Keith Frankish (2016), “Illusionism as a theory of consciousness”, Journal of Consciousness Studies 23(11–12), 11–39, Imprint Academic; reprinted as the opening essay of K. Frankish (ed.), Illusionism as a Theory of Consciousness (Imprint Academic, 2017). Read from the author’s eprint at https://keithfrankish.github.io/articles/Frankish_Illusionism%20as%20a%20theory%20of%20consciousness_eprint.pdf.
Key takeaways
- Three options for any apparent anomaly (§1.1): radical realism (real, needs new science), conservative realism (real, explicable with existing science), illusionism (not real; explain the appearance). Illusionism shares radical realism’s view that phenomenal properties are anomalous and conservative realism’s refusal of radical innovation.
- Strong versus weak illusionism (§1.2). A quasi-phenomenal property is a physical property introspection typically misrepresents as phenomenal. Strong illusionism says experiences have only these. Weak illusionism keeps real but deflated phenomenal properties. Unless weak illusionists can name a “phenomenal residue” stronger than quasi-phenomenality yet still conservatively explicable, which Frankish doubts, only radical realism and strong illusionism remain.
- What illusionism does not deny (§§1.6–1.7). Conscious experiences exist in an inclusive sense; what is denied is that they have phenomenal properties. Experiences are “like something” in a functional sense (introspective representation), and in that sense a zombie is like us. Everyday phenomenal talk can be read as talk within a fiction, like talk of Lear’s anguish.
- Arguments for it (§2): a presumption of conservatism, especially against pressure from introspection; epiphenomenalism threatens radical realism; conservative realism tends to explain the representation of phenomenality and so is covertly illusionist; the debunking argument (our phenomenal claims would be the same without phenomenal properties), which Frankish takes from Chalmers (1996, pp. 186–7) without Chalmers’s endorsement; and an abductive inference from anomalousness plus the uncheckability of introspection.
- Replies to objections (§3): denying the data (acquaintance makes the datum psychologically inert); no appearance–reality gap (true only in a weak sense compatible with illusionism); who is the audience (no inner observer, appreciation distributed over unintelligent subsystems); how non-existent properties are represented (functional-role or hybrid concepts).
Standing in the record
Filed on 2026-10-02 at the owner’s request, from a text an earlier reading fetched while filing Dennett. NOTE-384 is the close reading of the same day, and it placed the work: Active.
The record’s illusionist thread.
- Frankish’s later paper. “Closing the Cartesian Bureau de Change” (LIT-103, read in NOTE-143) applies this article’s position to Veit’s account of valence and leans on it: its “quasi-phenomenal properties” and its post-Cartesian paradigm are this paper’s terms, cited there as Frankish 2016/2017. This article is the argument LIT-103 assumes.
- Dennett’s reply. “Illusionism as the Obvious Default Theory of Consciousness” (LIT-434) appeared in the same JCS issue as a commentary on this article. It is filed unread. Frankish calls Dennett illusionism’s pre-eminent defender (p. 1) and draws on him throughout: the user illusion, the multiple drafts of sensation, heterophenomenology, and the storyteller’s authority over fiction. “Quining Qualia” (LIT-433, read in NOTE-357) is the source of the list of marks (ineffable, intrinsic, private, immediately apprehended) that Frankish says weak illusionists deflate.
- Chalmers’s meta-problem. “The Meta-Problem of Consciousness” (LIT-425, read in NOTE-354) takes this article’s illusionism as its target in §§6–7, and NOTE-354 records that Chalmers says he suggested the name. Two points of contact are checkable here. Frankish’s §2.3 debunking argument has the shape of the one LIT-425 §5 calls Argument A and answers: there is a consciousness-independent explanation of our phenomenal beliefs, so they are unjustified. LIT-425’s replies (acquaintance, and the constitutive role of consciousness in phenomenal belief) are the positions this article’s §3.1 attacks as making the datum “psychologically inert”. And LIT-425’s claim that only strong illusionism dissolves the hard problem agrees with this article’s argument that weak illusionism is unstable, for different reasons: Chalmers argues that only the strong form removes the problem, while Frankish argues that the weak form has no stable content. Where they part is LIT-425’s Moorean rejection of strong illusionism, which is the “denying the data” objection this article answers in §3.1.
- Taylor’s interpretation of Dennett (LIT-461, read in NOTE-379) reads Dennett’s view as realism about a real pattern and treats qualia eliminativism as peripheral. That is in tension with this article, which counts Dennett as the leading strong illusionist. NOTE-379 sets out the difference.
- Dennett’s 1980 eliminativism about pain (LIT-456, read in NOTE-371) is an early form of the view this article renames. Frankish’s §1.6 says “eliminativism” wrongly suggests a wider rejection of folk psychology; Dennett in 1980 holds exactly that wider view of pains, beliefs and desires, and rescues the latter two by the intentional stance.
On the society-of-mind bridge (THEORY-043). The article never discusses groups. It bears on THEORY-043 in one place. Its §3.3 answer to “who is the audience?” spreads the work of being taken in by the illusion over “numerous unintelligent subsystems”, with no unified audience smaller than the organism. That is Dennett’s discharge of homunculi applied to the illusion itself. On this view the THEORY’s narrowed question about groups becomes whether a group has the introspective representations that generate the illusion. That framing is mine, not Frankish’s, as it was in LIT-434’s entry.
On AI consciousness (THEORY-023). Not discussed by Frankish. His §1.7 has a consequence he does not draw: if what makes experience “like something” is introspective representation, then the question for a machine is whether it has the introspective mechanisms, not whether it has a further property. That sharpens the mimicry worry in THEORY-023 rather than settling it, since a system can produce phenomenal reports without such mechanisms; the meta-problem paper raises the same caution about puzzled phenomenal reports as evidence (LIT-425, fn 26).
It carries no instruction for machine-learning practice, and it does not belong in the anthology.