In a classic Monty Python sketch, scientists are embroiled in the age-old question of whether penguins are more intelligent than humans. Not surprisingly, the penguins failed human IQ tests. The penguins were then given cognitive tests in a naturalistic zoo enclosure, to which they gave no meaningful responses. Non-English-speaking Swedes in the same enclosure responded in a similar vein to the penguins. But how will we ever be able to ascertain whether species differences in cognition are real when the methodology used is automatically biased towards humans?

The cognitive differences between human and nonhuman animal minds suggested by Penn et al. are without exception impossible to quantify because of the reliance on language in experiments of human cognition. We are not aware of any cognitive tests given to adult humans that have been designed and implemented in exactly the same way as those given to animals. Boesch (Reference Boesch2007) addressed this issue with regard to comparisons between humans and other apes. For example, humans are given instructions in their appropriate language, and their performance is determined by their language competency; in tests of social cognition, humans are used as stimuli and caregivers are present when testing human infants. We desperately need more studies on the cognitive capacities of humans applying the same conditions as used for animals before we can begin to qualify impossible statements such as “evidence of absence.”

As with their previous critiques of the comparative literature, Penn et al. have been very selective in their choice of experiments to evaluate. The crux of their argument is that there is no evidence that any nonhuman animal represents unobservable states or relations of relations, so the choice of studies becomes important. Although we are glad that work on corvids is now being discussed, we would like to address a number of misinterpretations, absences, and misrepresentations of our work in the target article. We focus on two studies; the two trap-tube task in rooks (Seed et al. Reference Seed, Tebbich, Emery and Clayton2006) and experience projection by scrub-jays (Emery & Clayton Reference Emery and Clayton2001).

First, Penn et al. were quick to dismiss the results of the two trap-tube task without going into the essential details. Rooks do not use tools, and how could their performance on a tool-using task be due to “domain-specific expectations”? Seven rooks rapidly learned one of two configurations of the two trap-tube task, with a functional and a nonfunctional trap and transferred immediately to another, novel configuration. The two previously rewarded versions of the nonfunctional traps (i.e., no bottom so food falls through and the bottom raised so food passes over the trap) were then pitted against one another, so that either choice could lead to reward. However, a manipulation of the whole apparatus was performed which transformed one nonfunctional trap into a functional trap. Either the ends of the tube were blocked with rubber bungs so that the food would be trapped at one end or the whole apparatus was lowered so that food would fall into a now functional trap. At issue is the performance of one rook, Guillem, who instantly understood the problem, performing without errors on one version of the task and 90% correctly on the other.

We would like to discuss a second study investigating the ability of scrub-jays to re-cache food which they had previously cached in front of a conspecific, but only when they had previous experience of stealing another bird's caches; which we have suggested is an example of experience projection (Clayton et al. Reference Clayton, Dally and Emery2007; Emery Reference Emery and Watanabe2004; Emery & Clayton Reference Emery, Clayton, Rogers and Kaplan2004a). To selectively re-cache, we suggest that the birds will have had to retrieve a recent memory of whether they had been observed caching (as they do not re-cache if they had cached in private), which would be integrated with “knowledge” of what to do to avoid future pilferage, derived from their egocentric experience of pilferage. Penn and Povinelli (Reference Penn and Povinelli2007b) provided an alternative explanation, suggesting that the jays were behaving like monkeys who had been attacked and who subsequently redirected aggression to a third-party related to their attacker. We do not agree. In the monkey case, the victim receives aggression and then shortly afterwards attacks another, thereby reacting to an immediate past when in the same emotional state. By contrast, one group of scrub-jays sees another bird caching and then pilfers those caches, whereas a second group observes caching, but has no opportunity to steal those caches. Then months later, those same scrub-jays are given the opportunity to cache either when observed or when alone and then, three hours later, to recover those caches. Only the birds with pilfering experience re-hid their caches in new sites. We propose that re-caching functions to prevent future theft, and thus the most parsimonious explanation for why only those birds with pilfering experience do this is because they project this experience onto another bird and predict their future behaviour based on their “knowledge” of pilfering. Note that the scrub-jays cannot be basing their decisions on their emotional state, either three hours previously (as both groups should be in the same emotional state) or months previously.

Imagination and prospection are useful concepts to approach the problem of thought in nonverbal creatures. Indeed, it is impossible to think about the future (by its nature unobservable) without forming a “picture in the mind's eye.” As recent studies in scrub-jays and apes have suggested (Correia et al. Reference Correia, Dickinson and Clayton2007; Mulcahy & Call Reference Mulcahy and Call2006a; Raby et al. Reference Raby, Alexis, Dickinson and Clayton2007), nonhuman animals may think about alternative futures outside the realm of perception. We believe that these complex processes should not be neglected in the type of cognitive architectures discussed by Penn et al.; indeed, we have argued that planning, imagination, and prospection can be included in such models (Emery & Clayton, in press). Imagination and planning have recently been included in a neurally inspired cognitive architecture derived from global workspace theory (Shanahan Reference Shanahan2006; Shanahan & Baars Reference Shanahan and Baars2005) based on the mammalian brain which could be applied to other animals.