The target article (Phillips et al.) presents impressively diverse evidence for carving species-wide mental attitude attribution into two principal representational capacities: ascription of factive knowledge and non-factive beliefs. Although we find evidence for the primacy of knowledge compelling, we believe there is room to carve the theoretical space further into a diverse representational spectrum that characterizes the development and evolution of factive and non-factive theory of mind. Our strategy, similar to others' (Nagel, Reference Nagel2017; Phillips & Norby, Reference Phillips and Norby2021), is to start from the attributor's egocentric representation of reality and take minimal steps to build toward more charitable inclusion of the altercentric agent. We then address the case of great ape mindreading.
Knowledge attribution may pose the least extra demand on an attributor's egocentric representation. The attributor can get away with simply extending her personal representation to an altercentric agent. Presumably, noting the agent's presence and her epistemic contact (perceptual or inferential) with something in the attributor's egocentric reality is enough. From this, the attributor can generate a positive attribution of mental state with some content and, hence, a degree of separation between the ego and altercentric. The behavioral expectation for the agent will be based on this generalization of the egocentric reality.
Altercentric ignorance describes a situation in which the agent fails to witness something in the attributor's egocentric representation. The attributor has at least two ways of handling this situation. Agents' lack of epistemic contact with some event (e.g., “she did not see or could not know x”) may amount to no attribution and no behavioral expectation. Alternatively, the attributor can omit from her representation what the agent “missed” and attribute that modified representation (Phillips & Norby, Reference Phillips and Norby2021). Phillips and Norby (Reference Phillips and Norby2021) qualify ignorance as knowledge-like, with both requiring the attributor to hold the personal representation and an altercentric copy in parallel. Martin and Santos (Reference Martin and Santos2016) qualified this separation of the ego and altercentric as already akin to false beliefs, and suggested, in contrast, that the most basic response to ignorance is a simple failure to track the altercentric. Both characterizations of ignorance constitute forms of mindreading with different levels of representing the altercentric.
Egocentric ignorance constitutes cases in which the agent has privileged epistemic contact in comparison with the attributor – for example, a student, nervous about her exam results in an email, asks her friend to read it on her behalf. The student knows that her friend knows the results, but not what results. In this case, the attributor can no longer rely on her egocentric reality to handle the agent's. Hence, the egocentric ignorance situation requires a new way to represent the agent's altercentricity – the agents no longer succeeds or fails to know what the attributor knows but rather they have epistemic opportunities of their own. In egocentric ignorance cases, the content of the agent's attitude may arise independent of the attributor's reality but, critically, it is not specified (because the attributor does not know it). It is, therefore, not quite belief-like in the sense that there is no contrast in content of the egocentric and altercentric representation that could be interpreted as counterfactual (sensu Phillips & Norby, Reference Phillips and Norby2021). Egocentric ignorance, thus, constitutes an intermediate representational sophistication – more than knowledge, shy of belief.
False-belief representation is the first situation in which the altercentric representation must not only be held as independent, but the content of this altercentric state must also be updated dynamically. The attributor first tracks the agent as knowledgeable (that “object is in box A”; e.g., as they co-witness a hiding event), then ignorant (that “object is not in box A, object is in box B”; e.g., as the object is displaced in the agent's absence).
True belief representation, at least as exemplified in the Gettier case (Gettier, Reference Gettier1963), constitutes an even more demanding computational task. Here, agents co-witness an object's initial hiding and then, in the agent's absence, the object is temporarily removed and then returned to the same location (Horschler, Santos, & MacLean, Reference Horschler, Santos and MacLean2019; Kaminski, Call, & Tomasello, Reference Kaminski, Call and Tomasello2008). Consequently, the requisite updating process involves reconciliation of two contrasting altercentric mental states (knowledge and ignorance) with identical content (object is in location A).
We can now sketch a “scale” of cognitive complexity across the factive and non-factive mental attribution spectrum. Provided she can individuate agents, a hypothetical mindreading attributor can, at the most basic level, track epistemic contact or lack thereof. Egocentric reality alone is sufficient for tracking this minimal notion of knowledge and ignorance. Next, she can attribute copies of her egocentric information, and, in the case of altercentric ignorance, modify them. In egocentric ignorance contexts, the altercentric agent is granted a new epistemic status without the cognitive cost of attributing fully specified content to her. For belief tracking, the content of this altercentric representation is specified and updated. This process is even more computationally challenging in paradigmatic cases of true belief attribution. Finally, aspectuality may (or may not) impose further representational complexity (e.g., Perner, Huemer, & Leahy, Reference Perner, Huemer and Leahy2015; but see Rakoczy, Bergfeld, Schwarz, & Fizke, Reference Rakoczy, Bergfeld, Schwarz and Fizke2015).
If the present hierarchical characterization is correct, the predictions are clear. Children or animals proficient in more sophisticated abilities should master simpler ones (e.g., Fig. 1 of Krupenye, Reference Krupenye, Gilead and Ochsner2020). Great apes, for example, have shown competence on several recent false-belief tasks (Buttelmann, Buttelmann, Carpenter, Call, & Tomasello, Reference Buttelmann, Buttelmann, Carpenter, Call and Tomasello2017; Kano, Krupenye, Hirata, Tomonaga, & Call, Reference Kano, Krupenye, Hirata, Tomonaga and Call2019; Krupenye, Kano, Hirata, Call, & Tomasello, Reference Krupenye, Kano, Hirata, Call and Tomasello2016), raising the possibility that they may indeed track beliefs. Current evidence is tentatively consistent with the proposed complexity scale, suggesting that apes also track knowledge and altercentric ignorance (Hare, Call, & Tomasello, Reference Hare, Call and Tomasello2006; Karg, Schmelz, Call, & Tomasello, Reference Karg, Schmelz, Call and Tomasello2015), and potentially egocentric ignorance (Call & Tomasello, Reference Call and Tomasello1999; Krachun, Carpenter, Call, & Tomasello, Reference Krachun, Carpenter, Call and Tomasello2009), but perhaps not Gettier's true beliefs (Kaminski et al., Reference Kaminski, Call and Tomasello2008). Cases of egocentric ignorance, in particular however, deserve further, more targeted tests. Broader efforts in humans and nonhumans also demand new tasks that carefully tease apart attribution of knowledge and true belief, and of knowledge, egocentric ignorance, and false belief. Together, these developments will clarify the family, or hierarchy, of factive and non-factive theory of mind.
You have
Access
The target article (Phillips et al.) presents impressively diverse evidence for carving species-wide mental attitude attribution into two principal representational capacities: ascription of factive knowledge and non-factive beliefs. Although we find evidence for the primacy of knowledge compelling, we believe there is room to carve the theoretical space further into a diverse representational spectrum that characterizes the development and evolution of factive and non-factive theory of mind. Our strategy, similar to others' (Nagel, Reference Nagel2017; Phillips & Norby, Reference Phillips and Norby2021), is to start from the attributor's egocentric representation of reality and take minimal steps to build toward more charitable inclusion of the altercentric agent. We then address the case of great ape mindreading.
Knowledge attribution may pose the least extra demand on an attributor's egocentric representation. The attributor can get away with simply extending her personal representation to an altercentric agent. Presumably, noting the agent's presence and her epistemic contact (perceptual or inferential) with something in the attributor's egocentric reality is enough. From this, the attributor can generate a positive attribution of mental state with some content and, hence, a degree of separation between the ego and altercentric. The behavioral expectation for the agent will be based on this generalization of the egocentric reality.
Altercentric ignorance describes a situation in which the agent fails to witness something in the attributor's egocentric representation. The attributor has at least two ways of handling this situation. Agents' lack of epistemic contact with some event (e.g., “she did not see or could not know x”) may amount to no attribution and no behavioral expectation. Alternatively, the attributor can omit from her representation what the agent “missed” and attribute that modified representation (Phillips & Norby, Reference Phillips and Norby2021). Phillips and Norby (Reference Phillips and Norby2021) qualify ignorance as knowledge-like, with both requiring the attributor to hold the personal representation and an altercentric copy in parallel. Martin and Santos (Reference Martin and Santos2016) qualified this separation of the ego and altercentric as already akin to false beliefs, and suggested, in contrast, that the most basic response to ignorance is a simple failure to track the altercentric. Both characterizations of ignorance constitute forms of mindreading with different levels of representing the altercentric.
Egocentric ignorance constitutes cases in which the agent has privileged epistemic contact in comparison with the attributor – for example, a student, nervous about her exam results in an email, asks her friend to read it on her behalf. The student knows that her friend knows the results, but not what results. In this case, the attributor can no longer rely on her egocentric reality to handle the agent's. Hence, the egocentric ignorance situation requires a new way to represent the agent's altercentricity – the agents no longer succeeds or fails to know what the attributor knows but rather they have epistemic opportunities of their own. In egocentric ignorance cases, the content of the agent's attitude may arise independent of the attributor's reality but, critically, it is not specified (because the attributor does not know it). It is, therefore, not quite belief-like in the sense that there is no contrast in content of the egocentric and altercentric representation that could be interpreted as counterfactual (sensu Phillips & Norby, Reference Phillips and Norby2021). Egocentric ignorance, thus, constitutes an intermediate representational sophistication – more than knowledge, shy of belief.
False-belief representation is the first situation in which the altercentric representation must not only be held as independent, but the content of this altercentric state must also be updated dynamically. The attributor first tracks the agent as knowledgeable (that “object is in box A”; e.g., as they co-witness a hiding event), then ignorant (that “object is not in box A, object is in box B”; e.g., as the object is displaced in the agent's absence).
True belief representation, at least as exemplified in the Gettier case (Gettier, Reference Gettier1963), constitutes an even more demanding computational task. Here, agents co-witness an object's initial hiding and then, in the agent's absence, the object is temporarily removed and then returned to the same location (Horschler, Santos, & MacLean, Reference Horschler, Santos and MacLean2019; Kaminski, Call, & Tomasello, Reference Kaminski, Call and Tomasello2008). Consequently, the requisite updating process involves reconciliation of two contrasting altercentric mental states (knowledge and ignorance) with identical content (object is in location A).
We can now sketch a “scale” of cognitive complexity across the factive and non-factive mental attribution spectrum. Provided she can individuate agents, a hypothetical mindreading attributor can, at the most basic level, track epistemic contact or lack thereof. Egocentric reality alone is sufficient for tracking this minimal notion of knowledge and ignorance. Next, she can attribute copies of her egocentric information, and, in the case of altercentric ignorance, modify them. In egocentric ignorance contexts, the altercentric agent is granted a new epistemic status without the cognitive cost of attributing fully specified content to her. For belief tracking, the content of this altercentric representation is specified and updated. This process is even more computationally challenging in paradigmatic cases of true belief attribution. Finally, aspectuality may (or may not) impose further representational complexity (e.g., Perner, Huemer, & Leahy, Reference Perner, Huemer and Leahy2015; but see Rakoczy, Bergfeld, Schwarz, & Fizke, Reference Rakoczy, Bergfeld, Schwarz and Fizke2015).
If the present hierarchical characterization is correct, the predictions are clear. Children or animals proficient in more sophisticated abilities should master simpler ones (e.g., Fig. 1 of Krupenye, Reference Krupenye, Gilead and Ochsner2020). Great apes, for example, have shown competence on several recent false-belief tasks (Buttelmann, Buttelmann, Carpenter, Call, & Tomasello, Reference Buttelmann, Buttelmann, Carpenter, Call and Tomasello2017; Kano, Krupenye, Hirata, Tomonaga, & Call, Reference Kano, Krupenye, Hirata, Tomonaga and Call2019; Krupenye, Kano, Hirata, Call, & Tomasello, Reference Krupenye, Kano, Hirata, Call and Tomasello2016), raising the possibility that they may indeed track beliefs. Current evidence is tentatively consistent with the proposed complexity scale, suggesting that apes also track knowledge and altercentric ignorance (Hare, Call, & Tomasello, Reference Hare, Call and Tomasello2006; Karg, Schmelz, Call, & Tomasello, Reference Karg, Schmelz, Call and Tomasello2015), and potentially egocentric ignorance (Call & Tomasello, Reference Call and Tomasello1999; Krachun, Carpenter, Call, & Tomasello, Reference Krachun, Carpenter, Call and Tomasello2009), but perhaps not Gettier's true beliefs (Kaminski et al., Reference Kaminski, Call and Tomasello2008). Cases of egocentric ignorance, in particular however, deserve further, more targeted tests. Broader efforts in humans and nonhumans also demand new tasks that carefully tease apart attribution of knowledge and true belief, and of knowledge, egocentric ignorance, and false belief. Together, these developments will clarify the family, or hierarchy, of factive and non-factive theory of mind.
Financial support
KD was supported by UK Economic and Social Research Council studentship 2267016 and a St Leonards Research Scholarship from the University of St Andrews, and CK by European Commission Marie Sklodowska-Curie fellowship MENTALIZINGORIGINS.
Conflict of interest
None.