Contribution
The paper identifies a flaw in the no-report paradigm, the main method for deciding whether perceptual consciousness is prefrontal: it removes report but not the thinking about the stimulus that report would have expressed. It proposes the right target, a paradigm with no post-perceptual cognition that differs between percepts, and finds an instance in an existing binocular-rivalry experiment.
Key insight
The confound was never report as such. It was cognition about the percept, of which report is one product. A subject who is not asked to say which way the grating moves can still notice, think and wonder which way it moves, and that thought is correlated with the percept. The way to remove it is to make the perceptual change one that the subject cannot single out, so that nothing in thought follows it.
Assumptions
- Cognition means thought-based mentality: reasoning, deciding, evaluating, reporting and working memory, with transitions between propositional states that depend on their contents (Glossary).
- Consciousness means phenomenal consciousness throughout; the access distinction is noted in the Glossary.
- Correlational evidence only; Block cautions that causal interventions are better where available.
- Prefrontal regions at issue: dorsolateral, medial prefrontal, anterior cingulate and orbitofrontal. Both sides grant parietal involvement.
Key results
- The opposing sides. Prefrontalists: global workspace and higher-order theories, and the authors of the 2017 Science manifesto on machine consciousness. The back: IIT and recurrent processing. Noncognitivists can allow that the phenomenology of thought itself is prefrontal.
- Rivalry and replay. Early fMRI found stronger frontal differences at transitions; replay subtraction isolates transitions, not contents.
- No-report evidence on both sides. Frässle et al. 2014: with nystagmus as the readout and no task in the scanner, prefrontal differences were minor (“Frontal activity relates to introspection and action but not to perception”). Safavi et al. 2014 and later Utah-array work decoded rivalry contents from monkey prefrontal cortex without report, in monkeys never trained on a discrimination. Wang et al. 2013 decoded contents prefrontally, with report.
- The bored monkey problem. Hours of fixation with nothing to do invite thoughts aligned with the alternating percept. The same applies to the older human no-report studies (Lumer and Rees 1999).
- Inconspicuous rivalry (Brascamp et al. 2015, Fig. 2). Random-dot motion changing every 300 ms; rivalry switches with same-colour dots were detected at chance by d′. Density reports showed one eye’s array, not a merger, confirming real rivalry. Prefrontal differences appeared for different-colour dots and were “altogether undetectable” for same-colour ones.
- Box 1, rivalry versus merger. Attending to local features of a male and a female face gives rivalry; attending to gender gives a fused androgynous face. Locally compatible stimuli fuse even if globally different (Carlson and He).
- Box 2, predictive processing. Clark’s account of rivalry as competition between high-level hypotheses, because a face and a house cannot occupy one place, is undercut: low-level local features largely decide rivalry.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | No-report paradigms do not eliminate post-perceptual cognition correlated with the percept | strong as a methodological point | argument; the bored monkey |
| C2 | Prefrontal decoding in no-report rivalry therefore does not show that prefrontal cortex is part of conscious perception | moderate | follows from C1 |
| C3 | Brascamp et al.’s design is a no-differential-post-perceptual-cognition paradigm | moderate: switches detected at chance, rivalry confirmed by density and temporal regularities | Fig. 2; Brascamp et al. 2015 |
| C4 | That experiment gives some support to non-cognitive theories | weak: fMRI null result; Block concedes ECoG could support prefrontalism | Anticipating an Objection |
| C5 | Rivalry is driven by local features, against predictive-processing explanations | moderate: cited psychophysics (Carlson and He; Klink et al.; Freeman and Li) | Boxes 1–2 |
Concepts
- no-report paradigm: reports calibrate another indicator of consciousness (here nystagmus), then experimental subjects make none.
- no-(differential-)post-perceptual cognition paradigm: one in which no cognition that differs between the competing percepts follows them.
- prefrontalists: those who hold certain prefrontal circuits necessary for perceptual consciousness.
- optokinetic nystagmus (OKN): slow pursuit in the direction of perceived motion with fast returns, used as a no-report readout.
- bored monkey problem: uninstructed subjects think about what they see.
Connections
- Block (2007), LIT-646. Reference 1, cited for the problem of disentangling the basis of consciousness from the basis of report.
- Block (2005), LIT-629. The “front versus back” framing is the access NCC versus the phenomenal NCC under other names.
- Integrated information theory (LIT-445, NOTE-365) is classed as a “back of the head” theory and described in one line; the paper takes no view on it.
- Butlin et al. (LIT-056). Their indicators come from global workspace and higher-order theories, the prefrontalist side; the paper cites the related 2017 Science manifesto (Dehaene et al.), not the report.
Bearing on the record
- THEORY-102. That theory’s data come from no-report trials; this paper argues no-report trials leave post-perceptual cognition in place. Its finding of transient prefrontal content at onset is therefore ambiguous between perception and cognition about the image, while its sustained ventral temporal code is not. A candidate point for that theory’s discussion; my reading.
- THEORY-100. Its
promote_whenasks for a no-report paradigm in birds. This paper argues that would leave the access/phenomenal question open; a no-post-perceptual-cognition design would be needed. - No ML instruction.
Limitations
- The positive case is one fMRI null result, from a design built for a different purpose (avoiding attention to switches), and Block concedes the changes may be too fine for fMRI.
- The method is specific to rivalry; Block lists its extension as an outstanding question.
- The higher-order theory in which the higher-order state is post-perceptual is ruled out of court by the paradigm, as Block notes; versions in which it is a simultaneous pointer are not.
Open questions
- How could the no-post-perceptual-cognition principle be extended beyond binocular rivalry?
- If consciousness and cognition have entirely distinct bases, could a creature be radically wrong about its own conscious perception?
- Could some vegetative patients be conscious without even the cognition needed to imagine playing tennis?
Corrections
- none to a seeded skim (there was no seed)