The question of whether propositional attitudes (PA), especially beliefs, can be considered natural kinds becomes more pressing in the context of recent popularity of superficialist accounts of PA. Authors from the superficialist camp argue that concepts of PA could never prove to ‘carve nature at its joints’. That is, they could not be causally efficacious, concrete mental states held by subjects independently from observer. Instead, they are claimed to be context-dependent states attributed by interpreters depending on their pragmatic interests (Currie 2021; Schwitzgebel 2021; Posłajko 2024).

Before accepting the claim that PA cannot form natural kinds, one should consider more closely what conditions PA should fulfil in order to be regarded as legitimate candidates for natural kinds. This issue has not been often discussed explicitly (but see Perez 2004; Posłajko 2024). This paper is supposed to provide a more exhaustive analysis of possible readings of the view that PA form natural kinds.

I argue that, contrary to what superficialists claim, folk psychological ascriptions of PA carry substantial ontological commitments. The implicit assumption behind folk psychological practices is that the attributions of PA, if done accurately, provide explanations of the target’s observable behaviour in terms of concrete, unobservable mental causes. They assume, moreover, that these attributions are made on the basis of ceteris paribus generalisations that describe the nature of the mental states being ascribed. Since the theoretical purpose of psychology is to reveal the laws governing human mental states, the folk psychological categories can be considered as referring to the same mental states to which psychologists refer. Both lay people and psychologists talk about the same states even if they associate different descriptions with the terms they use. The former, presumably, defer to the latter the task of revealing the underlying nature of psychological kinds and, possibly, correcting the folk psychological generalisations. If such an account proves plausible, then they pose a threat for superficialist accounts of PA.

The proposed account of PA shall be situated in the context of competing theories of natural kinds. In the past decades there was a vivid discussion about natural kinds in philosophy of science. It started from radical physicalist (Lewis 1983) and essentialist (Putnam 1975) theories to give rise, later on, to more pluralistic theories. Lewis’s physicalist account was to the effect that only fundamental physical kinds can be considered natural kinds simpliciter. All the other kinds posited by special sciences can be more or less imperfectly natural depending on whether - and how smoothly - they can be reduced by means of definitions to the perfectly natural, physical kinds. Mental kinds are not good candidates for natural kinds when considered in light of Lewis’ criteria, chiefly due to their multiple realizability. However, there are independent reasons not to consider Lewis’ account plausible, having to do with questionable status of physicalist premises and non-reducibility of natural kinds posited by various special sciences, e.g. biology.

Natural kind essentialism in its traditional formulation does not have to be reductive, but it requires essential properties of a kind to be intrinsic properties. Since this requirement is problematic not only with regard to the posits of social sciences which define their terms by means of historical, relational and functional properties, but also to paradigmatic biological kinds, it has been claimed by many authors to be unduly restrictive. A possible strategy to preserve essentialism while admitting those latter kinds as natural kinds is to accept essentialistic definitions involving extrinsic, apart from intrinsic, properties (Okasha 2002). If this strategy proves plausible, then the prospects for accepting psychological natural kinds look significantly better.

More liberal, non-essentialist theories of natural kinds have also been proposed. Among the best developed ones is the ‘simple causal account’ by Khalidi (2013) according to which natural kinds are to be distinguished on the basis of projectibility, explanatory efficacy and predictive value (Khalidi 2013: 55). The account is pluralistic, since it rejects reductionist requirements and allows for non-intrinsic properties as distinctive properties of natural kinds. It also abandons the essentialistic requirement that the boundaries between kinds are to be clear-cut and discrete. The causal theory proposed by Khalidi does not give any substantial restrictions which would render mental states improper candidates for natural kinds. It admits, moreover, that some folk categories might prove to be referring to natural kinds.

It stems from the overview of the main theories of natural kinds that at least some of those theories do not exclude the possibility that PA form natural kinds. The more restrictive accounts like Lewis’ physicalist account or the intrinsic-properties essentialism do exclude this possibility, but there are good independent reasons not to adopt these accounts. The more pluralistic theories like extrinsic-and-intrinsic-properties essentialism and simple causal theory admit the possibility of PA being natural kinds (although neither of them settles if they actually are).

Superficialists about propositional attitudes argue as if the question of whether PA can be natural kinds was settled definitely. It seems, however, that this question is far from being settled. The answer to this question differs depending on more general issues having to do with metaphysics of natural kinds. If PA as they are used in folk psychology may refer to the mental natural kinds explored by psychology, then superficialists are in trouble.

References

Currie, D. S. (2021), How beliefs are like colors. Synthese, 199 (3-4), 7889-7918.

Khalidi, M. A. (2013). Natural Categories and Human Kinds: Classification in the Natural and Social Sciences. New York: Cambridge University Press.

Lewis, D. K. (1983). New work for a theory of universals. Australasian Journal for Philosophy, 61 (4), 343-377.

Okasha, S. (2002), Darwinian Metaphysics: Species and the Question of Essentialism. Synthese. 131 (2), 191–213.

Pérez, D.I. (2004). Mental concepts as natural kind concepts. Canadian Journal of Philosophy, 30 (sup. 1), 201-225.

Posłajko, K. (2024). Unreal beliefs. Bloomsbury Publishing.

Putnam, H. (1975). The meaning of ‘meaning’. Minnesota Studies in the Philosophy of Science, 7, 131-193.

Schwitzgebel, E. (2021). The Pragmatic Metaphysics of Belief. In. C. Borgoni, D. Kindermann, A. Onofri (Eds.), The Fragmented Mind. Oxford University Press.

I introduce, and highlight the explanatory benefits of, a knowledge-first theory of the folk psychology of belief. I make precise Williamson (2000)’s suggestion that “[...] to believe p is [...] to treat p in ways similar to the ways in which subjects treat propositions which they know” (pp.46-7). The resultant theory says that

1. To believe p is
   (i) to treat p in the fully specific way F such that
   (ii) F is relevantly similar to the fully specific way G such that
   (iii) in normal circumstances, anyone treats anything they know in G.

(i) is a habitual like ‘He seems to do ‘The Times’ crossword’ (Huddleston and Pullum, 2002, p.161) and does not require one to treat p in any way at the time at which the belief is held; instead, it ascribes a habit. Thus, my theory is dispositionalist (cf. Ryle, 1949). (iii) then specifies the dispositional stereotype for belief (cf. Schwitzgebel, 2002) by appeal to knowledge. This appeal to knowledge predicts the developmental trajectory of the folk psychologies of knowledge and belief summarized by Phillips et al. (2020) “Knowledge before Belief”. Moreover, since in a knowledge-first framework, knowledge is not reducible to belief, this appeal to knowledge provides an informative characterization of the dispositional stereotype. To arrive at this characterization, we can ask ‘In normal circumstances, how does anyone treat everything they know?’ Although the complete answer to this question is not straightforward to find, even partial answers, which are straightforward to find (using p as a premise in one’s practical reasoning normally accompanies knowing p; so do asserting p, judging p, and not wondering whether p), suffice for my theory to make concrete predictions.

Reference to similarity with a dispositional stereotype predicts that the verb ‘believe’ admits degree modifiers (‘fully/mostly’), comparatives (‘more than’), and equatives (‘as much as’). Given my theory, such constructions concern the degree to which a subject’s fully specific way of treating p resembles the dispositional stereotype. There are two added benefits of my theory’s account of the gradability of ‘believe’. First, it predicts that, just as Koev (2019, p.5) argues, ‘believe’ is associated with an absolute standard (cf. ‘flat’). Absent contextual manipulation and overt modification, the required degree of similarity to the dispositional stereotype is thus absolute. This explains intuitions to the effect that delusions and superstitions, among others, are not beliefs (Gallagher, 2009; Ichino, 2018). Second, my theory’s account of the gradability of ‘believe’ explains experimental results obtained by Buckwalter, Rose, and Turri (2015). They find that overt modification with ‘at least on some level’ leads significantly more participants to ascribe belief to protagonists who exhibit hardly any similarities with the dispositional stereotype specified by my theory. This is predicted by my theory: ‘at least on some level’ relaxes the required degree of similarity.

Reference to relevant similarity in (ii) allows my theory to also predict effects of contextual manipulation. For instance, in a context where interlocutors focus on whether a subject is in the habit of asserting p, this respect of similarity is relevant. Thus, in this context, my theory predicts even a subject that does not use p as a premise in much of their practical reasoning, but does assert p, to believe p. In line with natural case descriptions, patients suffering from the Cotard delusion, who assert that they are dead, are therefore classified as believing that they are, despite their failure to use the proposition that they are dead in much of their practical reasoning (cf. Bortolotti, 2010, p.163). The benefits of this context-sensitivity also generalize to other cases of delusion.

References
Bortolotti, L. (2010). Delusions and other irrational beliefs. International perspectives in philosophy and psychiatry. Oxford: Oxford University Press.
Buckwalter, W., D. Rose, and J. Turri (2015). “Belief Through Thick and Thin”. In: Noûs 49.4, pp. 748–775.
Gallagher, S. (2009). “Delusional realities”. In: Psychiatry as Cognitive Neuroscience. Ed. by M. Broome and L. Bortolotti. Oxford: Oxford University Press, pp. 245–266.
Huddleston, R. D. and G. K. Pullum (2002). The Cambridge grammar of the English language. Cambridge, UK; New York: Cambridge University Press.
Ichino, A. (2018). “Superstitious Confabulations”. In: Topoi.
Koev, T. (2019). “Strong beliefs, weak commitments”. In: Proceedings of Sinn und Bedeutung 23.2, pp. 1–18.
Phillips, J. et al. (2020). “Knowledge before Belief”. In: Behavioral and Brain Sciences, pp. 1–37.
Ryle, G. (1949). The Concept of Mind. London: Hutchinson & Co.
Schwitzgebel, E. (2002). “A Phenomenal, Dispositional Account of Belief”. In: Noûs 36.2, pp. 249–275.
Williamson, T. (2000). Knowledge and Its Limits. Oxford: Oxford University Press.