COMPETING FOR CONSCIOUSNESS

How Subconscious Thoughts Cook on the Backburner William H. Calvin [4.29.98]
Topic:
Introduction By: John Brockman

According to theoretical neuroscientist Bill Calvin, treating consciousness solely as awareness or attention greatly underestimates it, ignoring the temporary levels of organization associated with higher intellectual function. "The tasks that require consciousness," he says, "tend to be the ones that demand a lot of resources. Routine tasks can be handled on the back burner but dealing with ambiguity, groping around offline, generating creative choices, and performing precision movements may temporarily require substantial allocations of neocortex."

Recently, Calvin has proposed "a specific mechanism (consciousness as the current winner of Darwinian copying competitions in association cortex) that seems capable of encompassing the higher intellectual function aspects of consciousness as well as some of the attentional aspects. It includes features such as a coding space appropriate for analogies and a supervisory Darwinian process that can bias the operation of other Darwinian processes."

"Competing for Consciousness", derived in part from Calvin's 1996 book The Cerebral Code, is presented simultaneously on EDGE and as a plenary talk for the Tucson III consciousness forum.

Francis Crick likes to observe that people once worried about the boundary between the living and the nonliving. Today, the boundary seems meaningless; we instead talk about all the varied aspects of molecular biology. Today's brain researchers think it likely that much of the present scientific and philosophical concern about consciousness will soon become equally obsolete, that we will simply come to talk of the various physiological processes involved with attention and creative problem-solving. Dan Dennett called consciousness "the last surviving mystery." A mystery, Dennett said, "is a phenomenon that people don't know how to think about - yet." Here I will attempt to clarify the appropriate levels of explanation and then propose a candidate mechanism, a Darwin Machine¹ that seems, because its circuitry is found in many parts of neocortex, capable of encompassing the higher intellectual function aspects of consciousness as well as some of the attentional aspects.

Consciousness is the tip of the iceberg, in the sense that many other things are going on in the brain at the same time, hidden from view. There are subconscious trains of thought that vie for "attention." Though the obvious analogy is to the television viewer who surfs the channels (and our nighttime dreams often seem like switching between soap operas in progress), there need not be a central place where choices are viewed. The "best" channel need only temporarily win out over the others in the battle for access to output pathways such as speech and other body movements. Soon, another channel comes to dominate and we speak of "our attention shifting" - but there need not be an agent which makes the decision or performs the action. There is nothing in this overview that demands a central place: the "center of consciousness" could, instead, shift from moment to moment: from language to nonlanguage areas, from frontal to parietal lobe, from left to right hemisphere, and maybe even from cortical to subcortical structures - anywhere, I suspect, with the potential for generatingnovel patterns of movement. Routine tasks can be handled on the back burner but dealing with ambiguity, groping around, generating creative choices, and performing precision movements may temporarily require substantial allocations of neocortex.

Rather than place, I think that we need to concern ourselves with levels - levels of mechanism from the subsynaptic to the metaphorical - and how new levels can be temporarily formed as we think about what to say next. What's missing from most discussions of consciousness is, surprisingly, the whole concept that there are levels of mechanism, or levels of explanation. Douglas Hofstadter² gives a nice example of levels when he points out that the cause of a traffic jam is not to be found within a single car or its elements. Traffic jams are an example of self-organization, more easily recognized when stop-and-go achieves an extreme form of quasi-stability - the crystallization known as gridlock. An occasional traffic jam may be due to component failure, but faulty spark plugs aren't a very illuminating level of analysis-- not when compared to merging traffic, comfortable car spacing, driver reaction times, traffic signal settings, and the failure of drivers to accelerate for hills. -----

The more elementary levels of explanation are largely irrelevant to traffic jams. Such decoupling was emphasized by the physicist Heinz Pagels³, who noted:

"Causal decoupling" between the levels of the world implies that to understand the material basis of certain rules I must go to the next level down; but the rules can beapplied with confidence without any reference to the more basic level. Interestingly, the division of natural sciences reflects this causal decoupling. Nuclear physics, atomic physics, chemistry, molecular biology, biochemistry, and genetics are each independent disciplines valid in their own right, a consequence of the causal decoupling between them.... Such a series of "causal decouplings" may be extraordinarily complex, intricate beyond our current imaginings. Yet finally what we may arrive at is a theory of the mind and consciousness - a mind so decoupled from its material support systems that it seems to be independent of them - and "forgot" how we got to it.... The biological phenomenon of a self-reflexive consciousness is simply the last of a long and complex series of "causal decouplings" from the world of matter.

Closely related is the notion of emergent properties: traffic jams and crystals emerge from combinations, and we expect emergence to play a large role in the transient levels of organization involved with higher intellectual function (language, planning, games, etc.). In our search for a level corresponding to consciousness, it is well to recall that levels arise from what Jacob Bronowski called stratified stability:

Nature works by steps. The atoms form molecules, the molecules form bases, the bases direct the formation of amino acids, the amino acids form proteins, and proteins work in cells. The cells make up first of all the simple animals, and then the sophisticated ones, climbing step by step. The stable units that compose one level or stratum are the raw material for random encounters which produce higher configurations, some of which will chance to be stable.... Evolution is the climbing of a ladder from simple to complex by steps, each of which is stable in itself.

The tumult of random combinations occasionally produces a new form of organization. Some forms, such as the hexagonal cells that appear in the cooking porridge if you forget to stir it, are ephemeral (as, indeed, are the contents of our consciousness). Other forms may have a "ratchet" that prevents backsliding once some new order is achieved. Crystals are the best known of these quasi-stable forms.

If we take consciousness, as Karl Popper did, to be important for "solution of problems of the non-routine kind," then shaping up quality courses of action in thought is a key aspect of consciousness, one that goes well beyond mere awareness or shifting attention.-----

We no longer have to take it on faith that there are mechanisms capable of recursively bootstrapping random novelties into something of quality. For the last 160 years, there has been an existence proof, the Darwinian process. 6 The way in which quality is achieved using this process has long occupied the best minds in evolutionary biology. And the slow evolution of species, on the time scale of millennia, is no longer the only example: the immune response is now known to be another Darwinian process, operating on the time scale of days to weeks as a better and better antibody is shaped up in response to the challenge of a novel antigen. For decades, computer science has used a solution-finding procedure, called the genetic algorithm6, that mimics an expanded version of the Darwinian process on a time scale limited only by computer size and speed.-----

It would be surprising if the brain did not make some use of this fundamental principle for bootstrapping quality. Can this same well-known process operate quickly enough in the brain, on the time scale of thought and action? Can it account for much of what we call "consciousness"? I undertook to answer such questions in one of my 1996 books, The Cerebral Code, which analyzes the recurrent excitatory circuitry of mammalian neocortex. I showed that this widespread wiring principle was capable of running the classical Darwinian process:

a pattern (spatiotemporal firing pattern of a Hebbian cell-assembly, in this case) thatcopies with occasional variation, where populations of the variants compete for a limited work space, their relative success biased by a multifaced environment (both memorized and real-time, in this case), and with further variations centered on the more successful of the current generation (Darwin's inheritance principle). This full-fledged Darwinian process is what is associated with the recursive shaping up of quality; it should not be confused with mere selective survival of a single pattern and other "sparse sets" that utilize only a few of the "six essentials".7

The cortical circuitry that makes a full-fledged Darwinian process possible is not an obscure feature known only to a few neuroanatomists: it is easily the most prominent wiring principle seen in neocortex, that of the patterned recurrent excitatory connections between neighboring pyramidal neurons. It has just taken a while to realize one of the implications of it, an emergent property of the circuit not possessed by any of the individual elements: synchronized triangular arrays of pyramidal neurons, with nodes about 0.5 mm apart, is what you expect to observe.

Each pyramidal neuron has an axon that branches nearly 10,000 times. Some travel through the white matter but most of the branches never leave the cortical layers, terminating in a synapse within a millimeter or so. The axon travels sideways to excite other cortical neurons, mostly other pyramidal neurons. The deep-layer (V and VI) pyramidal neurons have such sideways axons that remain within the cortical layers, some terminating nearby and others more distantly.

It's the wiring seen8 in the branching of the axon of the superficial-layer pyramidal neurons (layers II and III), however, that is so striking. Their terminations are patterned: their axons are like express trains that skip a long series of intermediate stops, concentrating their synaptic outputs in zones about 0.5 mm apart. Like an express train, these axon skip the intermediate stops. That's what makes synchronized triangular arrays likely to form on occasion.

One consequence of the express-train axon is that cells 0.5 mm apart will tend to talk to one another: they will recurrently excite. While a chasing-their-tail loop is one possibility if synaptic strengths are quite high, even weak synaptic strengths have an important consequence: entrainment. Since 1665, when the Dutch physicist Christiaan Huygens noticed that pendulum clocks on the same shelf synchronized their ticks within a half hour, much additional work has been done on entrainment. A dramatic example from the Philippines was reported inScience by Hugh Smith in 1935:

Imagine a tree thirty-five to forty feet high, apparently with a firefly on every leaf, and all the fireflies flashing in perfect unison at a rate of about three times in two seconds, the tree being in complete darkness between flashes. Imagine a tenth of a mile of river front with an unbroken line of mangrove trees with fireflies on every leaf flashing in synchronization, the insects on the trees at the ends of the line acting in perfect unison with those between. Then, if one's imagination is sufficiently vivid, he may form some conception of this amazing spectacle.

Even small tendencies to advance the next flash when stimulated with light will suffice to create a "rush hour." Furthermore, you usually do not see waves propagating through such a population, except perhaps when the flashing is just beginning or ending.

Relaxation oscillators like neurons and fireflies will get in sync much more quickly than harmonic oscillators, and even weak interconnections will suffice.9 So, if several neurons 0.5 mm apart are firing for some reason (perhaps they both respond to the color yellow), there is a good chance that they will get in sync some of the time. What we have is a mechanism for forming a triangular array of synchronized cells, one that can extend its reach to wherever there are cells already firing, or close to firing. Most potential arrays will, of course, be silent; I tend to imagine fewer than a dozen actively firing, but the silent ones are likely also important.

What we now have is a hexagonal mosaic, formed of the active and silent triangular arrays. A hexagon is simply the largest cortical area that contains one, and only one, of the participating triangular arrays. The adjacent hexagons are nearly identical in their firing patterns.

It's a minimal Hebbian cell-assembly, potentially capable of recording the various features of an object or the details of a movement program. And a hexagon's spatiotemporal firing pattern is potentially a cerebral code, what represents an object or idea. Such a pattern is like a little tune (map each of the several hundred minicolumns to a note on a musical scale). There will be a different tune characterizing Apple than the tune forBanana.

The musical analogies also tell us that a hexagonal mosaic is like a plainchant choir, singing in the lockstep of a Gregorian chant. As the triangular arrays recruit followers on the edges, additional hexagons are added to the mosaic. Perhaps choosing between an apple and a banana for a snack is a matter of how big their choirs are.

Now imagine dueling choirs, abutting hexagonal mosaics singing different tunes, trying to recruit members at the expense of the other. Along the battlefront, there are hexagons that have both tunes superimposed, just as in a symphonic work. If the combination resonates well with the local neural network, we might speak of harmony, just as we do for the major and minor scales. Borderline superpositions (as well as more extensive ones that can be created by long corticocortical bundles) illustrate a powerful recombination principle, a way of doing associative memories that can represent relationships with the same 0.5 mm hexagonal code space as used for objects.

Another lesson of levels is that mechanisms that suffice at one level may prove to be shaky foundations, that other ways of doing the same thing are more extensible. Hexagonal codes are a much better foundation for superstructures (such as coding for analogies) than are the better-known associative memory mechanisms at the level of synaptic mechanisms for classical conditioning.

This isn't the place for showing the many implications of a cerebral coding scheme based on the spatiotemporal firing pattern within one of the recurrent-excitation-defined hexagons, a book-length project that I tackled inThe Cerebral Code. But with the notion of hexagonal mosaics that transiently compete for space in association cortex, you can now appreciate how a Darwinian process could operate in association cortex via the spatiotemporal patterns copying themselves sideways:

Like the classical examples of a full-fledged Darwinian process, there is a pattern that iscopied (indeed, what is reliably copied defined the hexagonal-shaped spatiotemporal pattern), variations occur (dropouts, off-focus nodes, superpositions), populations of the variant patterns compete for a work space, their relative success is biased by a multifaceted environment(current sensory as well as resonances with memorized patterns), and the more successful of the current patterns tend to produce more of the next round of variants (Darwin's inheritance principle is implemented because bigger mosaics have more perimeter, and the perimeter is where dropouts and off-focus nodes can escape the standardization enforced by six surrounding nodes all firing at the same time).

Though our recording methods currently lack sufficient resolution to see how often the various cortical areas actually utilize this Darwinian mode of operation (it could, for example, be restricted to a tune-up period early in life and seldom used thereafter), the time scale question is somewhat clearer

Unlike the generation times spanning days to decades of the usual Darwinian examples, cortex operates on a time scale of milliseconds to seconds, though its operations are biased by memories that span far longer time scales. Within seconds to minutes, neocortex ought to be capable of implementing all of the classical means of accelerating the rate of evolution (systematic recombination, parcellation, rapid "climate change," and refilling empty niches).

Because of its distributed nature and corticocortical connections between regions, cortex isn't limited to the standard Darwinian productions. It might well utilize some additional features, such as a supervisory Darwinian process that can bias the operation of other Darwinian processes

Yet it need not be some grand supervisor with even more intelligence. Until something fancier is clearly indicated, the default assumption ought to be that any regulatory process is essentially stupid, perhaps only chaotic phenomena on a grander or slower scale.

Indeed, there are some situations that might qualify for such two-level interactive evolution, such as the orbital frontal cortex role in monitoring progress on an agenda, a meta-sequence that seems to tick along on a different time scale than individual thoughts and sentences. There's no requirement that darwinian variations have to be random; a slow darwinian process could bias the general direction of the variants of a faster darwinian process that deals with lower-level matters, such as perception and movement on the time scale of seconds. There could be a cascade or web of such darwinian processes.

The neocortical Darwin Machine theory seems to me to be at the right level of explanation; it's not down in the synapse or cytoskeleton but up at the level of dynamics involving tens of thousands of neurons, generating the spatiotemporal patterns that are the precursors of movement - of behavior in the world outside the brain. Moreover, the theory is consistent with a lot of phenomena from a century of brain research, and it's testable (with some improvement in the spatial and temporal resolution of brain imaging or microelectrode arrays)

Composite cerebral codes, formed by superpositions and shaped up by darwinian copying competitions, could explain much of our mental lives. The codes themselves are suitably arbitrary, just as a century of argument about symbols has emphasized. Copying competitions suggest why we humans can get away with many more novel behaviors than other animals (we have offline evolution of nonstandard movement plans). It suggests how we can engage in analogical reasoning (relationships themselves can have codes that can compete). Because cerebral codes can be formed from pieces, you can imagine a unicorn and form a memory of it (bumps and ruts can reactivate the spatiotemporal code for unicorn). Best of all, a darwinian process provides a machine for metaphor: you can code relationships between relationships and shape them up into something of quality.

Resonances are better known these days as attractors; I imagine each hexagon's neural network as supporting a number of characteristic spatiotemporal patterns, just as spinal cord circuitry supports a number of gaits, the particular spatiotemporal pattern that you get depending on how you precondition the circuitry via the facilitation from other imposed patterns. And that may have something to say about the "stream of consciousness."

Manipulating the landscape of a basin of attraction is reminiscent of William James's train of thought, that series of mental states that preceded your current one, each one fading into the background but overlain on its predecessors - and all capable of contributing to what connections you're likely to make right now.

Just imagine those various fading attractors as like that Japanese technique of finely slicing some raw fish, then tilting the block sideways (fallen dominos are another analogy, if you are sashimi impaired). The bottom layer may be hardest to reach but it goes back furthest. Stage-setting with multiple layers of fading schemas may be handy for promoting creativity, getting the right layers of attractors in about the right order and so adjusting their relative strengths. (The Sashimi Theory of Creativity would, of course, be a suitably raw successor to all those half-baked right-brain schemes)

But such histories can also be distracting, and we often try to let them fade, try to avoid reexciting them with further thinking. There are various mind-clearing techniques; Donald Michael10suggests that forming large quasi-stable hexagonal territories might be what meditation with a mantra is all about, preempting the everyday concerns that would otherwise partition the work space and plate out new short-term attractors. By replacing it all with the mantra's nonsense pattern, and holding it long enough for neocortical LTP to fade, the meditator gets a fresh start (for things other than the mantra!)

An ordinary mantra won't, of course, wipe the work space clean: to prematurely erase those fading attractors, you'll need a fancier mantra that disrupts instead. Short of fogging with seizures, as in electroshock therapy, I don't know of any such eraser schemas - though one can imagine mental viruses11 that might preempt entry into those fading basins of attraction, more analogous to an obscuring coat of paint than to a true eraser.

Once they finish with things as basic as perceptual transformations and memory phenomena, theories of brain function must explain abstractions and associations as diverse as categories, abstracts, schemas, scripts, syntax, and metaphor. If we are to venture past the elementary notion of consciousness as mere awareness or shifting attention, we are going to need to account for all of higher intellectual function (language with syntax, structured planning ahead, logical chains of reasoning, games with arbitrary rules, music). That's the kind of coverage needed for a useful theory of consciousness (and this neocortical Darwin Machine enables predictions to be made, all across this spectrum). It may not have to explain all of two centuries of neurology, one century of psychology, and a half-century of neurobiology and cognitive neuroscience - but it can't be truly inconsistent with any of it. A theory of consciousness needs a lot of explanatory power, while still being specific enough to make experimental predictions

Let me turn now to how complex patterns might self-organize, using such Darwinian competitions to embed suitable resonances in the neural feltwork as we gain experience.

Our passion for discovering patterns seems to have a lot to do with our notions of consciousness. If we are to have meaningful, connected experiences - ones that we can comprehend and reason about - we must be able to discern patterns to our actions, perceptions, and conceptions. Underlying our vast network of interrelated literal meanings (all of those words about objects and actions) are those imaginative structures of understanding such as schema and metaphor, such as the mental imagery that allows us to extrapolate a path, or zoom in on one part of the whole, or zoom out until the trees merge into a forest.

Early childhood contains a number of pattern-finding challenges, and children seem extraordinarily acquisitive of ever-more-complex patterns hidden in the sounds and events that surround them. In our first year of life, we discovered phonemes within words. A year later, we were busy discovering schemas and syntax within sentences, and then we went on to discover narrative principles among more extended discourses. The hexagonal superpositions, so like the different voices of a symphonic performance, show us a way that new associations can be represented in the brain - and the Darwinian aspect suggests how quality could be shaped up via the usual variation, competition, and inheritance.

When we think seriously as adults, we think even more abstractly. We conjure up simplified pictures of reality called concepts or models. We can even discover patterns in speculative scenarios, as when we create a forwards-leaping chain of inferences (especially handy for speculating about consciousness!). As Paul Valéry once said, thought is all about "that which does not exist, that which is not before me, that which was, that which will be, that which is possible, that which is impossible."

Passive awareness (and its neural correlates) may be much simpler than the creative constructs implied by the James-Piaget-Popper levels of consciousness; a pop-through recognition of a familiar object may not need to utilize a cloning competition with alternatives in the manner of an ambiguous percept or a novel movement. Hexagonal mosaics surely aren't everything going on in the brain; indeed, they are probably just one mode of operation of some expanses of neocortex, and regulated by other brain regions such as hippocampus and thalamus. But here-this-minute, gone-the-next mosaics seem quite suitable for explaining many aspects of mind, aspects that have been difficult to imagine emerging from quantum mechanics, chemistry, neurotransmitters, single neurons, simple circuits, or even the smaller neocortical modules such as minicolumns. In some regions, at some times, hexagonal competitions might be the main thing happening.

There emerges from this view of our brain, with its relentless rearrangement from moment to moment, some glimpses of the neural foundations on which we construct our utterances and think our thoughts, some possibilities for implementing our kind of language and rational thought. Dueling choirs are at a level of explanation that looks as if it might be appropriate; we'll have to see just how far we can go with their Darwinian aspects as an explanation for talking-to-yourself consciousness.