Multiple Drafts...

Response to Glicksohn and Salter in Behavioral and Brain Sciences, vol. 18, no. 4, 1995, pp. 810-11.

Daniel Dennett and Marcel Kinsbourne

Multiple Drafts: An eternal golden braid?

We have learned that the issues we raised are very difficult to think about clearly, and what "works" for one thinker falls flat for another, and leads yet another astray. So it is particularly useful to get these re-expressions of points we have tried to make. Both commentaries help by proposing further details for the Multiple Drafts Model, and asking good questions. They either directly clarify, or force us to clarify, our own account. They also both demonstrate how hard it is for even sympathetic commentators always to avoid the very habits of thought the Multiple Drafts Model was designed to combat. While acknowledging and expanding on their positive contributions, we must sound a few relatively minor alarms.

Glicksohn assumes that later drafts are "more advanced," but this flirts with the pervasive mythology of the trip to the summit of consciousness, "advancing" all the way. At any time, we may suppose, there are just as many drafts that are, as it were, over the hill and degenerating. Better yet: there is no non-arbitrary way to identify a summit--a place where the "most advanced" drafts pass by for inspection.

The main thrust of Multiple Drafts is well captured by Salter in his penultimate paragraph. Enough information may often be available to fuel more than one version of reality. Then drafts compete in Pandemonium-like rivalry (Dennett 1991) and the rivalry is resolved in favor of one over the rest (the one that "makes most ecological sense")--but not for good. The competition is never- ending. There is no definitive or archival draft. As Salter remarks, update frequency, great in the early stages, declines as perception recedes into memory. Reinterpretation in memory may continue indefinitely--is this, perhaps Hofstadter's (1979) "eternal golden braid"?--depending on the contextual (cuing) conditions that from time to time arise.

As Glicksohn observes, percepts do not instantaneously arise in the mind in their full richness. Not only microgenesis, but any perceptual theory would acknowledge that, and would posit intermediary ("incomplete") stages or representations that precede in rapid succession the (definitive?) conscious percept. Microgenetic theory differs from current information processing stage theories in that it characterizes this progression as gradual differentiation (of initially global percepts) rather than stepwise integration (of fragmentary sense data). Glicksohn and Salter both wonder what might be the fate of preliminary stages (drafts). Once superseded, could they be recovered? As Salter contends, we "relegate their status to a relatively background category of residual brain activity". That makes them poor candidates for episodic remembering, but leaves them the potential of exerting "unconscious" priming effects. But before being superseded, Glicksohn asks, could a draft that is customarily preconscious under extreme circumstances be experienced? After all, as Salter points out, a draft can be meaningful although preconscious. Glicksohn raises the interesting possibilities that "contemplation" could simulate such a lesion effect and that hypnotic regression might reinstate in awareness a draft that long ago faded into oblivion. A radically altered context might so powerfully bias the competition between drafts that a new winner takes all.

The Multiple Drafts Model should not be taken to rule out these possibilities. We suggest, for instance, that if a transformation from draft to draft is slowed down by pathology, then before it is revised the earlier draft might persist long enough to enter into the patterned activity that underlies current experience (Kinsbourne 1988).

In spite of these useful clarifications, both authors exhibit lapses into Cartesian thinking--or at least ominously Cartesian ways of expressing themselves. There does not exist, to use Salter's words, a process such as "recruitment of consciousness"(into what?), nor any place where the "vehicle's arrival" is recognized (by whom?). As both authors acknowledge in other contexts, no central experiencer confers a durable stamp of approval on any particular draft.

Salter asks, "what is the source of editorial control"? We consider the "editing" to be accomplished top-down feedback from circuitry that embodies constraints garnered from experience, expectancy and context. The editing is not an exogenous act of supervision, but part of the self-organizing functioning of the network, and at the same level as the circuitry that conveys information bottom-up.

Salter suggests an average "editorial" period of about 200 msec. Perhaps this is about as long as a representation can be modified before it has been around long enough to be incorporated into consciousness' dominant focus? (Kinsbourne 1994). An event that is anticipated might more summarily pass through the editing process than an ambiguous one. Like Salter, we avoid the trap of treating editing time as a constant (the perceptual moment) and thereby conserve the flexibility of the Multiple Drafts Model.


Dennett, D. C, 1991, Consciousness Explained, Boston and New York: Little, Brown.

Hofstadter, D. R., 1979, Gödel, Escher, Bach: An Eternal Golden Braid, New York: Basic Books.

Kinsbourne, M., 1988, "Integrated Field Theory of Consciousness," in A. J. Marcel and E. Bisiach, eds., Consciousness in Contemporary Science, Oxford: Clarendon Press, pp.239-256.

Kinsbourne, M, 1994, "Models of Consciousness: Serial or Parallel in the Brain," in M. S. Gazzaniga, ed., The Cognitive Neurosciences, Cambridge, MA: MIT Press, pp.1321-1330.