviour’.
- Accepted January 6, 2016.
- © 2016 The Author(s)
Published by the Royal Society. All rights reserved.