How Creativity Manages the Mixups

Imagine a time when innovation was often nonsensical or even dangerous, because the brain couldn’t improve the coherence and quality of a novel course of action during “get set.”  While perhaps you could produce juxtapositions of “look” and “leap,” you couldn’t improve their quality and so might leap before you looked.  You’d be conservative, mistrusting innovation for good reasons.  You’d stick to slow groping, testing each stage along the way rather than being able to make multistage plans that usually worked the first time out.

            When Alexander Pope wrote about a little learning being a dangerous thing, and the need to drink deeply, he could have been writing as well about the good-enough thresholds for innovation.  The beginner’s mistakes could have made innovation have a negative payoff for a long time.

            Still, creativity is a good thing, right?  It ought to pay its way easily.  We make the same assumption about intelligence, and the record of the past raises the same two cautions.

            First, as Ian Tattersall said in Becoming Human when summarizing the archaeology, the behaviorally modern period “stands in dramatic contrast to the relative monotony of human evolution throughout the five million years that preceded it.  For prior to the Cro-Magnons, innovation was … sporadic at best.”

            Second, if creative brains are such a good thing, why aren’t there more of them?  Most animals get by perfectly well without being innovative off-line: it’s called “fumble and find.”  If they have to do something novel, where they don’t have a stored movement plan to call up, they just muddle through, slowly feeling their way.  (This is in contrast with the off-line “think first,” doing most of the innovation in your head before making your initial move.)

            If you have time to grope around, a goal plus some feedback along the way suffices nicely – and it mostly obviates the big problem with doing something for the first time, that of it possibly being dangerous.  “Feel your way” may be grossly inefficient but it has a lot of virtues and most animals stick to it when doing something they haven’t done before.

            There are only a few things before the transition where muddle-through might not suffice.  The ballistic movements are particularly interesting because feedback is too slow.  The last eighth of a second of a throw cannot be corrected because it takes about that long for sensations from the limb to travel back into the spinal cord, decisions to be made, and new movement commands to travel back out to modify muscle activation patterns.  A dart throw lasts only one-eighth of a second from start to launch, so you have to make the perfect plan during “get set.”  For a set piece target distance as in darts, you just call up with right commands from memory.  But if it is a novel task, you have to think first and get it right before starting the throw.  That kind of movement-command creativity had a payoff for at least two million years prior to the transition.

            Since the distinction of anatomically modern from behaviorally modern humans was made, the usual explanation of the transition has been that language arrived on the scene and changed everything, including creativity.  Yet language itself is just another example of creativity, once you move beyond stock phrases and start to speak sentences that you’ve never spoken before.  While language surely makes it easier to spread around the results of creativity and build atop what others have tested, what we really want is the source of both language and ballistic movement creativity.

            Here I will consider the demands that off-line creativity places on brain circuitry – and how “think first” creativity might have improved without any concomitant increase in brain size.

Structured stuff is all very nice, and it has some payoff for set pieces like well-practiced throws.  But that’s learning, with creativity only at the beginning when fumbling around will suffice to explore the combinations.  Coherence arises as you slowly get your act together.  Learning is what keeps it together.  Mammals play a lot while young, and much of play is good practice for making a living or discouraging competitors when adult.  But all of that is easily done, to judge from other mammals, and likely it didn’t require changes in brain size and organization for young hominids to learn in a similar way.

            I’ll restrict myself to creativity that is a “first of its kind” amalgamation, that involves stages, and with everything done off-line in the think first manner.  Finally, it all needs coherence – to hang together so well that it usually works the first time out.  It is that combination of requirements that makes it so difficult for the brain – yet children somehow manage it pretty well in the preschool years without being taught.

            Creativity involving stages that is done on the fly but off-line during “get set” – that four-part combination is what evolution produced somehow.  Eating meat regularly was an ancient payoff for it, even if it is no longer an everyday application for most of us.  I am fond of the evolutionary lessons of throwing for a number of reasons that you may not share, but just try to think of another task at which humans excel that requires all those things.  Now narrow down your list to those that have a big payoff at some point.  Narrow that list down to those with repeated payoffs, where each time you improve your technique, you eat even more high-quality food, chase off even more predators, or get even more mating opportunities.

            Also, I want to talk about the underpinnings of language and, while I have to use words to do it, I want simple mechanistic examples of antecedents – rather than the usual attempt to explain structured language in terms of its own obvious usefulness, once you have it.  Even if throwing were only another one of the mostly for-free secondary uses rather than a prime mover, I’d still pick it for my teaching example for all the reasons I mentioned when discussing structured stuff.  Throwing is structured, as in those nested commands for flipping the wrist while uncocking the elbow as you rotate the shoulder and lurch the body forward.  And, outside of set pieces and slow-motion bowling, it often requires a novel set of commands, assembled on the fly as you “get set.” 

            “The combinatorial engine underlying our number and language systems allows for a finite number of elements to be recombined into an infinite variety of expressions,” said Marc Hauser in Wild Minds.  But judging the coherence of a lumped-together assembly of concepts or movements has got to be very important.  And, since you will rarely be right when you first start to get set, you need a bootstrapping procedure for improving the assembly off-line until it is good enough to act on.  There may be various ways to do this but the known process that can achieve coherent results from incoherent raw materials is the impressive one that Charles Darwin (and later Alfred Russel Wallace) discovered.

            Darwin’s quality bootstrap is not only seen at work on the millennial time scale of species evolution but also on the days-to-weeks time scale of the immune response.  Can brain circuitry run a version of it on the time scale of thought and action, shaping up a good-enough movement program for ballistic movements?  And perhaps other creative sequences, such as novel sentences?  All in milliseconds to seconds?

One can summarize Darwin’s bootstrapping process in various ways.  A century ago, Wallace emphasized variation, selection, and inheritance.  (It reminds me of a three-legged stool:  evolution takes all of them to stand up.)  But as I explain at more length in A Brain for All Seasons (from which the next two pages are adapted), there are some hidden biological assumptions in that three-part summary.

            When trying to make Wallace’s list a little more abstract to encompass nongenetic possibilities like cognitive bootstrapping, I listed six ingredients that seem essential to turn the crank (in the sense that if you’re missing any one of them, you’re not likely to see much progress):

1.  There’s a pattern of some sort (a string of DNA bases called a gene is the most familiar such pattern, though a cultural meme – ideas, tunes – may also do nicely).

2.  Copies can be made of this pattern (indeed the minimal pattern that can be semifaithfully copied tends to define the pattern of interest).

3.  Variations occur, typically from copying errors and recombinations.

4.  A population of one variant competes with a population of another variant for occupation of a space (bluegrass competing against crabgrass for space in my backyard is an example of a copying competition).

5.  There is a multifaceted environment that makes one pattern’s population able to occupy a higher fraction of the space than the other (for grass, it’s how often you water it, trim it, fertilize it, freeze it, and walk on it).  This is the “natural selection” aspect for which Darwin named his theory, but it’s only one of six essential ingredients.

6.  And finally, the next round of variations are centered on the patterns that proved somewhat more successful in the prior copying competition.  So variation isn’t truly random; the starting place really does matter.  And the next generation’s starting place can, with success, shift a little.

Try leaving one of these out, and your quality improvement lasts only for the current generation – or it wanders aimlessly, only weakly directed by natural selection.

            Many processes loosely called “Darwinian” have only a few of these essentials, as in the selective survival of some neural connections in the brain during development (a third of cortical connections are edited out during childhood).  Yes, there is natural selection producing a useful pattern – but there are no copies, no populations competing, and there is no inheritance principle to promote “progress” over the generations.  Half a loaf is better than none, but this is one of these committees that doesn’t “get up and fly” unless all the members are present.

            And it flies even faster with a few optional catalysts.  There are some things that, while they aren’t essential in the same way, affect the rate at which evolutionary change can occur.  There are at least five things that speed up evolution (and here I’ll have to use species evolution examples; just remember that they can be translated into neural circuit equivalents).

            First is speciation, where a population becomes resistant to successful breeding with its parent population and thus preserves its new adaptations from being diluted by unimproved immigrants.  The crank now has a ratchet to minimize backsliding.

            Then there is sex (systematic means of creating variety by shuffling and recombination – don’t leave variations to chance!).

            Splitting a population up into islands (that temporarily promote inbreeding and limit competition from outsiders) can do wonders.

            Another prominent speedup is when you have empty niches to refill (where competition is temporarily suspended and the resources so rich that even oddities get a chance to grow up and reproduce).

            Climate fluctuations, whatever they may do via culling, also promote both island formation and empty niches quite vigorously on occasion, and so may temporarily speed up the pace of evolution.

            Some optional elements slow down evolution: “grooves” develop, ruts from which variations cannot effectively escape without causing fatal errors in development.  And the milder variations simply backslide via dilution, so the species average doesn’t drift much.  Similar stabilization is perhaps what has happened with “living fossil” species that remain largely unchanged for extremely long periods such as horseshoe “crabs.”

Are there brain circuits capable of running this sort of Darwinian process on the time scale of thought and action – say, milliseconds to minutes?  Do they have enough of the speedup catalysts to operate so quickly?

            That was the topic of my 1996 book, The Cerebral Code, and the answer appears to be yes for the recurrent excitatory circuits of the superficial layers of neocortex.  They have patterned connections rather like express trains that skip a lot of intermediate stops.  This should allow them to make clones of the spatiotemporal firing patterns, the codes that are used to represent a concept.  Concepts can compete for ascendancy by running a pattern-copying competition biased by a virtual environment of feelings, drives, and memories.  This can be as Darwinian as the copying competition between the bluegrass and the crabgrass in my back yard – except that what competes are codes for concepts and plans of action.

            The circuits that seem capable of copying spatiotemporal firing patterns are found in most areas of neocortex, not just in the areas that might be involved in ballistic movement planning or language.  But we still don’t know how much of the neocortex makes use of this ability and when.  Once, during a period of fetal or infant tune up?  All the time, in all areas of neocortex?  Where and when await experimental evidence, and the more interesting question – how to make subroutines out of the successful plans, so as to avoid running slowly through the whole Darwinian copying competition on subsequent occasions – hasn’t been elucidated at all.

            But let us assume that some brain circuits are capable of running such a process for making multistage coherent plans, and judging them for quality against your memory of what’s reasonable and safe, biased by your emotions, drives, hopes, and fears.  That gives us a prime candidate for the transition:  the secondary use of ballistic movement planning circuits for the novel structured tasks of higher intellectual function.

Now let me briefly address Steven Mithen’s 1996 notion of connecting compartments as a source of the mind’s big bang.  He proposes in The Prehistory of the Mind that before the transition, there were three different brain modules in the human brain that were specialized for “social or Machiavellian intelligence,” for “mechanical intelligence” as in tool use, and for “natural history” as in a propensity to classify objects.  These three modules remained isolated, suggests Mithen, from each other but around 50,000 years ago, some “genetic change in the brain” allowed them to communicate more effectively with each other, resulting in the enormous flexibility and versatility of human consciousness.

            I find Mithen’s notion appealing.  It works well with my neural-circuits theory for how long-distance cortical interconnections might improve dramatically (and without any further brain enlargement or reorganization, at that).  In the second half of my book that appeared that same year, The Cerebral Code, I explore the subject of rapid  communications between distant cortical areas.  The problem, from a brain theoretical point of view, is how to do it on the fly, without a slow learning procedure for each novel combination.

            I suggest that the improved interconnections between areas occur when the distorting corticocortical interconnections (lots of jumble and blur from the anatomy) finally become temporarily restored via a physiological workaround involving sufficient redundancy.  (It’s something like an error-correcting code.)  With such a plainchant chorus of sufficient size, you can recover a novel spatiotemporal firing pattern at the other end and pass it on (see chapter 7 of Cerebral Code) to a third cortical area, also unaltered.  That ability to handle novelty routinely, dependent on finally reaching a critical mass that allows recovery of the original pattern from the blur and jumble, might well have contributed to the aforementioned transition.

So that’s yet another transition candidate:  you could have had structured stuff that was novel with off-line planning, but it was just too slow and too dependent on slowly learned code equivalents – until highly efficient common codes finally developed to allow concepts to be passed around easily between cortical areas over the distortion-corrected pathways.  Even on the fly.

Our kind of high-order creativity – an ability to speculate, to shape up quality by bootstrapping from crude beginnings, yet without necessarily acting in the real world – is a recent thing, arriving well after the big brain itself.  Toolmaking creativity doesn’t look like the big actor in the ascent of humans – despite our usual notion of a versatile creativity being so important to being human.  We are forced to consider our off-line creativity as a late, perhaps fortuitous development after other, more important, things were finally in place.

            Of course, judging from the aforementioned malfunctions in structured stuff, you’d think that confusions were always a bad thing.  But remember that a little confusion helps you to escape popping into standard answers.  Being forced to pause, and sort out the possibilities for your confusion, can actually be helpful to creativity even though it slows down rapid decision making.

            Intelligent people can juggle a half-dozen concepts simultaneously and make good decisions rapidly – and many of them seldom have a creative moment.  They are so good at the standard answers and so eager to move on to the next decision that they never play around with nonstandard possibilities.  (Physicians often get caught in this “keeping busy” mental trap, moving on rather than contemplating.)  There is such a thing as being “too good” because, in much of life, there are no correct answers.  You have to invent new ones and contemplate them for some time.

            IQ tests tend to weigh heavily the speed of decision making and the number of concepts that you can juggle at the same time.  Measuring creativity with standardized pencil-and-paper tests is difficult to do, but creativity and versatility tend to be at the heart of our everyday impressions about whether someone is particularly intelligent.  Handling novel combinations and judging the coherence, redoing it to get a better innovation – that’s what intelligence usually involves.  It’s why I implied, at the beginning, that beyond-the-apes intelligence and creativity both had the same stumbling block:  judging off-line whether it all hangs together well enough to act on.