Contribution
It sorts the “binding problem” into several problems. The main cut is between binding as the brain’s recovery of a stimulus’s unity, which can happen in an anaesthetized animal, and binding as the construction of a unified experience, which can happen with no unified stimulus, as in random-dot stereograms and dreams. It then separates three levels of description and five kinds of phenomenal binding, and uses the scheme to say which empirical work on synchrony bears on consciousness.
Key insight
Stimulus-related and consciousness-related binding dissociate in both directions: synchrony to a real object under anaesthesia has no phenomenal unity to correlate with, and a hallucinated object is unified with no unity in the stimulus. So evidence for one is not evidence for the other, and only studies in which perceptual unity is illusory or hallucinatory, in conscious humans, isolate the second.
Assumptions
- Phenomenal unity is real and describable at its own level, “by describing actual contents of consciousness”.
- Lower levels matter to consciousness research only as mechanisms of phenomenal binding: cognitive and neural descriptions are “of interest to consciousness studies only insofar as they can be … related to integration at the phenomenal level”.
- A direct neural correlate of a content must differ when the content differs. This premise drives §5’s conclusion.
Key results
- Two conceptions (§1). Stimulus-related binding (Zeki’s formulation): “any neural processing that represents stimulus unity, regardless of whether the activity in question is conscious or unconscious”. Consciousness-related binding (Damasio’s): how “internal subjective percepts, as phenomenally unified entities” are constructed by the brain. They “do not necessarily have anything to do with each other”. “As far as I can tell, there is only one author in the literature, namely, John R. Smythies, who clearly distinguishes these two types”, as BP1 and BP2 (1994).
- Three levels (§2): phenomenal, neural mechanisms, cognitive mechanisms. “For consciousness studies, the binding problem is primarily the problem of integration at the phenomenal level”; for neuroscience it is integrating single-neuron activity into “functional neuronal groups”; for cognitive science, unifying the outputs of modules. Searle’s objection to Crick and Koch is read as a confusion of levels.
- Five kinds of phenomenal binding (§3), each with a dissociating deficit: property binding (achromatopsia shows a feature can be lost alone; illusory conjunctions show failure), part binding (some prosopagnosia), semantic-conceptual binding (semantic dementia; vague dream objects), location binding (Bálint’s syndrome, dorsal simultanagnosia) and serial or event binding (akinetopsia; identity changes in dreams). Spatial grouping (apperceptive agnosia) is given first and folded into property binding. Above them is the global unity of one world with one self in it, Metzinger’s “phenomenal Holon”.
- Synchrony (§§4–5). Human EEG studies of coherent stimuli show stimulus-related synchrony; illusory-triangle and Dalmatian studies are confounded with target detection. In his own study (Revonsuo et al., 1997; N = 8, 20 electrodes), viewing a random-dot autostereogram as coherent or incoherent gave no difference in 36–44 Hz power, while 40-Hz power rose 500–300 ms before subjects reported the gestalt. He concludes that 40-Hz synchrony is not the direct neural correlate of the content of visual awareness, but accompanies its construction, which “lends support” to Engel et al.’s claim in the same issue that temporal binding is a prerequisite for access to consciousness.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Stimulus-related and consciousness-related binding are distinct and dissociable | strong, conceptual | §1; anaesthesia and stereogram/dream cases |
| C2 | For consciousness studies the binding problem is integration at the phenomenal level | interpretive | §2, a proposal about the field’s priorities |
| C3 | Phenomenal binding is several mechanisms working partly independently | moderate | §3, single dissociations from neuropsychology |
| C4 | 40-Hz synchrony accompanies constructing a unified percept, not maintaining it | weak | one EEG study, N = 8, scalp power only |
Concepts
- Stimulus-related binding: neural recovery of a stimulus’s unity, conscious or not.
- Consciousness-related (phenomenal) binding: the integration of phenomenal contents into unified wholes.
- Levels of description: phenomenal, neural, cognitive.
- Property, part, semantic-conceptual, location and serial binding: the five local kinds of phenomenal unity; global unity is the Holon.
Connections
The cut is Smythies’s BP1/BP2, which Revonsuo credits and quotes. Metzinger’s “phenomenal Holon” (1995) is his source for global unity; the record holds Metzinger’s self-model account (THEORY-063), which keeps the unity of experience apart from the self-model. The property, part and location kinds are Treisman’s (1996). Engel, Fries, König, Brecht and Singer’s paper in the same issue is the synchrony thesis he tests.
Bearing on the record
- Which paper draws the distinction. This paper draws a distinction between integration that need involve no consciousness and phenomenal unity, and it is its main thesis, so the record’s identification of A27’s “Revonsuo’s 1999 treatment” with this paper fits. Two cautions. Its terms are “stimulus-related” and “consciousness-related”, not “functional” and “phenomenal”; “functional” appears only of neuronal groups. And Revonsuo does not claim the distinction as his: he credits Smythies (1994). Whether the issue introduction also draws it was not checked; that copy could not be fetched.
- CLAIM-355, “It does not solve phenomenal binding”. The claim quotes A27’s five bindings, of which “functional and temporal” are “the middle two”. Revonsuo’s five are a different list: property, part, semantic-conceptual, location and serial binding, all phenomenal. His serial binding is the nearest to A27’s temporal; nothing of his is “functional”. So A27’s five-part scheme is not Revonsuo’s; the record does not attribute it to him and should not. Revonsuo does support the claim’s separation: neural integration of a response can be real while saying nothing about phenomenal unity.
No instruction for machine-learning practice; nothing for the anthology.
Limitations
- A conceptual paper with one experiment, the author’s own, reported in summary. The null result on sustained viewing is an absence of a difference in scalp band power, which cannot rule out synchrony at scales EEG does not see; the paper does not discuss this.
- The dissociations are single, not double, for most of the five kinds, and the deficits are cited, not analysed.
- The distinction is not original, by the author’s own account.
- Errors and slips. None of substance found. “Reprsentative” and “entites” are typos in the published text.
Open questions
- Does the issue introduction (Revonsuo and Newman, 1999) draw a functional/phenomenal distinction in those words?
- Where does A27’s five-binding scheme come from?