Contribution
Critics from Block (1993, 1994) to Schwitzgebel (2007) have called Dennett’s theory of consciousness elusive or self-contradictory. Taylor gives one reading on which six of its central claims are consistent. The new move is to make “Real Patterns” the hub: consciousness is a real pattern in the drafts, and a user-illusion is a particular kind of real pattern. With that, the Orwellian/Stalinesque indeterminacy no longer needs Dennett’s “first-person operationalism”, and the conflict between fallibility and authority becomes a difference between literal and fictional readings of the same report.
Key insight
A real pattern is a description that trades predictive accuracy for compression. If consciousness is such a pattern, every question about it is a question about which description of the brain’s drafts best makes that trade. Questions with no such answer, such as whether the masked disc was briefly conscious, have no fact behind them, not because the fact is unverifiable but because both answers describe the pattern equally well. And a pattern coarse enough to be a user-illusion describes the brain only metaphorically, so its users are authors of their reports as much as observers.
Assumptions
- Content is taken for granted (n. 5). Taylor sets Dennett’s theory of content aside and assumes drafts are content-bearing.
- Dennett’s own description of the masking cases. The two readings are assumed to differ only in whether the disc was conscious, with no difference in downstream effects (p. 6). Taylor notes this is “somewhat simplistic” (n. 8, citing Haase and Fisk 2015). The argument for indeterminacy depends on it.
- Interpretive charity, not truth. The paper does not argue that the view is correct or that it is what Dennett meant (pp. 1–2).
- Six claims chosen as core (Box 1). Other claims Dennett has made are not reconciled, and his views changed over time (p. 2).
Key results
- The real-pattern reading (pp. 2–4): consciousness is a pattern in drafts. Three features of real patterns do the work: the compression and accuracy trade-off (Dennett 1991b p. 36), several patterns in one system (p. 35), and existence independent of anyone describing it (p. 34).
- Multiple drafts (Box 2; pp. 4–9). (1) No Cartesian theatre: a pattern is distributed, as a Sicilian dragon is not a place on the board. (2) It is an open question whether a draft is conscious: being part of a pattern depends on relations to other states. (3) No fact decides between Orwell and Stalin: the readings are equally good compressions with the same predictions. This replaces the argument from first-person operationalism, which Taylor sets aside.
- Global workspace (pp. 9–10). None of the three claims conflicts with GWT. A distributed model of working memory (Christophel et al. 2017; D’Esposito and Postle 2015) keeps sensory content in sensory cortex while it is broadcast, so nothing is “sent” to a theatre.
- Qualia (p. 10). Qualia are defined by Dennett’s four marks. Rejecting them is rejecting one picture, and it is peripheral: CE’s qualia chapter spends its length on Akins’s view of colour, and Dennett’s 2023 autobiography says he discussed qualia “grudgingly”.
- User-illusion (pp. 10–13). A user-illusion is a real pattern. It gives explanations accepted as complete (“I just looked, and saw them”) and it is so coarse that its descriptions lapse into metaphor, as with the people who call pencils plants (Dennett 2007).
- Authority (pp. 13–15). Reports are not immune to error read as literal descriptions of the brain (claim v). Read as unwitting metaphor they are fictions, over which the subject has an author’s authority (claim vi). Against Schwitzgebel (2007), it may be more charitable to read them as fictions.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Dennett’s theory can be read consistently as realism about a real pattern in drafts | moderate | the whole paper; one interpretation, with textual anchors (CE p. 460, 1991b), not shown to be Dennett’s |
| C2 | The Orwellian and Stalinesque readings of metacontrast are equally good descriptions of the pattern, so neither is correct | weak–moderate | pp. 7–8; depends on Dennett’s idealised description of the cases, which Taylor concedes is simplistic (n. 8) |
| C3 | Multiple drafts and GWT are compatible | moderate | pp. 9–10; the three core claims do not oppose GWT, plus a distributed working-memory model |
| C4 | Qualia eliminativism is peripheral to Dennett’s positive view | weak | p. 10; two textual hints (CE’s chapter proportions, “grudgingly” in the 2023 autobiography) |
| C5 | A user-illusion is a kind of real pattern | assertion | p. 11; argued by showing it has real patterns’ features (compression, aid to prediction) |
| C6 | Fallibility and dictatorial authority are two kinds of authority and do not conflict | moderate | pp. 13–15; reconstructs Dennett 2007 pp. 254–255, which Taylor calls “cryptic” |
Concepts
- Draft / content-bearing state — a brain state that encodes information. Dennett’s “content fixations”, “micro-takings” (with Kinsbourne) or “representations” (p. 4). Not ontologically loaded: processes and events count (n. 5).
- Real pattern — a description of a system more efficient than its “bit map”, whether or not anyone concocts it (pp. 3–4).
- User-illusion — a real pattern so coarse-grained that describing the system through it is metaphor (pp. 10–13).
- First-person operationalism — Dennett’s denial of “consciousness of a stimulus in the absence of the subject’s belief in that consciousness” (CE p. 132, as quoted, p. 7). Set aside by the paper.
Connections
- “Real Patterns” (LIT-220, NOTE-190). The paper’s foundation, used for three of its features. The record’s reading of “Real Patterns” covers the same passages (Life, chess, the bit map, the trade-off).
- “Quining Qualia” (LIT-433, NOTE-357). Cited for the definition of qualia. NOTE-357 reads the paper as an argument that the concept is confused; Taylor reads that as a negative thesis only. The two readings agree on what is rejected.
- “Time and the Observer” (LIT-431, NOTE-367). The source of the timing cases. Taylor discusses metacontrast only (n. 7 names colour phi and the cutaneous rabbit).
- Frankish 2016 (LIT-458, NOTE-384). Not cited. The two papers disagree about what is central in Dennett. For Frankish, Dennett is illusionism’s pre-eminent defender and the denial of phenomenal properties is the point. For Taylor, the denial is a sideline and the point is a “mild” realism. On Frankish’s §1.6 the two are compatible: illusionists grant consciousness in an inclusive sense, with only quasi-phenomenal properties. Taylor’s real pattern would be consciousness in that inclusive sense. Taylor does not draw this, and he does not discuss Dennett’s 2016 adoption of the word “illusionism” (LIT-434, unread).
- Dennett 1980 (LIT-456, NOTE-371). Not cited, and the closest precedent for the paper’s task. Stich charged Dennett with a “prima facie contradiction” about pain; Dennett’s reply holds that there are no pains and that people feel pain, compares pain with holes, voices and haircuts, and leaves it there. The CE line on which Taylor builds (“as real as haircuts”) is that comparison returned.
- Schwitzgebel 2007 (“No unchallengeable epistemic authority …”) is cited for the puzzle about authority and for the charity objection. It is not filed.
- Ladyman and Ross (LIT-301, unread) and Fallon 2020 are cited as readings of real patterns as an intermediate realism.
Bearing on the record
- THEORY-043. No direct bearing. The paper never discusses groups.
The real-pattern criterion it describes is indifferent to what the
parts are, which is consistent with THEORY-043’s point that Dennett’s
stance is “blind to composition and to consciousness by design”. The
paper does not say what makes a real pattern a conscious one rather
than, say, a belief pattern. Without that, it supplies no criterion of
the kind THEORY-043’s
promote_whenasks for. - THEORY-023. No direct bearing. On this reading, whether a machine is conscious would be whether its content-bearing states support the same coarse pattern. The paper does not say what fixes “the same”.
- No ML instruction. Nothing belongs in the anthology.
Limitations
- An interpretation, explicitly. “My goal is not to argue that the theory is correct” (p. 1); nor (p. 2) that it is exactly what Dennett meant.
- The masking argument rests on an idealisation that the author flags (n. 8). If the two readings predict different downstream effects (Haase and Fisk 2015), C2 fails for that case.
- “Which pattern?” is left open. A system has many real patterns (p. 3). The paper does not say which one consciousness is, beyond “a user-illusion”.
- Six claims only. Heterophenomenology, the self, memes and the origin of the metaphors are mentioned and set aside (p. 13).
Open questions
- Is the real pattern that is consciousness the same pattern as the one the intentional stance finds, or a different one in the same drafts?
- If authority is an author’s authority over unwitting metaphor, what makes a sincere report a fiction rather than a false belief? Taylor’s answer (the coarseness forces metaphor) is stated, not argued.
- Can the compression-equivalence argument be made for a case where the readings are known to differ in prediction, rather than stipulated not to?
Corrections
- No dossier existed, so there is no seeded skim to correct.
- Citation. The brief’s citation (Philosophical Psychology, DOI 10.1080/09515089.2024.2433526, CC BY) is right. Crossref now gives the issue: 39(3), 828–846, dated 3 April 2026.
- Two slips in the paper.
- The pencils-and-hammers passage is cited as “(Dennett, 2007, p. 25)” (p. 12). The surrounding citations are to pp. 254–255 of that article, which runs 247–270, so p. 25 cannot be right; p. 254 is likely.
- “Norman, L., Akins, K., Heywood, C., & Kentridge, R. (2013)” is given with Current Biology 24(23), 2822–2826 and DOI 10.1016/j.cub.2014.10.009, which is the 2014 volume. The year in the text and reference list does not match the volume.
- “Quining Qualia” pages are given as 43–77 in the reference list; Frankish (LIT-458) and NOTE-357 give 42–77.