In asking whether the representation of knowledge is more basic than the representation of belief, Phillips and colleagues presuppose that knowledge and belief are categorically distinct mental states. The authors also claim that what they call knowledge is clearly distinguishable from perceptual access to information, and that, therefore, the representation of another's knowledge is more than level-1 perspective taking. Although we remain unconvinced by this second claim, our focus rests on whether knowledge and belief are indeed categorically distinct.

The first two features put forward as essential for knowledge are that it must be (1) factive and (2) more than just true belief. In other words, you “know” something if you believe a truth for the right reasons. For instance, you can say you know that humans landed on the moon in 1969 if it is true and if you, say, watched the live transmission. If, however, you claim it was a spaceship headed for Orion that made an emergency landing on the moon, then your conviction that humans landed on the moon in 1969 would not count as knowledge, but as belief – and, well, rightly so.

We argue that the relationship between knowledge and belief is more complex and subtle. For instance, neither those who “believe” the 1969 lunar landing was accidental (because of an emergency en route to Orion) nor those who “know” it was pre-planned and intentional are knowing these things. We might know who told us about the moon landing, but we would have no way of knowing whether their account as such is true, and the “live transmission” some of us saw could have been faked. In fact, most of us cannot even know for sure that Earth is round. Although flat-Earth proponents at least have face-evidence to back them up, we others hold our conviction to be true simply because we believe that the people who told us knew. This trust in others' knowledge is our quintessential “evidence” not only for hard-to-verify facts such as moon landings, but for almost everything we take for granted (Bender & Beller, Reference Bender and Beller2019). The lion's share of our common knowledge is nothing else than belief – often plain, unverified belief – adopted from others, and hence culturally conveyed (Gatewood, Reference Gatewood, Kronenfeld, Bennardo, de Munck and Fischer2011). Distributed knowledge and cultural transmission are key mechanisms in the process that makes human cognition unique (Bender Reference Bender2020a; Tennie, Call, & Tomasello, Reference Tennie, Call and Tomasello2009), but they come at the price of us having to trust without personal verification that what we believe we know is actually true. On the contrary, cultural consensus on agreed-upon knowledge is never complete, and diminishing consensus is one driving force for the emergence of distinct (sub)cultural truths (Gatewood, Reference Gatewood2012).

Largely because of this cultural imprint, knowing and believing are part of a gradient rather than a simple dichotomy, contingent on the degree of uncertainty involved. Importantly, this gradient may be captured linguistically, with numerous languages even reflecting the role of cultural transmission in a much more nuanced way than English and French, the examples considered in the target article. As much as a quarter of the world's languages must qualify stated knowledge through a grammatical category called evidentiality. That is, speakers of these languages are obliged to specify, for every sentence they utter, the source of their information (Aikhenvald & Dixon, Reference Aikhenvald and Dixon2014; Chafe & Nichols, Reference Chafe and Nichols1986), for instance whether the speaker has gained the information personally or from someone else, through direct observation, by inference, from hearsay, or assuming (Aikhenvald, Reference Aikhenvald2004). Besides disqualifying English and French as sources of evidence for a universal distinction between belief and knowledge, the obligatory marking of evidentiality in many of the world's languages may also have implications for their speakers' willingness to engage in subjective activities (Luhrmann, Reference Luhrmann2011).

Junín Quechua, for instance, contains grammatical markers for indicating the source of information as being direct evidence (having seen it), conjectural, or hearsay. And, although speakers of this language make extensive use of vocabulary for talking about how things appear to be, mentalistic vocabulary is basically absent. In line with this distinct pattern of conversational topics, Junín Quechua children pass tests on representational changes and false beliefs significantly later than they pass tests on appearance-reality distinctions (Vinden, Reference Vinden1996; for more evidence of cross-cultural variability in the onset, unfolding, and pervasiveness of mental-state reasoning, see also Lillard, Reference Lillard1998; Luhrmann, Reference Luhrmann2011; Mayer & Träuble, Reference Mayer and Träuble2013, Reference Mayer and Träuble2015; Robbins & Rumsey, Reference Robbins and Rumsey2008; Träuble, Bender, & Konieczny, Reference Träuble, Bender, Konieczny, Wassmann, Träuble and Funke2013).

In other words, cultural conventions and linguistic practices for defining “knowledge” affect the readiness with which people ascribe distinct mental states to other people. This generates substantial variability both in behaviors indicative of theories of mind and in teaching (Bender, Reference Bender2019, Reference Bender2020b) and has implications for the authors' evolutionary account, according to which “the basic capacity for knowledge representation evolved for learning from others.” Although we agree that knowledge in the form of level-1 perspective taking is equally beneficial for human and nonhuman primates alike, the capacity critical for human teaching and cumulative culture would be level-2 perspective taking. Even when concerned with facts, efficiency of teaching increases with the ability to diagnose false beliefs in the learner. Unarguably, however, human teaching is even more strongly concerned with conveying beliefs, values, and practices, the high-fidelity copying of which serves to strengthen group cohesion (Legare & Nielsen, Reference Legare and Nielsen2015). To restate a claim from the target article more precisely, “we teach others (and expect them to learn) about what we know” – and especially about what we believe.

In conclusion, not only is the relationship between knowledge and belief more intricate than purported in the target article (see also Polanyi, Reference Polanyi1958), but humans have also evolved to appreciate the subtleties. Indeed, contemporary, fully enculturated humans have developed cultural as well as cognitive means to handle such subtleties with stunning finesse.