NOTE-117: Drayson — What motivates mental fictionalism?

Contribution

The paper separates two questions that Toon’s Mind As Metaphor (2023) runs together: whether folk-psychological discourse, read “Cartesianly”, posits internal states with intentional and causal properties, and whether those states must be the mental representations of the representational theory of mind (RTM). With them separated, Toon’s negative claim (no Cartesian truthmakers) does not follow from the falsity of RTM. It also shows that Toon’s positive claim, which buys the causal-explanatory advantage of fictionalism by letting pretence yield genuine assertions about a person’s overall causal state, reinstates the ontological commitment fictionalism was meant to avoid.

Key insight

Being a realist about beliefs is not the same as being a realist about mental representations. A belief can be an internal, theoretical, intentional and causally efficacious state of the whole person, individuated as a functional whole with no sub-state that carries its content (Loar 1982: attitudes “are not to be further explicated in terms of more fundamental ascriptions of content”). Once that option is on the table, skepticism about inner representations does not motivate fictionalism, and a fictionalist who wants mental talk to pick out real causes of behaviour has already conceded what realism needs.

Assumptions

Premises and authorities, since this is a philosophy paper.

• Mental fictionalism is coherent, assumed for the sake of argument (n. 2; cf. Wallace 2016).
• Motivating any fictionalism takes two steps (§1, pp. 1–2): a negative claim (the discourse lacks truthmakers) and a positive claim (it is worth keeping rather than eliminating or reinterpreting), on the model of moral (Nolan, Restall & West 2005; Joyce 2013) and mathematical (Balaguer 2018; Yablo 2001) fictionalism.
• Toon’s Cartesian reading is the standard one: folk psychology is a theory whose attitude terms are implicitly defined by functional role, and it attributes intentional and causal properties to the attitude (the believing), not to the proposition (pp. 4–5, n. 10). Drayson accepts this (“I do not intend to challenge it”, p. 5), while noting prescriptive or merely rationalizing readings as alternatives (n. 8).
• “Internal” means theoretical, not spatially inner: Toon takes it from Sellars’s Myth of Jones, where thoughts are “inner” because unobservable (p. 3, n. 7).
• RTM’s core commitment is that mental representations are individuated distinctly from attitudes, so that believing-that-p and desiring-that-p can token the same representation (p. 6; Mandelbaum & Quilty-Dunn 2018). Further structure (language-like syntax, compositionality) is inessential to the argument (n. 11).
• RTM is an empirical theory of the truthmakers, not an analysis of what folk-psychological sentences mean (p. 6, citing Fodor 1987 against Toon’s gloss at p. 34 of his book).
• Dispositions can be internal and causal (n. 19; McKitrick 2005).

Key results

What the paper argues, by section.

• §2, reconstruction (p. 4). Toon’s negative argument: P1 the Cartesian reading commits folk psychology to attitudes as internal, intentional, causal states; P2 that commitment is a commitment to RTM; P3 RTM is false because internal mental representation is incoherent; C on the Cartesian reading folk psychology lacks truthmakers. Validity needs P2 as an entailment, and Toon’s text only shows that Cartesians “tend to subscribe” to RTM.
• Against P2 (pp. 6–7). Non-representational functionalism (Lewis 1966, Putnam 1967, Loar 1982) individuates attitudes as functional wholes. Such states are internal in Toon’s (Sellarsian) sense, intentional at the personal level “even if none of its subpersonal realizers” are, and causally efficacious as brain states. So they meet P1’s truth conditions without RTM, and P2 is false. Making P2 true by definition would re-classify these theories as anti-Cartesian and weaken Toon’s positive argument.
• Against P3 (pp. 7–8). Toon’s case is metasemantic: intentionality belongs “first and foremost” to public, socially normed representations, and inner representations are “locked away” from convention. Drayson replies that RTM is not committed to a reductive, mind-first metasemantics (that is one motivation, e.g. Fodor’s), and that functional role semantics can let mental representations take content from social and communicative use (Block 1987, LIT-625). She also notes (n. 15) that Toon’s endorsement of Bennett & Hacker’s point that only persons represent would impugn most of cognitive psychology and neuroscience, not just RTM.
• Upshot of §2 (p. 9). “Even if there were no mental representations, we could still coherently claim that there were ‘Cartesian’ truthmakers.”
• §3, Toon’s positive argument reconstructed (pp. 9–11). Eliminativism would require a drastic change in language; anti-Cartesian readings (logical behaviourism, predictive instrumentalism) “deny the obvious fact that mental states are causes” (Toon p. 48). Fictionalism is better because pretence about an inner notebook lets us make genuine assertions that a person is in “a certain state”, a state such that the pretence is appropriate.
• §3, the dilemma (pp. 11–12). If the “overall state” is only a set of behavioural facts, fictionalism has no causal advantage over anti-Cartesian realism. If it is the state S that causes the behaviour (Toon p. 52), then the assertions commit Toon to S, and “why not identify state S as Mark’s belief?” Realist views on which attitudes are functional or dispositional states of the whole person with non-representational causal bases (Egan 2025; Langland-Hassan 2020) are then available. So the positive argument “provides Toon himself with a reason to be a realist”.
• §4 (pp. 12–13). Drayson shares Toon’s skepticism (for different reasons) about “an architecturally-innate and linguistically-structured” RTM and his worry that behavioural accounts fall short, but holds these are not the realist’s only options. She does not claim fictionalism is viable only once all realisms fail.

Claims

id	claim	strength	support
C1	Toon’s negative argument is valid only if P2 is an entailment, and his text supports only a tendency	informal argument (textual)	§2, p. 4, with quotations from Toon pp. 10–11
C2	One can accept internal, intentional, causal attitudes while rejecting RTM, so P2 is false	informal argument	pp. 6–7, 9: the non-representational functionalism of Lewis, Putnam and Loar, checked against each element of P1
C3	RTM is not committed to a reductive, mentalistic metasemantics, so Toon has not shown P3	informal argument	pp. 7–8: Fodor’s motivation distinguished from RTM’s core; Block 1987 as a counter-model
C4	Toon’s endorsement of Bennett & Hacker would impugn most of cognitive science, not only RTM	assertion	n. 15, one sentence
C5	If Toon redefines “Cartesian” to make P2 true, his positive claim becomes harder to establish	assertion with a forward reference	p. 7; the link is made only implicitly in §3
C6	Toon’s pretence account, by yielding genuine assertions about an overall causal state S, commits him ontologically to S	informal argument	pp. 11–12, quoting Toon p. 52; Yablo 2001 on assertoric fictionalism (n. 17)
C7	If S is merely behavioural, fictionalism has no causal-explanatory advantage over anti-Cartesian realism	informal argument	p. 11
C8	S can be identified with the belief on existing realist accounts (functional or dispositional whole-person states with non-representational bases)	informal argument, by citation	p. 12: Egan 2025, Langland-Hassan 2020; dispositions as internal and causal (n. 19, McKitrick 2005)
C9	The abstract’s “seems to ontologically commit him” is the paper’s actual strength: it shows a pressure toward realism, not that fictionalism is refuted	hedged by the author	abstract, §4 (“This is not to suggest that mental fictionalism is only a viable option once all forms of realism are rejected”)

Concepts

• mental fictionalism: engaging in truth-conditional mental discourse while having reason to believe it lacks truthmakers (abstract, §1).
• negative claim / positive claim: the two steps of motivating a fictionalism, lack of truthmakers and value in retention (§1).
• Cartesian interpretation (Toon’s term, no historical accuracy intended, n. 6): folk psychology is “an attempt to describe what goes on inside our heads”, made true only by internal states; physicalist views count as Cartesian.
• anti-Cartesian interpretation: truthmakers may be public behaviour patterns or predictive success (logical behaviourism, instrumentalism).
• internal: theoretical, posited to mediate between sensory input and motor output, not observationally definable (Sellars; n. 7, p. 7).
• representational theory of mind: attitudes involve tokening a mental representation that carries their content, individuated separately from the attitude type (p. 6).
• non-representational functionalism: attitudes individuated as functional wholes with no further content-bearing constituents (Loar 1982; p. 7).
• overall state: Toon’s “Mark’s overall state”, picked out via the folk-psychological metaphor and, on one reading, the cause of behaviour (p. 11).

Connections

Metaphysical thesis defended or presupposed. The paper defends a thesis about the ontology of mind via truthmaker reasoning: a discourse’s having truthmakers and its user’s ontological commitments are what is at issue, and the paper holds that realism about propositional attitudes as causally efficacious internal states is compatible with anti-representationalism. It presupposes that causal efficacy is the mark of reality in this debate (neither side questions that a state which causes behaviour exists), that functional states of the whole person can be causes (it does not address the exclusion worry that would say otherwise; see LIT-144 and LIT-147), and a fairly standard truthmaker framework. It is metaphysics of mind reached through philosophy of language, and the metaphysics tag is justified. cognition as primary topic (philosophy of mind) is right.

Links in this record.

• LIT-207 (Roberts, chatbot fictionalism): the nearest sibling. Roberts argues for a fictionalism about AI self-attributions of body-dependent states. Drayson’s diagnosis carries over as a test: a chatbot fictionalist who also wants chatbot talk to track real causes of output (as Toon wants for persons) inherits the same dilemma. Roberts’s anti-realism is conditional on embodiment and leaves belief open, which is where Drayson’s functionalist realism would sit.
• LIT-206 (Goldstein & Lederman, interpretationism about ChatGPT): an instance of the interpretivist “overall state” realism Drayson’s n. 20 grants to Curry and Williams, applied to LLMs.
• LIT-128 (affective fictionalism): another fictionalism about folk psychology, motivated normatively rather than by skepticism about truthmakers. Drayson’s two-step template (negative and positive claim) is a ready frame for asking what motivates it.
• LIT-149 (Martínez on representation) and LIT-148 (computational functionalism): positions on what makes an internal state a representation, bearing on whether RTM’s truthmakers exist, which is the premise Drayson leaves undecided.
• LIT-097 (The Extended Mind): Clark and Chalmers’s notebook is the same image Toon uses for his pretence (“memory-as-a-notebook game”, p. 10), used by them realistically.
• LIT-109 (Vetter, dispositions): relevant to n. 19’s claim that dispositions can be internal and causal.

Bearing on the record

It carries nothing for ML practice, and there is no Anthology document it should inform. For nucleation it is a precise statement of the realist reply to mental fictionalism, and its dilemma is directly reusable against fictionalist readings of LLM “beliefs” (LIT-207). The dossier’s suggestion that the middle position (realism with non-representational causal bases) is “a live option for interpreting model internals” is the reader’s extrapolation; the paper says nothing about artificial systems.

Limitations

• It evaluates motivations, not the position: coherence is assumed (n. 2), and fictionalism is not shown to be false.
• The functionalist alternative is sketched in one paragraph (p. 7) by citation. That such whole-person functional states are causally efficacious “as a state of the brain” is asserted, not defended against exclusion-style arguments.
• It engages only Toon 2023, and the page references are to that book; the argument’s force against other mental fictionalists (e.g. Demeter, Wallace) is not assessed.
• The P3 reply shows RTM is not committed to a mentalistic metasemantics, but it does not show that a socially grounded metasemantics for inner representations works; Block 1987 is named, not developed (the record now states it as THEORY-089).
• The positive-argument objection depends on reading Toon’s “overall state” causally; Drayson notes Toon “sometimes suggests” a behavioural reading (p. 11), so the objection is a dilemma, not a refutation of a fixed position.
• The preprint carries a residual “[reference removed for anonymous review]” (p. 12), presumably her Drayson 2022 chapter.

Open questions

• Can the fictionalist keep genuine assertion and avoid commitment, e.g. by a quasi-assertion account, and what would that cost Toon’s causal-explanation advantage? The paper flags the options (n. 17) but does not pursue them.
• Is a whole-person functional state a cause in the sense exclusion arguments demand? A worked answer (e.g. from the interventionist literature) would close the gap in C8.
• Has Toon replied? Unverified at the time of reading.

Corrections to the seeded skim

• The dossier and LIT-112 say Drayson “targets P2” only. She attacks two premises. Premise 2 she argues is false (pp. 6–7, 9). Premise 3 (the incoherence of internal representations) she argues is unsupported: Toon “mistakes one popular motivation for adopting the representational theory of mind for an essential property of the theory”, namely a reductive “mentalistic” metasemantics, and functional role semantics (Block 1987) could give mental representations socially conditioned contents (pp. 8–9). She does not accept Premise 3; she says Toon “would have to do more to defend” it (p. 8). This also answers the LIT’s check item on Premise 3.
• The dossier places the “why not identify S as the belief” move at p. 12 and describes the whole of §3 as addressing “fictionalism makes the most sense of behaviour”. The argument turns earlier, at p. 11: Toon’s claim that the pretence yields genuine assertions that the person is in “a certain state” (Toon pp. 23, 51–52) is what distinguishes him from quasi-assertion fictionalists, and it is this that brings the ontological commitment (with Yablo 2001’s general objection, n. 17). If S were only a complex set of behavioural facts, the fictionalist interpretation would lose its claimed advantage over anti-Cartesian ones (p. 11). So the objection is a dilemma, which the dossier does not state.
• The dossier’s check item asks whether Drayson’s realism “collapses into interpretivism (the footnote on Curry and Williams, p. 12)”. Footnote 20 answers it directly: she is not claiming Toon’s commitments push him toward Egan or Langland-Hassan’s realism, and she grants that interpretivist views (Curry 2025, Williams 2019) share Toon’s commitment to “overall mental states” without fictionalist semantics. The paper only needs that some realism is available, not which one.
• The dossier says the negative argument depends on “commitment to internal mental states does not require the representational theory of mind”. The paper also offers Toon an escape and closes it: he could make Premise 2 true by definition, counting all non-representational theories as anti-Cartesian, but that would make his positive claim harder to establish (p. 7).
• Minor: footnote 2 says the paper assumes “for the sake of argument that mental fictionalism is a coherent position”. The dossier’s framing of the debate (Toon “means to avoid” realism) is right, but the paper takes no view on coherence.
• metaphysics tag: justified — the paper is about whether folk-psychological discourse has truthmakers and what it ontologically commits its users to.