Exploring the role of individual learning in animal tool-use

Chimpanzees

Kummer & Goodall (1985) describe one of the first instances of spontaneous tool-use in wild chimpanzees from Gombe, Tanzania. The new behaviour involved the use of sticks as levers to open banana boxes introduced by the researchers (Kummer & Goodall, 1985). Although the previous tool-use knowledge of the Gombe chimpanzees is not reported, this observation was included in Table 2 because the banana boxes were artificial objects introduced by the researchers, making it unlikely that the chimpanzees had already encountered this specific problem in the wild. Indeed, many of the first observations of tool-use in animals were the product of individuals interacting with novel, artificial or introduced materials and objects (Köhler, 1925; Lefebvre, 1995). Subsequently, many early researchers interpreted new observations of animal tool-use from an anthropocentric point of view. Indeed, despite writing that the chimpanzees showed the behaviour ‘independently’ (i.e. without requiring social models), Kummer & Goodall (1985) also argue that the chimpanzees were learning the behavioural form from each other as several individuals demonstrated the same behaviour very quickly once they were exposed to the banana box. The authors interpret this observation as evidence that the chimpanzee must have copied the behaviour by watching others beforehand, rather than individually learning the behaviour¹ . Although it is possible that the chimpanzees observed each other, this fact alone might be rather irrelevant, as current experimental data suggests that unenculturated chimpanzees do not spontaneously copy actions in most contexts (see above, but see also Horner & Whiten (2005) for an alternative view). Therefore, an alternative explanation, which is more consistent with the interpretations of the majority of the studies in Table 2 and the current experimental data, may be that the chimpanzees converged instead on the same behavioural form by individually developing the most efficient solution to the common problem at hand. This is not to say that other variants of social learning played no role—it is indeed likely that the chimpanzees were drawn to the banana box through stimulus and/or local enhancement when they observed other individuals interacting with the box.

However, it is only with more controlled cases of reinnovations that learning mechanisms can be identified with (more) confidence in wild populations. Hobaiter et al. (2014) report one of the rare occurrences in which the emergence of tool-use behaviours in wild chimpanzees was tracked as the frequencies increased within the community. The authors describe the emergence of two tool-use behaviours in the chimpanzees of Budongo Forest, Uganda: moss-sponging and leaf-sponge re-use. The tools were used to retrieve water from a waterhole that had recently been flooded (Hobaiter et al., 2014). The authors followed the increases in frequency of these two behaviours within the Sonso chimpanzee community. Using network-based diffusion analysis for the first behaviour, Hobaiter et al. (2014) attributed at least 85% of the newly observed events of moss-sponging to social learning, arguing that for each new observation, naïve chimpanzees enhanced their chances of developing moss-sponging by a factor of 15. Thus, non-copying social learning most likely played a role in increasing the frequency of this behaviour within the community. Yet, even though the frequency of the behaviour increased through non-copying social learning, this leaves open the question as to whether the form of the behaviour was individually or socially learnt. This study was included in the database because, after the original innovation of moss-sponging by the alpha male, the alpha female also developed the behaviour independently before having observed the male moss-sponging (researchers were able to closely observe the chimpanzees during the study period, allowing for conclusions to be made on the background knowledge of each of the group members; Hobaiter et al., 2014). Therefore, it seems possible that although non-copying social learning facilitates the individual acquisition of the behavioural form by the rest of the group, the form of moss-sponging can be individual learnt, as it was reported in two separate individuals in the absence of any social models. In a follow-up study on the spread of moss-sponging three years after its innovation, Lamon et al. (2017) agreed with the individual learning approach to explaining the increase in frequency of this behaviour: ‘Of course, each moss-sponger has to individually learn the behaviour, but in all likelihood, this was facilitated by the social influence exerted by other group members that acted as models’ (Lamon et al., 2017, 6).

On the other hand, for the second behaviour, leaf-sponge re-use, the social network analysis failed to show a role of social learning in the increase in frequency of the behaviour, as eight naïve chimpanzees independently acquired the same behavioural form (Hobaiter et al., 2014). Thus, this study suggests that both moss-sponging and leaf sponge-reuse seem to be within chimpanzees’ individual learning abilities, but both behaviours are influenced to a higher or lesser extent by non-copying social learning.

Several similar reports of reinnovations of stick and leaf tool-use behaviours by naïve individuals were also found in captive chimpanzees. Wolfgang Köhler spent many years (1914–1920) investigating the cognition behind captive chimpanzee stick tool-use, and concluded that many of these behaviours, such as using a stick to retrieve out-of-reach foods, can emerge via ‘insight learning’² . Similarly, Kitahara-Frisch & Norikoshi (1982) examined the origins of sponge-making (a behaviour in which wild chimpanzees use leaves as sponges to absorb liquids such as water or honey; Kitahara-Frisch & Norikoshi, 1982, 42) and found that captive chimpanzees showed the same form of sponge-manufacture and use as their wild counterparts. The authors conclude that, contrary to previous claims, ‘the example of the mother is by no means necessary for the habit to appear in young animals. This observation raises the question of whether the acquisition of so-called proto-cultural habits does not rely as much, at least, on independent reinvention as on transmission through imitation learning’. The interpretation of the results of many of the studies included in Table 2 agree with those proposed by the ZLS³ and provide mounting evidence for the view that simple stick tool-use is, most likely, within the individual learning capabilities of both wild and captive chimpanzees (Bandini & Harrison, 2020; see also Gruber et al. (2009, 2011) for reports of wild chimpanzees not acquiring stick tool-use, although see the section on behavioural flexibility below for further discussion of these studies).

Gorillas

Although gorillas do not commonly use tools in the wild, sporadic cases of spontaneous tool-use by gorillas in the wild and in captivity have been reported. One such report comes from Kinani & Zimmerman (2015), who describe, for the first time, a female juvenile gorilla (Gorilla beringei beringei) using a stick to fish for ants in Volcanoes National Park, Rwanda. The authors describe the behaviour as being similar to chimpanzee ant-dipping (which was classified as a ‘putative cultural trait’; Whiten et al., 1999; for chimpanzees before it was discovered that the environment plays an important role, alongside other factors, in shaping the form of the behaviour; Humle & Matsuzawa, 2002, Humle, 2006, Schöning et al., 2008). This is the first report of ant-dipping (and indeed, any stick tool-use) in wild gorillas. Further evidence of gorillas’ spontaneous tool-use is offered by Lonsdorf et al. (2009), who exposed captive gorillas to an artificial termite mound in their enclosure. The authors report that although not all the gorillas in the group used tools, the alpha male fished for the bait using a stick on the first day of the study, demonstrating the same behavioural form as wild chimpanzees, despite being naïve (Lonsdorf et al., 2009). Similarly, Nakamichi (1999) observed a western lowland gorilla (Gorilla gorilla gorilla) at the San Diego Wild Animal Park throw sticks into the foliage of trees to knock down leaves and seeds (which were later consumed), and Fontaine, Moisson & Wickings (1995) describe how a group of gorillas in a zoo in Gabon spontaneously used sticks to reach objects outside of their enclosure and coconut fibres as sponges to absorb water, demonstrating similar behavioural forms as chimpanzees and other primates who regularly use tools (Fontaine, Moisson & Wickings, 1995). Although the behaviours described above are not particularly complex (following the current definition in the literature; Meulman et al., 2012), these reports demonstrate that gorillas are capable of spontaneously using some tools when motivated. Thus, despite rarely showing tool-use behaviours in the wild, gorillas have demonstrated a surprisingly extensive ability to innovate and reinnovate various tool-use behaviours via individual learning (Boysen et al., 1999; Fontaine, Moisson & Wickings, 1995; Lonsdorf et al., 2009; Manrique, Völter & Call, 2013; Nakamichi, 1999; Natale, Poti’ & Spinozzi, 1988; Neadle, Allritz & Tennie, 2017; Pouydebat et al., 2005; Tennie et al., 2008).

Capuchins

Capuchins exhibit one of the most extensive natural tool-use repertoires in primates (second only to chimpanzees), and a clear ability to individually learn these repertoires, including stone-tool behaviours. One example of stone-tool behaviour comes from wild bearded capuchins (Sapajus libidinosus) in Serra da Capivara National Park in Brazil. These capuchins have been observed deliberately pounding standing conglomerates with smaller hammerstones to break open fragments of the larger stones (Proffitt et al., 2016). This behaviour, named stone-on-stone (SoS) percussion, involves an individual selecting a smaller pounding stone and using it to strike the cobbles embedded in a conglomerate of standing stones using one or both hands (Proffitt et al., 2016). The purpose behind SoS percussion is still unclear, but it may be that the monkeys break stones open to ingest powdered lichens or quartz from inside (support for this possibility comes from a report of a captive capuchin (Cebus nigritus) that was also observed practicing SoS and then licking the lichens inside the stone; Bortolini & Bicca-Marques, 2007). An interesting by-product of this behaviour is that by pounding the stones together, the capuchins produce flakes superficially similar to those made by early hominins (according to the authors; Proffitt et al., 2016). This is the first observation of SoS behaviour, and the production of flakes, in wild capuchins. Although the learning mechanisms behind the acquisition of this behaviour in the wild remain to be identified, an earlier experimental test with captive capuchins (Sapajus apella) describes how a similar stone pounding behaviour, including the production (and use) of flakes, was reinnovated spontaneously by naïve capuchins (Westergaard & Suomi, 1994a). The unenculturated captive capuchins tested by Westergaard & Suomi (1994a; the unenculturated status of the capuchins was confirmed by G. Westergaard, 2019, personal communication) were also able to use the flakes they made as tools, and used them to cut open (by pushing and chiselling) the acetate top of a baited testing apparatus (similarly to the testing conditions the enculturated bonobo Kanzi faced; Toth et al., 1993). This study is particularly interesting as it demonstrates that naïve capuchins are not only able to reinnovate a wild behaviour (SoS percussion), but can also use the flakes they produced.

Ottoni & Mannu (2001) also describe a stone-tool behaviour in wild capuchins. The wild bearded capuchins of the Ecological Park of the Tiete River (São Paulo, Brazil) are provisioned daily with fruit and protein, but spend the majority of the day foraging for naturally occurring food sources (Ottoni & Mannu, 2001). During one of these foraging sessions, the monkeys were observed cracking open mature Syagrus nuts using a hammerstone on an anvil, similarly to how wild chimpanzees crack nuts. The authors conclude that nut-cracking is driven by ‘some kind of observational learning, albeit restricted to stimulus enhancement, as a starting point of a long process of individual improvement by trial-and-error’ (Ottoni & Mannu, 2001, 357). Similarly, in a study on nut-cracking in captive capuchins, Visalberghi (1987) provided naïve subjects with stones and encased almonds, but no social information on the behaviour. Over a period of ten trials, two males successfully used stone tools to crack open the almonds (Visalberghi, 1987). Although the rest of the group (40 individuals) had ample opportunities to observe the two males cracking nuts, no other individual from the group reinnovated nut-cracking (see below for further discussion). Nut-cracking has now also been observed in other species of capuchins, including wild Cebus libidinosus (Waga et al., 2006), Sapajus apella (Izawa & Mizuno, 1977), Cebus apella (Struhsaker & Leland, 1977), and in another captive population of Cebus apella in captivity (Antinucci & Visalberghi, 1986). These findings strongly support the view that nut-cracking is within the individual learning abilities of several capuchin species. Various other spontaneous tool-use behaviours have been observed in capuchins (Boinski, 1988; Fernandes, 1991; Soares Bortolini & Bicca-Marques, 2007; Visalberghi & Trinca, 1989; Westergaard & Fragaszy, 1987), making capuchins a promising species for the study of individual learning in primate tool-use.

Macaques

Some populations of Macaca fascicularis aurea (Mfa), a subspecies of long-tailed macaques, frequently use stone tools and two different methods (pound-hammering and axe-hammering) to pound open encased foods in Southeast Asia (Malaivijitnond & Hamada, 2008). The closely related subspecies, Macaca fascicularis fascicularis (Mff), have, however, never been observed to use tools, despite sharing an environment and even interbreeding with Mfa (Bandini & Tennie, 2018; Falótico et al., 2017; Gumert, Kluck & Malaivijitnond, 2009). Despite having access to social information on tool-use (i.e. Mfa who could act as demonstrators of the tool-use behaviours), Mff continue not to use stone tools in the wild. However, Zuberbühler et al. (1996) observed a captive Mff spontaneously using a stick to retrieve apples that had fallen just out of reach outside the enclosure. The behaviour was then also observed in other members of the group. Stick manipulation was found to slightly increase when the original innovator was raking in apples. However, the small increase of manipulation of sticks (perhaps via social facilitation) did not always result in tool-use by the other members of the group. The authors therefore conclude: ‘We cannot conclude from these data that stimulus enhancement is a necessary prerequisite to becoming a skilled animal, because MD, the inventor of the technique, most likely developed his skill by means of individual learning’ (Zuberbühler et al., 1996, 10). Naïve Mff have also been observed using human hair as dental floss (Watanabe, Urasopon & Malaivijitnond, 2007), and Tonkean macaques (Macaca tonkeana) have been observed spontaneously using sticks to retrieve honey from an out-of-reach apparatus (Anderson, 1985), to rake food into the enclosure (Ueno & Fujita, 1998), and making climbing structures out of sticks and browse (Ducoing & Thierry, 2005; Westergaard & Lindquist, 1987). Japanese macaques (Macaca fuscata) also use human hair as dental floss (Leca, Gunst & Huffman, 2010) and use stones (and infants) to push fruit out of a tube (Tokida et al., 1994). Lion-tailed macaques (Macaca silenus) use probes to extract syrup from a baited apparatus (Westergaard, 1988) and Rhesus macaques (Macaca mulatta) use cup-like containers to transport water around their enclosures (Parks & Novak, 1993). These reports demonstrate that, similarly to gorillas (who do not often practice tool-use in the wild), various species of macaques can reinnovate some tool-use behaviours when in the appropriate context.

Crows

In comparison to primates, the case for individual learning in bird material culture seems to be somewhat less debated. Indeed, several accounts exist of naïve birds in the wild and in captivity spontaneously acquiring wild behaviours (Auersperg et al., 2012; Bird & Emery, 2009a; Collias & Collias, 1964; Epstein, 1985; Jones & Kamil, 1973; Overington et al., 2011; Taylor et al., 2010; Tebbich et al., 2001, 2007). Some of the most impressive accounts of the individual learning of tool-use come from New Caledonian crows (Corvus moneduloides), who possess sophisticated stick tool-use repertoires, rivalling even those of non-human primates (Weir, 2002). Similarly to chimpanzees, some New Caledonian crow tool-use is subject to regional variation. In the chimpanzee case however, this variation has been used as evidence for the view that the behavioural forms are dependent on social learning (Whiten et al., 1999, 2001; Gruber et al., 2015). Yet this claim has not been made as often for wild New Caledonian crows (Weir & Kacelnik, 2006; although see also Logan et al. (2016) for an experimental test on the non-copying social learning abilities of New Caledonian crows and Hunt & Gray (2003) for claims that New Caledonian crows may have cumulative culture). Experimental studies with captive New Caledonian crows have demonstrated that naïve crows can spontaneously make and use some of the same tools as their wild counterparts. The authors state: ‘In the light of our findings, it is possible that the high level of skill observed in wild adult crows is not socially acquired’ (Kenward et al., 2005, 121). Naïve New Caledonian crows are also capable of developing tool bending (Weir & Kacelnik, 2006) and metatool-use (i.e. the ability to use one tool on another; Taylor et al., 2007; Whiten & Byrne, 1997). These findings led Kenward et al. (2005, 121) to conclude that: ‘the ability of this species to manufacture and use tools is at least partly inherited and not dependent on social input’.

Another strong example of individual learning in birds comes from Hawaiian crows (Alalā; C. hawaiiensis). Hawaiian crows are extinct in the wild and currently exist only in captivity, but it is likely that in their past natural state these crows also used tools (similarly to New Caledonian crows; Rutz et al., 2016). In an experimental study with captive Hawaiian crows, 78% of the population spontaneously used tools to probe for out-of-reach food, without social demonstrations (Rutz et al., 2016). Similarly to Kenward et al. (2005) conclusion for New Caledonian crows, Rutz et al. (2016, 405) also suggest that the observed behavioural repertoire is therefore a product of individual learning and/or genetic predispositions: ‘Alalā clearly possess a propensity to “discover” tool-assisted foraging solutions independently’.

Other birds

Similarly to primates, several species of birds that do not use tools in the wild spontaneously acquire tool-use when provided with the materials in captivity. One example of this phenomenon are captive born and raised blue jays (Cyanocitta cristata), who do not use tools in the wild, but were observed tearing up pieces of newspaper and using them to rake out-of-reach food pellets from outside their cage (Jones & Kamil, 1973). Although note that this is the only study in Table 2 in which the authors interpreted their observation as the product of imitation (‘the fact that to date we have found six jays in our colony demonstrating tool-using behaviour is more likely the result of […] observational learning or imitation than the result of the independent acquisition of this behaviour by each of the six jays’, Jones & Kamil, 1973, 1078). In the light of new understanding on the role of individual learning and non-copying social learning in bird tool-use, it is also possible that mechanisms other than copying drove the emergence of this behaviour, but this remains to be tested.

Most other cases of spontaneous tool-use in birds have been cited as examples of individual learning rather than of imitation and/or other variants of copying social learning. For example when captive Goffins cockatoos were observed making and using stick tools to rake in food from outside the enclosure, the authors argue that their ‘observations prove that innovative tool-related problem-solving is within this species’ cognitive resources’ (Auersperg et al., 2012, 2). When captive Rooks (Corvus frugilegus) used stones to collapse a platform to retrieve a worm, the authors similarly concluded that the reinnovation was an individually learnt solution to a new problem (Bird & Emery, 2009b). Similarly, when naïve pigeons solved a task inspired by Köhler (1925) work with chimpanzees, in which the pigeons had to use boxes to reach a banana hanging outside their enclosure, the author concludes that they ‘have on hand an instance of insightful problem solving’ (Epstein, 1985, 62). Indeed, albeit some minor exceptions, contrary to the primate case, the view that bird tool-use is the product of individual learning is pervasive throughout the literature (Laumer et al., 2017; Tebbich & Bshary, 2004; Tebbich et al., 2001, 2007).

Individual learning in other animals

Although most reports of animal spontaneous tool-use come from primates and birds, other animal species have also demonstrated the ability to individually acquire tool-use behaviours. For example a dingo (Canis dingo) was recorded moving objects around his enclosure, including a table, to climb on to reach food or to observe other animals outside his enclosure (Smith, Appleby & Litchfield, 2012). The dingo’s behaviour (using a table to access out-of-reach food) is reminiscent of Köhler (1925) early studies in which chimpanzees used boxes to reach hanging bananas. Similarly, reports exists of both wild and captive bears using tools, including one experimental study which examined whether captive bears would manipulate and re-orient objects to climb on to reach suspended food (Deecke, 2012; Waroff et al., 2017). Furthermore, Chevalier-Skolnikoff & Liska (1993) describe over 20 tool-use behaviours that emerged without social learning by captive African (Loxodonta africana) and Asian elephants (Elephas maximus). Behaviours included: ‘reach toward food with stick held in the trunk’; ‘Rub the body with a stick’; ‘probe musth gland with stick’ (Chevalier-Skolnikoff & Liska, 1993, 210). A recent observational report describes the use of two coconut halves as a protective shell by an octopus (Amphioctopus marginatus; Finn, Tregenza & Norman, 2009), and orphaned, captive juvenile sea otters (Enhydra lutris nereis) were found to develop the same stone tool pounding behavioural forms as observed in wild adult otters (Nicholson et al., 2007). Collectively, these studies demonstrate that tool-use is not restricted to primates and birds, but that many other animals possess the ability to use tools, and crucially, the ability to individually learn their behavioural forms.