Defining innovation is key to studying it; Ramsey et al.'s definition is based on: (1) what processes generate innovation, (2) whether innovations refer to behaviors new to an individual or to a population, and (3) a dichotomous key designed to guide field research.

1. Processes. Ramsey et al. need to clarify what makes innovation qualitatively different from novel learned behaviors that emerge in response to intrinsic (maturational) and extrinsic events. Is the goal to identify specific cognitive mechanisms in order to distinguish innovation from other novel learned behaviors? Higher-level cognition is implicated in most innovation research, but is this a necessary requirement? Ramsey et al. do distinguish between “cognitively simple innovations” produced by accident or trial and error and “cognitively complex innovations” produced by causal reasoning. Although specific learning mechanisms (e.g., trial-and-error, directed learning, goal-oriented behavior, or “insight”) have been intentionally barred from some definitions of innovation (e.g., Reader & Laland Reference Reader, Reader and Laland2003a) and requiring them creates challenges for field studies, mechanistic criteria can help clarify whether innovation is a subset of or equivalent to individual learning. This is analogous to the debate on whether any form of social learning constitutes culture (e.g., Rendell & Whitehead Reference Rendell and Whitehead2001). Does any form of individual learning qualify as innovation?

If, to identify innovation, one must exclude maturational and ecological causes as much as possible, the authors' approach suffers from the same shortcomings of the geographic method for identifying “culture” (e.g., Laland & Janik Reference Laland and Janik2006; Sargeant et al. Reference Sargeant, Wirsing, Heithaus and Mann2007). Not only would it be difficult to collect sufficient data to exclude ecological or maturational causation, these mechanisms interact to provide adaptive contexts for development of new behaviors and are thus critical to any study of innovation. Ramsey et al. acknowledge that these causes are not mutually exclusive and even suggest that novelty and “food scarcity, risk, and perhaps mobility (leading to ‘novelty’ when one returns to a location after an absence)” (sect. 5.2, para. 2) are factors underlying psychological processes of innovation, even though these seem to qualify as “environmentally induced.” Their approach differs from theoretical approaches that consider rates of environmental change essential factors in understanding selective pressures and adaptive benefits of various types of learning (e.g., Boyd & Richerson Reference Boyd and Richerson1985; Laland & Kendal 2003; Whitehead Reference Whitehead2007).

2. Individual versus population. The authors also define innovation as development of non-socially learned behavior novel to an individual, rather than a population. If the goal is to understand cognitive abilities, this point is vital because individual learning is more relevant to cognition than what occurs at the population level. For example, a population of one species exposed to anthropogenic change might exhibit more new, learned behaviors than a pristine population of a closely related species with equivalent learning ability. However, if one accepts this definition, innovation is, again, largely indistinguishable from the broader concept of individual (or non-social) learning. All learned behaviors are, by definition, novel to an individual. Alternatively, studies that define innovation as behavior novel to a population (e.g., Reader & Laland Reference Reader and Laland2003a) may aid in estimating (albeit conservatively) the rate of individual learning in wild populations (e.g., Reader & Laland Reference Reader2002) and also help identify conditions that favor niche expansion and facilitate documentation of the transmission of new and potentially cultural behaviors (e.g., Reader & Laland Reference Reader and Laland2003b). Thus, we consider “innovation” better defined as a subset of individual learning evidenced by behaviors new to a population, useful for studying conditions favoring the development, spread, or effects of new behaviors.

3. Dichotomous key. We applied Ramsey et al.'s operational definition to our long-term study of foraging behavior in wild bottlenose dolphins (Tursiops sp.) in Shark Bay, Australia. Bottlenose dolphins have obvious “innovative” abilities in that they readily produce novel behaviors in captive (Herman Reference Herman, Bekoff, Allen and Burghardt2002) and field settings in response to human activity (e.g., Chilvers & Corkeron Reference Chilvers and Corkeron2001; Mann & Kemps Reference Mann, Kemps, Gales, Hindell and Kirkwood2003). We have identified at least 12 distinct foraging behaviors and substantial individual specialization in our population (Mann & Sargeant Reference Mann, Sargeant, Fragaszy and Perry2003; Sargeant et al. Reference Sargeant, Mann, Berggren and Krützen2005). In applying the key to our data, several challenges were apparent. First, most foraging behaviors were correlated with ecological variation (Sargeant et al. Reference Sargeant, Mann, Berggren and Krützen2005; Reference Sargeant, Wirsing, Heithaus and Mann2007) and maturation (Mann & Sargeant Reference Mann, Sargeant, Fragaszy and Perry2003). Second, even in a well-studied population, relevant data were lacking. For example, to separate innovations from behaviors that were “environmentally induced” detailed knowledge of the environment at fine spatial and temporal scales is needed, something extremely difficult in studying dolphins that seek mobile patchy, highly variable, aquatic prey. Also, information on whether behaviors are innate or genetically determined or occur in captivity was not available, especially because captive dolphins are fed dead fish. Similarly, determining whether social learning was involved is controversial because standards of evidence and methodology are still hotly debated.

Finally, data from multiple populations might help, but are severely limited by diverse methodologies, observation conditions, and research effort. In using the dichotomous key, we had to disregard the clear social and ecological contributions for most foraging behaviors (Mann & Sargeant Reference Mann, Sargeant, Fragaszy and Perry2003; Sargeant et al. Reference Sargeant, Wirsing, Heithaus and Mann2007) and assume no exclusively genetic causes (e.g., Krützen et al. Reference Krützen, Mann, Heithaus, Connor, Bejder and Sherwin2005). Six foraging behaviors fit the criteria for innovation and several were inconclusive. Using Reader & Laland's (Reference Reader, Reader and Laland2003a) definition, two behaviors would be classified as innovations: sponging and trevally hunting. Trevally hunting involves an elaborate pursuit and processing of golden trevally fish (up to 1 m in length), and is used by one adult female (among hundreds of dolphins observed in the population). She catches trevally every few hours, and attracts many dolphins immediately following a catch, suggesting the availability of trevally to others. Sponging would also qualify largely because a small subset of individuals uses this foraging method and we have traced its development and habitat use (Mann & Sargeant Reference Mann, Sargeant, Fragaszy and Perry2003; Sargeant et al. Reference Sargeant, Wirsing, Heithaus and Mann2007). As expected, Reader and Laland's criteria were more conservative by identifying only relatively recent innovations of known origin.

The authors present an ambitious method for identifying novel learned behaviors that lays the groundwork for a standardized approach, but relevant data will likely be missing for many species. Although the authors attempt to identify ways to study potential innovations using more rapidly obtained cross-sectional data, their approach requires more effort in several areas (more sites, experimental or laboratory data) and it may still take substantial observation to determine which individuals exhibit a behavior. Thus, it is unlikely to “save time” relative to longitudinal study. In addition, their approach does not clarify what cognitive mechanisms are involved. Detailed developmental study on the emergence of behaviors might provide a better indication of maturational, parental, social, ecological, and other influences because these can be directly measured. Sample sizes are a limitation, but detailed developmental analyses can reveal processes involved in the emergence of novel behavior for the individual (e.g., Lonsdorf et al. Reference Lonsdorf, Eberly and Pusey2004; Thornton & McAuliffe Reference Thornton and McAuliffe2006). As with culture, developmental data are sorely needed to help identify intrinsic and extrinsic factors that favor novel behaviors.