Ramsey et al. make a much-needed attempt to clarify methodological issues for identifying innovative behaviour in animals. However, their effort is seriously weakened by an underlying presumption in their article that the behaviour of primates is generally learned whereas that of non-primates is generally “innate.” Three related reasons appear to explain this presumption: (1) There is a heavy emphasis on innovative behaviour by nonhuman primates, giving the false impression that innovative behaviour is rare in non-primates. This may reflect a continuing primate-centred view of animal cognition among many primatologists (Emery & Clayton Reference Emery, Clayton, Rogers and Kaplan2004a). (2) Potential innovations are excluded because of a flawed understanding of how learned behaviour is genetically assimilated. And (3) there is a poor grasp of the non-primate literature.

The authors exclude tool manufacture and use by New Caledonian crows (Corvus moneduloides, NC crows hereafter) as a potential source of innovations because they imply that the behaviour is “Baldwinian,” or “innate.” They base the exclusion on recent findings that captive naïve juvenile crows developed proficient stick tool use (Hunt et al. Reference Hunt, Lambert and Gray2007; Kenward et al. Reference Kenward, Weir, Rutz and Kacelnik2005). NC crows (and woodpecker finches Cactospiza pallida; Tebbich et al. Reference Tebbich, Taborsky, Fessl and Blomqvist2001) clearly have a strong predisposition for basic tool use. However, Ramsey et al. confuse the basic tool skills of naïve juveniles with the considerably more accomplished skills of free-living adult NC crows. NC crows manufacture a diverse range of tools that include hooked tools made from sticks and similar material and distinct designs (wide, narrow, and stepped) made from the leaves of Pandanus species trees (Hunt Reference Hunt1996; Hunt & Gray Reference Hunt and Gray2002; Reference Hunt and Gray2003; Reference Hunt and Gray2004). Although all the naïve crows mentioned ripped Pandanus leaves placed in their cages, only one bird made and used “rough” Pandanus tools. The shapes of these crude tools showed little resemblance to any of the three standardized shapes produced in the wild. The obvious mismatch between the basic tool skills of the naïve juveniles and the more complex skills of experienced free-living NC crows suggests that the tool skills used in the wild require learning. Four lines of evidence indicate that learning is involved in the development, maintenance, and evolution of NC crows' tool skills in the wild: flexibility, different tool designs, innovative behaviour, and a prolonged learning period in juveniles.

Behavioural flexibility is present in the tool skills of free-living NC crows. First, they manufacture three kinds of hook tools out of stick-like material: live twigs, fern stolons, and a thorny vine (Hunt & Gray Reference Hunt and Gray2002). Each kind of tool requires material-specific manufacture techniques for its production. Second, individuals in the same population can specialize in different tool types. At one site on Maré Island individual crows generally specialized in either stick tools or Pandanus tools (Hunt & Gray Reference Hunt and Gray2007). Preferences for a particular tool type were not closely associated with the sexual size dimorphism in NC crows (Kenward et al. Reference Kenward, Rutz, Weir, Chappell and Kacelnik2004). Third, we tested if two crows made tools of an appropriate length in response to holes of different depth (Hunt et al. Reference Hunt, Rutledge and Gray2006). On the first attempt in a trial both birds made tools of a consistent length regardless of hole depth. However, they generally made longer tools on the next attempt with deeper holes, which suggested that the crows relied on a heuristic strategy to solve tool problems. Last, the behaviour of a crow that we observed crafting hooked tools out of live twigs was organised in a flexible, hierarchical way (Hunt & Gray Reference Hunt and Gray2004).

A consequence of the diverse range of tool designs that NC crows manufacture is different sets of designs at sites (Hunt & Gray Reference Hunt and Gray2002; Reference Hunt and Gray2003). The differences are striking even when the main types of designs are included. NC crows on the island of Maré only make wide Pandanus tools and non-hooked twig tools. Crows at many sites on the mainland make stepped Pandanus tools and hooked twig tools. At certain other sites they make narrow and stepped Pandanus tools and hooked twig tools. In Parc Rivière Bleue, crows surprisingly make wide, narrow, and stepped Pandanus tools as well as hooked twig tools. There are no obvious ecological or genetic correlates associated with these site differences. Therefore, using the “geographic method,” which relies on circumstantial evidence (e.g., Whiten et al. Reference Whiten, Goodall, McGrew, Nishida, Reynolds, Sugiyama, Tutin, Wrangham and Boesch1999), the manufacture of these tool designs appears to meet the requirements for socially learned behaviour.

NC crows are capable of innovative behaviour. We recently documented innovative tool modification in the wild on Maré by two male NC crows (Hunt et al., submitted). The males used a novel folding technique to bend wide Pandanus tools into a boomerang-like shape. A practical consequence of the folding was that it reduced a tool's length without cutting the tool. This allowed the crows to hold longer tools near the working end and in line with their bills. Experiments with captive NC crows have also revealed innovative tool skills that are similar to, or even surpass, those of great apes (Taylor et al. Reference Taylor, Hunt, Holzhaider and Gray2007; Weir & Kacelnik Reference Weir and Kacelnik2006; Weir et al. Reference Weir, Chappell and Kacelnik2002). The positive correlations in birds and primates between innovation frequency, tool use, and brain size are also consistent with the evidence indicating that the tool skills of NC crows are learning-based behaviour (Lefebvre et al. Reference Lefebvre, Nicolakakis and Boire2002; Reader & Laland Reference Reader2002).

The final line of evidence is the prolonged development of tool skills in juvenile NC crows. Juveniles forage with their parents in family units for 10 months or more after they fledge and have ample opportunity to learn tool skills in a social context (Hunt & Gray Reference Hunt and Gray2007). It takes many months for them to become proficient in wide Pandanus tool manufacture and use (Holzhaider et al., in preparation). We also have experimental evidence that proficiency in these skills probably depends on appropriate behavioural strategies developed through associative learning (Holzhaider et al., in press).

The authors' belief that “Baldwinization” of behaviour eliminates learning, and therefore the potential for innovation, is based on a flawed understanding of the extent to which learned behaviour becomes genetically assimilated. The assimilation of adaptive learned behaviour is suggested to occur through two different mechanisms: the “Baldwin Effect” described by Baldwin (Reference Baldwin1896a; Reference Baldwin1902) and Waddington's (Reference Waddington1942; Reference Waddington1953) “genetic assimilation” (see West-Eberhard [Reference West-Eberhard2003] and papers in Weber & Depew [Reference Weber and Depew2003] for detailed discussion about these two mechanisms). “Genetic assimilation” in this commentary refers to both these mechanisms. The importance of the role that genetic assimilation has had in the evolution of animal behaviour remains unknown (Hall Reference Hall, Weber and Depew2003). Genetic assimilation of a learned behaviour may also lead to a “learned trait complex” in which supporting adaptive modifications (e.g., morphological and cognitive) refine and elaborate the trait in an ongoing cyclical process as long as it remains adaptive. The important point about this process is that although the adaptive behaviour becomes increasingly specialized and efficient through genetic assimilation, it is still learned behaviour. Genetic assimilation does not eliminate learning, and with learning the potential for innovation is retained (Bateson Reference Bateson2004; West-Eberhard Reference West-Eberhard2003).

Sterelny's (Reference Sterelny, Webber and Depew2004) scenario of how technological evolution in humans came about describes a learned trait complex. He suggests that only the basic elements of human technical capability were genetically assimilated (e.g., morphological attributes and cognitive skills). Sterelny stresses that “the total package would remain a gradually elaborating mix of learned and innate elements” (p. 298). Technological evolution would be initiated when individuals used the assimilated attributes and skills in association with learning to innovate and then transmit these innovations. If technological evolution of tools has occurred in NC crows, as we have suggested (Hunt & Gray Reference Hunt and Gray2003), it may have proceeded in a similar manner to Sterelny's scenario for humans (Hunt Reference Hunt2005). This is because only the basic elements of NC crows' technical capability have probably been genetically assimilated.