 |
A book by William H. Calvin UNIVERSITY OF WASHINGTON SEATTLE, WASHINGTON 98195-1800 USA |
|
THE CEREBRAL CODE Thinking a Thought in the Mosaics of the Mind Available from MIT Press and amazon.com. copyright 1996 by William H. Calvin |
|
Thinking a Thought in the Mosaics of the Mind
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. But these too are only intermediate goals for a theory of higher intellectual function. Any one worthy of the name also aspires, however sketchily, to explaining the unity of conscious experience and how it shifts (indeed, can be steered) among topics that had, shortly before, been gestating subconsciously. Consciousness, as I have discussed in both The Cerebral Symphony and How Brains Think, has so many different common connotations that discussions are often confusing, everyone talking at cross purposes. Even within medicine and neuroscience, the word means quite a few different things, and there is no reason to assume that they share common mechanisms. Many follow the long neurological tradition of defining consciousness quite narrowly, as mere awareness, but I think that we now have the conceptual tools to do better, to approach Piaget’s problem of what you use when you don’t know what to do. In the Karl Popper formulation of consciousness:
Much of our purposeful behaviour (and presumably of the purposeful behaviour of animals) happens without the intervention of consciousness. . . . Problems that can be solved by routine do not need consciousness. [The biological achievements that are helped by consciousness are the solution of problems of a non-routine kind.] But the role of consciousness is perhaps clearest where an aim or purpose. . . can be achieved by alternate means, and when two or more means are tried out, after deliberation.
|
|
We are unaware of most of the things that go on in our heads, and sometimes that’s better, as in
Zen archery. When we really learn a new movement sequence, it seems to become a subroutine
that no longer requires conscious attention: tying a necktie or hair ribbon required lots of
conscious attention in the beginning, but once established (perhaps at a subcortical level) we can
do it better
if we don’t try to think about it. What is, initially, consciously
mediated can become subconscious with practice. But our subconscious tasks run a spectrum from the expert to the amateur, even the random. Here we need to address both the expert subroutine that no longer requires conscious attention and the subconscious candidates that are being shaped up for quality — one of which eventually succeeds in replacing the current content of consciousness as it fades. A gedankenexperiment for consciousness in this broader sense is where this final chapter is heading, but let us first examine the security of the footing that this darwinian patchwork quilt has provided for such a necessarily ambitious extrapolation.
LONG before the reader has arrived at this part of my work, a crowd of difficulties will have occurred to him. Some of them are so serious that to this day I can hardly reflect on them without being in some degree staggered; but, to the best of my judgment, the greater number are only apparent, and those that are real are not, I think, fatal to the theory.
|
|
The first task of any theory is to economically account for the
descriptions, to cover a certain group of facts. The extent of
coverage of a theory is easily overestimated when the foundations
are not specifiable. Some theories are reminiscent of fortune
cookie wisdom: surely true at some place, at some time, in some
sense. The lack of specification of the when-where-why details
makes fortunes not terribly useful. We need details to avoid the
theoretical trap of premature closure, which leaves us holding a
fortune-cookie explanation that specifies little. For a brain theory, economical description is a particularly demanding task because of the need to span multiple levels of mechanism — from synapses to cells to circuits to modules, and more. And also span multiple levels of phenomenological explanation — such as attributes, objects, categories, analogies, and metaphors. Beyond this descriptive aspect, a theory tries to predict, whenever possible, features not yet observed. Prediction is a minor aspect of historical theories such as evolutionary theory, but it is always valued as a shortcut, its failures warning that it is perhaps time to try another formulation. Yet biology is full of exceptions to rules, and so theoretical predictions are more likely to be valued for the experimental strategies they suggest than they are for strategic tests in the falsification mode. A number of predictions have fallen out of neocortical hexagons, suggesting various experimental handles. If the reader will indulge me in another celebratory recapitulation, I’ll run through a selection of the theory’s predictions, together with some related descriptive successes.
|
|
As a descriptive theory, neocortical hexagons can explicitly
account for Hebb’s dual trace memory, with characteristic spatiotemporal patterns for immediate memory and attractors
embedded in the synaptic connectivity for the memories lasting
minutes and lifetimes. It also describes one aspect of the widely
distributed synchrony that has been observed in cortex. As a
predictive theory, it offers synchronized triangular arrays
extending their reach as the key prediction amenable to present
recording techniques. The hexagons theory describes why many short-term memories might not be successfully stored as long-term memories. It is consistent with the senile dementias that seemingly uncover long-ago memories during a period when no new short-term memories are being created. It predicts strategies for enhancing storage and recall, as in the sashimi staging of fading attractors. It describes redundancy of memory sites, describes the slow loss or modification of memories with age via the redundancy being reduced by overlain attractors. It predicts slowed access times when fewer pairs of adjacent hexagons remain that possess the same attractor within their repertoire. It describes simple associations of sensory schemas with movement program schemas. It predicts that this association could occur in convergence zones via superposition of active firing patterns, or melding of attractors — but, thanks to the faux fax, the linkage could even happen in sensory or motor cortices. It describes one aspect of cortical plasticity following stroke, via multifunctionality coexisting with specialization of cortical areas. It predicts that an expert region could, by simply altering the slant of the triangular arrays, avoid the local permanent connectivity specialties — and thereby function temporarily as a more general purpose work space. It predicts that a given cortical area could have multiple “personalities” corresponding to different slants of the triangular arrays. It describes some of the functional roles associated with minicolumns and macrocolumns. It predicts a new type of macrocolumn-sized, hexagonal-shaped functional structure in association cortex that, unlike other macrocolumns, may overlap with one another.
It describes why recurrent excitation among superficial pyramidal neurons should be associated with NMDA synapses, LTP, and standard distance gaps in the intrinsic horizontal connections. It suggests automatic gain controls that prevent runaway recruitment and, more uniquely, it predicts that the nodes of triangular arrays should shrink over time. It predicts that central regions should crystallize into closer approximations to equilateral triangles (remembering, of course, the “good-enough” caution, p. 36) as territories enlarge. It describes retrograde inhibition of memories, where a new telephone number may impede your access to the old one, via chaotic capture. It predicts that regional reductions in excitability could erase immediate memories, while leaving fading basins of attraction from which active patterns could be rekindled. It describes how seizures could fog recent memories, in the manner of road resurfacing filling in the gaps in the road that constitute the washboard pattern, or in the manner that edema can obscure characteristic facial wrinkles. It predicts that maintaining a widespread nonsensical “mantra” pattern for the duration of neocortical LTP could, by minimizing meaningful associations, serve to prevent other temporary attractors from forming. It describes both déjà vu and premature closure; it predicts that an agnosia could be created by a similar clone-too-quickly mechanism that preempts slower links from affecting the outcome. It describes the long reaction times (so improbable merely from conduction delays) of cognitive processing via the time it takes for synchronization, local recruitment, and cross-modality links via convergence zones. It predicts a shortening of such times via cross-modality priming. It describes neural equivalents of simple mutations and deletions (the “capacity to blunder slightly”) and predicts some equivalents of recombination via active superpositions by triangular array interdigitation. It predicts passive superpositions via successively-overlain attractors. It predicts geometrical arrangements of barriers and battle fronts whereby otherwise sterile hybrids might reproduce themselves. It describes the construction and deconstruction of attributes, objects, schemas, and even analogies; like Hebb’s cell-assembly theory applied to stabilized image fragmentation, it has features of both the holistic and specialization views of mental function. It predicts multiple levels of stratified stability, each able to use darwinian processes to improve quality and create novelties, some of which could constitute a heightened form of consciousness.
|
|
The different levels of explanation and mechanism are important to distinguish, if we are to avoid the confusions of level (and
all the arguing at cross purposes) that marked evolutionary
biology in the several decades after the rediscovery of Mendel’s
genetics in 1900. Eventually most came to see mutations and
selectionism as two sides of the same coin rather than competing
explanations, but it took a few decades. As Jonathan Weiner
wrote in The Beak of the Finch:
After Darwin’s death, many biologists found it easy to accept evolution and impossible to accept Darwin’s chief explanation for it. Evolution, yes; selection, no. William Bateson, the founder of modern genetics, wrote an elegy for Darwinism in 1913, calling it “so inapplicable to the facts that we can only marvel. . . at the want of penetration displayed by the advocates of such a proposition.”This was one of the inspirations for J. B. S. Haldane’s wonderful quip about “the four stages of acceptance of a scientific theory: (i) this is worthless nonsense; (ii) this is an interesting, but perverse, point of view; (iii) this is true, but quite unimportant; (iv) I always said so.” These days, confusing variation with selection is on a par with confusing genotype and phenotype, something that biologists (but few others) successfully avoid. It took three decades before the Neodarwinian Evolutionary Synthesis straightened out our thinking about the relationship between genetics and darwinism, between individuals and populations; hopefully we won’t have to go though it again when talking about milliseconds-to-minutes darwinism. However, ignorance about levels of explanation abounds, particularly in recent writings about quantum mechanics as a basis for consciousness, showing how attractive these confusions of level can be. In How Brains Think, I call the most grandiose confusion “the Janitor’s Dream.” (Hoping to leap from the subbasement of quantum mechanics to the penthouse of consciousness in a single bound!) For milliseconds-to-minutes darwinism, the most reasonable confusion is surely with the days-to-years selectionism that Changeux and Edelman describe so well for wiring up the nervous system in ontogeny, and for selecting useful modifications of that wiring by lifetime experience. I have tended to describe this level as involving modifications of the attractors; other terms associated with these long-term spatial-only patterns are bumps and ruts in the washboarded road, synaptic weights, and connectionism. Differential alterations of synaptic strengths are surely happening much as described, and Edelman’s reentrant interactions with “cortical appendendages” are likely a major mechanism for editing and emphasis, much in the manner of the sculpturing selectionism of William James a century earlier:
The mind is at every stage a theatre of simultaneous possibilities. Consciousness consists in the comparison of these with each other, the selection of some, and the suppression of the rest by the reinforcing and inhibiting agency of attention. The highest and most elaborated mental products are filtered from the data chosen by the faculty next beneath, out of the mass offered by the faculty below that, which mass in turn was sifted from a still larger amount of yet simpler material, and so on. The mind, in short, works on the data it receives very much as a sculptor works on his block of stone. Note that mine is not a rival theory to such forms of neural selectionism (except insofar as they claim to extend to the James-Piaget-Popper aspect of consciousness). Rather, the ephemeral copying competitions that I emphasize rest on the broad foundation of such longer-term selectionism, which forms part of the environment that biases cloning success (and faux faxing) on my short time scale of milliseconds to minutes. As such, my ephemeral copying competitions are one layer up from the connectionist layer, though they feed back to it when altering the synaptic strengths, just as Edelman’s reentry also shapes connectivity changes.
|
|
Hegemony is what establishes the topic of our current conscious
experience, in the present theory; the remaining mosaics of other
patterns constitute our subconscious. Leadership or dominance,
especially of one state over another,
seems applicable to cortical territories
as well as to nation-states. Eventually,
if no sensory input demands attention,
one of the nondominant patterns will
take over from the current winner, perhaps a pattern that is novel,
with no basis in existing memories, perhaps one that represents a
familiar worry. 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 in the manner of an ambiguous percept or a novel movement. Hexagonal mosaics surely aren’t everything going on in the brain;
Edelman quite reasonably forswears a treatment of thought itself at one point (this may seem surprising in a book claiming to be about consciousness, but remember those self-imposed blinders of the neurological tradition). He proceeds to define thought as the building of conceptual theories about the world. I too would emphasize process over product, though I would also include those events at the lower end of the quality scale — the process during the four hours of mulling-things-over each night (half of all sleep) that rarely makes any progress and that of the fantastic juxtapositions of our episodes of dreaming sleep. At the high end, I would emphasize the progressive shaping up of quality, the discovery of order among seeming disorder, and the creation of new levels of abstraction for relationships.
A major difference, however, is that thought may have to span many levels of explanation and locate an appropriate one. As we try to speak usefully about a subject, we are often torn between dwelling on rock-solid details and speaking in perhaps-too-abstract generalities. In a book-length work, one may range over the spectrum — though readers may not always synchronously co-vary in their preferences for details and overviews!
|
Although the “Janitor’s Dream” may be improbable, some leaping between nonadjacent levels is surely allowed within the neocortical representation levels, just from the nature of cerebral codes in the hexagonal theory. There’s really nothing to keep the cerebral code for Apple from competing with that of Fruit. We all make categorical mistakes, then try to weed them out. One simple pathology would be the failure to weed before speaking, as in the illogicality and non sequiturs of positive formal thought disorder. Should the neocortex be too excitable, few barriers to cloning will form. That means even fewer gateways; not as many variants will escape error correction, and so the solution space will be only briefly explored. You’d be able to colonize large areas without competition along the way. Besides making for results of poor quality, the spatiotemporal pattern that succeeds may have the large homogenous “choir” normally associated with successful memory recalls. Even situations that were unfamiliar might thereby seem familiar. Déjà vu experiences might be one result. Should the cortex be insufficiently excitable, routine tasks might still operate but tasks that required concentration or sorting through possibilities would be greatly slowed. Few new episodic memories would form, and amnesia might be a typical complaint. Patchy versions of too little or too much might give rise to dissociations and fugue states. The patient who finds himself in San Francisco, but is unable to remember why he traveled there, may have a substantial roadblock (an extensive barrier region, without U-paths available) in his cortex, or have dead end detours (attractor basins that divert attempts to approach those memories). Infrequent changes have an interesting pathology as well, even when average amounts are normal. Inability to form a large choir, as might occur when excitability fluctuations are minimized and allow stalemates to persist, would give rise to other types of thought pathology, such as indecision and inappropriate unfamiliarity judgments (suggesting jamais vu). If “climate changes” are an important part of the rapidity of recognition or decision making, some subcortical pathologies might have their effects on cortex by insufficient fluctuations, rather than from insufficient average levels of cortical input. Or their fluctuations might be too slow. Too frequent changes, of course, suggest accelerated variation and decision making — and bring to mind a classic form of mental disorder, the rapidity of thought associated with manic-depressive illness.
There is a particular kind of pain, elation, loneliness, and terror involved in this kind of madness. When you’re high it’s tremendous. The ideas and feelings are fast and frequent like shooting stars, and you follow them until you find better and brighter ones. Shyness goes, the right words and gestures are suddenly there, the power to captivate others a felt certainty. There are interests found in uninteresting people. Sensuality is pervasive and the desire to seduce and be seduced irresistible. Feelings of ease, intensity, power, well-being, financial omnipotence, and euphoria pervade one’s marrow. But, somewhere, this changes. The fast ideas are far too fast, and there are far too many; overwhelming confusion replaces clarity. Memory goes. Humor and absorption on friends’ faces are replaced by fear and concern. Everything previously moving with the grain is now against - you are irritable, angry, frightened, uncontrollable, and enmeshed totally in the blackest caves of the mind. You never knew those caves were there.
|
|
Rapidity of thought, in most of us, varies a great deal. The
phrase, however, has a special meaning in psychiatry. An expert
(someone who knows all the common mistakes and how to avoid
them) may be able to operate quickly in a way that the amateur
cannot. But the speed variations in someone with manic-depressive illness (or its less extreme relative, cyclothymia) may
have nothing to do with how well established the links are. Even
an expert, dealing with familiar material, may go from a fluidity
of making connections and decisions in hypomania to a slow,
labored train of thought in depression, lingering too long and
failing to make obvious connections with what is obviously there
(as too-late recall eventually proves). The effectiveness of the antidepressant medications suggests, of course, that depressed mood might involve imbalances in the major neuromodulators, especially the norepinephrine and serotonin systems. But the existence of mixed depression keeps us from making a simple equation of retarded thought with mood: in many bipolar and cyclothymic patients, racing thoughts and pressured speech can occur in both high and low states. In hypomania, every new idea may seem promising; in mixed depression, equally frequent thoughts may be systematically devalued and so one quickly thinks one’s way into a gloomy cave. While the cortical hexagons theory has nothing to say, at present, about global influences on mood or alertness in the manner of Hobson’s theory, it does offer a number of candidates for rapidity of thought (all of those analogs to climate change and island biogeography) and making connections (via the faux fax mechanism and those Ncc-link< Naction linkage requirements). One can easily imagine additional ones, equivalents of boredom or novelty-seeking, that might be associated with the number of simultaneous contenders or the “generation time.”
|
The experimental subjects in isolation saw, among other things, primitive animals in a prehistoric jungle and modern squirrels wearing snowshoes. . . . Charles Lindbergh in his solitary flight across the Atlantic was aware of “ghostly presences riding in my airplane” and “vapor-like shapes crowding the fuselage, speaking with human voices, giving me advice and important messages.” Solitary sailors, and survivors of shipwreck in lifeboats, report having visions. Even on land, in apparently normal circumstances, the monotony of long-distance car-driving on the Western plains of the North American continent may lead the driver to see things — jackrabbits big enough to step over the car in one case — and the long-distance truck-driver at night following that endless white line down the highway may wreck his truck trying to avoid collision with a nonexistent object on the highway before him.Thought disorders involve psychosis without a concomitant mood change. The two hallmarks of psychosis are hallucinations (not counting those of obvious cause, such as Lindbergh’s) and delusions. Hallucinations are where imagined events or memory recalls come to be mistaken for current sensory input: voices may be heard, burning sensations felt, shapes or people seen. Enthusiasts have been known to seek them out via dehydration, sweat baths, drugs, and sensory deprivation. Tuberculosis used to be a major hallucinogenic disease, via tuberculomas of the temporal lobe; the voices that Joan of Arc heard were likely due to this. Today, schizophrenia, manic-depressive illness, and epilepsy are the brain disorders producing most of the unprompted hallucinations (though we all, of course, experience psychotic symptoms with every REM-episode dream). To understand a mechanistic foundation for thought is, hopefully, to pave the way to appreciating and modifying the intrusive thoughts of schizophrenia. The hexagonal cloning theory suggests that the content of a hallucination is only the stuff of subconscious contention, that the pathology might consist of it being taken seriously, premature closure occurring before quality can be shaped up. Delusions are more subtle, long-term failures of reality testing than are hallucinations. The common delusions are of persecution, jealousy, grandiosity, sin and guilt — like hallucinations in dreams, we all experience them; the pathology is more a matter of persistence. In schizophrenia, they can extend to bizarre notions that have no basis in anyone’s experience, such as being controlled by men from Mars. And why is the delusion not overwritten by corrections, all those things that usually modify our erroneous judgments with further experience? Or discounted, as we regularly ignore our nighttime dreams? Why doesn’t an idée fixe fade, like my unused calculus theorems? A delusion often seems as intransigent as any innate tendency with which we are born, difficult to ignore or unlearn. The delusion (or, for that matter, the obsession or compulsion) could be from a particularly secure basin of attraction. A particularly broad catchment, so that many situations are distorted to fit the mold, their cerebral codes captured and standardized in the way that a black hole captures neighbors and renders them invisible, would result in delusionlike properties. Might a delusion’s setup involve unusually widespread clones of the attractor, many hundreds scattered over the connectivity of an entire Area, perhaps from some long-ago success that created a large choir that practiced too regularly? From particularly frequent recall, that served to embed the memory even more thoroughly, or obscured approach routes to other basins of attraction?
|
|
The meta-theory for hexagonal cloning competitions is, at
present, like talking about the weather. Subcortical supervisors
for quilting seem likely, but some neocortical regions might also
tend to supervise the cloning competitions of others, regulate their
habituation, and so forth. 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. As they say in many locales, “If you don’t like the weather, just wait an hour.” The marked seasons of the temperate zones have probably been important in species evolution and we may yet come to say, “If you don’t like the climate, just wait a decade.” As I argue in How Brains Think, the abrupt climate changes superimposed on the ice ages may have helped conserve an inefficient ape variant that happened to be a jack-of-all-trades, able to adapt to new diets within a generation. In cortex, changing the name of the game could happen on the milliseconds-to-minutes time scale. As such, the electroencephalogram may serve as a useful indicator. We are accustomed to eschewing any functional assignment to the various EEG rhythms, but hexagonal cloning may allow us to view the EEG in a new light — as drivers of territorial expansion and deme extinction.
|
|
If so many things are happening at once in the neocortex, why do
we have a unity of conscious experience? We speak with a single
voice, even if only talking to ourselves. We have a sense of being
at the center of a convergence of various narratives that we use to
explain the past, all while trying to choose between several
speculative scenarios about the future. There are some trivial answers to the unity question — as when we say that some things, such as sensing one’s blood pressure, are totally inaccessible to verbal reporting mechanisms — and so perhaps the unity is an illusion, simply a problem with what’s accessible to verbal report. In the context of cloning competitions, a more tempting answer is to say that we have a unity of consciousness because there can be only one winner of a competition — and that it’s simply the largest patch of the dozens currently to be found somewhere on the dynamically reforming patchwork quilt (or, at least, it’s the largest one with ready access to output pathways).
But if one phrases the consciousness question as a competition, then of course one gets a unitary answer. Perhaps we should rephrase the question: Can we possibly do two consciousness-level tasks at the same time? With a theory like spatially distributed cloning competitions, the answer is surely, “Why not?” That’s because the theory suggests various shaping-up competitions proceeding in parallel. If you can have simultaneous quarter-finals and semi-finals leading up to a brain-wide championship that we call the current content of consciousness, then why not simply simultaneously output both the global winner and the second-best from some other area, operating on a different train of thought?
|
|
We would not be easily persuaded that someone had a “two-track consciousness” if he merely claimed that he could attend to
two vigilance tasks simultaneously. Or keep two schemas in
mind, one visual and one verbal. We’d require sentence-like
tasks. Two simultaneous narratives would be the most persuasive. If consciousness involves generating novelty and selecting
between alternative courses of action (rather than consciousness
as mere awareness), the demonstration would need two such
tasks proceeding in parallel. But comprehension tasks are usually easier than production tasks: most of us can read a book while listening to the radio. What we need are two simultaneous production-task outputs from consciousness-demanding processes. Imagine, for example, a sign language interpreter who, when not interpreting spoken language into sign language, finds it possible to carry on two conversations at the same time, speaking to one person and signing to another. Or a touch typist conducting an internet dialogue with someone on another continent, while talking about something else with a person in the same room. It wouldn’t be interesting if the phenomenon turned out to merely be a matter of clever time-sharing, or of rote replays of memorized material. But, were we able to rule out such less-interesting explanations, a two-track success would suggest an interesting interpretation: that our seeming unity of consciousness is simply a matter of most of us having only one major output track available for unique representations (such as reporting on our current thoughts) — and that said pathway has a serial-order bottleneck that admits of only one completed thought at a time. Not having two independent output paths means that most of us get little experience in managing two trains of thought simultaneously; we can only use “on the other hand” tactics. But if we became experienced in managing two semi-independent tracks, thanks to several output paths to feed, then two internal voices might also be able to converse with one another.
|
|
Though independent simultaneous outputs may be infrequent,
simultaneous competitions that influence one another are probably common. We regularly evolve our sentences into “good
stories” at the paragraph or page level, applying the quality
criteria associated with narrative and epic to guide our productions toward satisfying endings. (I’m doing that right here — and
certainly in the forthcoming finale.) Why not also evolve them to
meet some nonstandard criterion that itself evolves, on a somewhat
slower time scale? Using a darwinian copying competition with
short-term memories that fade much more slowly? 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. There could be a cascade or web of such darwinian processes.
|
|
Thinking a thought, in this theory, is more than just a current
competition between the cerebral codes that have cloned mosaic
territories. It involves the recent history of such competitions,
plating the various work spaces with attractors that then fade. It
involves biases from moods and from agendas evolving elsewhere. And it involves the attractors of long-term memories,
different in different areas. Habituation processes in dominant areas may allow second-best areas to take over a moment later. Just as neuromodulators serve to sculpt one motor circuit from a network with many possible motor circuits, and so set up an action, the faster-acting synaptic modifications may also help move our attention from a present topic to a new one, from one second to the next. From the sashimi layering of fading attractors, we may get the appearance of a single-minded person inside, steering the train of thought and stage managing our purposeful behavior. Our current mental state is always unique because, even if we are thinking exactly the same thought as we did yesterday, those fading patterns underlying its spatiotemporal patterns differ in strength and spatial distribution from that of yesterday’s sashimi. They will lead the train of thought somewhere else. While I think that a Darwin Machine operates in the brain and can account for much of higher intellectual function, I am not equally certain that making frequent use of the darwinian algorithm is what elevated us from the ape level of mental abilities — or, for that matter, that it alone is what allows us to operate quickly enough to escape esprit de l’escalier. It may be that some shortcuts, cortical or subcortical, are absolutely essential in order for the darwinian process to operate quickly enough to produce useful results within the time span of short-term memory. Those shortcuts might structure agendas and the ascending levels of abstraction, or keep us from constantly backsliding into a hopeless muddle. I suspect that much of fleshing out this theory will involve the theoretical, experimental, and neurological identification of shortcuts, that much of its application to education will involve learning how to augment or avoid them. An example of a shortcut would be that set of constraints on word order known as syntax. I’m not thinking of plural and past tense formation so much as phrase structure. There’s a lot of embedding of phrases, as in a sentence such as What you see is what you get, where both the subject and the object are themselves sentences. Embedding is a key step up in beyond-the-apes language abilities.
|
|
Protolanguage is a simple form of language lacking the fancier
structure provided by syntax. It’s the language of the trained
animals, children less than two years of age, speakers of pidgins,
agrammatic aphasics, and American professors trying to communicate
with Greek shopkeepers. With structureless protolanguage, it takes a lot of time to relate who did what to whom, even when supplemented by gestures. The linguist Derek Bickerton suggests that there are no real intermediates between protolanguage and our full-fledged syntactic language. This raises the issue of what improved neural mechanism could make such a large difference — and I have one to suggest. This isn’t the place for serious linguistics or paleoanthropology, but, given eleven chapters of hexagons to warm up the reader, it is possible to develop a mechanistic outline for some of Universal Grammar (features common to all known languages except pidgins) in just a few more pages.
|
|
A noun and its modifiers, such as black shoe, could be implemented by simple borderline superpositions of hexagons. Superimpose
this modified noun upon the code for with and you have a prepositional phrase. Keep going and you might achieve a territory of
clones, each of which represents the tall blond man with one black
shoe. By this point, however, you are well out of the range for doing all the association via superpositions at a borderline between
territories; that would merely pour all of the words into a blender
and leave only a general impression of the intended topic, one that
becomes even less specific as more associations are added. The faux fax would seem, at first glimpse, to produce an even more ambiguous superposition. But bi-directional corticocortical links allow you to have your cake and eat it too. Back projections (six out of seven neocortical areas are reciprocally linked) can use the same code, and so immediately contribute to maintaining a chorus above a critical size (they are, presumably, always adapting and thereby falling silent). It would be like missing choir practice but participating via a conference phone call. The central chorus of What you see is what you get could have two subsidiary choruses for see and get, each implementing appropriate roles for its verb; if either subchorus falters, the top-level one stumbles. A backprojected spatiotemporal pattern might not need to be fully featured, nor fully synchronized, to help out with the peripheral site’s chorus. It could be more like that sing-along technique where a single voice prompts the next verse in a monotone and the chorus repeats it with melodic elaboration; some singing at a fifth or an octave above the others, some with a delay, and so forth. The backpath could include more code than the subchorus uses, just as choirmasters and folk singers manage to include exhortations with the desired text. Back projections provide a way of resolving any ambiguity associated with recursive embedding by maintaining an audit trail. (“Who mentioned X? Sing it again, the whole thing!”) With such structuring, there’s no longer a danger that the mental model of the eight-word amalgamation the tall blond man with one black shoe will be scrambled into a blond black man with one tall shoe.
|
|
Links can also implement the binding needed for words such as
he, himself, and each other, whose referents may be in preceding
sentences. That’s another of the linguists’ desiderata, along with
a neural mechanism for the Wh-question’s long-range dependencies. But at the very top of their Universal Grammar wish list
is recursive embedding, needed to nest one sentence inside
another. (I think I saw him leave to go home.) What keeps the top-level think verb’s hexagon happy enough to reproduce effectively in a copying competition with other variant interpretations? Presumably, a few alternatives assemble in parallel until one gains the strong “legs” needed to allow it to become robust enough to establish hegemony. If the leave link stumbles, the saw hexagons might not compete very effectively, and so the top level dangles.
A sentence tells a little story with characteristic roles. Most verbs require a subject-actor-agent but it can be a noun phrase such as The tall blond man with one black shoe or even a sentence itself such as What you see). Beyond that near generality, special cases abound. The intransitive verb sleep cannot tolerate an object-patient-theme (He sleeps it is a clanger). The verb give insists on both a recipient and an item given. I discuss such required and optional roles at more length in How Brains Think and suggest a mechanistic analogy for how such argument structure operates. So the “meaning of the sentence” is, in this model, an abstract cerebral code whose hexagons compete for territory with those suggesting alternative interpretations. Phrase structure is presumably a matter of the coherent corticocortical links to contributing territories, having their own competitions and tendencies to die out if not reinforced by backprojecting codes. Weblike crosstalk between subchoruses presumably occurs, and may be quite useful so long as it remains weak enough not to show up on the audit trail. Argument structure suggests multilobed attractors (p. 155). Surface structure, needed to actually speak one word after another, is a matter of unpacking the contributors in an order commensurate with a particular language’s surface structure conventions. The 18-to-36-month-old child seems to tune up this circuitry for the patterns of a particular language without much overt trial-and-error, merely by listening to the speech of others. In English, they include a subject-verb-object word order for simple declarative sentences, to which you add the inflections for plurals or tense, and perhaps some case markings (such as he-him, who-whom) to provide the listener with additional clues about whether the noun plays a subject or object role in the verb’s story.
|
That’s stratified stability at
work, and it may
be what enables a
series of simple
rules, at the level
of argument
structure, to
generate a proper
syntax. Each
verb has a
characteristic set
of links: some
required, some
optional, some
prohibited. The
conglomeration is
called a sentence
if all the obligatory links are
satisfied and no
words are left
dangling, unsupported by a structural role.
|
So coherent corticocorticals buy you some essentials of
phrase structure via a colonizing choir. What might degrade such
a nice system into a protolanguage? Incoherence will do. Were
the corticocortical’s error correction not well tuned, linkages
would be restricted to well-practiced special cases, perhaps only
the spatiotemporal firing patterns for a limited number of
vocabulary items, perhaps a few schemas and scripts. That’s because a cerebral code for an item is no longer the same, here and there. Jumble and smear means that a muddled version must be gradually learned by the target cortex; if sent back, a doubly-distorted version must be learned and an equivalence constructed to the original. The return path would be slow and chancy, and it is what allows a subchorus to be maintained and thereby permits an audit trail to resolve ambiguities. Lacking such structure would be very limiting, like signal flags on sailing ships, giving you a small shared vocabulary with only minor possibilities for novel associations. Embedding would probably be restricted to stock phrases. Relating who did what to whom would take a long time — just as in “protolanguage.” Adding faux faxing to a base that included regional darwinian copying competitions, each using memorized relationships to shape up quality, is thus a candidate for what made protolanguage into Language Itself. Corticocortical coherence that became good enough to convey even novel spatiotemporal patterns could have implemented the recursive embedding and long-range links aspects of Universal Grammar — and have done it in one step, without semistructured intermediate forms. Though there is much for linguists and archaeologists to sort out, some now think that a big language step accompanied the evolution of Homo erectus to early Homo sapiens about a quarter-million years ago. The arbitrary cerebral code linkages described here, in the context of a Darwin Machine at each end to augment quality, might provide the improved capabilities needed to convert the infrequently innovating Homo erectus culture (a million years of stasis) into our prolific one, capable of occasionally producing the incandescent mind of a Plato or Shakespeare.
|
|
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 suggestions for how thoughts might
miscue, and some possibilities for implementing the shortcut rules
which make possible our kind of language and rational thought. Since Darwin’s recognition of how biology could have been bootstrapped by natural selection operating on inherited variation, the immune response has been recognized as more than just an analogy: rather, it’s the same process, operating on the intermediate time scale of days to weeks (though running out of gas as the antigen is eliminated). By now, I think that we can recognize a darwinian process per se, what I have called a Darwin Machine, capable of operating on various time scales and in various media able to reproduce with inherited variation, as one of the key organizing principles of the universe. In the brain, such a process need not run out of challenges, as memes from a rich cultural life always provide another set of complex patterns to analyze for possible hidden structure, repeating the process that we used as two-year-olds to figure out the syntax of the utterances we heard. The extensive parallelism between Darwin’s 1859 formulation of principles, and the ones now suggested for hexagonal cloning competitions on the milliseconds-to-minutes time scale, can easily be seen by paraphrasing Darwin’s final paragraph of On the Origin of Species so as to emphasize the power of this process as it could be manifested in cerebral cortex. And borrowing Darwin’s finale framework seems only appropriate: it is, after all, his process.
|
|
It is interesting to contemplate a tangled mind, clothed with
memories of many kinds, with prototypical birds singing on the
hallucinated bushes, with various thoughts flitting about, and
with wormlike obsessions crawling through the damp cortex, and
to reflect that these elaborately constructed forms, so different
from each other, and dependent upon each other in so complex a
manner, have all been produced by laws acting around us. These
laws, taken in the largest sense, being Reproduction via clones of
cerebral codes; Variability from their interdigitation and escapes
from error correction, and from use and disuse; Inheritance which
follows from surface-to-volume principles at the perimeter of
cloned territories; a Ratio of Increase so high as to lead to a
Struggle for cortical space, and as a consequence to Natural
Selection from current and memorized environments, entailing
Divergence of Character and the Extinction of less-improved
forms. Thus, from the war of random thoughts initially no better
than those of our nighttime dreams, the most exalted process
which we are capable of conceiving, namely, the production of the
higher thoughts, directly follows. There is grandeur in this view
of mental life, with its ascending powers. From such a darwinian
ratchet for creating and refining ever more complex levels of
abstraction, there arose unbidden our own brain of unbounded
potential, able to discover the syntactic rules that nestle sentences
within sentences, able to invent new rules that extend mere
narratives into the long chains of rational thought. Transcending
its origins in tool use and social life, our reorganized brain can
now use stratified stability to explore the endless realm of memes.
Blind to our foundations, we nonetheless created poetry and
reason; with a clearer footing, we can perhaps contemplate how
our heightened consciousness evolved and is evolving.
|
You are reading THE CEREBRAL CODE.
The paperback US edition |