Ramsey et al. argue that behavioral innovation should play an important role in studies of animal social learning, evolution, ecology, and cognition. I agree entirely, and admire their work in establishing and moving the field forward. As there remains much to do, it is important, particularly at these early stages, to reach consensus on clear definitions and delimitation of phenomena. Ramsey et al. argue for revision of Reader and Laland's (2003a) definition of animal innovation, principally excluding environmentally induced novel behavior and defining innovation at the individual rather than population level. Both revisions are debatable. Although a group member can innovate even if another has made the same discovery, such innovations are difficult to discriminate from socially learned innovations, whereas first occurrences may be particularly relevant to identification of the individual and/or environmental variables that favor innovation. Ramsey et al.'s individual-level definition also raises problems regarding the distinction between innovation and individual learning (see Reader & Laland 2003a).

Ramsey et al. exclude environmentally induced innovations from consideration, but change in the physical or social environment may precede much innovation (Hauser Reference Hauser, Byrne and Whiten1988; Lee Reference Lee and Box1991; Reader & Laland Reference Reader and Laland2001), and innovative bird species are more likely to survive than less innovative species when introduced to novel environments (Sol Reference Sol, Reader and Laland2003). Moreover, many (but by no means all) reported innovations are responses to human-induced environmental changes (Lefebvre et al. Reference Lefebvre, Whittle, Lascaris and Finkelstein1997; Reference Lefebvre, Reader and Webster2001; Reader & Laland 2002). Innovation may also be prompted by social demands, where it may be key in outwitting rivals (e.g., innovative tactical deception; Byrne Reference Byrne, Reader and Laland2003), be frequency dependent, or be prompted or facilitated by existing innovations that create a need or opportunity for further innovation. All these would be considered environmentally induced innovations, but they are innovations nonetheless, and surely of both theoretical and applied interest (e.g., to conservation biology; McDougall et al. Reference McDougall, Réale, Sol and Reader2006). The exclusion of environmentally induced innovations, as Ramsey et al. advocate, will likely eliminate false positives, but may also exclude many interesting and functionally important “true” innovations. Innovation presumably carries maximal advantage as a response to novel circumstances where natural selection cannot have shaped appropriate responses.

Ramsey et al.'s new definition and classification scheme may carry other disadvantages for empirical studies of innovation (Kendal et al. raise complementary points in their commentary). Ramsey et al. propose several characters of potential utility for discriminating innovations from non-innovations, and present a useful exposition of how these characters can be considered together. However, the rate of acquisition, rate of spread, rarity, apparent cognitive complexity, and distribution within or across populations may provide misleading clues for identifying innovations. Such characters are not irrelevant to the study of proposed innovations, but should be interpreted cautiously: many genetic, social, environmental, and experiential influences can impact on the performance of a given behavior. Proper consideration of ecological explanations for behavioral differences between populations, for example, requires consideration of the availability, utility, and risks of alternative behavior patterns. The relative utility and performance of a given foraging behavior could be influenced by, for instance, the availability of other foods, the ease of identifying alternate foods, differences in need (e.g., disease promoting peculiar nutritional requirements), local predation risk, the activities of others (e.g., resource competition), or constraints on performance of alternative behaviors (Dewar Reference Dewar2004; Reader Reference Reader2004). Exclusion of such “ecological” causes for population differences may be extremely challenging with observational study alone (Laland & Janik Reference Laland and Janik2006). Moreover, as argued earlier, ecological causes would not eliminate the possibility that the behavior was an innovation.

Ramsey et al. suggest using geographic and local prevalence to identify innovations, but how informative are these variables? Innovation prompted by the peculiarities of local sub-habitats would result in a patchy distribution that would correlate with ecological differences. Conversely, with limited solutions to the same problem, an innovative species may repeatedly and independently invent the same behavior in many populations, resulting in high geographic prevalence. Nocturnal feeding by street lights in kingbirds, for example, has been repeatedly reported on several Caribbean islands, suggesting environmental induction in each population (Reader et al. Reference Reader, Morand-Ferron, Côté and Lefebvre2002a). Innovative species may also show high local prevalence of novel behaviors. As Ramsey et al. note, reported novel behavior is amenable to experimental investigation in the laboratory, in the field, and with wild-caught captives (e.g., Morand-Ferron et al. Reference Morand-Ferron, Lefebvre, Reader, Sol and Elvin2004; Pfeffer et al. Reference Pfeffer, Fritz and Kotrschal2002; Reader et al. Reference Reader, Nover and Lefebvre2002b), and such experiments provide a vital follow-up to observational data. Experimental investigations of celebrated cases of animal innovation – sweet-potato washing by macaques and milk-bottle opening by titmice – demonstrate that individual rates of discovery may be far higher than was generally thought (Kothbauer-Hellman Reference Kothbauer-Hellman1990; Sherry & Galef Reference Sherry and Galef1984; Reference Sherry and Galef1990; Visalberghi & Fragaszy Reference Visalberghi and Fragaszy1990). Where geographically or locally prevalent innovation suggests that a novel behavior is readily discovered, species comparisons may be informative: Why do some species apparently make these discoveries readily, while sympatric species do not? High local and geographic prevalence in particular species could result from several processes, including general behavioral flexibility, context-specific flexibility, or Ramsey et al.'s “Baldwinized innovations.” However, although Baldwin effects have significant theoretical impact for the evolutionary consequences of behavioral flexibility (Sol Reference Sol, Reader and Laland2003; Wyles et al. Reference Wyles, Kundel and Wilson1983), they lack clear empirical support (de Jong & Crozier Reference de Jong and Crozier2003).

Here are two final definitional points. First, Ramsey et al. introduce “behavioral flexibility” to mean behavior-level flexibility. However, current usage typically considers behavioral flexibility as part of phenotypic plasticity and thus repertoire, not behavior-level, flexibility (Bateson Reference Bateson, Halliday and Slater1983; Schlichting & Pigliucci Reference Schlichting and Pigliucci1998). Second, the suggested social learning-innovation continuum is unclear and open to several alternative interpretations: (1) Some individuals innovate while others socially learn these innovations; (2) some elements of a behavior are innovated and some are socially learned; (3) social learning involves some innovation, such as when stimulus enhancement attracts a bird to food but the processing methods are individually acquired; or (4) copying errors during social learning result in novel behavior.

An important question not raised by Ramsey et al. is whether innovative propensities are adaptations rather than side-products or indicators of other phenomena, such as general behavioral flexibility. Investigation of the genetic, hormonal, neuro-cognitive, and developmental mechanisms underlying innovation will help resolve this issue as well as the important questions raised by Ramsey et al. and Reader and Laland (2003a). There is clear potential for integration with research on temperament and behavioral syndromes (Réale et al. Reference Réale, Reader, Sol, McDougall and Dingemanse2007; Sih et al. Reference Sih, Bell, Johnson and Ziemba2004). Although some evidence suggests that innovative propensities may be consistent individual traits (Laland & Reader 1999; Pfeffer et al. Reference Pfeffer, Fritz and Kotrschal2002, but see Fragaszy & Visalberghi Reference Fragaszy and Visalberghi1990), it is not known whether innovatory propensities are heritable, are impacted by developmental experience, or whether animals can “learn to innovate.” A range of psychological processes may be involved in any one innovation, and may differ between different innovations. Progress will require investigation of the causes, consequences, and mechanisms underlying animal innovation; all areas ripe for study.