There are two kinds of scientific meetings I've attended in my short career: physicist's and engineer's.

Physicists would write an abstract of about 5 sentences on a Friday afternoon when the registration deadline for the meeting arrives, usually about 2 months away. They spend 10 minutes pondering what sort of catchy phrase will attract other colleagues to their talk, while not promising too much, since they haven't done the research yet. After struggling for a few minutes, they decide to recycle the talk they gave last year and add an extra sentence promising to extend their work to cover, oh, the next 15 minutes of data. Then they pay no mind to the talk until approximately 8 days before the meeting, at which point panic sets in. In the good old days, they would spend the entire flight to the meeting with the tray table down, a stack of clear acetate sheets, and a pocket full of permanent markers. Nowadays, the laptop is open and the PowerPoint clicking furiously, right up the chairman's call.  Engineers, on the other hand, require a type-set paper sent in months before the meeting, so that colleagues can read the paper before deciding whether to attend the talk. "So ancient", mutter the physicists, "don't they want up-to-date research?" Scientific progress: the thrill of anticipation, the agony of procrastination.

Well, I've agreed to give a paper at an engineering meeting, but the chairman of our session is a physicist. So I have to write a paper, but I can turn it in after the meeting. This will be a rough draft outline, and you, gentle reader, my first critics. Please take a moment to improve it by demolishing its poor logic and faulty assumptions.
[Note: here is the PDF of the talk as it was before the meeting. The meeting changed my emphasis, and required a whole new paper!]


Space-Time Design

Secure communication has always been of interest, from the time that your parents shooed you out of the room, and whose whispered secrets didn't carry through the door to your ear. But when communication channels became more public, with radio, and now the internet, secrets were harder to keep. Accordingly encryption has been employed by schoolkids and the military for millennia, and likewise decryption. One of the early favorites was a transposition cipher, where letters of the alphabet were replaced with cryptic tic-tac-toe symbols, or by shifting a certain distance along the alphabet. The German Enigma machine used in WWII was just a modern version of such a cipher, and Von Neumann's brilliant solution for deciphering that code gave birth to the modern computer, using its enormous patience to search for non-random patterns in the coded message.  And now in our modern world, the computer is routinely used to decipher Nature's encryption, looking for non-random patterns in the noisy confusion of global temperature trends, or cosmic background microwave radiation, for example.

You have undoubtedly already seen this magic when your digital camera photo came out too dark, perhaps because you were sitting too far away from the podium for your flash to illuminate the diploma your son received. But of course, the camera came with software to enhance photographs, or perhaps you had Photoshop handy, and you played with the contrast and brightness, and maybe even the "sharpen" tool. Presto! That nearly black photograph of the back of someone's head suddenly became your son, though perhaps a bit grainy.  What was the computer doing? It was suppressing the background, stretching the color tables, averaging over local pixels, trying to increase the signal-to-noise ratio. In fact, the retina at the back of the eye, between the pupil and the optic nerve, has 12 layers of neurons that do all these Photoshop functions, and are largely responsible for all those "optical illusions" that are the mainstay of 1960's OpArt. The eye processes images for the same reason that NASA does it on Mars landers, to compress the information while eliminating noise, so that the limited throughput of the transmitter (optic nerve) is getting all the best information available.

But here's the kicker: we adjust the Photoshop sliders while looking at the entire photo, but the eye does this without knowing the whole picture.  Most of those neurons in the 12 layers in the back of the eye connect adjacent "pixels". They perform all their functions with only local information.  Like the highway signs in France, the exits are only marked with the next nearest town, so to travel from Paris to Lyon with innumerable "traffic circle" intersections, one needs to know all the intermediate town names, possessing global information.  Yet the eye eliminates some 90-99% of the bandwidth collected by the retina, before passing back to the brain the best 1-10% without any such "big picture" information, how can it do that?

By knowing what the noise looks like. Random noise always looks the same, sort of like snow on those old analog TV's I grew up with, as I fiddled with the UHF tuner and rabbit ear antenna. Noise has little spatial correlation, so if I pressed my nose up to the curved glass of the TV's cathode ray tube, I could see little spots of white and black flicker randomly until the UHF tuner found the signal. But I couldn't really distinguish the ghostly figures behind the "snow" unless I stepped back to see the whole screen and let my eyes smear out the snow into a uniform gray. Nowadays, I just take off my glasses for the same effect, but the neurons in the eye can't "step back", they have to make the call by looking at just a few pixels to the right and the left of them.

However, by stacking up these neuron "comparators" into layers, then the output of one layer goes into the next layer, and after a few layers, the neurons have averaged the noise into a uniform gray, and they don't pass on that information (by "firing" a neuron) unless it changes suddenly. Mathematically, there are two separate processes going on in the eye, a spatial filter that looks for differences between pixels or edges, and a temporal filter that looks for differences in time or motion, though physically they are being done in the same place. The one eliminates spatial noise, which is uncorrelated in space, and the other eliminates temporal noise, which is uncorrelated in time (by being constant or random flickers).  I make it all sound easy, but as soon as one tries to implement "machine vision", say, to have a computer pick up a toy on an assembly line, then all sorts of complications arise, such as how does the algorithm compensate for camera motion or flickering lights?  But the basic point remains the same: space-time correlations equal information, and their lack equals noise.

While we are moderately familiar with the use of computers to process spatial information, we are less familiar with their use to process temporal information. In the past 10 years, computers have become the standard platform for making movies, sharing AVI files, and now YouTube, and all of these media are starved for bandwidth. So there has been considerable effort put into compression of movies, sending only the bits that demonstrate some sort of change. A similar effort has gone into music, with the result that instead of a single Beethoven's 9th on a CD, you can now put thousands of MP3 compressed songs into your iPod with apparently little degradation in the quality. The point is that songs and movies are time-redundant, containing more bits than are needed to convey the music or the video, and so finding better compression algorithms are essential in speeding up, or storing more of them, not just in your thumb drive, but in your brain.

Both of these examples, the eye and the brain, are intended to show how information is stored without wasting bits on noise. So when we say that chaos is the opposite of information, we are speaking quantitatively, with specific algorithms in mind, evaluated by their ability to eliminate spatial noise and time-redundancy while conserving the information, in both time and space. This distinction between information and noise, between order and chaos, between design and chance, has been at the core of modern science and metaphysics for 25 centuries, ever since Aristotle and Democritus argued about the essential character of nature.

Local Information

Over the past 3 millennia, there have been many attempts to understand the essence of matter and biology. The Aristotelean camp argued that biology is the correct paradigm, and matter showed proclivities that were similar to biology, so for example, gravity is the "desire" of matter for other matter. In contrast, Democritean materialism argued that physics is the proper paradigm for biology, so for example, the attraction of moths to light is a receptor process accounted for by non-living atoms bombarding a moth's brain. Over the last 25 centuries, the two paradigms have approached each other, with physics adding more and more "fundamental forces" showing long range behavior (like biology), whereas biology has become more and more reductionist chemistry, attributing more and more behavior to deterministic conditioning involving chemical or physical (un-living) causes.

In the early 20th century, these two fields overshot each other, with physicists discovering spiritual world of quantum mechanics (QM) with its dualistic, long-range, non-materialist behavior, and biologists discovering the materialist biochemical world of organic machines. Many have commented on this strange phenomena of physicists sounding like Cartesian gnostics (e.g., QM theorist David Bohm), while biologists have become ardent reductionists (e.g. neo-Darwinist Stephen Jay Gould). Since physics has made a round trip, from organic to material to organic interpretations, whereas biology has made only a one-way trip, from organic to material interpretation, we can use physics to predict the future for biology.

That is, physics was initially dominated by Aristotle (up until the high middle ages), with its organic, or biological explanation of elements and their elementary attractions, only recently converting to reductionist Materialism. As great a physicist as Ernst Mach in 1850 held firmly to the conviction that atoms were a Democritean metaphysical inventions without any real existence (since they had never been seen), despite their use by chemists to explain integer combinations of molecular reactions. Not until James Clerk-Maxwell applied the atom theory to gases (~1880) and then Ludwig Boltzmann calculated the statistics (~1890), did the skeptical physicists believe in their real existence. And of course, soon after came the first pictures of atoms, thereby putting to rest Mach's reservations.

One of the crucial points in Democritean atomism, is that atoms are all there is, there is no intelligence guiding them, no external influence affecting them, no creator adding or subtracting them, rather purely contact forces of collisions in the void could account for all manner of physical and chemical behavior.  So the success of 19th century atomism, the concrete predictions of Boltzmann and Maxwell, were that all physics phenomena were local.

Elaborating on this atomic analogy to physics, we can consider the effect of long range forces on a bottle of air. If all the air in the bottle is magically put into linear motion, say, by igniting a mixture of hydrogen and oxygen, then the bottle will experience an unbalanced force on the inside, the bottle will spontaneously leap off the table. Soon however, the collisions with each other will turn this motion into random directions, and the mixture will only get hot without moving. The whole magic of heat engines is to find a method of extracting information, or work, from a gas before the regular order dissipates. We can only extract work if the system is left in an even more disordered state afterwards, or else we would be violating the 2nd law of thermodynamics that says information isn't free. In physics jargon, the opposite of information is disorder or "entropy" (S), and this 2nd law states "in an isolated system the entropy must either stay constant or increase". Boltzmann is credited for explaining disorder as a mathematical quantity of a gas that could be counted, and was so proud of it that he had this equation engraved on his tombstone, S=k ln Ω, where Ω = the number of permutations or countable "states" of the system, "ln" is the logarithm function, and "k" is the units conversion called "Boltzmann's constant".  Which is to say, there is only one way to organize all the socks in a drawer, but a thousand ways to decorate the floor with them, so that a measure of disorder, S, would say that your teenager's room will naturally tend toward socks on the floor, and can only be put back in order with work (or information) to reverse the process.

Let's give a countable example. Suppose you have 2 dice, and roll them repeatedly, writing down the sum, and then binning the result as: 2-3-4-5-6-7-8-9-10-11-12.  Of course, you can just put them into Excel and use the histogram chart tool, but either way you should find a "bell-shaped" curve, with the most rolls occurring for bin "7", and the fewest for "2" and "12", because there are six ways to roll a "7", 1-6, 2-5, 3-4, 4-3, 5-2, 6-1, but only one way to roll a "2", 1-1. From Boltzmann's equation above, the entropy S of bin "7" is k ln 6 = 1.8 k, whereas bin "12" has entropy 0 k, so the entropy is maximized in bin 7. This bell-shaped curve, or Gaussian, is characteristic of random processes that are "fair", that have no outside influences on them. Should you instead discover a triangle-shaped histogram with most rolls at "12", then chances are you were given loaded dice (possessing internal information).

Now suppose you had as many dice as there are air molecules in your room, what would be the average if you added together all their pips and divided by the number of dice?

You're way ahead me, 3.500000000000000000000000 = Answer. Why so many zeroes? Because there are so many air molecules in the room. The more dice I average, the "sharper" the bell-shaped histogram curve becomes. It becomes so needle sharp that we can even start calling it a law. Perversely, the more random particles that get averaged together, the more non-random and predictable the result becomes. So if instead the average comes out 3.50000000001, we would know there was information added to the system at the parts per trillion level. Likewise, even if the average were 3.5 but the shape of the histogram were more triangular than Gaussian, we would know there was information added to the system, because the "wings" of the histogram have lower entropy (higher information) than the center.

Returning to atomism, the success of Boltzmann in describing the behavior of a gas as atoms, is just a restatement that atoms behave chaotically like dice rolls, that they lack direction, purpose, or inherent information (merely using up whatever information is put into to their original ordering). Now there have been some modifications to this word "local" to take into account magnetic and electric fields, which were the discoveries grudgingly acknowledged to non-materialist Michael Faraday, but these modifications were sweetened for the materialist by the invocation of "virtual particles" that mediate these long range forces. (Gravity alone remains stubbornly resistant to such particle analysis, with several famous scientists concluding that gravity will never be incorporated into a unified, materialist, particle theory.)  So in simplified terms, information is equivalent to long-range order, and conversely, anti-information or randomness is equivalent to short-range interactions.

But the QM discovery of long-range correlations (Einstein's 1935 EPR paper) has been a source of metaphysical friction in the community. These "wings" on the histograms may not be noticed when there are millions of atoms and the peak is sharp, but looking at atoms one by one makes the correlations obvious. Some take it to mean that materialism fails, viewing it as support for a dualist (gnostic) reality. Others view it as a math trick with no metaphysical connection to reality. Still others throw away the metaphysics supposing that there are an infinite number of realities. But whatever interpretation is chosen, the one interpretation that cannot be true, is that of Einstein's "naive realism", or simple materialism. Thus the past 50 years has seen a resurgence of "organic" interpretations of physics, because empirical measurements have shown a long term order that cannot be due to Democritean particles, though such theories are careful to reproduce the same average predictions of materialist physics.

Darwinism as a Chaotic System

Now when Darwin suggested that random mutations combined with natural selection would provide a chaotic explanation for apparent design and order in biology, he was making a mathematical statement about long term correlations. He was saying that species behave like larger versions of atoms, without time-ordering (purpose), and without spatial-ordering (e.g. Gaia). The opposite Aristotelean viewpoint was often called "vitalism", that there existed some long term order, either attractive ("destiny"), or pre-designed (British Deism and Bergson's "impetus"). Despite Darwin's apparent waffling on the issue, modern neo-Darwinists (Ernst Mayr, E.O. Wilson, Stephen Jay Gould) argue that all such "progress" in evolution is illusory, that there is no long-range correlation evident despite the archaeological record of increasing complexity through time. That is, the appearance of progress is driven by random, local processes, much as frost flowers form on a window, without any information beyond local, undirected, e.g., diffusion-limited growth.

Let me say that another way, the materialist atom knows of only local interactions. By following such rules, long range order is possible (as in crystals and frost flowers), but such patterns are either predictable or accidental, merely the accumulation of many random steps. Neo-Darwinists argue that the order, which is visible in living things, is much like that of a crystal, a long-range spatial order slowly, and randomly, accumulated over time. Similarly the historical ordering of evolution itself, from simple to complex organisms, is likewise a consequence of short-time interactions (mutations), accumulated over space (species).  This biological process is said to parallel the physical process of "emergence", or the appearance of complicated, large-scale properties based on simpler, small-scale physical laws; for example, the 12 ice crystal structures, the low freezing point, the high boiling point, and the hydrologic cycle, all result from the lowly hydrogen bond between water molecules.  Emergence or no, the Darwinian hypothesis of mutually reinforcing space-time correlations makes it quite difficult for theorists to mathematically tease apart the two dependent correlation functions--the spatial and temporal--which is the step needed to ascertain their long-range order.

For example, if we ignore long-range order, and plug in a simple mathematical model of evolution as the accumulation of random mutation steps, the diffusion of information (or progress) has no "arrow of time", no "rectifier", no "ratchet" that accepts only progress and rejects regress. I'm just repeating myself of course, because rectifiers are one result of long-range order.  Making the task more difficult, is the apparent sparseness of the data, which are predicted to be smoother than observed.  That is, rather than finding many organisms spanning the reptilian to amphibian transition, or the mammalian to marsupial transition, we find non-Gaussian distributed clusters of species. This is true whether we consider their shape (morphology) or their genes, and often the genes give a different grouping than the morphology (a complication we'll ignore, though highly revealing). This is even true when we express the differences chronologically, that long periods of apparent stasis are separated by short periods of extremely rapid change, i.e. Gould's famous Punctuated Equilibrium hypothesis. If the present large differences between species were the accumulation of small, random steps, then there should be smooth transitions in time and space ("missing links"), which are rarely observed.  This "clumpy" character of the evolutionary data has been seen before in physics, where it is indicative of long-range correlations, such as 2nd order phase transitions, critical points, or "Levy-flight" transport.  But if physics is happy to describe these long-range forces as emerging from short-range, materialist causes, then can biologists also find a convincingly materialistic long-range ordering parameter that can explain this clumpy, non-Gaussian behavior, starting with random, local interactions?

The Local Interaction Dilemma

Physicists often deal with interleaved problems, which may sometimes be resolved by considering "timescales". The random collisions in atomic theory are fast, taking little time, whereas the information/work extracted from a gas is noticeably slower, taking much more time. Long term, cumulative correlations are smaller and slower in effect, and must be examined and averaged (accumulated) carefully. So to see the randomness of a Darwinian process, we need to count the number of states of the system at a short enough timescale to catch the uncorrelated, "fast" motion, find the Boltzmann entropy, negate it to find the information, and plot that information as a function of time, all the while being careful not to average over long intervals that would distort the accumulation (which presumably would produce the emergence of long-term order).

Following this prescription, many have argued that unlike frost flowers, evolution shows a real increase of information, or the decrease of entropy, that cannot be accounted for by local, random, atomistic processes. It would be as if the frost on the window had formed the words "Stopping by Woods", which would be an unaccountable event. This conundrum of how to spontaneously get information out of random walk has stumped some of the brightest minds in genetics. Cornell geneticist John Sanford wrote Genetic Entropy & the Mystery of the Genome (2005) arguing that none of the proposed rectifiers actually worked in practice, but much to his dismay, random mutation processes observed in the laboratory invariably produced a loss of information, an increase of entropy, as predicted by thermodynamics. And rather than extrapolate toward inevitable progress of the species, such laboratory results extrapolate toward extinction.

This is not to say that evolution is impossible, since it is undeniable that there has been an increase in complexity over time, but that if there be emergent order, it must occur at a larger temporal or spatial scale than the supposed gene/codon level explored by these experiments.  While many neo-Darwinists fear that going beyond point mutations is a reintroduction of anti-science "vitalism", the entire methodology of these experiments is materialist, so that rather than reintroducing a non-materialist vitalism, they merely redirect Darwinian theory away from a mystical dependence on emergent physics, at least, at the codon level.

Hierarchical Evolution

Indeed, the past 10 years has seen many biologists positing a long-range order in evolution, so for example, Gould writes in The Structure of Evolutionary Theory (2003) that "hierarchy" in the action of evolution may act on more than one unit or one individual simultaneously, but on the whole organism or clade. Even earlier, U of Manitoba embryologist Richard Gordon writes in The Heirarchical Genome (1997), that evolution of the "cell-state splitter" (the mechanism taking a fertilized egg and converting it into a multi-cellular functional organism through a cell splitting and differentiation) can convert small changes of development into huge transformations of adult organisms. Thus evolution of higher order controls will take many fewer changes to create new species than counting the functional differences. By analogy with computers, there are many fewer steps needed in a heapsort than in a substitution sort, or by analogy with the military, an army is mobilized by a chain of command, rather than by a general barking orders to privates.  Therefore, argues Gordon, evolution progresses with far fewer (and bigger) steps than realized, and therefore more speedily and with less information, by evolving the embryo which develops the cellular machinery, rather than the cellular machinery directly.

In a similar vein, proteins are found to have an ordered structure larger than merely the sequence of their peptides. Using the Chinese written characters as a set with similar long-range order, a 2008 Biologic Institute paper by Axe et al, looks at how mutations in the vectors that compose a Chinese character are similar to mutations that affect proteins. The point of Gould, Gordon and Axe papers, is to argue that information is hierarchical, global, possibly even fractal. That is, a mutation at one point in a protein or in an embryonic development has long-reaching implications that are not local. Therefore whether we consider organisms, embryology or protein folding, information has global consequences, with mathematically distinct effects.  Can we use this observation to give material atoms the ability to create long range order, can we find a rectifier for the blind, localized steps of evolution?

Many mathematical biologists think so. Michael Deem writes in Physics Today (Jan 2007) that hierarchical evolution can even accelerate as it evolves more efficient "gene transfer" processes. That is, he not only proposes that entire genes can be swapped around, but even higher functional units like chromosomes. Evolution can function, he argues, on larger and larger scales so that the more complex and large-scale the organism, the bigger the evolutionary step permitted, and the faster evolution progresses.  It is an interesting "bootstrap" suggestion (with the added tingle that bootstrapping is very close to Lamarckian "self-directed" evolution), but is there any evidence that it is happening?

So we have a problem with neo-Darwinian paradigm: the posited introduction of information at the codon level to explain the historical increase in biological complexity (combined with the rectifier of population genetics) was contradicted by eighty years of experiment.  Can we rediscover this rectifier, this long range order in a more hierarchical level, at perhaps the gene, or chromosome or even genome level? That is, if we cannot introduce information by swapping codons, (or peptides), can we find some process that swaps genes, swaps chromosomes, perhaps even entire organisms, can we find the mechanism that powers Gould's hierarchical evolution?

Jumping Genes

In 1948 a maverick geneticist Barbara McClintock proposed that genes could move around on the chromosome, and thereby change their expression or performance. After 35 years, in 1983, she finally received the Nobel Prize for her pioneering work on transpositions. But by itself, jumping genes do not provide the hierarchical evolution that Gould dreams of. Reorganizing the contents of your file cabinet make it easier to find things, but they don't create information that wasn't already there. Gould, Deem and Gordon need a way to introduce large chunks of new information into the genome, preferably in the form of genes or bundles of genes. Of course, this is what goes on in sexual reproduction, which, unlike grafting, cloning, parthenogenesis, bacterial multiplication and other forms of asexual reproduction, results in a brand-new mixture of genes. But if we consider the gene complement of the entire species, the genome, rather than the chromosomes of a single individual, sex is not much different than jumping genes, rearranging the file drawers but not providing any new information. What we are looking for is something like sexual reproduction, but between species.

We've already found it, at least in bacteria, called "bacterial conjugation". If a bacteria contains a special loop of extra DNA called a plasmid, which is separate and in addition to its own DNA, then the plasmid can cause the bacteria to form a transfer tube to another bacteria, not necessarily the same species, and transfer itself and perhaps some host DNA as well. In rhizobacteria, this process can transfer bacterial DNA into a plant cell nucleus, making cell products not native to the plant. We might ordinarily think of plasmids as a virus, but often their presence is beneficial to their host, bringing it antibiotic resistance, or other symbiotic benefits. Notice also that this conjugation event does more than infect a new host with plasmids, but it can also transfer genes between species of bacteria.  Now this extraneous loop of DNA is not living, but it replicates, which is very close to the definition of a virus. Thus viruses should be seen as more than sub-living parasites, but as machines designed for horizontal gene transfer between species. Viruses are precisely the mechanism we needed to produce hierarchical evolution, and it would appear to be a very efficient one, evolved for this purpose.

Should I have called them a mechanism evolved for this purpose? The study of the most abundant of these viruses, the bacteriophage, would reveal a machine designed for the single purpose of finding a bacterium and injecting DNA through its cell wall. Its a sub-sub-miniature, flying hypodermic needle: no extraneous parts, self-assembling, minimalist design, hugely efficient. In fact, a recent study of sea water suggested that there are 10 times as many phages as bacteria in the sea water, or roughly an equivalent amount of biological matter in viruses as in bacteria. Phages are only by size, a minor constituent of the biosphere.

Now we come to an important point. If Darwin's natural selection were tuned to multiply phages (at the expense of bacteria), they should be tremendously efficient, which they are in most respects except one.  That is,with lifecycles shorter than a bacteria, with a multiplication factor of 100's, with replication errors much higher than a bacterium, natural selection should be brutal and swift for these phages, making them ruthlessly efficient. Yet despite this evolutionary pressure, phages have enormous DNA variations, even containing DNA that has no useful purpose to the virus. And it is more than merely mutations of the viral replication genes, but entire genes for things that a virus has no need at all, such as chlorophyll. Therefore more is going on with phages than merely natural selection at the codon level, and this provides ample support for the hierarchical model for phages. But most importantly, this can all be done mathematically.

The ease with which phages can have be analyzed and their DNA transcribed has resulted in a large and growing database of transcribed phage DNA. This database can be searched for particular genes as well as processed digitally, much as photographs or music can be compressed. And the results are quite surprising.

A simple exercise one can do, is to attempt to compress the database represented by all these phage DNA transcriptions. Obviously, some of the genes are identical, so the database should compress nicely. But compression algorithms using a hash function (Lempel-Ziff-Huffman) capture the amount of information at the local level, since hash functions usually have a limited number of digits. But compression algorithms that use larger units (fractal compression) seem to work better on this data base. Evidently, the information in the genome is stored both locally and globally.  This finding alone undermines the contention that the neo-Darwinian "local mutation rate" is responsible for the information in the genome, for such a process would not be able to encode this global information. So it would appear that these databases are supporting Gould, Gordon and Deem's contention that the genome is hierarchical.

Furthermore, it was reported that all the genes for the smallest genome cyanobacteria capable of photosynthesizing, are found in chopped up form in the phage database. This suggests that even should a catastrophic event, say, a meteor impact that boiled off all the oceans, exterminate photosynthetic cyanobacteria, nevertheless, phages plus an ordinary bacterium, say, one that lived in rocks a 100 meters below the ocean bed, could collaborate in replenishing the oceans with photosynthetic bacteria.  Another example is a recent Nature paper on the flora and fauna of a hyper-saline lake, that were found to all share similar genes for coping with the high salt content. It would seem obvious that some horizontal transfer of "salt-tolerant" genes was helping all the denizens cope with extended drought conditions. Phages are the obvious suspects.

And what both of these examples demonstrate, is that "survival of the fittest" is not even true as a tautology. Because the survival in both cases has nothing to do with their own fitness, but the beneficial intervention of a virus or viruses! Only if we consider the collective genome of all the species in the pond, could we invoke a "survival of the fittest ecology" but not "survival of the fittest specie". In other words, the fittest ecology is the one with the most collaborative species, the one with the highest viral infection, the one with the worst immunities. If you are okay with this transmutation from species to ecologies, then ponder that Darwin introduced this tautology to explain "The Origin of Species", not "The Origin of Ecologies". The differences are profound. The ecology has no desire to allow cyanobacteria to mutate into palm trees, if anything, the survival of the ecology is the stasis of the species. Thus ecological natural selection has the opposite effect hypothesized by Darwin, of stabilizing the species against extinction events or mutations. Saying this one more way, mutations cause rising entropy, so it is natural to discover that life, which battles entropy from birth to death, also battles entropic mutations over longer cycles of drying ponds, changing climates and millennial ice ages.

But if all this effort goes into keeping the genome viable, keeping the ecology stable, how then can hierarchical evolution occur? Have we not contradicted our earlier supposition that phages are capable of rapidly modifying the flora and fauna of an ecosystem, and thereby speeding evolution?

The obvious answer is nevertheless the most unsatisfying. The phages bring in genes external to the ecology. So, for example, when the Salton Sea, formed by a flood on the Colorado River in 1905, started to go from fresh to salty, salt-water diatoms from the ocean 80 miles away, started growing, apparently blown in by the wind. And with these organisms came a fresh lot of DNA for the phages to start dissecting and transcribing.  In the same way, when the Earth's oceans had converted the reducing atmosphere of the Hadean into the oxidizing atmosphere of the Pre-Cambrian, suddenly new life forms appeared on the scene. Where did they come from, blown in from what extra-terrestrial coast?

This is where the recent discovery of fossil cyanobacteria on carbonaceous chondrites (meteorites thought to be extinct comets) is most illuminating. In our previous paper, we argued that infected comets could spread cyanobacteria not only throughout the solar system, but more importantly, from comet to comet, making the collection of infected comets around the Sun or any star, a biosphere of greater size and mass than that of Earth.  Such a cometary biosphere should have its own complement of phages, just as we have seen from seawater. However, unlike cyanobacteria, phage fossils would be indistinguishable from crystals, and to date we have had no evidence of them. Yet these invisible phages would be capable of transporting quite advanced genes throughout the biosphere, regardless of their origin, or the organism that finds them useful. "The Selfish Gene" was intended to draw attention away from Darwin's species to the apparently independent existence of genes, but now it appears that a better title for the book would have been, "The Altruistic Gene".

Therefore the thesis of this paper, is that the punctuated equilibria observed by Eldredge and Gould is not due to some long-range modulation of the point-mutation rate caused by geographically isolated communities, but rather by the sporadic transport of new genes through cometary transport. If the transport is relatively unlikely, then long periods of stasis will be separated by rapid information injections, mediated by phages. For example, the chitin protein that forms the outer coat of invertebrates need not have evolved on Earth, nor require the transport of trilobites by comet, but merely the arrival of an interplanetary comet carrying phages with the chitin gene. Earth-based phages would then insert this gene into many hosts where it could form the basis of a new species or genus. Multiple genes could be moved in this way, leading to the evolution of a new family, and so forth. We are speculating, of course, on the mechanics of hierarchical evolution, and really are only presenting a framework for the introduction of information into the empirically observed evolutionary process. (In point of metaphysics discussed later, the very concept of hierarchical evolution is problematical, since the hierarchy includes itself, which poses serious definitional paradoxes introduced by recursion.)

That is, we have converted the evolutionary history of the Earth's biosphere, the temporal progress of life on Earth, into a spatial series, by multiplying by the speed of comets. For example, the introduction of cyanobacteria in the Earth's oceans 3.8 Gy ago, converts via a 2 km/s cometary speed to a minimal sphere of cyanobacterially infected galaxy of 25,000 ly radius. Likewise the introduction of body-plans in the Cambrian explosion 530My ago converts to a minimal sphere of galactic bugs about 3,500 ly in radius. Applying then these temporal/spatial constraints to the evolutionary history of Earth, it is possible to represent evolution as a diffusive equilibrium of information from the galaxy. Then the sporadic or punctuated nature of evolution represents the "graininess" of the diffusive process, the density of information and its temporal transport.

If then the clumpiness of the data is not an artifact of collection, and cannot be attributed to unlikely events such as asteroidal impacts and sterilization of the planet, then it should be a result of the transport physics. That is, it may represent Levy-flight transport of low density information through the galaxy. The following model is an attempt to characterize this information transport process, and use a minimal set of adjustable parameters to match the data.

[Insert Model Here]

Philosophical Addendum

Does hierarchical evolution save neo-Darwinian theory? The immediate objection arises from the existence of the machinery in the first place. For if there exist machinery for accomplishing random swapping of genes, what is its origin? If the probability of a million monkeys creating Shakespeare by randomly moving around paragraphs of text in Microsoft Word is so much better than typing it out a letter at time on a typewriter, what then is the likelihood of the text and the tools, of Shakespeare + IBMPC versus Shakespeare + Royal? It would seem to me that the entropic cost of a PC far outweighs the entropic gain of moving around paragraph-sized chunks of text. Like Darwin's original theorizing, extrapolation is no excuse for avoiding quantitative estimates.

But there are more serious objections. Even supposing that truly clever paragraph processors can be made with low entropic cost, making it profitable for monkeys to leave their Royals behind, there remains a second metaphysical problem. For a hierarchical mechanism must also evolve itself, as Deem has argued, since the machinery of this gene swapper is also made by genes. This produces an evolutionary feedback far stronger than natural selection, for the organism can now directly change its progeny. An electrical analogy is an op-amp with feedback that is now capable of oscillatory behavior completely absent in a feed-forward networks. A computer analogy is the unpredictability of a Turing machine computation when using feedback. A psychology analogy is the irreducible concept of self-awareness.  Nor is it possible to save biology from this fate by finding some physical explanation that limits this feedback to a subset of the biology, for Gould has argued that hierarchical evolution would include more than individual organism units. Imagine, the microbes in a shrinking pond decide globally to evolve salt-tolerance, or the humans on the planet Earth decide globally to reduce carbon emissions. Thus this behavior Gould posits is experimentally indistinguishable from Aristotelean teleology, the very thing Gould is careful to exclude!

And this was the major conclusion of Goedel's incompleteness theorem, that the mere inclusion of recursion prevents the system from being described by simple axioms. It becomes impossible to exclude teleological statements from the logic system, thereby destroying Bertrand Russell's program of removing metaphysics by logic. That is, the mere inclusion of feedback permits teleology, if not demanding it (since co-operating self-determining evolution should be orders of magnitude more efficient than random evolution, and dominate the gene pool). Despite Darwin's best efforts, Lamarck lives again.

Nor is this teleology restricted to the organism or clade, for we have argued it extends to the entire ecology of a salt pond. And by extrapolation, to the entire ecology of the Earth, of the Solar System and of the Milky Way. The entire galaxy may be involved in the evolution of life on Earth, making the evolutionary paradigm indistinguishable from creation narratives.

So from a scientific standpoint, hierarchical evolution would solve a lot of conundrums, but from a philosophical standpoint, it would introduce a great many more. Just as Darwin's great contribution was a way to simultaneously solve the scientific and philosophical problems of materialism, so the future will salute the one who simultaneously solves the materialist and philosophical problems of science, or more precisely, the scientific and religious problems of creation.