In the target article, Ramsey, Bastian, and van Schaik (Ramsey et al.) propose a new definition of animal innovation and a 55-step key to help field workers decide whether a putative case is a true one. We applaud any attempt to improve upon the first papers in what is a very young field, but doubt that Ramsey et al.'s effort will be useful to many researchers.
Why study innovation? The authors propose four reasons for the study of innovation: ecology, macroevolution, culture, and intelligence. The ecological reasons do not require the subtle categorization provided here. In fact, behavioral diversity is blind to the origin of the innovation; it could be genetic or environmentally induced and so most ecologists will be unconcerned about its precise origin. The same holds for its macroevolutionary interest. The extent to which a species is insulated against environmental challenges depends on its ability to adapt its behavior to modified conditions, suggesting that environmental induction is a much more important factor than innovation to students of macroevolution. Ramsey et al. state that innovation can sometimes contribute to culture, but most of their article confuses the two concepts and in a somewhat circular logic, eventually considers culture as an indication of innovation. Finally, the relevance of innovation to intelligence is decreed, not demonstrated. There is no a priori reason to believe that intelligence is associated with innovativeness; that association must be the subject of empirical scrutiny, not a declaration of self-evidence.
What is an innovation? The approach to the concept of innovation that is presented here is to decide a priori on the characteristics that it should possess if it were to be recognized. This leads to a regression of finer and finer details of classification to make sure the category is as pure as possible. So, one must decide that the behavior is learned individually, not part of normal maturation, not an accident, and not induced by an environmental change. Whatever is left then is a real innovation, which can be categorized as an invention, or a weak versus a strong innovation, depending on whether the innovation results from simple or complex cognition – categories that appear entirely tautological and far from being amenable to strong quantitative study.
Is the operational definition really operational? Going through the steps of the proposed 55-step key, it becomes clear that the definition does not do away with guesswork. In fact, it makes guessing a systematic process where one must decide whether one's study falls into one or another dichotomous category, many of which can only be decided by subjective decision (e.g., “behavior common in focal population,” box 21 – what exactly is “common”?). A great deal of the key requires knowing ecological properties that have posed significant measurement problems to ecologists (e.g., environmental heterogeneity). Moreover, some possibilities will be difficult to reject. For example, while it is possible to detect an environmental change responsible for the appearance of a novel behavior, the non-detection of such a change cannot be taken as meaning the absence of change. Given the potential costs of eating novel foods, entering new habitats, or trying novel foraging techniques, environmental induction is likely to be a powerful driver of innovations. We thus feel that Ramsey et al.'s exclusion of environmentally induced innovations is too restrictive. The idea behind this exclusion seems to be borrowed from the ethnographic approach to animal culture. In that context, the exclusion is relevant. However, we think more harm than good can be done with this criterion in the study of innovations.
The operational definition of innovation relies on a series of assumptions that are not all self-evident. To recognize innovations in the field, for instance, you must first accept that innovations tend to arise more rarely than environmental induction. The number of individuals exhibiting the innovation tends to rise more slowly because the spread of the behavior depends, at least in part, on social learning. Moreover, innovation requires a longer delay from the moment the environment changes than environmental induction. Here, the more rapid the spread the more likely individuals have adopted the behavior independently, because social learning is assumed to be slower than the individual learning that is involved in environmental induction. These considerations ignore that the speed of acquisition of a novel behavior also depends on costs and benefits to the behavior and its alternatives, not simply on its means of diffusion. The assumptions used by Ramsey et al. are perhaps correct, but they are far from self-evident and must be the object of scientific scrutiny. Rarity is also important in identifying an innovation, but rarity must be defined as a function of species-specific traits such as degree of sociality and the amount of exploration, clearly not easy empirical measures. The use of such subjective criteria will prevent comparisons of data from different research programs and different taxa, thereby hampering comparative work on innovativeness.
How to avoid tautologies: The use of a neutral definition. In birds, accounts of innovation were first encouraged by Thorpe (Reference Thorpe1943; Reference Thorpe1956), who used the terms “unusual” and “original” to qualify these innovations. In primates, Goodall's section of the pioneering paper by Kummer and Goodall (Reference Kummer and Goodall1985) defines innovation as “a solution to a novel problem, or a novel solution to an old one”; a communication signal “not observed in other individuals in the group”; “a new ecological discovery such as a food item not previously part of the diet of the group” (p. 205, our italics). The first comparative studies of innovations used the terms italicised here as inclusion criteria in their databases. All subsequent comparative tests of neural, cognitive, ecological, or evolutionary correlates of innovation were based on these criteria.
In order to test a scientific prediction, it is important for the definitions of the variables involved in the test to be as neutral as possible with respect to the hypothesis. If this is not the case, the result can be tautological. We believe that the many conditions required by Ramsey et al. for a behavior to be considered innovative will lead to taxonomic biases, rendering invalid many of the tests the authors cite in their section 2 to justify the importance of innovations. Ramsey et al. recognize this danger when they say that “cognitively complex innovations . . . may largely be limited to the great apes” (sect. 5.2, para. 4). This may very well be true, but a statement like this should be the result of an objective test, not a possible artifact of a definition. Terms like “cognitively complex” can be vague and difficult to use objectively; the emphasis on intelligence can create a circular relationship with brain size.
We believe that the descriptors initially proposed by Thorpe and Goodall and used since then by others, are much more neutral and unbiased than the criteria proposed by Ramsey et al. What precision might be gained in the study of a few cases by the application of Ramsey et al.'s criteria will be lost through the biases introduced in comparative trends. The use of neutral key words and the reliance on the judgment of the initial observers of the innovation is bound to introduce some type 1 error. However, if this error is randomly distributed and/or can be quantified and removed (as population size, research effort, observer bias and other variables have been in studies of birds and primates), valid tests of predictions can still be made. This is unlikely to be the case with Ramsey et al.'s criteria.
