NOTE-367: Dennett & Kinsbourne — Time and the Observer

Contribution

The paper names a hidden assumption, Cartesian materialism, and shows it at work in the interpretation of four temporal illusions and in the arguments of both sides of the Libet controversy. It offers the Multiple Drafts model in its place. The new tool is the distinction between the time a content is represented and the time it represents, applied to the brain’s timing. With it the paper argues that some questions about when an experience happened have no answer at fine scales.

Key insight

A letter dated before the battle can arrive after news of the battle, and a reader who sees the date still knows the order. The brain can likewise represent “A before B” without representing A first. So the experienced order of events is a content that some process concludes, not the order in which the vehicles of content arrive anywhere. Once there is no place where arrival counts, “was this conscious before or after the revision?” can lose its answer.

Assumptions

• Materialism, and no central gateway. “Nothing in the functional neuroanatomy of the brain suggests such a general meeting place” (p. 185). This is asserted with the state of neuroanatomy in 1990, not argued.
• Subjective sequence is accessed through report. “Only through the subjects’ verbalizations … do we gain access to a perspective from which the anomalies can appear” (p. 187). Note 2 sets aside the solitary experimenter and languageless animals.
• Causes precede effects throughout: what the model rules out is “magical” or precognitive filling in (p. 190).
• Scale. The claims about indeterminate timing are restricted to intervals “of a few hundred milliseconds” (p. 200). At ordinary scales the experienced order of thoughts is granted to be their actual order.

Key results

• Four phenomena (§1.2).
  • Colour phi (Kolers and von Grünau 1976): a red spot followed by a green one 4° away is seen to move and change colour in mid-path.
  • The cutaneous rabbit (Geldard and Sherrick 1972): taps at the wrist and elbow are felt as evenly spaced hops up the arm.
  • Libet’s backward referral: skin stimuli are felt as earlier than cortical stimuli that precede them.
  • Libet’s delay of intention: reported intentions lag the readiness potential by 350–400 ms.
• Representing vs represented time (§2.1). The Battle of New Orleans example (the battle was fought on 8 January 1815, fifteen days after the truce). “Content-sensitive settling”, by analogy with synchronising a soundtrack to film and with random-dot stereograms, shows how order can be inferred without date stamps. Mellor’s claim that “perceptions of temporal order need temporally ordered perceptions” is rejected, with what is right in it kept: timing matters at the start, when it fixes content, and at the deadline, when content must be in time to control behaviour.
• The temporal control window (§2.1). Within the window bounded by the earliest arrival of information and the latest time it can still control a behaviour, the vehicles’ timing can be moved freely.
• Orwellian vs Stalinesque (§2.2). Two accounts of colour phi:
  • Stalinesque: an editing room before consciousness waits for the green spot and splices in intermediate frames.
  • Orwellian: the subject is briefly conscious of the stationary red spot, and memory is rewritten. The button-press latency favours the Orwellian story. The Stalinesque theorist replies that the press was unconsciously triggered. Both fit all verbal, behavioural and first-person data, and metacontrast gives the same standoff. So “there is really only a verbal difference between the two theories” (p. 193).
• No filling in (§2.2). The brain “just adjusts to the conclusion that is drawn”; it need not construct the intervening frames, “for no one is looking” (p. 194).
• Libet re-read (§3). His account of cortical stimulation is called Stalinesque, and Churchland’s “go” reaction-time critique Orwellian; both assume an absolute time of experience. In the intention experiments the clock-position judgment is “an artifact of the experimental situation” (p. 198). A cortex-wide “mode shift” model of consciousness is “still the Cartesian Theater” if its objective timing is taken as definitive (p. 199).
• Grey Walter’s carousel (§3.3). Patients whose motor-cortex signal advanced the slides felt the projector anticipate their decisions. The paper reads this as a violated expectation of feedback, not as a timing of the conscious decision.

Claims

id	claim	strength	support
C1	There is no place in the brain where contents must arrive to be conscious	moderate	neuroanatomy asserted (p. 185); the homunculus-regress argument (p. 185)
C2	The brain can and often does represent temporal order by content rather than by the timing of vehicles	moderate	engineering argument; Efron (1973) on order-dependent discriminations below the temporal-order threshold (p. 190)
C3	Stalinesque and Orwellian accounts of colour phi and metacontrast are empirically indistinguishable, including from the first-person perspective	moderate for the cases analysed	§2.2, case by case
C4	So at sub-second scales the distinction between them has no fact of the matter	weak–moderate	the step from C3 to C4 is verificationist; Harnad’s dissent (note 11) is that it is a measurement problem, not a metaphysical one
C5	Libet’s backward referral and subjective delay of intention are artifacts of conflating time represented with time of representing	moderate	§§3.1–3.2; Libet’s own concessions quoted (pp. 197–198)
C6	The editorial processes seen at short scales continue indefinitely in memory	weak	§4, a sketch from the simplification of remembered events
C7	A portion of the world that composes “a skein of narratives” is an observer	assertion, offered as “a fundamental metaphysical category”	p. 196, one sentence

Concepts

• Cartesian materialism / Cartesian Theater. The view left when Descartes’s dualism is dropped but a central “place where it all comes together” is kept (p. 185).
• Multiple Drafts model. Distributed, once-only discriminations, continually revised; probing at different times yields different narratives; no “final” or “published” draft is canonical (p. 185).
• Orwellian revision. Post-experiential falsification of memory.
• Stalinesque revision. Pre-experiential falsification, a staged “show trial” that is then accurately remembered.
• Content-sensitive settling. Inferring order by aligning contents, as an editor slides a soundtrack against film (p. 189).
• Temporal control window. The interval within which a representation’s timing does not matter, bounded by causation at one end and the deadline of the behaviour it controls at the other (p. 190).

Connections

• Minsky. The Society of Mind (LIT-429, filed unread) is cited once, as “Minsky 1985, p. 61”, for the “which finish line” problem (p. 184). The memo the record has read (LIT-399, NOTE-344) shares the denial of a central controller. The paper’s units are discriminations and editing processes, not agents, and it never describes the brain as a society.
• The Emotion Machine (LIT-432, NOTE-348). Its C5, the “Immanence Illusion”, quotes Dennett and Kinsbourne for the claim that no resource can know what another is doing right now.
• Chalmers, “Facing Up” (LIT-406, NOTE-336). It gives the multiple-drafts model a one-sentence verdict: another theory of awareness, silent on experience (its C3). This paper would answer that the same verdict applies to any model that times an “onset of consciousness”, and that the hard problem’s “further question” has the same form as the question this paper calls verbal.
• Block (LIT-407). Not mentioned in the text. The paper’s regress argument against a central conscious subsystem is the converse of Block’s homunculi-head: a conscious part doing the whole’s work, not unconscious parts doing it.
• Kammerer (LIT-409, NOTE-362). Postdates the paper. See Bearing.
• Damasio. “Damasio 1989a” is cited for there being no single place where causal trains deposit their contents (p. 184). Which 1989 paper that is could not be checked, since the reference list was not in the file. The record’s LIT-385 is one candidate.
• Libet, Churchland, Goodman, Kolers, Grey Walter, Efron, Mellor. None is held in the record.

Bearing on the record

• THEORY-043. Three points, all my inferences, since the paper does not discuss groups or nesting:
  • Not a society of agents. The model shares the denial of a centre with Minsky, Selfridge and Brooks, but its parts are not agents and it says nothing about whether they could be minds. It places it on the downward side of the THEORY without adding to it.
  • A regress argument against one kind of nesting. A conscious central subsystem would duplicate the work of the whole’s other parts (p. 185). That is close to Kammerer’s functional principle: what is excluded is a part that consciously represents for the whole in a role nothing else could fill. It does not exclude minded parts that do not play that role. So it gives the THEORY’s first refusal a second, independent motivation for the intra-brain case, the motivation Kammerer’s version lacks. It still does not reach the United States. This is a candidate for the THEORY’s text, since the promotion condition asks for “an anti-nesting principle justified by something other than the absurdity of group minds”.
  • Against definite borders in time. IIT’s exclusion postulate is derived from an axiom that experience is definite, and the record’s reading of IIT 3.0 (NOTE-365) has it pick a definite spatio-temporal grain, the one at which integration peaks. This paper’s C3–C4 deny that there is a fact about which content is conscious at an instant within a few hundred milliseconds. If C4 holds, the axiom fails for time, and with it one of the THEORY’s anti-nesting routes. C4 is the paper’s weakest step.
• THEORY-023. No bearing beyond the general point that any architectural indicator timed by onset is, by this paper’s argument, the Cartesian Theater again.
• No new THEORY. The regress argument would strengthen THEORY-043, if its author judges it independent of the group verdict; it is not enough for a THEORY of its own.
• No ML instruction. Nothing here belongs in the anthology.

Limitations

• Commentary not read. The issue carries open peer commentary and a response to p. 247; I read one page of one commentary. Antony’s opening page argues that the most charitable reading of the paper commits it to eliminativism about consciousness, and that other passages say consciousness exists. How the authors answer that is unread.
• The C3-to-C4 step. Indistinguishability by every available test is taken to show that there is no fact. Harnad’s contrary view is reported fairly in note 11 and is not refuted.
• Libet’s data. The paper accepts Churchland’s technical criticisms of Libet’s cortical-stimulation studies (“very few subjects … inadequately controlled … not replicated”, p. 196) and does not re-analyse them.
• No experiment. The Grey Walter follow-up (a variable delay before the slide advances) is proposed, not run.

Open questions

• What is the delay at which Grey Walter’s “precognitive carousel” effect disappears? The paper says the experiment was never done.
• Does Antony’s charge of eliminativism stand after the authors’ response? Reading pp. 201–247 would settle what the paper is committed to.
• Can the Orwellian/Stalinesque standoff be broken by a measure the paper did not consider, for instance a no-report paradigm? The paper’s argument covers every measure that the subject’s judgments mediate.

Corrections

• none to a seeded skim (there was no seed)
• The brief’s citation (BBS 15(2):183–201, 1992, with open peer commentary) is correct; Crossref dates the issue June 1992.
• The brief calls Multiple Drafts “the multiple-drafts model” of consciousness. In this paper it is a model of subjective timing; the conclusion restricts its main claim to “temporal properties of experience” (p. 200) and defers other implications to Dennett 1991b.