Multiple drafts model

From Scholarpedia
Daniel Dennett and Kathleen Akins (2008), Scholarpedia, 3(4):4321. doi:10.4249/scholarpedia.4321 revision #89051 [link to/cite this article]
Jump to: navigation, search

The multiple drafts model of consciousness (Dennett, 1991, 1996, 1998, Dennett and Kinsbourne, 1992) was developed as an alternative to the perennially attractive, but incoherent, model of conscious experience Dennett calls Cartesian materialism, the idea that after early unconscious processing occurs in various relatively peripheral brain structures "everything comes together" in some privileged central place in the brain–which Dennett calls the Cartesian Theater --for "presentation" to the inner self or homunculus. There is no such place in the brain, but many theories seem to presuppose that there must be something like it. For instance:

  1. They postpone indefinitely the task of saying where and when the results of all the transformations and discriminations are “made available to conscious awareness,” which suggests sending their products to some higher center (for what purpose? what would happen there?), or
  2. They argue that a decision to move a limb takes several hundred milliseconds to “rise to consciousness,” creating an ominous picture of human agents as deluded about their ability to make a conscious decision (e.g., Libet, 1985, Libet et al., 1999, Wegner, 2002), or
  3. They suppose that the transduction by sense organs of light and sound and odor and so forth into an unconscious neural code must be followed (somewhere in the brain) by a second transduction into some other “medium”, the medium of consciousness (e.g., Mangan, 1993).


Escape from the Cartesian theater

Why is the Cartesian Theater such a seductive idea? Mainly, Dennett thinks, because since we become conscious of various features of our experience, there must be some kind of transition: if not arrival at a place or crossing of a boundary or translation into a new format, then a change of functional state of one sort or another. Our conscious experience is of events that can usually be objectively timed quite precisely; lights going on, words being spoken, clock chimes, and gunshots are salient examples, and in general we believe, for good reason, that our conscious, subjective experience tracks objective events with impressive temporal accuracy. It seems at first that there must be a quite specific moment at which each item makes its entrance in our experience. If there is no such second transduction, occurring at a particular time and place in the brain shortly after the objective event has been transduced by our peripheral sense organs, how can we understand the difference between those brain processes that involve contents but are somehow “outside of consciousness” and those that somehow subserve conscious experience? The short answer is that if we scrupulously honor the distinction between the timing represented in consciousness and the timing of the conscious representing, we can see that the difference between unconscious and conscious is, like the difference between night and day, huge but gradual. A temporally punctate event need not make the transition from unconsciously discriminated to consciously experienced in a temporally punctate moment. The multiple drafts model is an attempt to show how this can be true, and how it makes sense of otherwise paradoxical findings. Since it is primarily concerned with subverting the powerful intuition that no possible model could avoid a commitment to a Cartesian Theater, it is a deliberately noncommittal sketch–not tied to the specifics of neuroanatomy yet to be determined. A wide variety of quite different specific models of brain activity could qualify as multiple drafts models of consciousness if they honored its key propositions:

  1. The work done by the imaginary homunculus in the Cartesian Theater must be broken up and distributed in time and space to specialized lesser agencies in the brain.
  2. Once these specialists have done their work, that work doesn’t have to be done again in a central re-presentation process. That means that the content involved doesn’t have to be perceived again, discriminated again, enjoyed again, abhorred again (if it is, for instance, a pain) nor does it have to be moved somewhere and presented again in order to be stored in memory.
  3. There is a massively parallel process in the brain–in the cortices and subcortical structures they interact with–in which multiple (and often incompatible) streams of content fixation, transformation, influence, suppression, enhancement, “binding”, memory-loading, etc., take place simultaneously (and asynchronously). These are the multiple drafts out of which the appearance of a ‘final draft’–the imagined draft of consciousness enacted on the imagined stage of the Cartesian Theater–is created by the occurrence of “probes” that retrospectively elevate some drafts at the expense of others. In the absence of such probes, the question of whether or not a content was conscious is ill-posed. (This will be explained below.)
  4. “Since you are nothing beyond the various subagencies and processes in your nervous system that compose you, the following question is always a trap: ‘Exactly when did I (as opposed to various parts of my brain) become informed, aware, conscious, of some event?’ (Dennett, 1998, p105) It is a trap in the sense that it may not have, or need, an answer because it has false presuppositions.

Fame in the brain

In later work (1996, 2001, 2005) Dennett replaced the metaphor of multiple drafts with the metaphor of fame in the brain. Just as becoming famous is not a precisely datable event like being transduced into a medium (like being televised), so achieving fame (or ‘clout’) in the brain is not a precisely datable transition in the brain. It is a competitive phenomenon --not all can be famous--and it is only retrospectively determinable since it is constituted by its sequelae. (As he never tires of insisting, one must always ask the Hard Question “And then what happens?”) One cannot, logically cannot, be famous for fifteen minutes; that would not be fame. And a content cannot be conscious for fifteen milliseconds and utterly forgotten afterwards; that would not be consciousness. The fact that the hypothesis seems to make sense is due to the mistaken imagery with which we adorn the everyday concept of consciousness.

The transition from unconscious to conscious is real, but since it involves the accumulation of a wide variety of sequelae which may become undone and redone, overruled and then rehabilitated over a longish period of time (by the standards of experimental cognitive science) it is only in special and extreme cases that one can motivate the precise timing of the transition. Remember: the work that would be done by the homunculus in the Cartesian Theater is distributed in time as well as space. The tennis player returning serve does not have to wait 300 milliseconds to become “fully conscious” of the serve’s trajectory before shaping her return, and a quick-witted pun can be on your lips before you are “fully conscious” of the words you are responding to. You can’t issue clever retorts to speech acts you are never conscious of, but your consciousness of them is not a pre-condition for framing your response. And if, for one reason or another, your emergent pun is aborted before it is uttered (it would have been in bad taste, perhaps) it may not even “cross your mind”–consciously. Were you conscious of thinking of it and then inhibiting its expression? That is just one possibility; it may have been considered and rejected so briefly and inconsequentially (it has no sequelae) that no retrospective account of what it was like to have that conversation would include it. In normal conditions (when our brains are not given tasks beyond their competence), there is a consilience of contents and actions that encourages us to see the work done by the distributed agencies as the work of a single, rational ego residing in headquarters, but at other times our content-determinations can become disjointed and delayed with comic effect, as when you pull back the bedcovers and see–“Eeek!”–and run out of the room screaming “It’s only a plastic snake!” Indeed it is, so why are you running and why are you screaming? One team in your brain has taken charge while another team is still sorting out the implications.

There is probably still a temptation to insist that there just has to be an answer to the question of whether or not the content in question “reached the level of consciousness.” Doesn’t an objective science of consciousness have to discover a “neural correlate of consciousness” or NCC (Koch 2004, and many others) that is either present or absent in such instances, or else concede defeat? No. Like the transition from night to day, the transition from unconscious to conscious can be temporally smeared, leaving us with no motivated, non-arbitrary way of drawing the called-for line. It can be objectively true that you were conscious of an event (as subsequent events make manifest) without there being any objective truth about precisely when you became conscious of the event!

The onset of consciousness can only be roughly timed in retrospect–just like speciation

A more instructive parallel than night and day—since day always follows night, and consciousness doesn’t always follow unconsciousness—is speciation, in which the same curious retrospective status can be transparently observed. Every birth in every lineage is a potential speciation event but not one in a million turns out, in the fullness of time, to have been a speciation event. It is logically impossible to discern which births are speciation events at the time they occur, because nothing about them at the time distinguishes them; that status depends on whether or not they lead to further births and still further births, and the eventual triumph of that lineage of births over others. Speciation events are discriminable only by “retrospective coronations” (Dennett, 1995). In other words, there is no temporally local feature to be the BCS, the biological correlate of speciation. Similarly, those content-fixations in the brain that turn out to be have been conscious are those that happen to have the sequelae that ensure their fame–and this depends not just on their temporally and spatially local properties, but on the subsequent competition for fame with their rivals. Precisely when did this content or that become conscious? That is an ill-posed question, for exactly the same reason that the parallel question is ill-posed in biology.

Consider a thought experiment about a remarkable possibility: at this time, there is just one species of hominid on the planet, Homo sapiens. But suppose that fifty years from now all but a lucky handful of our descendants are wiped out by a virus that leaves just two groups of survivors: a thousand Inuit living on Cornwallis Island off Greenland, and a thousand Andaman Islanders living in the Indian Ocean. These two populations continue to be both geographically and reproductively isolated, we can well suppose, and are physiologically quite different. Imagine that they go on for another thousand years, eventually repopulating the globe with two species–as they learn when they finally encounter each other and discover that their efforts at mating are fruitless. So this would be a standard case of allopatric speciation, produced over time by geographical isolation. They might wonder: when precisely did the speciation occur? It is probably the case that their last common ancestor lived more than thirty thousand years ago, but speciation didn’t occur then and there, and not for some time afterwards. Did speciation occur before the dawn of agriculture or after the creation of the Internet? There would be no motivatable, non-arbitrary answer. The offspring of their last common ancestor could turn out to have been the founders of two different species, but it still isn’t a settled fact today, one way or the other, about whether a speciation event started then. Here we would have an event, a birth that could turn out to have played a pivotal role in human history, that occurred in a precise time and place but didn’t acquire its special status until millennia of sequelae had fixed that role, which was never a foregone conclusion. All it would take to prevent that birth from ever having been a speciation event would be a single boatload (or planeload) of islanders making a journey that led to the “premature” reuniting of the branches. One can imagine that some precise if unknowable moment in the interval between separation of the lineages and demonstration of their status as species is the moment when enough chromosomal divergence had accumulated in the two gene pools so that any cross-lineage matings would have been infertile, a necessary but insufficient condition for speciation, but surmises about such counterfactuals are of scant significance. In a parallel spirit we could say, if we felt the need, that Arthur P. Snerdly (who?) of Los Angeles, California would have become famous on June 6, 1968 if Sirhan Sirhan hadn’t shot Robert Kennedy there the day before. (Arthur set a new record for eating pizza on June 5, unlike his twin brother, Rufus, who spent the day in bed and would never have become famous for anything. So what? Arthur still wasn’t famous, was he, in spite of his feat? Still, Arthur had a necessary condition for fame that his brother lacked, a notable difference. We can expect to find, and time the onset of, necessary conditions for fame in the brain as well, but when sufficient conditions ripen slowly and uncertainly over longer periods of time, identifying these onsets of necessary conditions as the onset of consciousness is at best arbitrary and misleading, since in many critically similar instances, these onsets will not lead to the sequelae we use as our anchoring symptoms of consciousness.)

We can discern a parallel puzzle arising for the timing of consciousness in the phenomenon of coming to notice the chiming of a clock. Only on the fourth peal, perhaps, did you become aware of the clock chiming but you discover you can retrospectively count the chimes in conscious memory. But when were you first conscious of the first chime? When it happened, five seconds ago, or just now, when you experienced it in recollection just as the fifth chime pealed? Or at some time in between? If the clock had chimed just three times, would you have ever been conscious of those chimes? We could no doubt discover a host of processes initiated in your brain by the discrimination of those three chimes at the time they occurred, but would those processes have been “enough for consciousness” (of a sort)? How much is enough? How much influence is enough for fame? How many descendants are enough for a species? There is nothing either mysterious or “verificationist” about the observation that events can come to have been conscious without there being a moment when some threshold was crossed marking the onset of consciousness. (And this allows that in the normal course of events, we can roughly time the onset of consciousness, just as we can roughly time speciation events.)

Beyond these analogies, the literal claims of the multiple drafts model are so bland as to seem to some critics to be a denial of the very existence of consciousness–nothing dramatic happens:

Contents arise, get revised, contributed to the interpretation of other contents or to the modulation of behavior (verbal or otherwise), and in the process leave their traces in memory, which then eventually decay or get incorporated into or overwritten by later contents, wholly or in part. This skein of contents is only rather like a narrative because of its multiplicity; at any point in time there are multiple drafts of narrative fragments at various stages of editing in various places in the brain. . . . Probing this stream at various intervals produces different effects, precipitating different narratives—and these are narratives: single versions of a portion of ‘the stream of consciousness.’ If one delays the probe too long, the result is apt to be no narrative left at all. If one probes ‘too early’ one may gather data on how early a particular discrimination is achieved in the stream, but at the cost of disrupting the normal progression of the stream. (Dennett, 1991, p135-6)

That is the point of the multiple drafts model: consciousness is not what it seems to be (a magic show illuminating an inner stage). It is the differential influence of various contents on the processes that control the body of an agent composed of those processes and capable of telling us, retrospectively, about some of them. What we can later tell about is very much a matter of how we are probed—or how we probe ourselves—as the multiple drafts churn along.

Probes, the ability to recollect, and the personal level of explanation

What, then, is a probe? It is not some special sort of event unlike all the other events occurring in the brain (how could it be?) that somehow creates consciousness when it happens. It is just whatever event in the brain happens to boost some aspect of the current content-fixations into prominence, fame in the brain. In the simplest case, a probe is a new stimulus that draws attention (resources) to a particular area of visual space or a particular segment on the auditory stream, for instance, thereby promoting the influence (the fame, the clout) of whatever is occurring there and rendering it reportable and recollectable–if the other drafts competing for this influence permit it. They may not, for any of a hundred reasons, in which case the content in question, in spite of its alliance with a probe, fails to achieve enough fame to enter the rolls of history in memory. An endogenously generated probe (such as the direction of attention that ensues when you decide to try to study the content of your parafoveal vision) elevates contents in that “background” to prominence, but this is not the prominence, the influence or clout, those contents would have had anyway in the absence of the probe. The idea that there is a probe-independent standard of intensity or salience that guarantees “entry into consciousness” is an undesirable byproduct of Cartesian materialism, and more particularly of the popular but seldom stated background image of consciousness as something very special that is somehow kindled when some critical mass or threshold is reached. When we replace that image with a model that stresses the competitive nature of the dynamic processes that shape the multiple drafts, we can appreciate that when the competition is stiff, intensities or saliences that under other circumstances would ensure fame, may fail to make an impression, and when competition is relaxed, relatively non-urgent non-starters can sidle into consciousness more or less by default.

Let’s consider a standard case in more detail than usual. What were you conscious of during that half-hour commute home in your car? Retrospectively, all you can report now, as you turn off the ignition, is your concentrated musings about the essay you are working on (which you can recall with considerable detail), oh, and noticing some trash by the side of the road as you turned the last corner, and, now that you think about it hard, a few vague recollections of your hands on the steering wheel at some point. (That’s why you’re not actually surprised to find yourself sitting in your car in the garage; you do remember, sort of, driving home.) So you were certainly conscious of all that, since now you’re recalling it, but if you hadn’t worked so hard just now trying to recall details, there would soon be no recoverable trace of those details. The question of whether you had been, nevertheless, really conscious of them “in the background” all along suggests itself. What could give this question meaning? Only some tell-tale talent, some temporary, unactualized disposition that distinguished these contents from those that were truly or more deeply unconscious–analogous to the accumulated chromosomal incompatibilities in the gene pools that marked a necessary but not sufficient condition of speciation. And of course the temporary presence of this disposition would have to be indirectly confirmed by, for instance, evidence of neural activity of a sort established to subserve the competence when the disposition is actualized. Thus Lamme (2003) has argued that recurrent feedback processes between earlier and later visual areas are a necessary but not sufficient condition for the sorts of high-level discriminations that we can make with contents that are definitely conscious. Discovering that such recurrent processes are not in evidence when people drive “on auto-pilot” would pretty much clinch the case that their driving was unconscious. Finding the recurrent processes (which is likely, given the adroitness exhibited by those who drive on auto-pilot) would still leave open the question of whether to call those contents conscious or merely potentially conscious.

What do two conscious contents have to have in common? Only the historical property of having won a temporally local competition with sufficient decisiveness to linger long enough to enable recollection at some later time. They may have other properties in common, and it could emerge that the presence of some such property explained--in normal circumstances--the competitive advantage enjoyed by the winners, in which case we would have a strong candidate for a neural correlate of consciousness, and, hence, third-person grounds for overriding first-person authority about what was conscious and what wasn’t. We should not be worried, however, if no such tell-tale property arises. Our scientific understanding of the difference between night and day is not jeopardized by our inability to give a non-arbitrary dividing line between the two.

What, then, is the importance of (verbalizable) recollectability? Intuitively, the ability to report a content is conclusive evidence of consciousness. Why should this be? Not because it is an infallible sign of what is going on in one’s mind–after all, a subject’s later recollections can fade or become distorted–but because our interpersonal communications, our discussions and comparisons, generate both the terms and the topics of consciousness. The personal level of explanation is defined by the limits of our abilities to respond to queries about what we are doing and why. What is off-limits to such inquiries, however cognitive or sophisticated, is sub-personal and unconscious. A reported episode or nuance, current or recollected, has left the privacy of the subpersonal brain and entered the interpersonal public forum of consideration and comparison. How are we able to do this, when we cannot similarly “introspect” the private processes occurring in our kidneys or immune systems? The details of the answer are still to be worked out, but the main point to make is simply that any evolved and matured capacity to frame and utter speech acts must identify a domain of topics or contents about which such speech acts can be controllably formed. What we can talk about in public if we choose is, ipso facto, what we are conscious of. In species that lack anything functionally analogous to this “publication” competence, it remains an open question whether anything like the personal/subpersonal distinction can be drawn, and that is why the attempts to extrapolate claims anchored in human consciousness to claims about the consciousness or lack thereof in other species are currently so imponderable. Until we have developed accounts of the specific sorts of competitions that sort out the use of resources in the brains of these species, we won’t have any leverage for settling such questions.


  • Dennett, D.C. (1991) Consciousness Explained. Boston: Little Brown.
  • Dennett, D.C. (1995) Darwin’s Dangerous Idea. New York: Simon & Schuster.
  • Dennett, D.C. (1996) Consciousness: More like Fame than Television (Bewusstsein hat mehr mit Ruhm als mit Fernsehen zu tun). In Die Technik auf dem Weg zur Seele. Christa Maar, Ernst Pöppel, and Thomas Christaller, eds. Rowohlt.
  • Dennett, D.C. (1998) The Myth of Double Transduction. In the volume of the International Consciousness Conference, Toward a Science of Consciousness II, The Second Tucson Discussions and Debates, S. Hameroff, ed., A.W. Kaszniak, and A.C. Scott, MIT Press, 1998, pp. 97-107.
  • Dennett, D.C. (2001) Are we explaining consciousness yet?. Cognition, 79, pp. 221-37.
  • Dennett, D.C. (2005) Sweet Dreams: Philosophical Obstacles to a Theory of Consciousness. MIT Press.
  • Dennett, D.C., and Kinsbourne, M. (1992) Time and the Observer: The Where and When of Consciousness in the Brain. Behavioral and Brain Sciences, 15, 183-247.
  • Lamme, V.A. (2003) Why Visual Attention and Awareness are Different. TRENDS in Cognitive Sciences. 7(1): 12-18.
  • Libet, B (1985) Unconscious Cerebral Initiative and the Role of Conscious Will in Voluntary Action. Behavioral and Brain Sciences, 8, pp.529-66.
  • Libet, B, Freeman, A., and Sutherland K. (1999) The Volitional Brain: Towards a neuroscience of free will. Thorverton, UK: Imprint Academic.
  • Mangan, B. (1993) Taking Phenomenology seriously: The fringe and it implications for cognitive research. Consciousness and Cognition, 2 89-108.
  • Wegner, D. (2002) The Illusion of Conscious Will. Cambridge, MA: Bradford Books/MIT Press.

Internal references

  • Howard Eichenbaum (2008) Memory. Scholarpedia, 3(3):1747.
  • Almut Schüz (2008) Neuroanatomy. Scholarpedia, 3(3):3158.

Recommended reading

  • None

External links

See also

Attention, Consciousness, Models of Consciousness

Personal tools
Focal areas