Phillips et al. argued that knowledge attribution is “evolutionary more foundational” than belief attribution, with the former present in monkeys and apes and the latter occurring chiefly in humans. Additionally, Phillips et al. argued that these two theory of mind (ToM) skills are independent from one another. These propositions may seem sensible to comparative psychologists as many (but not all) previous studies in this field have produced evidence for knowledge attribution but not belief attribution in nonhuman animals. However, recent evidence suggests that such a characterization might be too simplistic and overstated. Below, we examine this evidence and its implications by focusing on two meanings of evolutionary foundations of knowledge and belief and conclude that it is conceivable that great apes and macaques have both knowledge and belief representations.

One meaning of “evolutionary more foundational” refers to the temporal emergence of the skills in evolutionary time. Phillips et al. propose that knowledge attribution, which humans share with nonhuman primates, is more ancient than belief attribution, which only humans possess. However, four recent studies with nonhuman primates cast some doubt on this idea (Buttelmann, Buttelmann, Carpenter, Call, & Tomasello, Reference Buttelmann, Buttelmann, Carpenter, Call and Tomasello2017; Hayashi et al., Reference Hayashi, Akikawa, Kawasaki, Egawa, Minamimoto, Kobayashi and Hasegawa2020; Kano, Krupenye, Hirata, Tomonaga, & Call, Reference Kano, Krupenye, Hirata, Tomonaga and Call2019; Krupenye, Kano, Hirata, Call, & Tomasello, Reference Krupenye, Kano, Hirata, Call and Tomasello2016). For instance, in the so-called anticipatory-looking false-belief (AL-FB) tests, apes anticipated that an agent will go to the location where the agent falsely believed an object to be. Phillips et al. minimized those findings from AL-FB tests because of the low-level alternatives proposed by some authors, namely that apes “submentalize” (Heyes, Reference Heyes2017) or “see the last location that the agent saw” (Scarf & Ruffman, Reference Scarf and Ruffman2017). However, these alternative explanations were carefully examined and ruled out by subsequent studies with great apes (Kano, Krupenye, Hirata, Call, & Tomasello, Reference Kano, Krupenye, Hirata, Call and Tomasello2017; Kano et al., Reference Kano, Krupenye, Hirata, Tomonaga and Call2019). Notably, Hayashi et al. (Reference Hayashi, Akikawa, Kawasaki, Egawa, Minamimoto, Kobayashi and Hasegawa2020) recently showed that macaques also pass this AL-FB test, and that inactivation of the macaque medial prefrontal cortex, one of the key regions that support human-adult ToM, disrupted their performance (Hayashi et al., Reference Hayashi, Akikawa, Kawasaki, Egawa, Minamimoto, Kobayashi and Hasegawa2020). Thus, Phillips et al.'s claim about belief attribution phylogenetically preceding knowledge attribution based on primate data needs to be reconsidered. Phillips et al.'s misgivings about AL-FB data may be partly motivated by recent replication issues in the AL-FB tests with human infants (Kulke & Rakoczy, Reference Kulke and Rakoczy2018) and because the primate studies were inspired by studies with human infants (e.g., Southgate, Senju, & Csibra, Reference Southgate, Senju and Csibra2007), they have also come under scrutiny (Horschler, MacLean, & Santos, Reference Horschler, MacLean and Santos2020). However, the primate work departed from the original versions by introducing several key methodological changes aimed at optimizing the test for nonhuman primates, and the results have been replicated with two different groups of apes and one group of monkeys (Kano, Call, & Krupenye, Reference Kano, Call and Krupenye2020).

Another meaning of “evolutionary more foundational” refers to one skill being simpler than the other (i.e., less cognitively demanding) and this could explain why in comparative studies is easier to obtain positive results in the knowledge-ignorance than false-belief conditions in traditional nonverbal ToM tests. However, the vast majority of comparative findings come from variations of only two main paradigms: the food-competition in the laboratory (e.g., Kaminski, Call, & Tomasello, Reference Kaminski, Call and Tomasello2008) and the violation-of-expectation tests in the wild (e.g., Marticorena, Ruiz, Mukerji, Goddu, & Santos, Reference Marticorena, Ruiz, Mukerji, Goddu and Santos2011). Although those two paradigms have shown that apes and monkeys attribute knowledge to others, they might not be suitable to detect false-belief attribution because of certain inherent design limitations (as in any single paradigm). How stimuli are presented, how responses are measured as well as the task's motivational substrate can impact performance. For instance, primate violation-of-expectation tests have presented agents performing relatively simple actions, whereas primate AL-FB tests have presented stories in videos illustrating dynamic social interaction between an agent and an antagonist, which may be more intuitively appealing to highly social primates (Kano et al., Reference Kano, Call and Krupenye2020). Studies that have abandoned the two main paradigms and their inherent limitations have started to produce different results. In fact, in a recent study, apes not only pass false-belief conditions, but also did not find them harder than true-belief or knowledge-ignorance conditions (Buttelmann et al., Reference Buttelmann, Buttelmann, Carpenter, Call and Tomasello2017; also see Buttelmann, Carpenter, & Tomasello, Reference Buttelmann, Carpenter and Tomasello2009), which casts some doubt on the idea that false-belief conditions are invariably harder than other epistemic conditions. We are not suggesting that AL-FB tests can better capture nonhuman primate ToM in general, but we think it unlikely that a couple of paradigms would be sufficient to capture the full range of socio-cognitive skills that primates deploy in social interaction, particularly knowing that paradigm changes have historically brought substantial empirical and conceptual advances. In keeping with this progress, future studies should investigate whether the notion of aspectuality is part of belief attribution in apes (Low & Watts, Reference Low and Watts2013).

We liked the authors discussion about the potential functions of knowledge and belief attribution, with the former being particularly useful for learning from others, and the latter for predicting others' behavior, particularly when, unbeknownst to the subjects, the situation has changed. For nonhuman primates, both learning from others and predicting others' behavior should be important, and therefore both knowledge and belief representations can be adaptive. Perhaps for young (human and nonhuman) infants that are dependent on adults, learning from others is more important than predicting others' behavior, but one cannot argue that, for (human and nonhuman) adults, the latter is less important than the former. Imagine, for example, the situation in which an orangutan mother fails to anticipate her child's travel path in a dense forest where visibility is limited; when the child is traveling as usual and did not see (but the mother saw) a branch on which he usually walks was broken. It may be precisely such a situation that critically matters to nonhuman primates – that could happen in their natural lives and affect their fitness. Future studies should endeavor to make the test situations even more ecologically (or ethologically) valid to uncover further elements of belief attribution in primates.