We are sympathetic to the author's focus on “understanding others' minds in relation to the actual world” (p. 2), because it recognizes that references to worldly facts rather than mental states are primary in explaining actions. However, the empirical evidence cited does not support their main claim “knowledge before belief.” Our criticism can be put in terms of a dilemma: (1) If the authors do not employ the same notion of knowledge throughout the paper, then they do not compare like with like. Although the intended focus of the argument is on declarative knowledge-that, some passages (sects 4.1 and 6.1) employ a wider notion including less demanding knowledge-how. Consequently, the authors should also consider the possibility of a wider, non-propositional form of belief. (2) If they do only employ knowledge-that, then their claim is not supported by the evidence, which is better explained by more basic means. (3) A superior teleological account grounds action explanation in an appreciation of an agent pursuing a goal.

1. (1) The authors specify four features “essential to knowledge” (p. 4), but do not elaborate on their operative conception of belief. Because they assess ascription of belief via understanding false belief, they assume a demanding notion of belief throughout, without considering a wider notion setting weaker constraints on belief. The features of knowledge mentioned are not characteristic of knowledge-how. Knowledge-how to swim can be mastered better or worse but does not amount to knowing a fact. By analogy, a wider notion of belief could be characterized as a minimally structured informational state that can be systematically connected to motivational states (see Newen & Starzak, Reference Newen and Starzak2020). Comparing like with like suggests a parallelism between knowledge and belief. If we ascribe propositional knowledge-that, we employ a concept that has “merely” believing-that as a fallback option, such that both unfold as a package from more basic roots.

2. (2) Although the authors formulate their central claim concerning the primacy of propositional knowledge-that, most of the evidence mentioned in section 4 can be explained in terms of perceptual access and/or knowledge-how. Understanding others in terms of perceptual access is simpler because it is immediately situation-based. To understand that Eve reaches for something because she sees it, I can rely on her line of sight. This explains the study by Krachun, Carpenter, Call, and Tomasello (Reference Krachun, Carpenter, Call and Tomasello2009) without recourse to an attribution of knowledge. The chimpanzees learn to look for the food where their human competitors look for it because they recognize that their competitors could observe both where the food was placed and the switch of the containers. Similarly, the experiment by Luo and Johnson (Reference Luo and Johnson2009) modulates an agent's perceptual access to information and does not warrant attribution of knowledge-that. Furthermore, the studies by Behne, Liszkowski, Carpenter, and Tomasello (Reference Behne, Liszkowski, Carpenter and Tomasello2012) and Kovács, Tauzin, Téglás, Gergely, and Csibra (Reference Kovács, Tauzin, Téglás, Gergely and Csibra2014) do not license the conclusion that infants attribute to others knowledge-that which they lack. Infants may simply recognize that the other agent “encountered” an object (Butterfill & Apperly, Reference Butterfill and Apperly2013) or that she knows how to acquire, interact with, or refer to an object. This suggests that there is a basic capacity for representing knowledge-how without yet involving knowledge-that. In the first instance, we learn from others how to do things, not facts. This is borne out by the evidence provided in section 6.2: Nonhuman primates gain knowledge-how to forage or to solve problems; and in order to learn children turn to competent adults, that is, to adults with know-how.

3. (3) Regarding the question of how we explain others' actions, there is an even simpler alternative, ignored by the authors. “Teleology” postulates a basic way of understanding simple actions. Developmental evidence suggests that young infants ground their expectations about people's actions in perceived objects and facts and that adults continue to use such explanatory strategies wherever it proves sufficient (Gergely & Csibra, Reference Gergely and Csibra2003; Perner & Roessler, Reference Perner, Roessler, Aguilar and Buckareff2010). Facts are reasons for action (Alvarez, Reference Alvarez2010). The structure of the information that the interpreter attributes to the agent is not yet differentiated into attributing states of believing and states of knowing (see Perner & Esken, Reference Perner, Esken and Barrouillet2015; Perner & Roessler, Reference Perner and Roessler2012; Roessler & Perner, Reference Roessler, Perner, Baron-Cohen, Tager-Flusberg and Lombardo2013). By understanding an agent's action as a means of pursuing a goal, the interpreter recognizes facts motivating the agent's actions without representing her mental states. Only at around the age of four do children learn to appreciate that other agents can relate to facts differently from their own perspective, manifest, for example, in the attribution of false beliefs (Wimmer & Perner, Reference Wimmer and Perner1983).

If the authors employ a wide notion of knowledge including knowledge-how, then teleology is more informative by bringing in facts and goal-directed actions. If the authors employ a demanding knowledge-that account, then teleology offers more parsimonious explanations of the relevant data, without relying on a demanding understanding of either belief or knowledge.

The teleological approach also has more specific advantages over a knowledge-first account:

1. (a) It captures both informational and motivational aspects of action understanding (Glock & Schmidt, Reference Glock, Schmidt and Schumann2019) and allows for further developmental steps toward a more sophisticated understanding of others.

2. (b) It offers a realistic conception of how humans eventually arrived at ideas such as “belief” and “knowledge.” Animals can act for reasons – on account of facts – without understanding that others have such reasons too (Glock, Reference Glock2019). Similarly, one need not represent an agent's mental state in order to understand that she acts on account of facts.

3. (c) It allows action predictions in cases where appealing to mental states (knowledge or belief) lacks warrant. Where will my colleague be at 4pm today? No idea what goes through her mind, but at 4pm is our faculty meeting. Because this fact provides reasons for her to attend, I predict she'll be there.

4. (d) Teleology provides an explanation for the “reality error,” which is more economical than that of the knowledge-first approach. Children anticipate that someone will look for the object in its real location because he or she has objective reasons to do so.

In short, instead of “knowledge first,” we suggest “teleology first,” that is, sensitivity to facts is fundamental.