In their commentary, Rosati and Warneken (2016, p. 2) state the following about their original article (Warneken & Rosati, 2015):
Based on this whole sequence of experiments, we argued that chimpanzees show several of the psychological capacities that are necessary to cook food: motivation or desire to pursue cooked food; patience to wait temporal delays to acquire cooked foods as well as pay some additional energetic cost to transport food so it would be cooked; the self-control to give up food in one’s hand to have it be cooked; a basic causal comprehension of how these devices transformed raw foods after little experience; and finally the ability to save raw food for future cooking opportunities.
We agree that all of the points above were demonstrated, except one. We continue to disagree that the chimpanzees needed a “basic causal comprehension of how these devices transformed raw foods after little experience.” It remains as plausible that the chimpanzees associated one stimulus (the cooking device) with high preference foods, and were attracted to it (whether choosing it, putting things into it, etc.), without needing an understanding of the transformation process. As we noted in our original review, there is no clear reason to prefer a causal-understanding interpretation over an associatively learned explanation. This position does not diminish the “smart-looking” choices of these chimpanzees, it simply means that we remain agnostic as to the mechanism (or mechanisms) underlying their behaviors. The chimpanzees may fully understand that this device “cooked” food (in whatever way they might think about cooking), but these data in total do not more strongly support that conclusion than one grounded in more basic forms of learning, at least without some evidence of the capacity to create or innovate a novel technique to transform the food.
Rosati and Warneken (2016, p. 3) also state that “To our knowledge, there is no overarching theory that specifies the necessary and sufficient behavioral skills for organisms to cook. Therefore, our approach was to suggest and test for basic cognitive and behavioral skills whose absence would preclude cooking behaviors.” Cooking can be objectively defined as requiring some suite of behavioral and cognitive capacities, and some definitions do exist (e.g., Short, 2003, outlined the mechanical skills, perceptual abilities in terms of taste, color and texture of food, and knowledge of physics required to cook). We agree that chimpanzees demonstrate some of those capacities, but we do not agree that such premises necessarily lead to the conclusion that chimpanzees understand or appreciate cooking as a process that transforms the same item of food from one state (uncooked) to another (cooked), even if they can learn to obtain cooked foods through their behavioral responses.
In the present case, the argument goes like this: Take a behavior (such as cooking), break it into constituent pieces that are all necessary, and then use evidence of those components to argue that there is a capacity for that behavior (such as for cooking). The problem is that this is the logical error of affirming the consequent. Having some, or even all, of the requisite capacities that are part of a larger behavioral capacity or concept does not tell us anything about the necessary capability for that behavior or for knowledge of that concept. This argument only becomes invalidated when one finds a species that cannot exhibit self-control or planning, thereby necessarily being excluded from being capable of cooking (a point raised by Rosati & Warneken, 2016). Indeed, Rosati and Warneken acknowledged that simply showing constituent abilities that, when combined, could generate a complex behavior does not mean those constituent abilities would generate a complex behavior. Beyond which, we note (as do Rosati & Warneken, 2016) that it is unlikely that Warneken and Rosati (2015) identified all, or even the most important, factors that precipitated fire use and cooking in early hominids: As we noted in our original review, an unusual (even by primate standards) predilection toward information-seeking and curiosity, and exceptional episodic memory capacities, would also undoubtedly rate high on this list, as would factors such as food competition, tolerance, and cooperation, which would be necessary to delay the immediate consumption of captured food in wild chimpanzees.
A second point of disagreement between us and Warneken and Rosati (2015) comes from statements such as that chimpanzees have “a practical understanding of this basic cooking transformation after minimal experience.” Understanding a process—or even a “practical understanding,” which we assume means in a “folk psychology” kind of way—and learning about contingencies for behaviors are not the same thing. There can be very different interpretations of observed behaviors, and the burden is on experimenters to distinguish between them. This is not a criticism of the design of the original study or the effort given to execute it—which we know was monumental, given how hard it is to do this kind of research with chimpanzees—but a difference of opinion on the interpretation of the data.
We have outlined why eating high-preference foods, and why taking more familiar and low-preference foods to the cooker over taking unfamiliar things, are both explained by competing response strengths (Beran, Hopper, de Waal, Sayers, & Brosnan, 2015). Rosati and Warneken (2016) suggest that the speed with which the chimpanzees learned to use the “cooking pot” to get preferred cooked food might argue against an associative-learning interpretation, and this is a reasonable claim. However, we do not know how many trials chimpanzees need to learn such new associations or whether other species could pass these tests just as quickly. We are inclined to accept that chimpanzees can generalize what they understand about test apparatuses rather quickly. But that does not mean that such flexible and objectively “smart” behavior must rely at all on understanding the idea of cooking (in whatever form) versus knowing how to get the best thing possible and then structuring behavior to get that. Note that there is no disagreement about self-control or planning abilities; the issue is whether those demonstrations were sufficient for claiming an understanding of cooking behavior. Indeed, Rosati and Warneken argue that although each of our individual points is true, we lacked an overarching theory to explain their entire suite of data, a condition met by cooking. We hope we have made clear that we think that the best overall explanation for the (impressive) suite of skills demonstrated in the original study is the chimpanzees’ intelligent, flexible, and rapid learning, but not necessarily a “practical understanding of cooking” per se. The domain-general cognitive capacities required for cooking proposed by Warneken and Rosati (2015) could collectively allow for cooking—as they also would for many other behaviors (e.g., hunting and foraging)—but as we discussed above, simply demonstrating evidence for each (even collectively) does not allow for the conclusion that control of fire is the only missing element from chimpanzees cooking.
In tests of self-control, it is well-established that when presented with two options, whether quantitatively or qualitatively different, chimpanzees will point to the best thing (e.g., Boysen & Berntson, 1995). In the test given to chimpanzees by Warneken and Rosati (2015), the choice of the larger amount (three items over one item) was also what was called the self-controlled response. There is no way to distinguish in that choice whether the chimpanzees were choosing to wait, or wanted more food with no aspect of delay factoring into that choice. However, Warneken and Rosati reported a difference between one raw versus three raw items and one raw versus three cooked items. The reverse-reward paradigm (Boysen & Berntson, 1995) has consistently shown that the greater the discrepancy between options, the worse the subject performs. Given this, it would have been surprising to see anything other than the given outcome. We agree that the chimpanzees would more often point to three cooked pieces versus one raw piece than they would point to three raw pieces versus one raw piece, but we do not see this as necessarily reflecting anything about the degree of self-control they exhibited as a function of what they could obtain. Finally, we leave it to the reader to decide whether Clever Hans effects should no longer be considered a concern in comparative methodology (see Beran, 2012). We argue that it is important to control for possible inadvertent cues given to subjects, even if the use of such cues is not expected to occur in some tests or with some species.